Talk to the Veterans Crisis Line now
U.S. flag
An official website of the United States government

VA Health Systems Research

Go to the VA ORD website
Go to the QUERI website

1999 HSR&D Annual Meeting Abstracts

51. Determinants of House Staff Time Use in the Inpatient Setting

Timothy Dresselhaus MD. Center for the Study of Healthcare Provider Behavior, San Diego, CA. D Timberlake. Jeff Luck MBA, PhD. UCLA School of Public Health, Los Angeles, CA. B Wright and R Spragg. Samuel Bozzette MD, PhD, Center for the Study of Healthcare Provider Behavior, San Diego, CA.

Objectives: As patient care in the VA shifts to the ambulatory setting, the inpatient time of house staff will be constrained. To maintain the value of inpatient time, training programs must preserve activities contributing directly to patient care and education. Data on time allocation and the determinants of time use are needed to inform the redesign of inpatient training. We used automated random sampling to measure house staff inpatient work and developed prediction models to identify factors influencing house staff time allocation in the inpatient setting.

Methods: We developed and validated a computer assisted self interview (CASI) survey and assigned house staff rotating on our VA medical service hand-held computers for periods of five consecutive weekdays. House staff were prompted randomly within consecutive intervals to complete the work survey. Fifty-one medical house staff (30 interns, 21 residents) were sampled during 480 weekdays. We analyzed activities according to three content domains: direct patient care (time with patient), indirect patient care (time supporting patient care but not with patient), and education. In predicting the proportion of time allocated to these domains, the following independent variables were included in three multivariate models: gender, house staff type (intern or resident), program director ratings of overall clinical performance, attending ratings of clinical performance for the concurrent rotation (rotation rating), type of day in call schedule (long call, post-call, short call, pre-call), scheduled clinic day, and patient load. Two separate random-effects linear regression models were used to identify the significant predictors of the proportions of time that house staff participated in indirect patient care and education. For direct patient care, a logistic regression model incorporating robust variance estimates for clustered data was used to identify the predictors of days in which sampled activities included direct patient care versus days in which sampled activities did not include direct patient care.

Results: In the 480 days of observation, 3,762 complete responses were obtained. House staff spent more time in indirect (56%) than direct patient care (14%) or educational activities (45%). The proportion of time spent in indirect patient care was negatively associated with rotation rating (p<0.01) and long call (vs pre-call, p<0.01). Time spent in direct patient care was positively correlated to long call (vs pre-call, p<0.001) and short call (vs pre-call, p<0.01), but negatively correlated to scheduled clinic day (p<0.05). Time spent in education was negatively associated with female gender (p<0.05), short call (vs pre-call, p<0.01), and intern status (vs resident, p<0.01) but positively associated with rotation rating (p<0.05).

Conclusions: House staff at our VA spend more of their workday in activities indirectly supporting patient care than in time with patients or in educational activities. Elements of program structure correlate to time spent in direct patient care, while house staff factors, including ratings of clinical performance, correlate to the amount of time in educational or indirect patient care activities.

Impact: Planners must consider the influence of structural change upon house staff time allocation to patient care and education when redesigning house staff inpatient work. Further research is needed to understand how time use is related to the clinical performance of house staff.

52. Effects of Race, Income, and Psychological Well-being on HIV Patient Decisions about Emergency Department (ED) Use.

Allen Gifford, MD. San Diego VA Medical Center. San Diego, CA. R Collins and D Timberlake. Samuel Bozzette, MD, PhD, Center for the Study of Healthcare Provider Behavior, San Diego, CA. M Shapiro, M Schuter, and D Kanouse.

Objectives: Veterans with HIV often choose the ED for care of both non-urgent and urgent symptoms, however ED care is expensive, and quality of care may suffer due to poor integration of ED care with primary care. To understand what determines HIV patient decisions to go to the ED or to their usual source of primary care (USOC), we interviewed a probability sample of both veteran and non-veteran HIV patients in the U.S. to determine how they would seek care in response to several symptom scenarios.

Methods: A nationally representative probability sample of all U.S. HIV+ adults in care underwent detailed structured interviews. Three brief clinical scenarios describing headache, respiratory, and other HIV-related symptoms were asked of each subject. Different scenarios were asked of early stage (non-AIDS) and late stage (AIDS) patients. Scenario response options included "I would go to the emergency room" and "I would go to the doctor's office the same day." Sociodemographic (SD), access to care, health status, psychological, and knowledge/attitude/belief (KAB) variables were also assessed. Weights were used to adjust for the sampling design, and non-response; linearization methods corrected for weights and multistage sample design. Proportional logit models to identify independent predictors of propensity to seek ED care were estimated separately for early and late stage patients. A hypothesis-driven regression approach was used to evaluate the relative contributions of SD, access, health, and KAB variables to propensity to choose ED care (among those who would seek immediate care).

Results: Data were collected from 1245 non-AIDS and 1612 AIDS patients representing 86,800 and 126,420 HIV patients respectively nationwide. In a full multivariate model of those with AIDS, African-American race was the strongest independent predictor of propensity to use the ED (Adjusted odds ratio [AOR] compared with white 2.87, p<.0001), even with adjustment for all other SD, access, health, psychological, and KAB variables. Other independent predictors were low income (AOR 1.64, p<.04), low psychological well-being (AOR 1.08, p<.02), cognitive denial coping (AOR 1.01, p<.05), and several access (travel time to USOC, usually see the same person at USOC, length of time with same USOC provider) and attitude (prefer doctor to make decisions) variables. Among HIV+ patients without AIDS, African-American race was again the strongest independent predictor of propensity to use the ED (AOR 3.73, p<.0001), again with adjustment for all other variables; Hispanic ethnicity was also a predictor (AOR 2.17, p<.005), as was female gender (AOR 1.45, p<.03). Other predictors were again low income (AOR 2.00, p<.009), low psychological well-being (AOR 1.11, p<.02), and access (wait time for appointment date, wait time at appointment) variables. Health insurance category was not associated with propensity to use the ED in either model.

Conclusions: African-American race is a significant, independent predictor of propensity to use the ED across stages of disease, even after adjustment for access, income, insurance, and other variables. Income, psychological well-being, and access variables are also important.

Impact: Cultural preferences or habits that go beyond socioeconomic and access issues may influence propensity to use the ED among African-Americans.

53. Cost-effectiveness of the Primary Care Provider Program at the West Los Angeles VAMC

Hwai-Tai Lam, PhD. VA Greater Los Angeles Healthcare System, Los Angeles, CA. D Norman, S Cretin, G Sun, M Gee, and M Wong.

Objectives: Primary Care Provider Program (PCP) was implemented at the West Los Angeles VA Medical Center in 1996. This study evaluates the cost-effectiveness of that program. Patients' total medical costs and admission rates for a period of eleven months, from October 1997 through August 1998, were used as outcome measures to answer the following questions: (1) Does the coordination of a PCP reduce patients' overall medical costs? (2) Does the coordination of a PCP reduce patients' admission rates?

Methods: This study includes patients who have utilized the West Los Angeles health care system during the study period, and who had at least three outpatient visits. Patients referred from other VA facilities and psychiatric patients who did not utilize surgical or medical clinics were excluded. Cost data were downloaded from the Decision Support System (DSS). Patients were categorized into three groups based on their contact with their primary care provider teams. Group one (N=5,444) was patients who were seen by their PCP teams in more than 50% of their primary care clinic visits. Group two (N=5,452) was patients seen by their PCP team in less than 50% of their primary care visits. Group three patients (N=15,081) had no primary care clinic visits during the study period. A linear regression model controlling for risk factors assessed the effect of contacting with primary care provider teams on patients' overall medical costs.

Results: Group one patients, who had the most contact with their PCP, had the lowest average total cost ($8,298) per patient when compared to Group two ($9,470) and Group three ($9,451). Group one patients also had the lowest inpatient cost ($2,731); Group two and Group three's costs were $4,022 and $6,341, respectively. It had an admission rate of 0.27 while Group two and Group three's rates were 0.34, and an average length of hospital stay of 5.64 days (Group two: 7.41 days, Group three: 6.69 days). However, Group one patients had the highest outpatient costs of $5,567 (Group two: $5,447, Group three: $3,110). Although Group two patients had the highest average total costs, after controlling for age and principal diagnoses, the linear regression model showed that Group two's costs were lower than Group three's (p=0.0001), and Group one's cost were lowest of all.

Conclusions: Use of a PCP team had a significant effect on reducing patients' total costs and length of hospital stay. The effect would be greater if patients were seen at the primary care clinics by their assigned PCPs or by other providers of the same team. The setting of services provided by PCPs seemed to shift from inpatient to outpatient.

Impact: Currently, at the West LA VAMC, the PCP program and the patient self-coordinated system coexist. Patients can have access to sub-specialty care for a long period of time without going through the PCP program. The results of this study attest to the cost-effectiveness of a PCP program and support its reinforcement.

54. The Impact of Involving LPNs in Colorectal Cancer Screening

Nancy Thompson, PhD. University of Iowa, Iowa City, IA. Michael Chapko, PhD. VA Puget Sound Health Care System, Seattle, WA.

Objectives: The objective was to determine if involving nurses in the delivery of fecal occult blood tests (FOBT) for early detection of colorectal cancer would improve the utilization of FOBTs without negatively impacting the patient return rate in a VA primary care clinic.

Methods: The study was conducted between January and March 1998. The intervention involved having licensed practical nurses (LPNs) assess need for and order FOBTs during the time they interacted with 50-69 year old patients scheduled to see primary care providers in the General Internal Medicine Clinic at the Seattle VAMC. The orders were to be made on the basis of a predefined protocol. As the clinic is organized as a firm, it was randomly divided into experimental and control units. In the experimental unit FOBT orders were placed by the primary care providers(physicians/nurse practitioners MD/NP)or the LPNs. In the control unit FOBT orders were placed by only the primary care providers. Data were collected regarding orders, and returns as well as other characteristics of both the patients and providers. A 90-day window of time was allowed to elapse prior to categorizing a test as "returned" or "not returned." Descriptive and regression analyses were conducted.

Results: A total of 1109 patients were included in the study (545 experimental, 564 control). The experimental and control units were found to be comparable in terms of the variables of interest. The experimental unit ordered significantly more FOBT tests than the control unit for all patients (52 vs 15%, p<.000) as well as for FOBT eligible patients (72 vs 19%, p<.000).

Within the experimental unit, the LPN orders were returned as frequently as the MD/NP orders (LPN 42% vs MD/NP 59%, p =.0951)and were less likely to be placed inappropriately (LPN 6% vs MD/NP 31%, p <.000. Finally, the LPNs' involvement in ordering FOBTs did not decrease their ability to provide other expected services.

Conclusions: The use of LPNs in ordering FOBTs significantly increased the rates of ordering without compromising the potential for its return rate or other expected LPN services. Efforts to improve the delivery of preventive services should consider changes to the system that support use of nurses in delivering preventive services.

Impact: This study has demonstrated that LPNs ordering FOBTs in a VA medical facility can produce nearly a threefold increase in the number of veterans receiving and returning the test with minimal cost.

55. A Standardized Method for Assessing Urgency Among Walk-in VA Patients

Donna Washington, MD, MPH. Assistant Professor of Medicine, UCLA, Los Angeles, CA. Paul Shekelle, MD. West LA VA Medical Center, Los Angeles, CA. CD Stevens and RH Brook.

Objectives: Faced with the challenge of delivering maximum medical value within a fixed budget, the Department of Veteran Affairs has begun adopting managed care principles. Both cost and quality concerns dictate the need to shift from unscheduled, episodic care in hospital emergency departments to a strengthened primary care system. However, few systematic methods exist for identifying walk-in patients who are safe for triage to primary care settings. We developed, operationalized and validated a method to direct walk-in VA patients to emergency, urgent, or deferred (by up to 1 week) care.

Methods: Using the scientific literature and expert opinion, a 17-member multidisciplinary physician panel rated the safety of deferring care for 365 detailed clinical scenarios representing common complaints (abdominal pain, musculoskeletal pain, and respiratory infection symptoms) of ambulatory adults presenting to emergency departments. We converted ratings into explicit, standardized triage algorithms for use by emergency department nurses. Following training in their use, nurses at the West Los Angeles VAMC applied the algorithms to 1,187 consecutive walk-in patients who had conditions covered by the guidelines. Patients meeting deferred care criteria were offered the option of an appointment within 1 week in the ambulatory care clinic. All other patients received same day care in the emergency department or urgent care clinic at the study site. We measured nurse triage times using our guidelines in comparison to usual implicit judgment methods. We also assessed non-elective hospitalizations at all southern California and Nevada VA facilities within 7 days of triage and 30-day mortality using the VA National Health Exchange and Beneficiary Identification Records Locator System databases respectively.

Results: The mean nurse triage time using the criteria was 9.95 minutes (95% C.I. 8.1 to 11.8) in contrast to 9.52 minutes (95% C.I., 7.9 to 11.2) without the criteria. 226 (19%) patients met deferred care criteria, and of this group, 154 (68%) had their care deferred by up to one week. Transportation difficulties represented the most common reason for declining a deferred care appointment. Patients meeting deferred care criteria experienced zero (95% C.I., 0% to 1.2%) related non-elective VA hospitalizations within 7 days of evaluation, and none died within 30 days. By contrast, 7% (95% C.I., 5.5 to 8.9%) of patients who did not meet deferred care criteria were hospitalized non-electively for related conditions, and five (0.5%) died.

Conclusions: We developed standardized, clinically-detailed triage guidelines for deferring care that apply to a significant proportion of patients with common ambulatory conditions. Trained nurses applied the criteria to a large group of patients without adverse impact on triage time or work flow. No patient classified as safe for deferred care using the guidelines was hospitalized within a week at a VA facility or died within 30 days.

Impact: This explicit triage approach may allow facilities to manage their acute care resources more efficiently by safely diverting a significant proportion of walk-in patients from the emergency department to primary care settings. The safety and reliability of achieving this goal with the implicit triage methods currently used at most facilities is unknown.

56. Determining the Cost of VA Health Care

Paul Barnett, PhD. Center for Healthcare Evaluation, Menlo Park, CA.

Objectives: Cost data are needed to conduct cost-effectiveness analysis and to increase the policy relevance of health services research. Since VA does not routinely prepare patient bills, it is difficult to determine the cost of VA provided health care. This workshop will introduce the data sources and methods needed to determine the cost of VA health care for use in health services research. Cost finding will be illustrated from examples drawn from health services research projects and clinical trials conducted by the Cooperative Studies Program.

Target: The workshop is intended for health services researchers and VA managers with research problems requiring cost data. Attendees need not have a health economics background.

Methods: Methods for estimating VA costs will be described, including micro-cost and average-cost methods. Micro-cost methods include time activity analysis and the use of detailed VA utilization data and cost estimates from non-VA settings, either by preparing a pseudo-bill, or by estimating a clinical cost function from non-VA data. Average cost methods combine VA cost and utilization data, and resource weights obtained from the non-VA sector. Data sources that are useful for cost-determination will be described. VA data sources include the Cost Distribution Report, the Financial Management System, Fee Basis Files, centralized VA utilization data bases such as the Patient Treatment File and National Ambulatory Care Database, and decentralized utilization data from the VISTA clinical data system. The workshop will also describe extraction of cost data from the VA Decision Support System. Non-VA data sources, including Medicare DRG weights and reimbursement schedules, will also be described. Cost determination methods will be illustrated by example. Examples will include a health services study of costs incurred by patients treated for substance abuse disorders, and clinical trials of interventions in cardiology and primary care. The workshop presenters will lead a discussion of the advantages and drawbacks of the different methods, along with suggestions for the methods appropriate to different types of research.

57. Using VHA Administrative Data to Measure Performance: Methodological Issues of HEDIS

Ann Hendricks, PhD, Debra Jones, PhD, Cheryl Hankin, PhD, James Rothendler, MD, Bei-Hung Chang, ScD. Bedford VA Medical Center, Bedford, MA. Catherine Comstock, MPH. Carlisle, MA. Amy Rosen, PhD and Mark Prashker, MD, MPH. Center for Health Quality, Outcomes, and Economic Research, Bedford, MA.

Objectives: The primary goal of this workshop is to explore the methodological issues faced by managers and researchers trying to construct standardized VISN-level performance measures. The topic is timely and consistent with the conference theme (research at the interface) because it relates to policymakers' attempts to monitor performance in the delivery of health care. Activities: The workshop presenters include researchers and clinicians who will illustrate the issues using experience with six HEDIS measures applied to VHA and then lead discussions regarding possible solutions for research and management. Specific aims are to give a general overview of moving from theory to practice in applying HEDIS measures in VHA and to discuss five methodological areas and our methods in resolving issues: 1) definitions and construction of denominators and numerators using different VA databases (illustrated by rates of eye exams for people with diabetes), 2) defining levels of severity for psychiatric illnesses (illustrated for rates of follow-up care and readmissions for people hospitalized for mental illness), 3) risk adjusting for underlying prevalence of disease in the denominator population (illustrated for rates of cardiac procedures), 4) comparisons across providers (Bayesian techniques for VISN comparisons), and 5) veterans use of VA and Medicare. Materials based on the project will be disseminated.

Audience: Health services researchers with an interest in either developing or evaluating VA performance measures, and medical and mental health care managers who use such measures in assessing VISN-level or facility-level performance.

58. Implementing Ambulatory Care Case-Mix Measures in the VA: From Theory to Practice Amy Rosen, PhD, Arlene Ash, PhD, James Rothendler, MD,and Susan Loveland, MAT, Bedford VA Medical Center, Bedford, MA.

Objectives: Ambulatory care case-mix measures are currently being used by health care organizations to describe the illness burden of their populations, evaluate the content of ambulatory practice, assess the costs and quality of care, and predict resource utilization. As the VA evolves into an integrated health care delivery system, methods that take account of the unique issues related to ambulatory care and that can classify patients into clinically homogeneous groupings are critical for accurately assessing VA's effectiveness as a managed care organization.

Activities: This workshop will discuss the theoretical and operational frameworks useful in constructing population-based ambulatory care case-mix measures with VA administrative data. We will lead a series of presentations that will describe: 1) the importance of using case-mix measures to evaluate health care delivery in VA; 2) the leading case-mix measurement systems, Ambulatory Care Groups (ACGs) and Diagnostic Cost Groups (DCGs) that are currently being used; 3) the data and file requirements necessary for implementing the ACG and DCG software; and 4) the special considerations in constructing input and output measures using VA inpatient and outpatient data. Examples of how specific outcome measures were constructed from VA files will be presented by the speakers; however, as this is an interactive workshop, participants are encouraged to bring examples using ambulatory care case-mix measures from their own research studies.

Target Audience: This workshop is intended for researchers interested in using administrative databases to evaluate changes in health care delivery using case-mix measurement systems, as well as managers and clinicians who need to use the results of these analyses for evaluating health care delivery.

Audience's Assumed Familiarity with the Subject: Some familiarity with the subject of risk adjustment as well as the different VA databases would be helpful but is not required.

59. Are Clinical Practice Guidelines Implemented in Clinical Practice?

Mark Bauer, MD. Mental Health and Behavioral Sciences Service, Providence, RI.

Objectives: Clinical practice guidelines have been proposed to fill many roles from improving quality of care to reducing health care costs. Several major efforts have been undertaken to develop guidelines for mental health. However, we have little information regarding the extent to which guidelines are actually implemented in medical, surgical, or mental health practice. This paper reviews the available evidence regarding guideline implementation in medical, surgical, and mental health general clinical practice settings.

Methods: All available articles on adherence to guidelines, as defined by IOM, were located through Medline search supplemented by review of bibliographies of located articles. Abstracts and articles were then reviewed and the study categorized as positive or negative according to the authors' principle interpretation.

Results: This review found 36 articles distributed across three types of studies: cross-sectional, pre/post, and intervention trials. 33% of cross-sectional and pre/post trials and 67% of intervention trials demonstrate adherence to practice guidelines, for an overall adherence rate of approximately 45%. Additional studies indicate that when intervention trials finish, rates of adherence return to baseline.

Conclusions: Thus the substantial expenditure of effort among leading academics and major professional organizations may actually be having little effect on > general clinical practice.

Impact: The issue of how to facilitate the adoption of guidelines in general clinical practice must become a focus of research; otherwise we run the risk of putting tremendous effort into documents that will have primarily archival value, and negligible public health impact. In addition, some data indicate that there may be value in investigating this issue from the perspective of "diffusion research." This decades-old multidisciplinary approach has been successfully applied to understanding the adoption of new technologies in many fields, but has been applied to health care delivery issues only sporadically.

HSR&D Funded: DEV 97-015

60. VA Cooperative Study ##430: Reducing the Efficacy-Effectiveness Gap in Bipolar Disorder

Mark Bauer MD. Mental Health and Behavioral Sciences Service, Providence, RI. E Dawson, N Shea, L McBride, and WO Williford.

Objectives: The VA Cooperative Study has funded a 12-site randomized controlled trial of an easy-access program for bipolar disorder from 1997-2003. This study proposes that increased access and provider and patient education will reduce the "efficacy-effectiveness gap" for bipolar disorder. We specifically hypothesize that compared to usual VA care (UVAC) the easy-access Bipolar Disorders Program (BDP), will improve clinical, functional, and economic outcome.

Methods: Patients who are admitted to a acute psychiatric ward with a primary or co-primary admission diagnosis of bipolar disorder type I or II are screened for the study. One hundred and ninety-one patients are being randomized to each of the two treatment groups (BDP and UVAC). All enrolled patients are followed up for three years.

Results: All 12 sites have been trained to criterion and are fully functioning. Statistics regarding training to criterion are presented. As of October 15, 1998, 996 patients have been screened and 124 patients have been randomized (67 BDP and 57 UVAC), which is 107% of the current randomization goal. Follow-up data flow is also presented with quality of follow-up evaluation.

Conclusions: Complex treatment interventions can be implemented validly across multiple sites. Monitors based on "CQI" principles can provide important process information. Reliability of outcome data can be assured by training and ongoing monitoring.

Impact: In terms of impact, this study will have impact on both the private and the government healthcare sectors. It is designed to evaluate the basic principle that augmenting ambulatory access for major mental illness will improve outcome and reduce overall treatment costs. If results are positive, this study will provide reason to reconsider the prevailing trend toward limitation of ambulatory service that is characteristic of many managed care systems today. Further, the study will provide specific direction with regard to how to structure such ambulatory services.

HSR&D Funded: CSP ##430

61. Consistent Satisfaction Measured by American Board of Internal Medicine Patient Satisfaction Questionnaire Following Initial Evaluation of Inpatients by VA Medical Center Interns

MJ Bittner, MD. Omaha VA Medical Center, Omaha, NE. EC Rich, RL Recker, PD Turner, and MW Lubeley.

Objectives: As academic medical centers face more competition and as inpatient care in VA medical centers evolves, concern heightens about maintaining patient satisfaction. Among Lexington VA inpatients with chronic obstructive pulmonary disease studied January-May 1995, greater intern workloads were associated with diminished patient satisfaction. (Acad Med 1998;73:427-429) We looked for this finding among patients with a variety of diagnoses, studying patients shortly after admission.

Methods: We studied patients of Creighton medical interns assigned to non-intensive care unit inpatient floors at the Omaha VA Medical Center July 6-October 12, using the American Board of Internal Medicine's Patient Satisfaction Questionnaire (PSQ). This ten-item instrument assesses physicians' interpersonal skills. Each item has five responses, from 1 (excellent) to 5 (poor). We analyzed responses from patients who completed the questionnaire in a face-to-face interview the morning after admission; however, for patients admitted after 5 pm, the interview occurred the second day after admission. At the interview we noted these variables based on the Lexington study: patient age, intern gender, intern census size, and severity of illness (modified APACHE II score). Additionally, we noted the number of admissions to the intern in the previous 24 hours and whether the admission occurred after inaugurating stricter utilization review (August 24). Using multiple linear regression, we studied mean PSQ response as the dependent variable, assessing the effect of intern workload (census, admissions) and controlling for the effect of patient factors (age, illness severity), intern gender, and utilization review.

Results: Of 254 assigned patients, we excluded those initially admitted to intensive care (48), unavailable for interview (11), not competent (12), or not completing all items (59). Because some patients were excluded for multiple reasons, 182 remained for analysis. Satisfaction scores obtained shortly after admission (mean 1.79, standard deviation 0.74) resembled those from Lexington patients just before discharge (1.7, s.d. 0.9). Fewer of our patients (19.2%) rated all ten items as excellent (Lexington, 37%). The mean age was 65.8 (s.d. 13.9), modified APACHE II score 7.27 (s.d. 3.53), intern census 3.23 (s.d. 1.84), and intern admissions 0.77 (s.d. 1.03). Of the admissions, 37.4% were to women interns and 42.9% occurred after instituting stricter review criteria. Only 2.7% of the admissions involved interns with more than three admissions in the 24-hour period before the questionnaire administration. In the multiple linear regression model, neither census, admissions, nor any of the other explanatory variables were significantly associated with PSQ response.

Conclusions: Greater workload adversely affected satisfaction among Lexington COPD patients, and greater age was associated with more satisfaction among Lexington patients. In a setting with a limited number of intern admissions and limited intern census, we observed neither finding; nor did we find an association between satisfaction and intern gender, modified APACHE II score, or institution of stricter utilization review.

Impact: An academic VA medical center can achieve a consistently high level of patient satisfaction with physician interpersonal and communication skills, even among inexperienced house officers, in an environment where intern workload is controlled.

62. The Effect of Depression on Change in Stroke Patients' Physical and Mental Health

Hayden Bosworth PhD and Ronnie Horner PhD. Center for Health Services Research in Primary Care, Durham, NC. David B. Matchar

Objectives: This study examined the effect of depressive symptoms on change in acute ischemic stroke patients' physical and mental health using a subset of data from the VA Acute Stroke (VASt) study, a nationwide prospective cohort of 881 patients admitted for acute stroke at any of nine VA sites between 4/1/95 and 3/31/97.

Methods: Patients were interviewed at 1, 6, and 12 months post admission via telephone regarding depressive symptoms (short form CES-D), activities of daily living (ADL), and preference for current health state (measured by the time-tradeoff method). Analyses were confined to the 246 patients (M=66 years of age; SD=10 years) who had three measurements and complete data (28%). Proxies (n=451) were not included because they cannot provide valid responses regarding patients' psychological health. Besides the proxies, there were 184 (21%) had incomplete data at baseline. Patients who were able to participate were on average younger, had more education, were less likely to be married, and have less physical problems in contrast to those patients that required a proxy to compete the questionnaire for them or had missing data.

Results: A repeated ANOVA controlling for age, gender, level of education, race, and ADL at baseline indicated that increased time was positively associated with increased patients' perceptions of their current health state over the 12-months post hospital admissions (F[2, 204]=3.13, p<.04). After adjusting for related factors, number of depressive symptoms

(F[1, 204]=10.89, p<.001) was inversely related to patients' perception of their current health state at 1, 6, and 12-months post-stroke (i.e., willingness to trade more years of life to live in perfect health), despite improvements in physical function. In a similar model that examined change in physical function, increased number of depressive symptoms was also associated with worse function (F[1, 238]=9.88; p<.002). There were no significant interactions between depressive symptoms and any independent variable for either model.

Conclusions: It has been suggested that depression may be a reactive process to the physical limitation effects of stroke. We found that depressive symptoms continue to be independently related to both short (1 month) and longer-term (12-months) perceptions of quality of life in this

sample of veteran stroke patients, despite increased improvement in physical function. Presence of increased depressive symptoms was also independently associated with both short and longer- term physical function. Further studies are necessary to determine if the effect of depression is even more profound in more debilitated stroke patients.

Impact: Depressive symptoms aggravate veterans' perceptions of quality of life, thereby possibly reducing the effects of the recovery process and inhibiting the process and the benefits of rehabilitation. Health care providers may need to focus more attention on the mental health needs of stroke patients.

HSR&D Funded: SDR 93-003 "Clinical Management of Patients with Stroke at VAMCs

63. Implementing Depression Screening in Primary Care: An Effective Strategy?

Edmund Chaney PhD, Nicole Hasenberg MPH, and Susan Hedrick, PhD, VA Puget Sound Health Care System, Seattle, WA .

Objective: Although increasingly recommended or mandated in quality improvement efforts, the utility of screening for depression in primary care remains controversial. This study addresses one critical issue on which there is little data: what is the actual gain in detection of unrecognized, untreated major depression for which the patient is at least amenable to discussing treatment alternatives?

Methods: Preliminary data are drawn from a four-month period in which patients responded to a comprehensive primary care clinic-wide health promotion survey as they checked in for their appointment. Clinical records for patients screening positive on the two survey depression items were examined to determine whether patients were being actively treated for depression in Mental Health Clinic. Those who were not were contacted by telephone and, if willing, were administered a structured computer-assisted interview based on the Prime-MD by a trained evaluator. Those positive for major depression or dysthymia were asked whether they were interested in treatment (and if interested, that information was conveyed to their primary care provider.)

Results: The General Internal Medicine primary care clinic of the Seattle Division of the VA Puget Sound Health Care System in which the survey was conducted currently has an enrollment of 9100 patients and averages 850 visits per week. Patients are predominantly male (95%), Caucasian (90%), elderly (mean age 65), and not working (86%). During the four- month period, 837 patients were positive on the depression screen and 508 have been reviewed to date. Chart review found 45% to be in mental health treatment, either in the primary care setting, or in specialty care. After interview we found that 15% met MDD criteria and 2% had a mood disorder secondary to alcohol abuse. Of the rest, 11% could not be contacted, 2% could not be interviewed because of hearing or cognition, 12% declined to be interviewed and 3% we are still attempting to reach. Of those interviewed, 10% did not meet depression criteria. Of these currently untreated patients, almost all of the former group (and some of the latter) were amenable to discussing treatment alternatives.

Conclusions: In this population, a high percentage of patients screened were either already in treatment or met depression criteria, suggesting a short screen is an effective method of identifying depression. However, following up on those patients not in treatment requires a significant investment of staff time, even using computer-assisted technology.

Impacts: This study provides information on one component of depression treatment guidelines implementation in primary care which can be used by planners in deciding how to distribute available resources.

64. Coronary Artery Stent Outcomes in Department of Veterans Affairs Medical Centers

Michael Chapko, PhD and Nathan Every, MD, MPH. VA Puget Sound Health Care System. Seattle, WA. C Maynard and JL Ritchie

Objectives: Randomized trials of coronary stents versus conventional balloon percutaneous transluminal coronary angioplasty (PTCA) have demonstrated improved short and long-term outcomes for selected patients receiving stents. PTCA caseloads in Department of Veterans Affairs (VA) medical centers have increased from about 5,000 per year in fiscal year (FY) 1993 to 6,000 per year in FY 97 In FY 96, the first year the ICD-9 stent code was used, there were 3895 conventional PTCA and 2573 stent cases (40%). In FY 97, the number of stent cases increased 39% to 3588 and accounted for 58% of all PTCAs. The purpose of this study is to compare outcomes in patients with stents with outcomes in patients with conventional PTCA in the VA in FY 97.

Methods: Data on the outcomes of conventional PTCA and PTCA with stents were obtained for all VA medical facilities from the Austin Automation Center. For patients with more than 1 procedure, only the index case is considered, and results are reported separately for men with and without the primary diagnosis of acute miocardial infarction (AMI).

Results: For both AMI and no AMI patients, the stent (N: AMI = 607; No AMI = 2389) and no stent (N: AMI = 485; No AMI = 1999) groups were comparable with respect to age, race, and medical histories including diabetes, hypertension, myocardial infarction, and chronic heart failure. For both AMI and no AMI patients there was no significant difference in mortality rates (< 11 days after admission) between patients receiving (AMI = 2.8%; No AMI = .8%) and not receiving stents (AMI = 3.3%; No AMI = 1.2%). However, patients receiving stents (AMI = 0.2%; No AMI = 0.8%) had significantly lower rates of coronary artery bypass graft surgery (CABG) compared to patients who did not receive stents (AMI = 2.1%; No AMI = 3.0%). This association did not change after multivariate adjustment. In VA medical centers, the use of stents has increased, and the rate of same admission CABG has decreased.

Conclusions: The observational data from the VA with regard to the reduction in CABG rates is consistent with previous clinical trails that demonstrated the positive outcomes from the use of stents.

Impact: This study has the potential to encourage the continued use of stents and thereby reduce CABG rates within the VA.

HSR&D Funded: IIR 94-044

65. Symptom-Based Predictors of a Ten-Year Course of Treated Depression

Ruth Cronkite, PhD and Rudolf Moos, PhD. Center for Health Care Evaluation, Menlo Park, CA

Objectives: The primary objective of this research is to predict the likelihood of a long-term chronic course of depression from a set of risk factors that reflect three symptom domains associated with poor outcome: severity of specific depressive symptoms, lack of self-confidence, and a tendency toward social isolation and avoidance coping. We also examine the intensity of the index episode of treatment as a moderator of the influence of baseline risk factors on long-term outcome.

Methods: In a sample of 313 unipolar depressed patients, 20 potential symptom-based risk factors were assessed at treatment intake in the following domains: depressive symptoms, self-concept, and social functioning and coping. These patients were followed for ten years and were categorized into two groups: (1) those who were considered to have followed a chronic course, and (2) those who were considered to have followed a course of remission or partial remission.Chi-square analyses were used to select a set of baseline risk factors that were most strongly associated with a chronic course. Scores reflecting the number of risk factors present at intake were used to identify the relationship between the number of risk factors and chronic course. Logistic regression analysis was used to examine the influence of the amount of treatment on the association between the number of risk factors and long-term chronic course.

Results: The prototypic chronically depressed patient was an individual who at baseline experienced more severe symptoms of fatigue, loss of interest in usual activities, trouble sleeping, and thoughts about death or suicide; was not calm, successful, or self-confident; did not socialize with friends outside the home, and frequently coped with stressors by avoiding other people. A larger number of risk factors was associated with a higher likelihood of experiencing a chronic course. High-risk patients who received more psychological treatment during the index episode were more likely to experience a long-term course of remission or partial remission.

Conclusion: Symptoms that change more slowly during treatment and that are more closely tied to behavioral concomitants of depressed mood may be better predictors of outcome than depressed mood itself. Frequent thoughts of death and suicidal ideation are especially serious symptoms and are strong predictors of chronicity. Lack of self-confidence and social withdrawal are also important predictors of long-term chronic course and point to the heterogeneity of the mechanisms that influence chronicity.

Impact: Information about specific risk factors may help to direct treatment efforts and point to the need for continuing care. When patients report ongoing severe symptoms that predict a chronic course, a longer and more intensive course of treatment may be indicated. When the patient has a low self-concept, is socially isolated, and tries to manage stressors by avoiding people, these issues should be specifically addressed in treatment. By alerting the clinician to the patient's need for additional care, feedback obtained from a risk factor index can help to address the pervasive under- treatment of depression.

66. Family Satisfaction with VA Hospice Care

Elaine Czarnowski, RN, Elaine Hickey, RN, MSN, Cheryl Hankin, PhD. Bedford VA Medical Center, Bedford, MA. J Anderson. Carol VanDeusen Lukas, EdD. Boston VA Medical Center, Boston, MA. L Leonard

Objective: VA's emphasis on patient-centered care requires the evaluation of satisfaction with services. We present data regarding family satisfaction with hospice care as a component of a Congressionally mandated study to evaluate VA hospice programs.

Methods: As part of The Veterans Hospice Care Study, The National Hospice Organization (NHO) Family Satisfaction Survey was completed by a random sample of 528 family members of recently deceased veterans who received VA hospice care. Family members were asked to rate levels of satisfaction with eight aspects of hospice care. These aspects were: 1) satisfaction with the patient's pain control after hospice admission; 2) satisfaction with other symptom control; 3) quality of family and patient education; 4) responsiveness of the interdisciplinary care team to patient and family needs; 5) staff assistance in managing patient and family stress and anxiety; 6) staff support of patient and family spiritual needs; 7) perceived timeliness of patient's referral to hospice; 8) staff efforts to support the patient's quality of life. Levels of satisfaction were measured on a scale of 1 (very dissatisfied) to 5 (very satisfied), so that higher mean scores (closer to 5.0) correspond to greater levels of satisfaction. VA survey findings were compared with responses from 17,510 family members from private (non-VA) hospices. In addition, family members of veterans were asked to identify what they believed were the strengths and barriers of VA hospice programs.

Results: Both VA mean satisfaction scores and NHO reported scores ranged from satisfied (4.0) to very satisfied (5.0). In comparison to the NHO results, the levels of satisfaction reported by all VA hospice patients were similar but slightly lower than NHO scores. (VA mean satisfaction scores ranged from 95% to 100% of the NHO mean scores.) The most frequently cited strength of VA hospice programs was the commitment and skill of VA hospice staff. Family members specifically commented on staff's willingness to individualize care to meet the changing needs of patients and families. The most frequently cited barrier to effective hospice care was a perceived delay in veterans' referrals to hospice. Some family members stated that they believed the veteran would have benefited from hospice services earlier during his or her disease.

Conclusions: Study results show that overall, family members of VA hospice patients were satisfied to highly satisfied with the care that veterans received. Levels of satisfaction with VA hospice care were comparable to those reported in the NHO survey of non-VA hospices. The average length of stay in hospice for the veterans in this sample was 65 days compared to 58 days for NHO patients. Although veterans' length of stay was longer than those reported for non-VA hospice patients, some family members would have preferred earlier referrals.

Statements: Satisfaction with care is an essential part of any program evaluation. High levels of satisfaction among family members of veterans who received hospice care are an indicator of the effectiveness of VA hospice programs. Family comments indicate that care must be taken to assure timely referral to hospice.

HSR&D Funded: MMR 97-004

67. Mental Health Services Delivery in the Department of Veterans Affairs: Treatment in Psychiatric, Primary Care, and Specialty Medical Settings

Benjamin Druss, MD, MPH and Robert Rosenheck, MD. Northwest Program Evaluation Center, West Haven, CT.

Objectives: Epidemiological surveys suggest that half of patients with mental disorders in the US are treated in general medical settings. This paper examines delivery of mental health services in psychiatric, primary care, and specialty medical clinics in the Department of Veterans Affairs (VA), the largest public sector health care system in the US.

Methods: Using national VA encounter files, we examined all outpatient visits to the VA during an 18-month period: October 1996-March 1998. During this time, VA policy promoted a general shift to a primary care-based delivery system. Among veterans with any visit for a primary diagnosis of a mental or substance abuse disorder, we compared the locus, diagnoses, and procedures performed in each of three settings -- specialty mental health clinics, primary care medical clinics, and medical subspecialty clinics.

Results: Of 437,035 veterans treated for a mental disorder during the final 6 months of the study period, 83.0% were seen for their mental disorders only in specialty mental health clinics; 5.0% were seen exclusively in primary care medical clinics; 1.7% in specialty medical clinics; and 10.3% in multiple settings. Over 90% of individuals with serious Axis I disorders (Schizophrenia, major depression, bipolar disorder and PTSD) received mental health care only in specialty mental health settings. There was no change in the distribution of care between medical and mental health settings over time, although there was a shift in mental health services delivery from specialty medical to primary care settings.

Conclusions: A far greater percentage of mental health care in the VA than in community or HMO populations is provided through specialty mental health clinics. In contrast to private sector samples, the high levels of psychiatric morbidity seen in VA patients may make medical treaters reluctant to assume a large portion of their patients' mental health care.

Impact: These findings may help inform debates regarding mental health staffing projections both in the VA and in the larger "de facto" US mental health care system. Estimates of mental health staffing needs based on "benchmarking," or extrapolation from current managed care staffing patterns, have generally projected a surplus of mental health providers in the US. However, mental health's two-tiered public/private system may make such techniques problematic to apply to mental health services. Developing mental health staffing models that can be generalized to both the private and public sectors will ultimately require better methods of adjusting for case-mix and estimating need in these very different populations.

68. Provider Advice and Walking for Exercise in Elderly Primary Care Patients, and Speciality medical Setting

Patricia Dubbert, PhD. GU "Sonny" Montgomery VA Medical Center, Jackson, MS. ER Meydrech, KA Kirchner, KM Cooper, and DE Bilbrew.

Objectives: Recent studies have confirmed that physical activity is beneficial for elderly men and women. Epidemiologic studies and health guidelines suggest that many elders could decrease risk of disability and premature mortality by increasing moderate intensity activity such as walking. The purpose of this study was to examine prevalence and potential predictors of walking for exercise among patients being screened for the Seniors Telephone Exercise Primary Care Study (STEPS), an exercise nurse counseling intervention trial.

Methods: Methods:Participants were veterans, 60-80 years of age, enrolled in VA primary care clinics, and whose health status permitted increased walking. From 352 identified by medical record and PC provider review, 280 completed a phone prescreening which included questions about health habits and recall of provider advice to exercise. 182 subsequently completed a screening visit which allowed collection of demographic and additional health data.

Results: 164 (59%) of the 280 phone prescreen participants reported some walking for exercise. Of those already walking, 56% walked <= one day ago, 68% within the past 3 days, and 72% within the past week. 68% reported walks >= 20 minutes duration. The frequency distribution of walking was bimodal, with about 40% of participants walking 2-3 times per week and 27% reporting daily walking. 58% reported they had been walking for 6 months or longer. 16% of patients reported doing regular exercise, but not walking. Information obtained from the 182 participants who completed a screening visit found the mean age was 68.8 (S.D. 4.5) years, 26% were minority, 79% married, 22% employed, 55% lived in a rural setting, and 47% had < HS education. Participants had on average 3 major medical diagnoses. 143 (51%) of the 280 prescreening participants recalled being given exercise advice by their health care provider. Estimated in a model that included ethnicity, number of medical conditions, and urban vs. rural dwelling, recall of provider advice was significantly related to walking for exercise (O.R. =2.27, C.I. 1.23-4.17, P<.01). None of the other covariates was significantly related to walking.

Conclusions: More than half of elderly VA primary care patients with chronic illness whose providers thought they could increase physical activity were already doing some walking for exercise. Notably, those who recalled a provider advising them to exercise were more likely to be walking. Speed/intensity of walking was not assessed, and it is likely that the health benefits of walking could be improved if some participants walked faster or for longer periods of time.

Implications: Although these data are observational, the results support previous research with younger, healthier populations which indicate that provider counseling can increase physical activity in sedentary patients. At least 1/3 of participants reported no regular physical activity. Providers should continue to encourage patients to be active as their health status permits. Future studies should address barriers to activity and cost-effective methods of preventive counseling including physical activity

HSR&D Funded: NRI 95-022

69. Evaluating Screening Criteria for Adverse Outcomes in Medical Patients

Ron Evans, MSW. VA Puget Sound health Care System, Seattle, WA.

Objectives: The purpose of the current study was to identify variables near hospital admission that could effectively discriminate patients at risk for nursing home placement, lengthy hospital stay, or readmission. The specific goals were to determine the sensitivity and specificity of differing screening strategies in predicting adverse outcomes.

Methods: We evaluated the screening criteria, including those available through hospital billing and resource usage data, to determine if a subset of generic screens might efficiently identify outcome. Risk criteria reported in the literature were used to predict discharge destination and duration of care for 1,332 admissions.

Results: Factors that discriminated outcome included: comorbidity, mental status, living arrangement, transfer to a special care unit, prior admission within 1 year, iatrogenic trauma, and pending litigation. Sensitivity and specificity of individual screens varied widely. Prior hospitalization was the most sensitive [64%] but least specific [47%]. Transfer to a special care unit was specific [91%] but not sensitive [11%]. Transfer to another hospital was intermediate [sensitivity 32%, specificity 71%]. Combinations of screens were compared, including some using only resource usage data. The most sensitive strategy using billing data detected 44% of adverse outcomes and cost only $4 per admission [$51 per adverse event], versus $14 per admission [$94 per event] when all records were reviewed.

Conclusion: We concluded that use of cumulative risk scores can result in accurate prediction of hospital outcome, which may be useful in targeting patients for intervention.

Impact: Using screens available through billing and resource use data, although insensitive, would be the most cost effective strategy.

70. A Pilot Study: Telecare in the Management of Diabetes.

Stephan Gaehde, MD, MPH. Boston VA Medical Center, Boston, MA. BG Fincke. Dan Berlowitz, MD, MPH. Bedford VA Medical Center, Bedford, MA. J Clark and J Anderson.

Objectives: Recent advances in telemedicine may lead to substantial changes in the outpatient management of chronic illness. Changes in computer technology and the development of sophisticated disease management software that is both physician and patient-friendly may lead to reduced distance between doctors and patients and improved patient self-management of complex health problems, such a diabetes.Here we present the results from a pilot study to assess the feasibility of a telemedicine program that links VA physicians with outpatients with diabetes, through a system of personal computers in physician offices and their patients' homes connected by ISDN lines that enable data transfer and videoconferencing. Specific study objectives were to: 1) Characterize how the technology is adopted, accepted and integrated into the process of care by patients and providers. 2) Determine effect size of the intervention on satisfaction with care, degree of glycemic control, health-related quality of life, and utilization of health care resources.

Methods: In this descriptive pilot study using a pre/post intervention design, we enrolled 20 patients with type 2 diabetes mellitus requiring insulin from 5 provider panels at the Boston VAMC. Study outcomes include degree of glycemic control, patient satisfaction with care, health-related quality of life, and utilization of resources. Qualitative data regarding acceptability and adoption of the intervention were collected in structured interviews conducted by a medical sociologist. Use of the system was characterized by analysis of descriptive data. Outcomes were assessed using a pre/post intervention design.

Results: Patient interviews prior to having the computers set up in their homes and at study termination revealed a varied response. Some patients found the system facile to use and said that the routine of interacting with the program and reporting the results of glucose and symptoms served to reinforce their motivation for self-management while several patients had used the system very little, either because they could not get the system to work or they lost interest. Glycemic control improved from a mean value of 9.4% to 8.7% (p=0.24) from the start of training to 4.5 months. We also report how the intervention was used and adopted by patients as well as its effect on health-related quality of life, satisfaction with care, utilization of resources, and outpatient visits.

Conclusions: This project has demonstrated that: 1) elderly veterans without previous computer experience are able to use a technically sophisticated telemedicine intervention 2) successful deployment of a sophisticated telemedicine system in actual clinical care. Telemedicine may reduce the distance between physicians and their patients, who are responsible for following a complicated self-management regimens at home, with patients experiencing the closer attention as supportive.

Impact: The findings of this study will provide important insights into how such interventions are adopted by patients and providers and how they may be best positioned in the process of clinical care in a VA setting. These finding will be very useful in designing effective future interventions. Further, the findings will allow estimation of effect sizes of several outcomes including glycemic control, health-related quality of life and patient satisfaction with care.

HSR&D Funded: SDR 94-11

71. Assessing VA's National Formulary Policy by Physician Survey

Chester Good, MD, MPH. Pittsburgh VA Medical Center, Pittsburgh, PA. Peter Glassman, MD. West Los Angeles VA Medical Center. Los Angeles, CA. M Kelley, M Bradley, M Valentino, J Ogden, and K Kizer.

Objectives: The VA National Formulary (VANF) was implemented in June 1997 to improve equity of access to pharmaceuticals across VHA and to improve accountability of VA's 1.6 billion dollar pharmaceutical program. We surveyed physicians approximately 1 year after implementation to help assess VANF's effects on patient care, access to drugs, physician workload and VA's ability to train resident staff.

Methods: Questions, scored on a 5-point Likert scale, addressed general issues about the VANF and specific effects of choosing selected drugs within six drug classes. Respondents also provided demographic information. The sample population (n = 4640), based on the circulation files of The Veterans Health Journal, included all listed general internists (n = 2824) and convenience samples of neurologists (n = 238), psychiatrists (n=997), general surgeons (n = 429), neurologists (n = 238) and urologists (n = 152). Non-responding physicians received a second survey approximately 1 month after the first. A total of 104 physicians were declared ineligible, leaving a final sample of 4536. Comparisons across groups were by Chi-squared analysis.

Results: Overall response rate was 45% (2041/4536). Physicians were all attendings, average age 49 years, with 11 years of VA service, and averaging 5 half-day outpatient clinics per week; 73% were full-time employees; 20% practiced in other health systems with formularies, and 13% were on VA Pharmacy Committees. Most physicians (63%) agreed that they could prescribe needed drugs; 66% agreed that patients could obtain non-formulary drugs, when necessary. About one-third (32%) of physicians disagreed that access to prescription pharmaceuticals had increased over the past year. Although 29% stated the VANF impinged on providing quality care to their patients, fewer physicians (24%) felt that it impinged on providing quality care to other VA patients. Thirty-eight percent (38%) felt that the VANF was more restrictive than private sector formularies but only 16% felt that the VANF diminished the ability to train residents for managed care. Most physicians (60%) did not agree that the Formulary added substantially to their workload. Regarding questions on drug class selections, the overall perception was that selections had nominal effect on patient care. For example, choosing lovastatin and simvastatin as formulary drugs was felt to have a positive effect on patient care by 29% of physicians, no effect by 46% and a negative effect by 8%. We noted significant differences among physicians on many issues. For example, VA physicians who worked in other health care systems with formularies were less likely to agree that the VANF added substantially to workload (28% vs. 38%, P < .001).

Conclusions: Most participating VA physicians did not perceive that the VANF negatively affected quality of patient care, access to pharmaceuticals, physician workload or resident training.

Impact: VHA is the first national health care system to assess the impact of formulary policy by surveying physicians. Survey results are being used to increase understanding of the VANF and to improve its functioning, with further policy evaluations now being directed towards ensuring access to non-formulary products and towards understanding effects on physician workload. Future surveys are planned.

72. Does Military Status Influence Use of VA Ambulatory Care?

Nancy Harada, PhD, Donna Washington, MD, MPH, and JoAnn Damron-Rodriguez, PhD. West Los Angeles VA Medical Center, Los Angeles, CA. T Makinodan, H Liu, and S Dhanani.

Objectives: Military experiences have been documented to influence a veteran's subsequent use of health care services. Specifically, these military experiences may serve as predisposing, enabling, or need variables affecting VA ambulatory care utilization. The objectives of this study were to identify military status factors predictive of VA ambulatory care use, and to determine whether these factors remain significant after controlling for race/ethnicity, sociodemographic, and health status characteristics.

Methods: The source of data was the 1992 National Survey of Veterans (NSV). Since the NSV sample is 97% male, our analyses were limited to this gender group. Bivariate analyses were conducted to describe the demographic and military status characteristics of VA ambulatory care users and non-users. A hierarchical logistic regression analysis was conducted with use of VA ambulatory care as the dependent variable. The initial logistic model included several variables descriptive of military status, followed by 3 additional models controlling for race/ethnicity, sociodemographic, and health status.

Results: Of the total sample (n=7,851, mean age=56 years), 41% had used VA ambulatory care services in the previous year. Sixty-four percent of the VA users came exclusively to the VA for ambulatory care, and the remaining had used the VA in conjunction with non-VA ambulatory care. The first model explored the contribution of military status variables to ambulatory care use without controlling for race/ethnicity, sociodemographic, or health status. Significant military status variables included retirement from the military for disability (OR=1.6), combat exposure (OR=1.1), service in the Marines (OR=1.3), Army (OR=1.3), or during the Korean Conflict (OR=1.3), service-connected injury (OR=1.6), 6 or more years of military service (OR=1.3), and retirement from the military after 20 years of service (OR=.58). Controlling for race/ethnicity in the second model did not alter the list of significant military status variables. The inclusion of sociodemographic characteristics in model 3 decreased the list of significant military status variables to service in the Marines (OR=1.6), service-connected injury (OR=1.2), and 6 or more years of service (OR=1.4). After controlling for race/ethnicity, sociodemographic, and health status in the final model only military service for 6 or more years remained significant (OR=1.4). However the final model showed that veterans who are minorities, of lower socioeconomic status, uninsured, have poor health status and a service-connected disability have a greater likelihood of being a VA ambulatory care user.

Conclusions: The findings demonstrate that the VA serves as an important source of health care for veterans with service-connected disabilities and long term service in the military. In addition, veterans who are minorities, of low socioeconomic status and poor health status also use the VA for ambulatory care.

Impact: Military status characteristics describing veterans who use ambulatory care highlight the importance of health care entitlement for Americans who served their country. The VA serves as an important source of ambulatory care for veterans with service-connected disabilities and long term service in the military. As the VA seeks to restructure its delivery of ambulatory care, it must create policies to benefit this deserving population.

HSR&D Funded: ECV 97-028

73. How Primary Care Providers Treat Depression: Attitudes, Skills, Barriers and Personal Experience

Nicole Hasenberg, MPH, Edmund Chaney, PhD, CE Hansen, and Susan Hedrick, PhD. VA Puget Sound Health Care System, Seattle, WA.

Objectives: We wanted to learn more about primary care providers (PCP) attitudes regarding the diagnosis and treatment of depression with the eventual goal of targeting different educational interventions toward PCPs and improving medical care for patients who are depressed.

Methods: We obtained cross-sectional data from PCPs as part of a larger study about depression treatment at the General Internal Medicine Clinic at the VA Puget Sound Health Care System in Seattle. We asked them to evaluate their attitudes, skills, and barriers to treating their patients with major depression. We also were interested in whether having personal experience with depression would influence self-reported skills and comfort level with treating depression.

Results: At the end of two GIMC Journal Clubs and several Resident Post-Clinic Case Conferences during January, February, July, and August of 1998, we notified PCPs about the depression study and asked them to complete an anonymous 5 to 10-minute questionnaire. During that time, 90 Primary Care Providers were on staff, of whom 31(34%) were Attending

Physicians or Medical Fellows, 18(20%) were Nurse Practitioners and 41(46%) were Medical Residents. Of those on staff, 61(68%) PCPs completed the survey.

Almost all PCPs (91.8%) feel comfortable writing referrals for further depression treatment. Overall comfort level with skill-based treatment was good; 40(78.7%) PCPs said they felt somewhat or very skilled at making a diagnosis; 39(63.9%) felt somewhat or very skilled about writing a prescription for anti-depressant medication; and 28(45.9%) believed the same about counseling and education regarding depression.

Many providers reported the following barriers to care limit them a great deal in treating depression: no time for education and/or counseling (54%); preferred medication difficult to obtain (43%); viewing other medical problems as more pressing (36%); perceiving the patient as reluctant to accept diagnosis or treatment (30%); inadequate time for follow-up (28%); and lack of availability of mental health professionals (25%).

In terms of provision of care, 59% think depression usually should be treated in primary care settings compared with 31% who think those with depression should be referred to specialty clinics.

In this population 48(79%) providers report that they or a close friend or family member had personal experiences with depression. We tested whether there was an association between providers who had personal experience with depression and all the mentioned skills and barriers to care. None of the variables were statistically significant, however small numbers in each cell may influence such results.

Conclusions: Primary care provider education may be more effective if it is tailored toward counseling and educating a person who is depressed. Individual VAs can focus on particular systems barriers (such as getting different medications on the formulary or providing more time for care of depressed patients) to improve the treatment of depression. Finally, initial analysis suggests that personal experience with depression does not seem to significantly influence skills and barriers to treatment.

Impact: These results can be used for educating primary care providers and for planning and delivering depression treatment in ambulatory care settings.

HSR&D Funded: 95-097

74. The Use of Ambulatory Care Sensitive Hospitalization as a Qualitative Indicator of Primary Care

Laura Hechtel, PhD, Margaret Byrne PhD, and Carol Ashton MD, MPH. Houston VA Medical Center, Houston, TX.

Objectives: Ambulatory care sensitive (ACS) conditions have been proposed as an indicator of access to, and quality of, primary care. In this study, we examine the validity of using such an indicator as it relates to primary care enrollment.

Methods: From the Outpatient Clinic file (OPC, FY 1996), users were considered to be enrolled in primary care if they had at least one primary care clinic stop. From the Patient Treatment File (FY 1996), ACS and non-ACS hospitalizations were determined for each OPC user and for each hospital facility. In addition, the 10 most frequent DRG's for primary care (PC) enrollees and non-primary care (NPC) enrollees were collected.

Results: 31.0% of PC enrollees were hospitalized for any condition as compared to 21.7 % for NPC enrollees. PC enrollees had a significantly higher (p=0.0001) hospitalization rate (1130 per 1000 individuals) than NPC enrollees (1068 per 1000 individuals). The proportion of hospitalizations that were for ACS conditions was significantly greater (p=0.0001) in PC enrollees (12.8%) than in NPC enrollees (7.5%). This outcome was consistent across hospital facilities (n=153). Among 10 most frequent DRG's, 4 were ACS conditions while the other 6 were psychiatric diseases. All 4 ACS conditions were more common than expected in PC enrollees, while 4 of the 6 psychiatric diseases were more common than expected in the NPC enrollees.

Conclusions: The hospitalization rate for ACS conditions is higher in VA users enrolled in PC compared with those not enrolled in PC. That this finding was consistent across all VA facilities makes it highly unlikely that the ACS hospitalization rate is measuring quality of care. It is more likely that ACS hospitalization rates are reflecting systematic differences in the prevalence and severity of certain diseases between the PC and NPC VA user populations. If hospitalization rates for ACS conditions are to be used as an indicator of quality of care, more research is needed into their validity.

Impact: Using hospitalization rates for ACS conditions as a quality of care indicator assumes that the populations under study are clinically similar and utilize VA hospital facilities in the same manner. However, this study shows that this may not be true. Health science research must take into consideration population characteristics in order to more reliably assess quality of care. Our findings show that it is inappropriate at the present time to use hospitalization rates for ACS conditions as a quality indicator with VA system.

75. Treatment can Enhance the Effectiveness of Substance Abuse Self-Help Groups

Keith Humphreys, PhD. VA Palo Alto Health Care System, Menlo Park, CA. Rudolf Moos, PhD. VA Palo Alto Health Care System, Palo Alto, CA.

Objectives: Affiliation with Alcoholics Anonymous (AA) and other 12-step self-help groups is becoming more common at the same time as professional substance abuse treatment services are becoming less available and of shorter duration. As a result of these two trends, patients' outcomes may be increasingly influenced by the degree to which professional treatment programs help patients take maximum advantage of self-help groups. The present study of 3018 treated veterans examined how the theoretical orientation of a substance abuse treatment program affects (1) The proportion of its patients that participate in self- help groups, and, (2) The degree of benefit patients derive from participation in self-help groups.

Methods: A 1-year longitudinal study at 15 VA facilities nationwide.

Results: Patients treated in 12-step and eclectic treatment programs had higher rates of subsequent participation in 12-step self-help groups than did patients treated in cognitive behavioral programs. Further, the theoretical orientation of treatment moderated the outcome of self-help group participation: As the degree of programs' emphasis on 12-step approaches increased, the positive relationships of 12-step group participation to better substance use and psychological outcomes became stronger.

Conculsions: Hence, it appears that 12-step oriented treatment programs enhance the effectiveness of 12-step self-help groups. Findings are discussed in terms of implications for clinical practice and for future evaluations of the combined effects of treatment and self-help groups.

Impact: These findings present a practical way for VA clinicians to improve outcomes at little additional costs.

76. Do Oral Health-Related Quality of Life Measures Relate to Use of Dental Care?

Judith Jones, DDS, MPH, Nancy Kressin, PhD, A Spiro III, PhD, Donald Miller, ScD and Lewis Kazis, ScD. Bedford VA Medical Center, Bedford, MA. RI Garcia.

Objectives: Valid dental outcome measures should vary with the use of dental services. The purpose of this analysis is to examine the relationship of oral health-related quality of life measures to past use of dental care in two populations.

Methods: We examined the retrospective relationships of self-reported oral health measures to self-reported use of care in two contrasting samples of veterans, the Veterans Health Study (VHS, N=538, mean age=62) and the VA Dental Longitudinal Study (DLS, N=278, mean age =71). Self-reported oral health measures included a single-item self-report measure of oral health (OH1), the 3-item Oral Health-Related Quality of Life measure (OHQOL, Kressin, et al, 1996) the 12-item Geriatric Oral Health Assessment (GOHAI, Atchison & Dolan, 1990), and the 49-item Oral Health Impact Profile (OHIP, Slade & Spencer, 1994). Use of care was categorized into <=1year, >1year; and <=2years, >2years. Reason for last visit was divided into emergency and routine care (exam and cleaning, fillings, other).

Results: In the VHS sample, better oral health (OH1) was associated with recency of dental visit, i.e., better oral health was associated (p<0.05) with more recent use (in last year and last 2 years). Better scores on the OH1, OHQOL and OHIP were significantly associated with reason for last visit, with approximately 0.5 sdev lower scores, on average, in persons who used emergency as compared to routine care. In the DLS sample, there were no significant differences in mean self-reported oral health scores by recency of use or reason for last visit; however trends were in the expected directions.

Conclusions: The validity of these self-report measures of oral health is suggested by the association with recency of dental care and reason for last visit in VA health care users. However, no significant associations were observed in the DLS, most of whom use routine dental care.

Impact: Self-reported oral health measures may be useful to monitor the effects of dental care on patients' quality of life in users of VA health care.

HSR&D Funded: IIR 93-025

77. Health Outcomes of Veterans Using SF-36V: 1998 National Survey of Ambulatory Care Patients

Lewis Kazis, ScD. Bedford VA Medical Center, Bedford, MA. N Wilson, W Rogers, and A Lee. Xinhua Ren, PhD, Katherine Skinner, PhD, Alfredo Selim, MD, MPH, and Donald Miller, ScD. Bedford VA Medical Center, Bedford, MA.

Objectives: The Veterans Health Administration (VHA) is now implementing patient centered measures of functional status as part of its performance measurement system to set goals and standards for the 22 VISNs (geographically based groups of hospitals), individual hospitals and providers. To monitor the changes in health status over time, we established a cohort of veteran users of the VHA who were administered the SF-36V in 1996 and again in 1998. These changes may be related to the processes of care.

Methods: The SF-36V (Short Form Health Survey for Veterans) was mailed to a national probability sample of 25,040 veterans between January and February 1998 who had been administered the same questionnaire, 17 months earlier in 1996. For the cohort, the response rate was 85.2% using a modified Total Design Methodology approach developed by Dillman. The SF-36V, developed in the Veterans Health Study is a patient based questionnaire designed specifically for use among veterans who are in ambulatory care. The SF-36V is a reliable and valid measure of health status with increased precision over the MOS SF-36. Eight concepts of health are summarized into physical (PCS) and mental (MCS) component summaries standardized to the U.S. population with a mean of 50. Each of the patient scores were computed as the difference of the 1996 score from the 1998 score for PCS and MCS, respectively. A negative score denotes worsening. Overall trends in the VA were compared to the Medical Outcomes Study (MOS), a general population in civilian managed care for the same duration of follow-up. Comparisons of the trends over time among VISNs were made with multivariable adjustments for case mix using age, gender, and diagnoses with a previously validated approach using ICD-9 codes.

Results: Rates of physical decline in the VA were about a third of that observed in the civilian sector (MOS) (p<0.01). For PCS, the 17-month physical decline was -0.39. The MOS study showed a much greater rate of decline equivalent to -1.08 over the same time period. There were no important differences in the trends among the VISNs for PCS. For MCS, the rate of mental decline in the VA compared to the MOS study was comparable (p>0.20). However, we found considerable variation in mental health trends across the VISNs. Trends varied from 0.69 to -0.93, an overall range of 1.62 (p< 0.05). After discounting sampling errors, regression to the mean effects and attrition, trends in PCS for the VA compared to the civilian sector remained significant and clinically important. For MCS, differences in the trends observed among VISNs, are larger than case mix explanations would suggest even after taking random sampling errors into account. The differences in the trends for MCS are equivalent to treating the psychological impact of two medical conditions.

Conclusions: Substantial variation in MCS trends among VISNs were observed and for the most part, were not explained by case mix or methodologic factors. These findings suggest that processes of care are likely candidates to explain these differences.

Impact: These results have important implications for quality assessment and for setting goals for the VISN performance measures.

HSR&D Funded: SDR 91-006

78. Multi-site Research: Overcoming Hurdles

Anne Keane PA-C, JD, Nathan Every MD, MPH, and Anne Sales PhD, RN. VA Puget Sound Health Care System, Seattle, WA.

Objectives: In this paper, we describe the process and costs of obtaining Institutional Review Board (IRB or Human Subjects Committee) approval and access to electronic data systems at each of 30 VA Medical Centers (VAMCs) in five Networks for a study of patient outcomes and clinical integration in cardiology. We then propose changes to some of the processes of obtaining necessary approvals that would enable research to be done at less cost while still preserving adequate reviews for patient safety at each medical center and facility providing patient care.

Methods: As part of the proposal process, we gained approval from our local IRB for the protocol. This submission and required revisions are not included in our estimates of time and cost. Beginning in January 1998, we contacted the IRBs for all of the VAMCs in the five VISNs included in our study: VISNs 13, 18, 19, 20, and 22. We contacted the people who had supported the study at each of the medical centers, and obtained the contacts for the local IRB. We then submitted the application required for the local IRB, as well as complete copies of the protocol, human subject submission and approval from our own IRB. During the same time period, we began to contact Information Resource Management (IRM) at each medical center to obtain the applications for gaining access to electronic records. Once we had obtained IRB approval at each site, we requested IRM access.

Results: We found a wide variety of IRB mechanisms in the 35 sites we initially contacted. We began the process in January 1998, and are still completing IRB negotiations at one site in November 1998. Most sites had given us approval by June 1998. We had to drop some sites from the study because of the difficulties of negotiating IRB approval. The project manager and secretarial support staff spent the most time involved in obtaining IRB approval, for a combined total of approximately 800 hours of effort, at a cost of approximately $20500. Several staff members were involved in obtaining IRM access, with a combined total of 560 hours, for a total cost of $15600. The two Principal Investigators were involved to a lesser extent, for about 32 hours, at a cost of approximately $1100. The total cost of this piece of the project was approximately $37200. The final number of VAMCs participating in the study is 30, for an average cost of $1240 to obtain both IRB approval and IRM access to each medical center.

Conclusions: Obtaining IRB approval and IRM access for a large, multi-site study is time-consuming and expensive. The process worked best when there were dedicated staff, and when we could link medical centers together in a geographic region, generally the VISN.

Impact: These results are likely to be of interest to VA HSRD Management, researchers in the field, and VISN staff who have a vested interest in VISN-wide research studies. Obtaining high quality comparative data at relatively low cost across medical centers is a high priority for these groups.

HSR&D Funded: ACC 97-001

79. Toward Gender-Aware VA Health Care: Staff Ideology, Sensitivity, and Knowledge

Lynda King, PhD and Daniel King, PhD. National Center for PTSD, Boston, MA. P Miller and V Savarese. J Wolfe

Objectives: Despite VHA efforts to better accommodate the healthcare needs of a growing number of women veterans, an historical orientation towards treating male patients may have resulted in a less than optimal care environment for female patients. Our guiding hypothesis is that this environment is influenced by three overlapping personal characteristics of VHA employees: (1) the tendency to work without gender stereotypes, (2) sensitivity to female patients� special needs, and (3) knowledge of women veterans and their healthcare. Together these comprise what we have termed gender awareness. For this three-year project, we seek to develop a reliable and valid self-report measure of gender awareness for use with all VHA personnel. The project has four objectives: (1) to refine the definitions of the gender awareness components and create the instrument, (2) to establish the instrument's psychometric properties, (3) to obtain normative data, and (4) to make the instrument available for large-scale use.

Methods: The design is observational and cross-sectional, with multiple waves of data for reliability and validity analyses. Participants are VHA employees selected using proportional stratified random sampling from the VA New England Healthcare System. In the first year, 621 employees received a preliminary version of the instrument; 371 (60%) were completed. The sample's mean age was 46.01 years (SD = 11.58); 60% were female; and 25% were military veterans. Analyses included computation of descriptive statistics and frequency distributions for all questionnaire items, probabilities of endorsement, item-total correlations, coefficients of internal consistency reliability (alpha), and interscale correlations.

Results: The 18-item scale assessing the ideology component of gender awareness had a mean of 72.38 (SD = 11.77; range = 39-90); alpha = .90. The 19-item sensitivity scale had a mean of 75.72 (SD = 8.08; range = 56-94); alpha = .75. On the 36-item knowledge test, the mean number of correct responses was 19.60 (SD = 4.92; range = 0-29). The most-missed knowledge items were related to demographics of women veterans, eligibility for sexual trauma services, and VHA utilization rates.

Conclusions: The project is ongoing. Year 1 was successful, with the more affective components of gender awareness, ideology and sensitivity, having established discriminant validity, correlating .38. Presently, the instrument is being further refined, as the sensitivity scale is seeing additional adjustments to improve its consistency.

Impact: The gender awareness assessment instrument will provide a means to identify and measure factors that contribute to the quality of VA healthcare for women veterans, and will provide a mechanism to pinpoint training needs for the education and remediation of employees. Moreover, the instrument will afford a means to monitor organizational improvements in gender awareness over time. Our long-term goal is to have a product that has justifiable transportability on a large scale. At an organizational level, we anticipate that the proposed project will contribute to the contemporary emphasis on outcomes measurement. The gender awareness assessment package can fit nicely within a program to monitor patient satisfaction among the growing number of women veterans, as that satisfaction relates to features of organizational climate.

HSR&D Funded: GEN 97-014

80. Is Depression Associated with Oral Health-Related Quality of Life?

Nancy Kressin, PhD, Avron Spiro III, PhD, Katherine Skinner, PhD and Judith Jones, DDS, MPH. Bedford VA Medical Center, Bedford, MA.

Objectives: The health-related quality of life (functional status, emotional well-being) of patients with depression is often as low as, or lower than, that of patients with chronic medical conditions. However, we do not know whether depression has a similar effect on oral health- related quality of life (oral QOL). VA dental policymakers, clinicians and researchers are increasingly relying on oral QOL ratings to evaluate dental treatment needs and outcomes of care. Thus, it is important to understand what factors influence such ratings.

Methods: We examined the association between depression (measured by the CES-D) and oral QOL, using two different indices: the Geriatric Oral Health Assessment Index (GOHAI) and the Oral Health-Related Quality of Life measure (OHQOL). Using data from 3 veteran samples: male VA patients in the Veterans Health Study (VHS), female VA patients in the VA Women's Health Project (WHP), and male community dwelling veterans who do not use VA care (Normative Aging Study (NAS)), we examined whether individuals who screened positive for depression (scoring above the standard cutpoint) had worse oral QOL than those who were not, controlling for sociodemographics (age, education, marital status), and self-reported oral health.

Results: In bivariate analyses, being depressed was associated with worse OHQOL scores in both the VHS and WHP veteran patient samples, as well as in the NAS. Depressed individuals had worse GOHAI scores in the VHS and WHP, but not in the NAS. After controlling for self-reported oral health, age, income, marital status and education, depression remained significantly associated with both oral quality of life measures in all samples, and the independent and control variables together explained between 15 and 30% of the variance.

Conclusions: These results suggest that there is a strong association between depression and oral quality of life, suggesting further negative health impacts of depression in addition to those already quantified with regard to physical health. However, these cross-sectional data cannot prove causality. Future research should further explore the mechanisms of the association of depression and oral quality of life through the use of longitudinal data.

Impact: The understanding of psychosocial and other factors which influence patients' ratings of quality of life is crucial to the accurate interpretation of findings by researchers, clinicians, and policy makers. Recognizing that depression is a significant correlate of oral health outcomes improves the measurement of oral quality of life and provides a potential avenue for interventions to improve oral health outcomes.

HSR&D Funded: 93-05, SD 91-006, SDR 93-101, HFP 91-012

81. An Overview of the Decision Support System

Joseph Kubal, MA, Denise Hynes RN, PhD, Diane Cowper, MA, Michael Kerr, MS, MA and J Palmer. VA Information Resource Center, Hines IL.

Objectives: The primary objective of this informational poster is to present summary information on the currently evolving Decision Support System (DSS). DSS is a database that provides integrated clinical and financial data to help managers make informed tactical and strategic decisions. The DSS software was developed and tailored for the VA by Transition Systems Incorporated (TSI). The information for this poster was developed by the VHA Chief Information Office and the DSS Program Office and is being presented by VA Information Resource Center (VIREC) staff to inform HSR&D researchers of this new VA-wide system which ultimately has ramifications for health services research.

Methods: All the data in the DSS come from existing VA databases resident at the Austin Automation Center (AAC). These include: 1) Personnel and Accounting Integrated Data (PAID), 2) Consolidated Memorandum of Receipt (CMR), 3) Financial Management System, 4) National Patient Care Database (NPCD), 5) Patient Treatment File (PTF), and 6) Patient Assessment Instrument (PAI).

Results: DSS: 1) Allows budgeting and budget modeling for medical centers, VISNs and VHA based on case-specific workload, 2) provides for resource distribution to the medical centers and VISNs based on performance, 3) Supports implementation of managed care for the VHA system, 4) Allows equitable comparisons of medical centers, 5) Supports VHA funding request to OMB and Congress, 6) Supports quality management and quality improvement initiatives, 7) Supports the development of an itemized patient bill for the Medical Care Cost Recovery (MCCR) program and 8) Provides productivity analysis and patient specific costs.

Conclusions: The VA is now realizing what private hospitals have known for some time � becoming and remaining competitive ensures success and survival. The health care industry is rapidly evolving; the VA must keep up and compete with other health care organizations to remain viable. DSS is designed to provide the information that is needed to make those business decisions that are required in a competitive marketplace.

Impact: DSS will provide the critical tools necessary for VA hospitals to remain cost effective and efficient in the dynamic U.S. health care environment. The VA strategy for implementing DSS emphasizes an interdisciplinary approach, integrating clinical and administrative cost accounting systems, and now senior leadership. Consistent use of this approach in DSS will permit VHA to become and remain competitive now and in the future. Finally, DSS information will prove useful to HSR&D, CSP, and other VA researchers in their ongoing quest for accurate clinical and cost data.

HSR&D Funded: SDR 98-004

82. Establishing Clinical Equivalence: An Example from a Study of Self-Care Center versus Full-Care Hemodialysis

Martin Lee PhD, CA Landis, Mingming Wang MPH, Elizabeth Yano PhD, Lisa Rubenstein MD, MSPH, Center for the Study of Healthcare Provider Behavior, Sepulveda, CA.

Objective: In most studies, the goal of the research is to demonstrate that one or more of the intervention groups is superior in some sense to other arms of the study. These comparisons may be dealt with using standard statistical approaches. However, on some occasions, the objective may be to demonstrate that an intervention is no worse than another, since such a finding might be helpful if this new approach is easier to use, more cost-effective, etc., and could be used if the clinical effectiveness is basically equivalent to existing methodologies. A failure to reject the usual statistical null hypothesis of no difference is not grounds for reaching this conclusion because of power considerations arising from small sample sizes. Our objective here is to demonstrate how to reach the conclusion of clinical equivalence using a proper statistical framework within the context of a study on two methods for the care of chronic dialysis patients.

Method: An observational retrospective cohort study was conducted on 135 patients undergoing hemodialysis patients at the West Los Angeles VA Medical Center between June 1977 and December 1994. Forty-nine subjects used a self-care approach (at the center) and 86 received traditional full-care hemodialysis. Patient mortality was the primary endpoint insofar as demonstrating clinical equivalence of these two methods for care delivery. The Cox proportional hazards model was used to generate the adjusted the 1, 2, and 5-year survival probabilities with respect to patient age and presence of pulmonary disease (the only significant baseline covariates). The comparison of these probabilities was based on the usual two-sample Z-statistic with the hypothesis testing paradigm modified whereby the null hypothesis was one of inferiority (defined as a difference in survival probabilities of more than 10%) and the alternative hypothesis indicating clinical equivalence.

Results: The estimates of the survival probabilities were .976 for self-care and .928 for full-care at 1 year, .891 for self-care and .752 for full-care at 2 years, and .719 and .466 at 5-years. The test of inferiority was significant in each case (p<.001), providing statistical evidence of a lack of inferiority for the self-care approach.

Conclusion: Using an appropriate statistical strategy, it was possible to demonstrate that a self-care center approach to hemodialysis is no worse than the traditional method for delivery of this care. Given the potential for cost-savings and improved patient quality of life, this approach may be a reasonable alternative.

Impact: This study indicates that alternatives to full-service clinical care at a VA center are reasonable to consider, and may in fact, be potentially superior. This is particularly relevant as the VA healthcare system further embarks on the examination of practice patterns to identify and promulgate best practices through the QUERI process.

83. Evaluating the Effect of Primary Care Clinic Visits on Survival for Hospitalization

Martin Lee PhD, Mingming Wang MPH, Elizabeth Yano PhD, and Lisa Rubenstein MD, MSPH, Center for the Study of Healthcare Provider Behavior, Sepulveda, CA.

Objectives: Objective: From an examination of patients hospitalized at the Sepulveda and West Los Angeles VA Medical Centers, we evaluated the possibility that more frequent primary care users had a greater likelihood of surviving a hospital stay for one of a set of significant clinical situations.

Methods: Method: All patients hospitalized during the 1993 fiscal year at Sepulveda VA (SVA) and the West Los Angeles VA (WVA) (two academically-affiliated VA medical centers with active primary care programs in Southern California) were identified. The subset of individuals identified as being admitted for major cardiac, cerebrovascular and respiratory conditions based on ICD-9 codes were specifically utilized in this study, as these conditions appeared to be the most likely to be affected by frequent primary care. We classified each case according to whether the patient survived (survivors), died in hospital or within thirty days of discharge (early death), or died between thirty days and one year after discharge (late death). We obtained demographic information and the mean number of primary care clinic visits during the twelve months prior to hospitalization for patients in each of these three categories. We used a polytomous logistic regression model to determine whether there was a relationship between primary care visitation and survival after adjustment for key independent survival indicators such as age. We also utilized a Cox proportional hazards regression model to examine the same variables as a function of the actual survival time (potentially censored twelve months after hospital discharge). Initial univariate analyses were considered to compare the three study groupings with respect to mean primary care visitation (using the Kruskal-Wallis test).

Results: The SVA sample included 3,459 inpatients (88.8% survived, 5.9% late deaths, 5.3% early deaths) and the WVA sample had 5,344 individuals (87.4% survived, 6.2% late deaths, 6.4% early deaths). On a univariate basis, both institutions' survivors had significantly more primary care visits (average 6.7/yr: SVA, 5.3/yr: WVA) than early death (5.1/yr:SVA, 2.7/yr:WLA) or late death (4.1/yr:SVA, 2.7/yr:WLA) patients.This significance held up in both multivariate models.

Conclusions: Patients admitted with serious major medical conditions who had been seen more often in primary care clinics prior to their hospitalization were more likely to survive their hospital stay. The significance of the primary care visitation pattern may be to hospitalize the patient earlier in the course of an illness episode or result in better peri-hospital management. It is also possible that the more frequent users had different casemix characteristics than less frequent users, although we did attempt to adjust for many of these potential covariates. At the very least, investigators evaluating the significance of primary care should consider visit patterns into consideration.

Impact: With the emphasis in the VA healthcare system on the delivery of primary care, this study suggests that such care may have a direct impact on future hospitalization survival. This provides further impetus for the continuing efforts to adequately deliver this care.

84. The Economic Impact of Automated Primary Screening for Cervical Cancer: Use of a Markov Chain Model

Martin Lee, PhD. Sepulveda VA Medical Center, Sepulveda, CA. BL Smith, S Leader, and P Westlake.

Objective: The Papanicolaou (Pap) smear detects precursor changes to cervical cancer enabling therapeutic interventions to potentially avert invasive cervical cancer. Its widespread use has resulted in more than a 70% decline in cervical cancer-related morbidity and mortality in the United States over the past forty years. Yet, cervical cancer continues to be prevalent. Newly developed technologies such as automated primary screening have been introduced to reduce the incidence of false negative Pap exams. We have examined one such device, the AutoPap� Primary Screening System (Neopath, Inc., Redmond, WA) with respect to its economic impact on the healthcare system, since this device has been previously shown to be more sensitive and specific than manual examination of Pap smears. This study serves to address the concern as to whether a clinically effective diagnostic procedure can also be cost effective.

Method: To evaluate the economic impact of AutoPap, we developed a model of the progression of cervical disease that involves six stages of pathology from a healthy state through death. This traditional model was expanded to include substances that allow for treatment (or its absence) and the success thereof within each stage. During a fixed time interval (say, one year), patients are allowed to move between stages or remain in the same stage with certain fixed probabilities. This was represented by a basic first-order Markov model. The probabilities that populated this model were determined from the literature or from consultation with medical practitioners. Ranges (determined by a modified Delphi process) were use to assess model sensitivity. The model was run on a hypothetical group of 18 year old women whose mortality pattern represented that for the general population of women of that age in the US. The program DATA (TreeAge Software, V3.0.17) was used to run the simulation.

Results: Annual screening with AutoPap produced a meaningful increase in life expectancy of 32.1 days relative to manual screening at a marginal savings of $628 per person (or -$7,144 per life year saved). Less frequent screening yielded lower positive savings.

Conclusion: Automated screening for cervical cancer has the potential to significantly improve healthcare outcomes and reduce cost. The use of appropriate mathematical models for the evaluation of the economic impact of such new interventions can be of great value in these determinations.

Impact: Given that as of the end of 1996 there were 1.2 million female veterans, an evaluation such as that reported here may prove to have a significant economic impact for the VA system.

85. Patient-Centered Alternative to Psychiatric Hospitalization

James Lohr, MD. VA San Diego Healthcare System. San Diego, CA. W Hawthorne, B Green

Objectives: The purpose of this study was to assess the effectiveness of an alternative to acute psychiatric hospital treatment, the Short-Term Acute Residential Treatment(START) model.

Methods: A total of 376 START program clients participated in the study, which used a repeated measures design and assessed participants on multiple standardized measures of symptoms and functioning at admission, discharge, and at a 4-month follow-up interval. Ex post facto comparisons were made with data from 186 psychiatric hospital patients from previously conducted outcome studies utilizing the same instruments and procedures. Measures included the Brief Symptom Inventory, the Behavior and Symptom Identification Scale-32, the Medical Outcome Short Form 36, and the Client Satisfaction Questionnaire-8, as well as demographic and other data.

Results: The results of this study indicate that START facilities and psychiatric hospital programs admit persons with similar levels of acute distress, demonstrate comparable levels of improvement by discharge, and produce an equivalent degree of short-term stability of treatment gains.

Conclusions: This study supports the START program model as a less costly yet similarly effective alternative to psychiatric hospitalization for many voluntary adults.

Impact: During the 1997 fiscal year, the VHA spent 1.5 billion dollars on mental health services for veterans, of which inpatient psychiatric hospital services accounted for 50% of those costs. The development of START programs could have significant impact on this very costly aspect of psychiatric care.

86. VA Utilization by Level of Diagnostic Cost Group (DCG) Predicted Risk

Susan Loveland and Amy Rosen. Bedford VA Medical Center, Bedford, MA. A Ash Cheryl Hankin, PhD, James Rothendler, MD, Dan Berlowitz, MD, MPH, and Jennifer Anderson, PhD. Bedford VA Medical Center, Bedford, MA. Mark Moskowitz, MD. Boston University School of Medicine, Boston, MA.

Objectives: As VHA adopts managed care practices, it needs to understand the disease burden and resource utilization of its population. Risk adjustment methods, such as Diagnostic Cost Groups (DCGs), can be used to profile and predict the health care resource consumption of population subgroups across VHA facilities. This study examines the relationship between levels of DCG predicted risk (DPR) and resource utilization among a sample of VA health care users.

Methods: We used VA administrative data, specifically the outpatient, inpatient, extended care, and census files from FY97. We selected a random 1% sample of all veterans (N=26,165) with any utilization during that period, retaining diagnostic information assigned during "face-to-face" provider encounters and deleting diagnoses associated with lab, x-ray and telephone stops. We divided the sample into two patient subgroups based on the setting in which they received care: outpatient care only (N=21,765), or those with any inpatient care (N=4,400). Patients with any nursing home care (N=484) were separately identified as were those who received inpatient care for substance abuse problems (N=632). The DCG model profiled the disease burden of the population by summarizing the diagnostic information (i.e., ICD-9-CM codes) available for each veteran during the study period and produced a risk score (DPR) to predict utilization during the same period. We ranked patients into deciles of DPR, where the first decile contains patients with the lowest DPR, and the 10th, those with the highest DPR. We examined the composition of the deciles and computed the mean number of service days (the sum of outpatient visit days and inpatient days) within each decile.

Results: Ten percent of the outpatients fell into the highest two deciles of DPR, in contrast to 69.4% of the inpatients. Similarly 70.1% of the outpatients fell into the lowest 6 deciles compared to only 10.1% of the inpatients. As DPR increased, the mean service days of the outpatients increased from 3.6 to 27.6. A similar pattern of increasing utilization with level of DPR was found for inpatients in the highest risk deciles, with mean service days increasing from 46.5 to 65.8. Furthermore, while only 1.8% of all patients were identified as nursing home residents, 60.5% fell into the highest DPR decile. Similarly, 95.9% of substance abuse patients fell into the two highest risk deciles, with mean service days in the 9th and 10th deciles of 60.7 and 102.6, respectively.

Conclusions: Higher levels of predicted risk were associated with increased levels of utilization among specific subgroups of the VA population. In addition, inpatients and nursing home residents, who would be expected to have higher utilization were, in fact, classified in the highest deciles.

Impact: This study validates the usefulness of the DCG methodology in measuring the disease burden of the VA population. Reliable risk stratification can lead to a better understanding of the factors that drive resource consumption, an understanding which is critical in managing health care delivery.

HSR&D Funded: MPC 97-009

87. Mortality and Days of Survival for Medicare Beneficiaries in the Fee-for-Service and HMO Systems

Matthew Maciejewski, PhD. VA Puget Sound Health Care System, Seattle, WA.

Objectives: To compare the mortality rates and days of survival for elderly Medicare beneficiaries enrolled in fee-for-service (FFS) health plans with enrollees and disenrollees of health maintenance organizations (HMOs) under capitation.

Methods: Data for this study are based on 1993 and 1994 Medicare Patient Activity Record and enrollment files, as well as county-level data on enrollment in Medicare HMOs under capitation (TEFRA-risk HMOs) and Medicare capitation payments. The sample of Medicare beneficiaries is based upon FFS and HMO enrollees in the 124 counties with the greatest enrollment in TEFRA-risk HMOs. The first part of the analysis compared a cohort of FFS and HMO enrollees, and the second part compared a cohort of FFS enrollees and HMO disenrollees. The mortality rate for Medicare beneficiaries in the two delivery systems was estimated using logistic regression. For beneficiaries that died between April 1, 1993 and April 1, 1994, the number of days of survival in each system was estimated using ordinary least squares. Selection bias was partially controlled using variables for age, gender, and institutional status. The analyses also included several county-specific variables of interest.

Results: HMO enrollees had lower mortality rates and higher days of survival than FFS enrollees. Conversely, HMO disenrollees had higher mortality rates and lower days of survival than a comparable group of FFS enrollees.

Conclusions: Movement of elderly Americans into HMOs under capitation is associated with lower mortality rates and higher survival times until death. Although this analysis did not completely control for selection bias, this study provides some evidence that alternative delivery systems may yield better health outcomes for elderly veterans than conventional, FFS systems. However, some elderly beneficiaries in poor health or likely to be in poor health may find HMOs to be incompatible with their needs and will disenroll. These beneficiaries may be able to obtain better outcomes by remaining in the FFS system.

Impact: The VA may be able to improve patient outcomes by introducing new organizational structures and financial incentives applied successfully in other settings. Greater enrollment of veterans into community-based outpatient clinics is one such alternative delivery system that contracts with providers under capitation in some sites. The VA may be able to capitalize on some of the lessons learned from Medicare's experience with TEFRA-risk HMOs in its own efforts to promote veterans' health.

88. Recidivism among Veterans with Schizophrenia Living in Board and Care: An Outcome Evaluation of the Community Residential Care Program

Alvin Mares MSPS, MSW, and J McGuire. Alhambra, CA.

Objectives: Recidivism-the repeated use of inpatient services--among veterans diagnosed with schizophrenia is an important problem. In FY96, veterans with schizophrenia consumed 21% of all inpatient bed days while representing only 4% of the total inpatient population. Health providers are increasingly using intensive community-based case management interventions such as Assertive Community Treatment (ACT) and Intensive Psychiatric Community Care to reduce the risk of recidivism. While much is known about intensive community interventions (especially ACT), little is known of the effectiveness of less intensive community interventions, such as the Community Residential Care (CRC) Program, in reducing the risk of recidivism. CRC provides schizophrenic veterans and others who live in privately operated board and care homes with monthly home visits. The purpose of this study is to estimate the effect of CRC home visits on the risk of recidivism among patients at West Los Angeles VAMC.

Methods: A retrospective cohort study design was used. A group CRC patients (N=214) was followed from their first CRC home visit until 8/31/98 for first psych/SA and med/surg admission. DHCP Patient File was queried on 9/17/98 for all patients matching the street address or facility name of one of the 27 participating board and care homes. A total of 321 matches were made-214 CRC patients and 107 non-CRC patients. All 214 CRC patients (52% of the total CRC patient roster) were included in the study. This method of selecting subjects was chosen to allow for future comparisons of inpatient lengths of stay among CRC and matched comparison subjects. Five secondary data sources were used. DHCP Patient, Outpatient Clinic Visit, and Patient Treatment Files provided data on socio-demographics, CRC home visits, and hospitalizations, respectively. The CRC program database provided additional licensing data on board and care homes. Psychiatric/substance abuse (psych/SA) and medical/surgical (med/surg) hospitalizations were defined based on discharge ward. Twenty inpatient wards were designated psych/SA wards and forty wards as med/surg. All domiciliary and nursing home admissions were excluded. Cox regression was used, in concert with the Andersen Behavioral Model, to identify recidivism risk factors. Twelve dichotomized covariates were included: visit duration, worker profession, home size, age, marital status, race, income, service connected percentage, family support, psychiatric diagnosis, symptoms, and alcohol/substance abuse diagnosis.

Results: Psych/SA recidivism risk factors included: home visit duration fewer than four years (RR=2.86, p=.0002), minority status (RR=1.94, p=.0273), and schizophrenia diagnosis (RR=1.79, p=.1056). Med/surg risk factors included: home visit duration fewer than four years and med/surg (RR=3.88, p=.0055), residence in a facility having 80+ beds (RR=1.97, p=.1103), and limited family contact (RR=1.82, p=.1126).

Conclusions: Minority status, schizophrenia diagnosis, and shorter time receiving CRC home visits were associated with increased psych/SA recidivism. Living in a larger board and care, having limited social contact with family, and shorter time receiving CRC home visits were associated with increased med/surg recidivism.

Impact: Additional efforts may be warranted to reduce recidivism among minority, those with schizophrenia, residents of larger facilities, those isolated from family, and recently enrolled CRC patients.

89. Patterns of Medical Treatment for Congestive Heart Failure (CHF) in a Veteran Population

JL Meier, PharmD. VANC Health Care System, Martinez, CA. JR Lopez, RM Moskowitz, and D Siegel

Objectives: Evidence of the beneficial effect of angiotensin converting enzyme (ACE) inhibitors in reducing morbidity and mortality in patients with CHF has accumulated over the past decade. Although ACE inhibitor use in patients with CHF has increased, only 36% of patients studied by Smith et al (Arch Int Med, 1998) in 1994-1995 received treatment with ACE inhibitors. Hydralazine and isosorbide dinitrate in combination also reduce morbidity and mortality in patients with CHF, and angiotensin II antagonists are being studied for this purpose. To determine the extent of ACE inhibitor and other medication use in patients with CHF in the VA Northern California Health Care System (VANCHCS).

Methods: ICD-9 codes from patient encounter forms identified patients with CHF in the VANCHCS database between 4/1/97 and 3/31/98. Researchers related this data to prescriptions for drugs categorized in the VA Cardiovascular Series, including ACE inhibitors, angiotensin II antagonists, hydralazine, isosorbide dinitrate, nitroglycerin patches, digoxin, loop diuretics, hydrochlorothiazide, and potassium-sparing diuretics filled over a six month period (1/1/98 to 6/30/98) using Access=99 and Excel=99.

Results: Of 1518 patients with CHF, the percentage receiving a drug included: diuretic (78%), ACE inhibitor (63%), digoxin (50%), nitrate (34%) angiotensin II antagonist (10%), and hydralazine (5%). Seventy-four patients 5%) received prescriptions for both hydralazine and a nitrate, and 840 (55%) received a diuretic and an ACE inhibitor. Thirty-two percent of patients received a combination of a diuretic, digoxin, and either an ACE inhibitor or an angiotensin II antagonist.

Conclusions: ACE inhibitor use in this analysis is higher than published data from other populations that have been studied. Greater use of ACE inhibitors or hydralazine with nitrates may further improve patient outcomes.

Impact: A higher proportion of patients with CHF are treated with ACE inhibitors at VANCHCS. However, there is potential to further increase use and improve CHF management.

90. Comparison of Health Care Use and Outcomes for HIV-Infected Patients in VA Versus Non-VA Settings

Terri Menke, PhD, Linda Rabeneck, MD, MPH, and Nelda Wray, MD, MPH. Houston, VA Medical Center, Houston, TX.

Objectives: To compare VA and non-VA health care utilization and outcomes for HIV-infected patients.

Methods: For VA users, we collected data on both VA and non-VA health care obtained during 1994 for 470 patients at five VAMCs: New York, Miami, Houston, Los Angeles, and San Francisco. Data on VA health care came from VA's HIV Registry, which electronically extracts data on all HIV-infected patients from VISTA. Information on the sampled patients� use of non-VA health care and patient characteristics was obtained from patient interviews. We used the questionnaire from the AIDS Costs and Service Utilization Survey (ACSUS), a study of non-VA HIV-infected patients conducted by the Agency for Health Care Policy and Research. The ACSUS study provided the non-VA sample with which our VA sample was compared in terms of utilization and outcomes. We estimated regression equations for seventeen utilization measures. We used logistic regression to estimate the probabilities of any use for hospitalizations, physician visits, emergency department visits, nursing home stays, mental health visits, substance abuse visits, home health visits, dental visits, and anti-retroviral medication. We used Poisson regression to estimate equations among users of each type of health care for the amount of use (e.g., number of hospitalizations among those with inpatient stays). We used linear regression to estimate an equation for inpatient length of stay in logarithm form. The covariates included measures of HIV illness severity, age, ethnicity, HIV transmission route, education, social support, health insurance, employment status, income, health status, physical function, role function, social function, and mental function. Dummy variables were used to estimate differences in use among three patient categories, controlling for the covariates: (1) VA only users; (2) dual VA and non-VA users; and (3) only non-VA. We used Cox proportional hazards modeling to compare progression from early to late stage HIV disease or death among the three patient groups.

Results: There were no significant differences among VA-only, dual VA and non-VA, and non-VA only users in the use of inpatient hospital or nursing home care, or in the number of physician, emergency, mental health, substance abuse, or home health visits among users. Dual VA and non-VA users had higher probabilities of physician, mental health, substance abuse, and dental visits than the other patient groups. Compared to those who only used non-VA care, both dual users and VA-only users had lower probabilities of emergency department visits and anti-retroviral medication use, but obtained more dental visits per user. VA-only users had a lower probability of home health use than the other patient groups.

Conclusions: Depending on the type of health care, HIV-infected patients in VA obtained the same or less care compared to patients in non-VA settings. Outcomes were the same for VA and non-VA patients. These results imply that VA care for HIV-infected patients is more efficient than non-VA care.

Impact: In this era when VA must compete with non-VA settings in terms of providing cost-effective care, this study demonstrates that VA is more efficient at least for HIV-infected patients.

HSR&D Funded: IIR 95.107

91. Developing Algorithms to Define Episodes of Care

Terri Menke PhD, Nelda Wray MD, MPH, Carol Ashton MD, MPH, and Linda Rabeneck MD, MPH. Houston VA Medical Center, Houston, TX.

Objectives: To define episodes of care that include hospitalizations, including the time frames pre and post-hospitalization, and the health care to be included.

Methods: An expert panel of physicians selected high-volume diagnoses or surgical procedures to examine, and developed criteria for defining episodes of care. Among the conditions selected were: coronary artery bypass, colon cancer procedures, total hip replacement, bleeding ulcer, and congestive heart failure. For each condition selected, computer printouts using VA administrative data from FY1997 were generated of the frequency of: (1) visits by clinic stop, by month, from 6 months prior to admission to 6 months following discharge; (2) outpatient tests and procedures by CPT code for relevant organ systems, by month, from 6 months prior to admission to 6 months following discharge; (3) readmissions by ICD-9 code of the primary diagnosis, by week, up to 6 months following discharge from the index hospitalization; and (4) nursing home stays, by week, up to 6 months following discharge. The expert panel used their clinical knowledge and the computer printouts to determine how long prior to the hospitalization the episode began, how long after discharge the episode ended, and which specific clinic stops, tests, readmissions, and nursing home stays should be included in the episode.

Results: Episodes of care for surgical hospitalizations took the following general form. Episodes of care for medical diagnoses were defined analogously. (1) Pre-admission care: (a) started 90 days prior to admission for most surgeries; (b) included visits to principal medicine clinics, relevant medical specialty and surgery clinics, nursing, relevant behavioral counseling, and clinics designated as pre-operative; and (c) included outpatient tests or procedures specific to the evaluation of the extent or indication for surgery. (2) Post-discharge outpatient care: (a) included visits to principal medicine or nursing clinics within 4-6 weeks after discharge, depending on the surgery; and (b) included visits to all other clinics specified in (1) and relevant rehabilitation visits up to 6 months following discharge. (3) Readmissions: (a) were all included up to 1-6 weeks after discharge, depending on the surgery; (b) were included in the episode based on containing a relevant surgical procedure or primary diagnosis if they occurred between the time specified in (a) and 6 months after discharge. (4) Nursing home stays were included if they immediately followed the initial hospitalization, or a readmission that was included in the episode.

Conclusions: A process combining clinical knowledge and relevant data was developed for defining episodes of care that include a hospitalization. The algorithms developed can be applied by managers or researchers who want to examine the costs, efficiency, or outcomes for episodes of care.

Impact: Current organizational features of health care have shifted the focus of quality and efficiency evaluation from the health care event (e.g. hospitalization) to the entire package of services required to treat an episode of illness. This shift is evident in VA's adoption of the VERA system, which allocates funding to VISNs on a capitated basis.

HSR&D Funded: IIR 95-139

92. Designing Guidelines for Successful Dissemination and Implementation: The California Guidelines for Alzheimer's Disease Management

Brian Mittman, MD. VA Greater Los Angeles Health Care System, Sepulveda, CA. Saliba, DA Lang, and BG Vickrey.

Objectives: Efforts to improve the quality and outcomes of healthcare through clinical practice guidelines have generally been disappointing: while hundreds of guidelines are developed and disseminated yearly, studies continue to show low levels of guideline use and minimal impact on clinical practice. The Executive Committee of the California Workgroup on Guidelines for Alzheimer's Disease Management in Primary Care (a broad public-private coalition including VHA representatives) sought to avoid this problem by commissioning the Sepulveda Field Program to form a guideline dissemination/implementation team. Our overall goal was to enhance acceptance and impact of the guideline, which was then still under development. We sought to accomplish this goal by (1) collaborating with the guideline development Workgroup to ensure that the guideline's content, format and other characteristics reflected current thinking regarding attributes facilitating guideline use and acceptance, and by (2) developing a comprehensive dissemination and implementation plan containing recommendations for the Executive Committee and for provider organizations scheduled to receive the guideline.

Methods: Our efforts to influence guideline content and format began with presentations at Workgroup and subcommittee meetings regarding dissemination and implementation challenges and solutions. We presented results of published research and reviews, and detailed examples from our own work. We illustrated our recommendations by revising specific guideline sections and by drafting recommended additions (e.g., purpose and scope statements). Our content/format recommendations addressed language (e.g., clear, declarative recommendation statements rather than unstructured text or broad, ambiguous statements), overall guideline structure (e.g., development of one-page recommendation list and accompanying narrative summarizing relevant evidence) and other issues. Our dissemination/implementation report included a brief analysis of the guideline's features relevant to implementation success and strategies, assessment of the guideline's likely impact on quality, cost and other outcomes, and detailed recommendations for dissemination and implementation activities, accompanied by reviews of relevant literature and other evidence. We addressed (1) guideline development (e.g., recruitment of key stakeholder groups as collaborators in guideline development, to ensure buy-in and subsequent dissemination support), (2) publicity (collaborative press conferences and guideline release activities), (3) publication (preparation of a manuscript for journal publication in conjunction with the stand-alone guideline document, preparation of summaries for inclusion in guideline directories and compilations), (4) guideline tools (e.g., medical record checklists, patient brochures), (5) implementation support packages (sample educational program outlines and resources, detailed plans for provider organization implementation projects), and (6) funding strategies (recommendations regarding private foundations, for-profit industry organizations such as pharmaceutical companies, and government agencies).

Results: The guideline development Workgroup, subcommittees and Executive Committee accepted nearly all of the dissemination/implementation team's recommendations regarding guideline content and format, and is currently implementing many of the dissemination/implementation report's recommendations.

Conclusions: Guideline dissemination and implementation remain key challenges to success of clinical practice guidelines as quality improvement tools. Attending to these challenges during the development of a guideline, rather than after its release, should help meet many of these challenges.

Impact: Development of the guideline was enhanced by the emphasis placed on dissemination and implementation needs and the resulting awareness and attention the guideline development Workgroup devoted to implementation barriers and challenges. Acceptance and use of the guideline are likely to be enhanced as well.

93. Health and Functional Status Differences Between Veterans Admitted to State Veterans Homes Compared to Community Nursing Homes

Joan Penrod, MSW, PhD. University of Nebraska Medical Center, Omaha, NE. M Vandivort, L Zhao, D Shutzer, and J Potter.

Objectives: To examine factors associated with admission of veterans to state veterans homes (SVHs) compared to community nursing homes (CNHs) in Nebraska.

Methods: The Minimum Data Set (MDS) was obtained for veterans admitted to SVHs and CNHs in Nebraska over a 12 month period. Demographic, health, and functional status differences in the two admission cohorts were examined. Multiple logistic regression was used to estimate the effects of admission functional and cognitive status, age, gender, social support, marital status, presence of behavior, mental health, and incontinence problems, and number of medications used on the probability of admission to a SVH versus a CNH.

Results: Veterans admitted to Nebraska SVHs (N = 188) are significantly less functionally impaired at admission (p= .003) compared to those admitted to CNHs (N = 126). However, a history of mental problems (p< .001) and use of anti-psychotic medication (p = .001) is more prevalent in the SVH admission cohort. Moreover, the results of the multiple logistic regression indicate as functionally disability increases, veterans' odds of admission to a SVH as compared to a CNH decrease (p = .01). Also, veterans with cognitive impairment and mental health problems are more likely to be admitted to a SVH rather than a CNH. Finally, veterans living alone, admitted from home rather than a hospital or nursing home, and those with daily contact with family are less likely to enter a SVH compared to a CNH. Age, gender, incontinence, and marital status had no effect of the odds of admission to SVHs.

Conclusions: The SVHs in Nebraska appear to admit from a different population of veterans than the CNHs. In particular, SVHs are more likely to accept veteran residents with mental health problems compared to the CNHs. However, Nebraska SVHs attract a less functionally disabled admission cohort. Nebraska SVHs have a policy of not accepting ventilator dependent residents. This may account for part of the higher level of functional disability in the CNH veteran cohort. Furthermore many CNHs are reluctant to take residents with significant mental health problems and by rule, can not accept applicants if a psychiatric illness is the primary diagnosis at admission. Thus, the SVHs may serve veterans who would otherwise have difficulty with traditional institutional long-term care arrangements.

Impact: Reorganization, consolidation, and closer management of VA nursing home resources is a major focus at the Federal, VA Network, and State level. This study provides relevant information to assist in that effort. Specifically, moves to reduce reliance on SVHs will need to consider alternative care arrangement for veterans with mental health problems.

HSR&D Funded: DEV 97-019

94. Access to Care Among Adults with Diabetes in VA and County Clinics

John Piette, PhD, Center for health Care Evaluation, Menlo Park, CA.

Objectives: In a prior study, we found that VA patients with diabetes were substantially more satisfied with their access to care than similar patients in a nearby county health care system. In this study, we sought to determine the actual prevalence of access problems perceived by patients in the two systems.

Methods: Data were collected via structured telephone surveys of patients participating in two randomized trials of automated diabetes management with nurse follow-up. Participants in the trials were recruited at the time of visits to general medicine clinics in each health system. Data for the current study were from follow-up surveys for 108 patients recruited through 4 VA clinics and 132 patients recruited through 2 county clinics who all were randomized to the control groups.

Results: Compared to county patients, VA patients were older, more likely to be married, and had better glucose control. In the six months prior to their surveys, between 9% and 10% of county patients reported that they failed to refill a prescription, seek care for an urgent health problem, or call an ambulance because of cost concerns. However, less than 2% of VA patients perceived similar barriers (each p <.01). Compared to county patients, fewer VA patients failed to seek medical advice by phone because they did not know the number or thought that they would be put on hold (14% versus 23%, respectively; p = .07). Fewer VA patients failed to seek urgent care because they thought that the clinic would not be helpful (2% versus 11%; p < .01). Roughly equal numbers of patients in both systems failed to refill a prescription because they "didn't know how" (13% for VA patients and 11% for county patients, p = .66). None of these differences were substantially changed when we adjusted for sociodemographic factors that could affect perceived access (e.g., social support, education, and age).

Conclusions: Overall, the proportion of VA patients with diabetes who perceived financial barriers to accessing medical care was small on an absolute level and relative to county clinic patients. Relative to county patients, VA patients perceived the system's telephone services and emergency care services as more accessible. Nevertheless, a significant minority of VA patients perceived problems accessing telephone care services and medication refills.

Impact: VA policy-makers can be assured that few patients with diabetes perceive financial barriers to receiving care. The problems reported by county patients suggest that VA patients' perceptions might be worsened if they faced significant co-payments. VA should focus on improving the perceived accessibility of telephone care and medication refills.

HSR&D Funded: IIR 95-084

95. Ambulatory Care Groups and SF-36 Mental and Phusical Component Summary Scores

Kenneth Pietz, PhD, and Carol Ashton, MD. MPH. Houston VA Medical Center, Houston, TX. J Tuchschmidt, G Twedt, and J Bellah. Nelda Wray, MD, MPH. Houston VA Medical Center, Houston, TX.

Objectives: Determine the relationship between Adjusted Clinical Groups (ACGs) and SF-36 physical and mental component summary scores. Determine whether these SF-36 summary scores can be used as case-mix adjustment variables in resource utilization studies, as proxy for patients' frailty and illness burden. Determine how much of the variation in per-patient resource utilization is explained by the SF-36 summary scores.

Methods: The data consists of 24,726 patients in the Northwest Network (VISN 20) who voluntarily completed an SF-36, who had inpatient or outpatient care during the study period from September 1, 1997 to August 31, 1998 and who had a primary care provider assigned. The amount of utilization of each of nine resources for the study period was known for each patient. The resources were: bed days of care, hospital stays, medical encounters, laboratory tests and cost, radiology tests and cost, prescriptions, and pharmacy cost. The ACG of each patient was computed using he/her demographic information and inpatient and outpatient ICD-9 codes. A significant amount of the variation in patients' resource utilization is explained by the patient's ACG. For each patient the Kazis physical and mental component summary scores (PCS and MCS, respectively) were calculated from their SF-36 health self-assessment. The PCS ranged from a minimum of 3.1 to a maximum of 70.6. The mean and standard deviation were 33.4 and 11.4, respectively. The MCS ranged from a minimum of 3.8 to a maximum of 78.9. The mean and standard deviation were 44.8 and 13.9, respectively. The effect of using PCS and MCS to explain variation in resource usage was evaluated. Also, these scores were added as independent variables to the regression models which model resource utilization as a function of the patient's ACG. Finally, regression models were generated which evaluated the ability of the ACGs to explain variation in PCS and MCS.

Findings: The amount of variation in the resource utilization explained by the PCS and MCS was highest for prescriptions. The amounts of variation explained in the logarithms of prescriptions and pharmacy cost explained by PCS and MCS combined were 12.3% and 8.9% respectively, compared with 27.4% and 19.7% explained by ACGs. For the other resources the amount of variation explained by the PCS and MCS was less than 7%. The change in $-squared by adding the summary scores to the ACG models of resource utilization was negligible except for prescriptions and pharmacy costs. Finally, ACGs explain 8.1% of the variation in PCS and 5.5% of the variation in MCS.

Conclusions: The PCS and MCS summary scores add to the ability of ACGs to predict prescriptions and pharmacy cost in this population of veterans. PCS and MCS add little predictive ability to ACGs for the other utilization outcomes. The summary scores cannot replace ACGs as case-mix adjustment covariates.

Impact: SF-36 physical and mental component summary scores are not usable as case-mix adjustment covariates as a proxy for frailty and illness burden in a population of veteran patients.

96. Development of a Case Mix Adjustment Model for Hemoglobin A1c

Leonard Pogach, MD. Houston VA Medical Center, Houston, TX. QW Zhang, S Dhar, H Cheng, G Hawley,and D Repke.

Objectives: The current VA health outcome performance measure for glycemic control, the proportion of veterans exceeding an unadjusted glycohemoglobin of 10 percent, was chosen because of the absence of laboratory standardization and case mix adjustment. The objective of this study was to validate a case mix model that would permit accurate cross sectional comparisons of mean Hemoglobin A1c (HbA1c) values among administrative units.

Methods: A cohort of 53,343 patients with documented HbA1c values was identified from a larger cohort of 203,987 veterans receiving oral agents, insulin and/or blood glucose monitoring supplies from 130 participating stations in the FY1996 National Center for Cost Containment Diabetes Cost and Outcomes Report using a previously validated data extraction program based upon the DHCP pharmacy file. A hierarchical, mixed effects models was utilized that employed a nested data structure to estimate random variance and covariance components among dependent measures to increase the model's explanatory value. High-normal adjusted HbA1c levels (subtracting the upper normative limit from each individual value) were predicted by linear age (centered at the median of 66 years), quadratic age, gender, race/ethnicity, patient's financial means classification (temporary surrogate of SES), treatment modality (insulin, oral agents, insulin and oral agents, and neither insulin nor oral agents), blood glucose monitoring (yes/no), total number of VA clinic visits (natural log transformed), and the interactions of demographic variables with treatment modality and blood glucose monitoring. The last HbA1c record was utilized. Fixed parameter estimates (beta) were determined for each predictor.

Results: The intercept for the cohort mean adjusted HbA1c was 1.93 [percent above high normal value]. No difference in HbA1c level was found between gender groups or individuals receiving monitoring supplies. HgbA1c level was higher in younger than in older patients. Compared to those who used oral medications only, patients treated by insulin had higher HgbA1c results (beta=.546, p=.0001), while those treated with both insulin and oral agents during FY96 had even higher HgbA1c (beta=.96, p=.0001) while patients receiving only monitoring supplies showed a lower level of HgbA1c (beta=-1.068, p=.0001). There was a complex interaction between ethnicity and treatment modality, and Black Hispanic/African American patients had higher HgbA1c (beta=0.28, p=0.0001) than White patients for those who received oral or oral plus insulin therapy. The number of outpatient visits to a VA clinic was associated with a significantly lower level of HgbA1c (beta=.115, p=.0001). The predictor betas were almost identical whether adjusted or unadjusted means or threshold proportions were utilized. However, the modality of reporting HbA1c results determined which facilities were classified as outliers from the national means. After adjusting for the upper limit of the normative ranges, a significant (p<0.05) difference in mean adjusted HbA1c occurred in 26 of 48 participating facilities.

Conclusions: Age, ethnicity, treatment modality and clinic visits, but not gender, socioeconomic status or blood glucose monitoring were statistically significant predictors of HbA1c values.

Impact: Future diabetes performance measures for HbA1c could utilize case mix adjustment to compare mean HbA1c values of populations as well as proportions of individuals exceeding a threshold measure.

97. Extending the Use of the Self-Administered Quality of Well-Being to Patients with Depression

Jeff Pyne, MD. North Little Rock VA Medical Center, North Little Rock, AR. WJ Sieber, K David, and DK Williams

Objective: We examined the sensitivity of the self-administered Quality of Well-Being (QWB-SA) Scale to changing levels of depression severity. The original QWB Scale was designed as a generic measure for use in cost-effectiveness analyses and provided the recommended health policy effectiveness unit of quality-adjusted life years (QALYs). The QWB-SA was developed in response to concerns that the original interviewer version of the QWB took too long and was too costly to administer. The QWB-SA takes approximately 7 minutes to complete, has been used to estimate QALYs in migraine and arthritic patients, and has well-established validity and reliability.

Methods: A total of 67 subjects participated in the study: 51 male (39 inpatient and 12 outpatient) and 16 female (7 inpatient and 9 outpatient). All subjects were diagnosed with a current major depressive episode using the Structured Clinical Interview for DSM-IV and were simultaneously administered the QWB-SA, the interview version of the QWB Scale, 17-item Hamilton Rating Scale for Depression (HRSD), and Beck Depression Inventory (BDI). The output from both QWB scales is a single score between 0.0 (death) and 1.0 (perfect health). The study design was observational. Acute treatment response was determined using weekly QWB and depression ratings during 4 weeks of medication treatment or until criteria for a 50% improvement from baseline HRSD was met (defined as the mean HRSD score from weeks 1 and 2), and longer term response was determined at 4-months post-treatment.

Results: There were no differences found between male and female subjects on any measure and therefore the data were combined for all reported analyses. To place the QWB-SA scores in context, the reported mean score (SD) for migraineurs without headache was .619 (.15), with headache .484 (.15); patients with rheumatoid arthritis .509 (.14); in the current sample depressed outpatients .479 (.12), inpatients .382 (.12). Baseline correlations between QWB-SA, HRSD and BDI were -.38 (p<.01) and -.30 (p<.05), respectively. Acute change score correlations between QWB-SA, HRSD and BDI were both -.27 (p<.05). Four month change score correlations between QWB-SA, HRSD and BDI were -.51 (p<.01) and -.57 (p<.01), respectively. Interviewer QWB change score correlations with HRSD and BDI were non-significant for acute treatment response and less robust for the 4-month response.

Conclusions: The QWB-SA scores for depressed subjects were similar or lower than QWB-SA scores for other chronic disorders, namely migraine and rheumatoid arthritis patients. The baseline QWB-SA correlation with depression severity is significant and consistent with a medium effect size. The acute and 4-months post-treatment change score correlations are significant and consistent with a medium and large effect size, respectively. The QWB-SA appears to be more sensitive to acute change in depression severity than the interviewer QWB.

Impacts: The QWB-SA is an inexpensive and low staff impact method for documenting the cost-effectiveness of medical interventions and for comparing the cost per QALY across specialties. The QWB-SA may be a valuable measure to inform healthcare resource allocation decisions.

98. The Relationship between Functional Status and Satisfaction of Care among Patients Served by the Veterans Health Administration

Xinhua Ren, PhD and Lewis Kazis, ScD. Bedford VA Medical Center, Bedford, MA. A Lee and W Rogers. Susan Pendergrass, DPH. Department of Veterans Affairs, Ridgeland, MS.

Objectives: As the Veterans Health Administration (VHA) places high priority on quality of care, comprehensive frameworks for quality accountability have been established to manage performance within the VHA. To address this accountability, there is an increasing need for the VHA to assess the outcomes of health care. In this presentation, we assess the relationship between two domains of VHA health care value, patient satisfaction of care and functional status, in the context of sociodemographic characteristics such as age, gender, race, marital status, educational attainment, and periods of service (WWII, Korea, Vietnam, and Persian Gulf).

Methods: We analyzed both the cross-sectional data from the 1996 National Survey of VA Ambulatory Care Patients and the cohort data from the 1998 National Survey of VA Ambulatory Care Patients. The 1996 Survey included 32,631 patients, among whom 26,424 patients were followed in the 1998 Survey. Participants in the study were mailed a health status (the SF-36V) and a Customer Service Standards (or patient satisfaction) questionnaire. Using the cross-sectional data, we first conducted Pearson's product moment correlations to examine the relationship between functional status and patient satisfaction. Then using ordinary least squares regression (OLS), we examined the effects of sociodemographics (such as age, gender, race, marital status,, education, and periods of service) and functional status on domains of patient satisfaction. Using the cohort data, we conducted "cross correlations" between patient satisfaction and functional status in wave 1 (1996 Survey) and patient satisfaction and functional status in wave 2 (1998 Survey) to examine the causal relationship between functional status and patient satisfaction.

Results: In the cross-sectional data analysis, the SF-36V scales had significant, negative correlation coefficients with the satisfaction scales (p < 0.001), indicating that patients with better functional status were less likely to be dissatisfied with the care they received within the VHA. Furthermore, compared to physical health (or PCS), mental health (or MCS) seemed to be more strongly correlated with patient satisfaction. Patients who were older, white, married, and more educated had significant, negative regression coefficients on satisfaction scales (p < 0.01), indicating that those patients were less likely to be dissatisfied with the care they received within the VHA. In the cohort data analysis, the relationship between physical (PCS) or mental (MCS) health and domains of patient satisfaction was rather stable over time; however, the effect of PCS or MCS on patient satisfaction seemed to be more important than that of patient satisfaction on PCS or MCS.

Conclusions: The study results suggest that after adjusting for sociodemographics such as age, gender, race, marital status, education, and periods of service, patients with poorer functional status are more likely to be dissatisfied with the care they received from the VHA than patients with better functional status.

Impact: Results have implications for the use of the SF-36V as an important adjuster in future work. The physical and mental component summary scales from the SF-36V are both correlated with the satisfaction scales, suggesting that comparisons across VISNs should consider both unadjusted and adjusted satisfaction scale scores.

99. VISN 10 Automated ICU Severity Adjustment Tool: SISVISTA

Marta Render MD. Cincinnati VA Medical Center, Cincinnati, OH. SH Timmons, R Hayward, D Welsh. Timothy Hofer, MD, MSc. Ann Arbor VA Medical Center, Ann Arbor, MI.

Objective: Proprietary intensive care unit (ICU) risk adjustment tools are costly. We developed an automated ICU risk adjustment tool using only those variables from APACHE III found in the computerized database of the Department of Veterans Affairs (VistA).

Methods: We included the first admission of each patient to any intensive care unit in three Ohio Veterans Affairs hospitals from 2/1/98 through 7/31/97. Customized programming in the VistA database at each facility identified eligible patients and extracted relevant variables: diagnosis, age, comorbidity, admission source (direct or transfer) and a modified Acute Physiology score, hematocrit, white blood count, albumin, sodium, glucose, bilirubin, blood urea nitrogen, creatinine, partial pressure of oxygen/fraction of inspired oxygen, partial pressure of carbon dioxide and pH. A window surrounding admission (+ 24 hours) identified variables and applied APACHE III weights. Direct admissions were those patients admitted to the ICU from the emergency area or transferred to the ICU within 1 day of surgery; all others were considered transfers. APACHE III comorbid diagnoses were defined using ICD.9.CM coding from the inpatient treatment file (PTF). The ICD.9.CM ICU discharge code from the PTF file was sorted first into APACHE III diagnostic groups, then into 10 groups of organ system dysfunction.

Statistics: Odds ratios for predictor variables, summary statistics including the C statistic, and goodness of fit testing are reported for the logistic regression model predicting 30 day mortality.

Results: 4626 of 4851 patients (99.5% of the cohort, 477 deaths) had complete datasets. Of the 4626 patients, 1389 were classified as post-operative (30%). Patients were 63 + 12 years and predominantly male (97.5%). 74% of patients were directly admitted to ICU from the emergency area, and 26% were transferred into the ICU. Mortality was increased in patients transferring to the unit (21%) compared to those directly admitted (6.5%, p = 0.001); and increased as the length of time from hospital admission to ICU transfer increased (mortality at 1 day 15.6%, at 2-4 days 19%, and at > 4 days 25.2%, p = 0.002). The most common diagnoses were cardiovascular diseases (MI, CHF). In multivariable analysis, age score (OR [95%C1] = 1.05 [1.02, 1.07]), some diagnostic systems (Hematology, neurology, orthopedics, respiratory, sepsis and trauma, p<0.05), comorbidity score (OR [95%C1]=1.11[1.08, 1.1]), direct admission (OR [95%C1]-0.4[0.3, 0.5]), and the sum of the laboratory scores (OR [95%C1]=1.07[1.06,1.08]) were significant. The model had excellent discrimination and calibration (C statistic = 0.85, Hosmer &##150; Lemeshow, p = 0.07).

Conclusion: This relatively economical automated ICU risk adjustment system has discrimination similar to existing risk adjustment systems (APACHE II, SAPS II).

Impact: An automated risk adjustment system has broad potential administrative and research applications in the VHA locally, regionally and nationally, and in community health care systems with advanced information systems. These results are being validated in a national database.

HSR&D Funded: DEV 97-032

100. Psychogeriatric Interventions Saves Inpatient Costs: Preliminary Findings

Joel Rosansky, LCSW, R Anderson, R Bastani, R Gould, D Huang, L F Jarvik, G Kominsk, A E Maxwell, J Sanchez, and S Taylor, UPBEAT Program Staff. VA Greater Los Angeles Healthcare System, Los Angeles, CA.

Objectives: The Unified Psychogeriatric Biopsychosocial Evaluation and Treatment (UPBEAT) Program is a multi-center, interdisciplinary, randomized, six year outcome-based clinical demonstration project sponsored by the Department of Veterans Affairs to improve health care and reduce utilization by geriatric veterans. Specifically, UPBEAT Care is expected to reduce inpatient utilization and unscheduled hospital visits, and increase consumer satisfaction.

Methods: UPBEAT staff at each of the nine participating VA Medical Centers use the Veterans Health Information System and Technology Architecture (VISTA) to identify patients 60 years or older, admitted to acute medical/surgical inpatient services. Patients are invited to participate in the program evaluation unless they meet exclusion criteria. Exclusion criteria include: psychiatric appointments during the preceding or subsequent 6 months, diagnosed with Post Traumatic Stress Disorder, psychosis, dementia or other cognitive impairment, spinal cord injury, rehabilitation plan, outside catchment area, homeless, admission from nursing home, chemotherapy, hospice care, or previous assignment to UPBEAT or Usual Care. Eligible patients providing informed consent are interviewed at bedside with the anxiety and depression subscales from the Mental Health Index (MHI-38), the Alcohol Use Disorder Identification Test (AUDIT) and the RAND 36-item Health Survey Short Form (SF-36) for patients meeting cutoff scores for anxiety, depression, or alcohol misuse symptomatology. Qualified patients are randomized to UPBEAT or Usual Care. UPBEAT Care patients receive in-depth diagnostic assessments, including the following scales: Geriatric Depression, Hamilton Anxiety and Depression, Lubben, Affect Balance, Activities of Daily Living, Cumulative Illness Rating, Mini-Mental State Examination, Multi-Axial Diagnostic Assessment. Each UPBEAT Care patient is assigned to an UPBEAT Care Coordinator from an interdisciplinary team. UPBEAT Care teams, with expertise in psychogeriatrics, are staffed from psychiatry, nursing, social work, and at some sites psychology. Based on the diagnostic assessment, UPBEAT Care Coordinators provide patients with referrals to community and hospital services, or if necessary, direct interventions utilizing educational, psychosocial, psychotherapeutic or psychopharmacologic approaches. Outcome evaluation includes follow-up interviews of both UPBEAT and Usual Care patients at 6, 12 and 24 months after enrollment. In addition, utilization and cost data are obtained for both groups from VHA databases.

Results: Based on preliminary data from the Allocation Resource Center for 420 UPBEAT Care and 462 Usual Care patients, the net changes in inpatient utilization between 12 months pre- and 12 months post-enrollment yielded a first-year savings of 5.4 inpatient days/patient for UPBEAT Care over Usual Care. This resulted in a savings of over $2,184,000 based on the Cost Distribution Report average inpatient cost of $963/day.

Conclusions: The UPBEAT Program, i.e. skilled clinical assessment with identification of symptoms of depression, anxiety and alcohol misuse followed by community-based mental health care management by an interdisciplinary team with expertise in psychogeriatrics, resulted in reduced inpatient utilization and thereby, cost savings.

Impact: UPBEAT is expected to demonstrate that skilled clinical assessment and community-based mental health care coordination (primary and secondary mental health and other health care intervention and prevention) by an interdisciplinary team with expertise in psychogeriatrics can yield enhanced delivery of integrated health care and decreased utilization of expensive acute inpatient services (medical/surgical/ psychiatric).

Questions about the HSR website? Email the Web Team

Any health information on this website is strictly for informational purposes and is not intended as medical advice. It should not be used to diagnose or treat any condition.