HSR&D Home » Research » RRP 12-479 – HSR&D Study
Predicting Pressure Ulcer Risk in VHA Facilities
Linda J. Cowan, PhD MSN ARNP
North Florida/South Georgia Veterans Health System, Gainesville, FL
Funding Period: December 2013 - February 2016
This project proposal was submitted in response to solicitation for proposals that would support evaluation of implementation of the VHA Handbook 1180.02, "Prevention of Pressure Ulcers" (issued in July 2011). The handbook provides comprehensive procedures for the assessment and prevention of pressure ulcers (PUs). In March 2012, 14 PU nurse investigators were asked to participate in the 1st VA Nursing PU Summit sponsored by the VA Office of Nursing Services (ONS). The group developed a strategic plan for research related to PUs and implementation of the new VHA handbook to address key areas of importance to VA PU prevention. The PI and Co-Is were tasked with addressing PU risk, risk factor identification and, specifically, to further the preliminary pilot research that the PI had previously conducted on PU risk assessment.
Critical to PU prevention is the accurate identification of risk. The Braden Score for Predicting Pressure Sore Risk (Bergstrom, Braden, Laguzza, & Holman, 1987) is the most commonly used tool in VHA facilities. This tool is a six-factor tool developed in 1987, based on nursing observations, and is embedded in the electronic record (nursing documentation template) for all admissions, transfers and discharges, as well as daily and weekly skin assessments. Current research identifies strong PU risk factors in acute care such as advanced age, specific medical conditions affecting tissue perfusion, oxygenation and cellular function, as well as physiological status changes. These risk factors are not accounted for by the Braden tool (Cowan, Stechmiller, Rowe & Kairalla, 2012). PU prevention guidelines (WHS, 2008; NPUAP, 2009; WOCN, 2010) stipulate that if significant risk factors are not accounted for by current risk assessment tools, effective preventive interventions may not be implemented.
The development of a more accurate PU risk predictive model consisting of the best predictive risk factors will provide clinicians with improved methods to identify Veterans at highest risk for developing PUs and guide future clinical strategies for preventing PU within our Veteran population.
The purpose of this study is to determine the adequacy of the use of Braden Scale Pressure Ulcer Risk documentation by nurses for Veterans in acute care settings. The secondary purpose of this study is to develop a predictive model to identify Veterans in acute care that are at highest risk of developing a PU. Specifically, we hope to pinpoint which additional PU risk factors listed in the July 2011 VHA PU Prevention Handbook are the strongest predictors of PU risk for Veterans in acute care, and develop a predictive model accordingly. Our overarching goal for this program of research is the scientific support for the development of a new tool that identifies high risk and links to appropriate PU preventive interventions. This proposed study is the first step in the planned program of research.
The study design is a nested case-control that includes a retrospective chart review of electronic medical records aimed at building a predictive model of pressure ulcer risk. Independent variables included demographic data and hospitalization information such as age, race/ethnicity, and length of stay (LOS) for hospital admission. Medical diagnoses were also included as independent variables. The diagnoses were reported as categorical variables with YES/NO (1=yes, 0 = no) to indicate if these diagnoses are present in the chart during the time of hospitalization (prior to PU formation for the PU group) and will also be identified using International Classification of Diseases version 9 (ICD-9) for discharge diagnoses recorded for all subjects. These diagnoses include: severe nutritional compromise (or malnutrition); pneumonia/pneumonitis; candidiasis/mycoses; CVA/paralysis including spinal cord injury, immobility (ALS, MS, MD); decreased strength/muscle mass (muscular atrophy/'frail' elderly); acute renal failure; anemia; infection (including gangrene/bacterial skin infection, urinary tract infection, osteomyelitis and sepsis); acute respiratory failure; diabetes mellitus with complication; alterations in consciousness or awareness (including senility/dementia); alterations in sensation (peripheral neuropathy; paresthesias, as well as pain >4 on scale of 1 to 10); drugs (NSAIDS, steroids, chemotherapy agents, continued nicotine use); devices/equipment (documented CPAP, BIPAP, catheters, restraints; adaptive equipment/splints); end stage renal disease; congestive heart failure; peripheral vascular disease, vasculitis and other collagen vascular disorders; immune deficiency states, malignancies; chronic obstructive pulmonary disease; friction/shear, contractures/spasticity, poor posture while seated (pelvic obliquity/sacral sitting); bowel/bladder incontinence, abnormal fluid accumulation (edema); and dehydration. Medical factors such as surgery during the hospitalization and a diagnosis or documented history of previous PU (prior to reviewed hospitalization) were also included as a categorical independent variable (YES/NO). Braden Scale total scores and sub-scores were recorded as independent variables with the most recent Braden total scores and sub-scores that were recorded prior to the development of PU (in the PU group) or the lowest Braden total scores (with sub-scores) documented during the admission of interest (in the NO PU group).
The outcome or dependent variable is reported as a dichotomous variable (DID develop PU / DID NOT develop PU) and will be identified by any diagnosis of PU stage II or greater (including deep tissue injury discoloration & unstageable PUs) during the acute inpatient admission. Frequencies by stage of PU are also reported. Frequencies are reported for race, history of previous PU, and each predetermined high-risk diagnoses for the total sample.
Inclusion/Exclusion Criteria: Inclusion criteria included the following: Veterans age 18 or older hospitalized for at least 3 days in an acute care facility during 2011 with at least 2 recorded Braden Scale scores during the hospitalization period; and who had no report of PU in 2010 (to define a cohort at risk for development of new HAPU). From this primary pool, all Veterans who developed PU stage II or greater (including Deep Tissue Injury and Unstageable PU) anytime during 2011 were identified as potential cases (HAPU group) and those remaining were identified as potential controls. Exclusion criteria for the eligible pool include the following: Veteran records were excluded if the Veteran was not hospitalized in an acute care Veteran facility during the identified date span or whose records were missing 3 or more required pieces of information, or did not have any recorded Braden Scores. If records were missing more than three pieces of required information on the data collection sheet, the record was excluded from analysis. From a pool of 17,000 Veteran's electronic health records nationwide, 500 Veteran charts were identified who had a diagnosis of pressure ulcer during a 3 day or longer acute hospitalization during 2011, and 500 Veteran charts were identified who did not develop pressure ulcers during acute care hospitalizations (in similar locations) during 2011. Random selection of charts from these two groups resulted in a total of 355 Veterans charts who met the inclusion/exclusion criteria (by preliminary data query). However, after chart review, 246 of these charts were excluded from analyses for the following reasons:
82 with no acute care hospitalizations in 2011 (data mining identified these as having acute care hospitalizations in 2011 for > 3 days but it was not true) - 16 of these Veterans were identified by the data "pull" as being hospitalized in 2011 but they died before 2011 (10 in the "PU group" and 6 in the "No PU group")
3 with hospitalizations did occur 2011 but were not over 2 days, as suggested by the data pull
14 were mental health admissions and not acute care medical admissions (2 in "PU" group and 12 in no PU group)
54 had pressure ulcers present on admission (POA) and these were not acquired during the hospitalizations
47 in the "PU" group did not actually have any PU noted in any documentation during the entire hospitalization or any other acute care hospitalization for 2011- however, they DID have other types of wounds (11 DFU, 9 VLU, 5 surgical, 6 cellulitis, 1 incontinence associated dermatitis, 17 'other')
29 in the "PU group" had no PU and no wounds of any kind documented during the entire hospitalization
17 missing more than 5 required data pieces - they had less than 2 Braden scores documented during the entire hospitalization (14) or CAPRI would not allow access to the chart (2) or the electronic health record was that of a spouse and not a Veteran (1) Summary descriptive statistics were computed as means and standard deviations for numeric data and percentages and frequencies for categorical data. Chi-square (or Fisher's exact) tests were used to examine categorical factors possibly related to having a pressure ulcer. T-tests were used to examine numerical factors possibly related to having a pressure ulcer. To address the question of the strongest predictors of a PrU among the Braden subscales and Braden total score, a stepwise logistic model was performed which included the six subscales and the total score. To address the question of the strongest predictors of a PrU among patient characteristics, a stepwise logistic model was performed which included patient characteristics which were significantly related to presence of a PrU in one of the bivariate tests (i.e., the chi-square, Fisher exact, or t-tests). To address the question of the strongest predictors of a PrU among the Braden subscales, Braden total score, and patient characteristics, a third stepwise logistic model was performed which included the six subscales, the total score, and the patient characteristics which were significant in the bivariate analysis. The level of significance was set at .05 and two sided tests were employed for all analyses. SAS software version 9.4 (Cary, N.C.) was used for all data analyses.
Final analysis includes 109 Veteran charts: 26 with hospital acquired PU and 83 without PU.
Summary descriptive statistics were computed as means and standard deviations for numeric data and percentages and frequencies for categorical data. Chi-square (or Fisher's exact) tests were used to examine categorical factors possibly related to having a pressure ulcer. T-tests were used to examine numerical factors possibly related to having a pressure ulcer. To address the question of the strongest predictors of a PrU among the Braden subscales and Braden total score, a stepwise logistic model was performed which included the six subscales and the total score. To address the question of the strongest predictors of a PrU among patient characteristics, a stepwise logistic model was performed which included patient characteristics which were significantly related to presence of a PrU in one of the bivariate tests (i.e., the chi-square, Fisher exact, or t-tests). To address the question of the strongest predictors of a PrU among the Braden subscales, Braden total score, and patient characteristics, a third stepwise logistic model was performed which included the six subscales, the total score, and the patient characteristics which were significant in the bivariate analysis. The level of significance was set at .05 and two sided tests were employed for all analyses. SAS software version 9.4 (Cary, N.C.) was used for all data analyses. Some data were not recorded on some charts, which is why the sample size fluctuates slightly by variable.
Univariate measures: The demographic characteristics and comorbidities of Veterans in the total sample: Male (n=104), Female (n=1); Black (n=18), Hispanic (n=1), White (n=75). Race/ethnicity was not recorded on 15 charts included in analysis; Mean age of sample was 66.5 years (SD12.0); Mean BMI for total sample was 27.6 (SD 7.5). Diagnosis frequency in total sample of 109 Veterans: Altered Awareness (n=26); Altered sensation (n= 35); Anemia (n=64); CHF (n=30); COPD (n=21); CVA (n=16); Complications from implanted mechanical device (n=3); Contractures (n=2); DM complication (n=16); DM without complication (n=37); Decreased muscle (n=33); Dehydration (n=12); External devices (such as Foley catheter) (n=21); Drugs (NSAIDS, CHEMO agents) (n=41); ESRD (n=10); Gangrene (n=18); Immune deficiency (n=4); Incontinence (n=21); Malignancy (n=25); Malnutrition (n=32); Mycoses/candidiasis (n=8); Pneumonia/Pneumonitis (n=17); Renal failure (n=21); Sepsis (n=15); UTI (n=11); Vascular disease (n=27).
Bivariate measures provide results of testing for the association of patient demographic factors with PU and the association of patient demographic factors with mortality. Race (with a higher proportion of white Veterans than black), age (those with PU were older), and BMI (those with PU had lower BMI), were significantly related to PU. Race, age, and BMI were not found to be significantly related to mortality. The comorbidities of altered awareness, altered sensation, anemia, complication due to implanted device, DM complication, decreased muscle strength (diagnosis of ALS/MS/MD), use of device/equipment (CPAP, BiPAP, catheters, restraints, or adaptive equipment), gangrene/skin infection/osteomyelitis, incontinence (fecal or urinary), and presence of vascular disease were each found to be significantly related to presence of PU (p<0.05). The lowest albumin (lower = at greater risk), length of ICU stay (longer stay = increased risk), Braden total score (lower score = increased risk), and each Braden sub-score (lower score = higher risk) were all significantly related to PU (p<0.5).
The results of the stepwise logistic regression which modeled presence of PU using the six Braden subscales and the Braden total score demonstrated only Friction and Sensory subscales emerged as predictive in model. In the stepwise logistic regression including patient characteristics found to be significant in the bivariate analyses (but without Braden score variables), only altered awareness and incontinence were retained as significant predictors of PU. In the stepwise logistic regression including all Braden subscales, the Braden total score, and the patient characteristics found to be significant in the bivariate analyses, altered awareness, incontinence, and the Braden Sensory subscale were retained as significant predictors of PU.
Findings from this study indicate the current Braden tool was able to identify some risk for pressure ulcers within the sample, especially incontinence and sensory deficits. However, other important factors may be overlooked (such as altered awareness) and should be considered when assessing patients at risk of pressure ulcers. Findings from this study agree with the VHA Handbook 1180.02. The VHA Handbook stresses using a "Pressure Ulcer Risk Scale" and encourages the additional identification (and documentation) of other potential risk factors. The factors listed in the Handbook in Appendix C (especially the first 5) were found to be significantly associated with PU development in this study (Alteration in Consciousness or Awareness; Alteration in Sensation; Alteration in Mobility; Strength, and Muscle Mass; Medical Devices, Alteration in Nutrition; Friction/Shear; and Incontinence). The Handbook also stresses an interprofessional "Pressure Ulcer Prevention Program must include a plan for ongoing education (and documentation) for staff and Veterans and/or the Veteran's designated family members, surrogates, or authorized decision makers."
The high number of charts excluded from analysis for missing three or more important data points and the small sample size (particularly in the "with PU" group) are limitations for this study.
The challenges encountered during this study provide valuable information related to documentation of pressure ulcer risk in current VA electronic health record systems. Inconsistency related to the documentation of PU presence (especially related to 'present on admission' PU versus hospital acquired PU or other wounds such as skin tears being documented as PU), PU risk assessments, daily skin assessments, PU staging, preventive interventions ordered, patient/caregiver education, and discharge planning related to pressure ulcers within the electronic health record was a problem. Findings from this study can help direct a larger program of research to address these inconsistencies.
External Links for this Project
NIH ReporterGrant Number: I21HX001011-01
Dimensions for VADimensions for VA is a web-based tool available to VA staff that enables detailed searches of published research and research projects.
If you have VA-Intranet access, click here for more information vaww.hsrd.research.va.gov/dimensions/
VA staff not currently on the VA network can access Dimensions by registering for an account using their VA email address. Search Dimensions for this project
DRA: Other Conditions
MeSH Terms: none