Veterans with schizophrenia consume a disproportionate amount of health services in the Veterans Health Administration (VA) yet still have poor psychiatric and medical outcomes. The mean lifespan of patients with schizophrenia is 10-20 years less than that of the general population. Because of this mortality profile, age 50 and older may be considered "late-life" for these veterans. The literature is mixed as to the quantity and quality of care older schizophrenia patients receive, relative to patients with the same medical illnesses absent schizophrenia. Cross-sectional study design or rigorous (but poorly generalizable) inclusion criteria (such as studying only patients well engaged in care) may explain mixed findings.
Patients with schizophrenia experience elevated rates of several medical conditions, including cardiovascular disease, respiratory problems, and diabetes (DM). We proposed comparing the care and outcomes of VA patients with schizophrenia, diabetes, or both diagnoses. Diabetes is common among VA patients (~20% of patients), incurs significant morbidity and mortality, is treated with an array of medications, and has several indicators of care recorded in VA databases (clinic visits, blood glucose and HbA1c, blood pressure control). It is a compelling disease to examine in schizophrenia patients because (a) schizophrenia patients have limited insight into medical or psychiatric conditions and are frequently non-compliant with treatment recommendations; (b) schizophrenia itself appears to be an independent risk factor for DM; (c) the medications used to treat schizophrenia increase the risk of blood glucose abnormalities/DM; (d) common comorbidities are the same illnesses monitored among DM patients: hypertension and hypercholesterolemia; and (e) both DM and schizophrenia require high levels of involvement with the health care system. Overall, we hypothesized that lack of persistent engagement in care was an independent predictor of premature mortality among veterans with schizophrenia.
OBJECTIVE 1: Compare patterns of care among patients aged 50+ with schizophrenia, diabetes, or both schizophrenia and diabetes. Hypothesis for Objective 1: Patients with schizophrenia are more likely to reduce their use of VA care than diabetes patients, variably over time. OBJECTIVE 2: Compare prescription profiles (e.g., use of antipsychotic or antidiabetic agents) and clinical indicators (e.g., blood glucose, A1c testing, lipids). Hypotheses for Objective 2: Patients with schizophrenia are less likely to fill prescriptions for appropriate medications than DM-only patients. Schizophrenia patients have fewer assays and greater fluctuations in glucose, HbA1c, and lipid levels over time than DM-only patients. OBJECTIVE 3: Assess the impact of patterns of care and intermediate processes and outcomes on mortality among the three groups of patients. Hypothesis for Objective 3: Patients with both schizophrenia and diabetes experience higher mortality than age-matched patients with diabetes or schizophrenia alone. OBJECTIVE 4: Analyze impact of delayed diagnosis/ treatment of diabetes on mortality among patients with schizophrenia without identified diabetes at baseline (FY02), hypothesizing that longer delay from an abnormal blood glucose or A1c test to diabetes diagnosis/ treatment was associated with higher mortality, adjusting for covariates. OBJECTIVE 5: Assess frequency of assessment of cardiovascular risk factors and the relative predictive value of these factors on mortality. Cardiovascular risk factors include: high blood glucose or hemoglobin A1c, triglycerides, blood pressure, and body mass index; and low high-density lipoproteins. We hypothesized that fasting glucose and A1c tests would maintain significant predictive value, controlling for other cardiovascular risk factors, in models of mortality for patients with schizophrenia.
Study design: retrospective cohort study. Major characteristics: veterans with schizophrenia, diabetes, or both conditions age 50 or older receiving VA care nationwide; no non-VA sites. Major variables/sources of data: inpatient medical and psychiatric stays; outpatient primary, specialty, psychiatric care, laboratory and medication histories, height-weight assessments, and demographic and service variables derived from VA administrative databases (VA-Medicare data inclusion was proposed but not achieved) for fiscal years FY02 through FY05. Main types of analysis: generalized linear models appropriate to the distribution of the outcome measures (e.g., Poisson regression for count data, logistic regression for dichotomous outcomes), controlling for demographic and clinical correlates as well as facility-level clustering of patients as needed. Random effects accounted for within-subject correlation of repeated measures for some models.
In the total sample of 241,466 VA patients who met eligibility criteria after exclusions for (a) no valid VA priority status, or (b) utilization more than 30 days after death date, there were 53,134 with schizophrenia, and 188,332 patients with diabetes but no schizophrenia (age-matched 4:1 to schizophrenia patients). About 25% of the schizophrenia patients had comorbid diabetes at the start of the study (fiscal year 2002; n=13,025) leaving 40,109 with schizophrenia only. The mean age for the sample was 61 years (+/- 10) with a range of 50-104 years. Mortality over the 4-year study was 14.5% overall, with differential rates by diagnostic sub-group: 13.7% of diabetes-only patients died, 16.3% of schizophrenia-only patients died, and 20.6% of schizophrenia patients with comorbid diabetes at baseline died. Race data were missing on 26% of the sample while the remainder were 53% white, 15% African-American, 6% Hispanic. Women comprised 2.2% of the sample.
Patterns of primary care utilization included Increasing over 4 years, Consistent, High and Decreasing, or Low and Decreasing. The interaction of primary care pattern with diagnostic group was significant in the model predicting survival. Adjusting for age, race/ethnicity, marital and priority status, other morbidity and medication use, Low-decreasing primary care was associated with 4-fold risk of death for both schizophrenia and diabetes patients; the effect was 5-fold for patients in the schizophrenia+diabetes group. A pattern of High-decreasing primary care was associated with a less dramatic increased relative risk of death, about 2-fold.
In terms of baseline (FY02) prescription profiles and clinical (lab) indicators, schizophrenia-only patients were less likely than their diabetic age-matched counterparts to be receiving antihypertensives (36% vs 83%) although having schizophrenia with comorbid diabetes was associated with greater receipt of antihypertensives (72%). Having co-occurring schizophrenia with diabetes was also associated with decreased receipt of hypoglycemic prescriptions relative to having diabetes alone (76% vs 88%). The combination of schizophrenia plus diabetes also appeared to be associated with slightly decreased receipt of antipsychotic medication (83% vs 87%). These relative patterns recurred each of the subsequent years of the study, FY03-FY05.
Among the patients with schizophrenia only at baseline (n=39,825 after excluding those prescribed hypoglycemic medication absent a diabetes diagnosis), we examined available blood glucose and hemoglobin A1c tests. To proxy fasting glucose, we used blood glucose results where there was a same-day low-density lipoprotein test result, because lipid panels require fasting (medical record extracts do not record fasting status). Using cut-points from the literature and the American Diabetes Association, 32% of older patients with schizophrenia had either a proxy fasting blood glucose or A1c test during the year, in spite of multiple risk factors for developing diabetes. Combining the tests, 5,353 patients had pre-diabetic dysglycemia per single test (13%) and 1,291 patients had likely diabetic dysglycemia per single test (3%). Only 12% of pre-diabetic patients were diagnosed with or treated for diabetes during the next three years (deaths and lost-to-follow-up excluded), although 25% of dysglycemic patients might be expected to progress to diabetes over 3-5 years per the literature. Mortality was higher among dysglycemic patients if their condition remained undiagnosed/untreated over the next three years (13% vs 8%).
Analysis of lab data, blood pressure assessments, and height-weight measures yielded a model of mortality over the study period (FY02-FY05) that also adjusted for demographic and service measures. The sample was again restricted to patients with schizophrenia but no diabetes at baseline, to allow examination of the effect of developing comorbid medical disease. Interestingly, while baseline high blood pressure and elevated blood glucose were associated with mortality, elevated lipids and obesity showed "protective" effects; however, the most telling part of the analysis was in the estimated effects of missing data. Indicators for missing data on blood glucose/A1c, on triglycerides, on HDL and LDL, and on height-weight were included in the model and three of these six indicators were significantly associated with increased mortality.
VA patients with chronic conditions that are psychiatric, specifically schizophrenia, continue to lag behind those with chronic medical disease, such as diabetes, in terms of receipt of primary care. Subsequently, survival remains foreshortened for these patients. Deficits in patient self-management, such as appointment-keeping, prescription-filling, illness insight, and pill-taking, may be primary drivers of the adverse outcomes observed in the data, suggesting the need to shift the burden of care management to social contacts (caretakers) or system processes (such as case managers or reminder and outreach efforts). Some low-cost interventions may yield survival benefits, such as obtaining lab measurements of lipids and blood glucose, prescribing hypoglycemic medication for older schizophrenia patients, and assessing height and weight to monitor weight management. Results have been disseminated to the Mental Health and Diabetes QUERIs through our QUERI advisors and, it is hoped, will inform the development of future interventions to provide effective chronic illness management for these complex patients.
External Links for this Project
- Copeland LA, Zeber JE, Wang CP, Parchman ML, Lawrence VA, Valenstein M, Miller AL. Patterns of primary care and mortality among patients with schizophrenia or diabetes: a cluster analysis approach to the retrospective study of healthcare utilization. BMC health services research. 2009 Jul 26; 9:127. [view]
- Copeland LA, Parchman ML, Zeber JE, Lawrence VA, Downs JR, Miller AL. Prediabetes assessment and follow-up in older veterans with schizophrenia. The American journal of geriatric psychiatry : official journal of the American Association for Geriatric Psychiatry. 2010 Oct 1; 18(10):887-96. [view]
- Lawrence VA, Noel PH, Yoder LH, Gaylord K, Cornell JE, Johnson AJ, Hsu JR, Wolf SE. Health status of combat-injured military personnel at hospital discharge. Poster session presented at: Society of General Internal Medicine Annual Meeting; 2012 May 1; Orlando, FL. [view]
- Parchman ML, Copeland L, Zeber J, Hosek B, Lawrence V. Identifying pre-diabetes among veterans with schizophrenia. Poster session presented at: AcademyHealth Annual Research Meeting; 2008 Jun 1; Washington, DC. [view]
- Copeland LA, Miller AL, Lawrence VA, Hosek B, Parchman ML, Zeber JE. Pre-diabetes and diabetes assessment in veterans with schizophrenia. Paper presented at: International Congress on Schizophrenia Research; 2009 Mar 28; San Diego, CA. [view]
- Copeland LA, Zeber JE, Parchman ML, Downs JR, Miller AL, Lawrence VA. Prediabetes Assessment and Follow-up in Older Veterans with Schizophrenia. Poster session presented at: VA HSR&D Field-Based Mental Health and Substance Use Disorders Meeting; 2010 Apr 27; Little Rock, AR. [view]
- Wathen P, Johnson M, Lindner P, Lawrence VA. Waxman Clinical Skills Center: Lumbar Puncture. Paper presented at: American College of Physicians Annual Meeting; 2012 May 1; New Orleans, LA. [view]