1063 — Risk Factors Predicting Pressure Ulcers in Veterans with Spinal Cord Injuries and Disorders: An Application of Generalized Ordinal Logistic Regression
Smith BM (COE Hines (MCHSPR)) , Guihan M
(COE Hines (MCHSPR)), LaVela SL
(COE Hines (MCHSPR)), Garber SL
(DeBakey VA Medical Center)
Pressure ulcers (PrUs) are a major cause of morbidity for veterans with spinal cord injury/disorder (SCI&D). The objectives of this study were to identify risk factors for PrUs (eg, patient characteristics, health behaviors) and one or more than one PrUs in the previous year. It provides an example of how the generalized ordinal logistic regression model can inform clinical care.
A cross-sectional survey incorporating questions from the CDC’s Behavioral Risk Factor Surveillance System was mailed to Paralyzed Veterans of America members in October 2003 (n=2,561). Because some variables (i.e., race, service connected status, education) violated assumptions underlying a proportional odds ordinal regression model, a multivariable partial proportional odds ratio model was used to examine the association between patient characteristics and the self-report of one or more PrUs.
Patient characteristics significantly associated with having one or more PrUs included diabetes (OR=1.83, CI: 1.47 – 2.28, p=0.000), current smoker (OR=1.31, CI: 1.06 – 1.62, p=0.014), injury duration >30 years (OR=1.52, CI: 1.18 – 1.97, p=0.001), and reporting frequent days of mental distress (OR=1.63, CI: 1.33 – 1.99, p=0.000). Veterans service connected for their SCI (OR=0.57, CI: 0.47 – 0.69, p=0.000) or who only received care at the VA (OR=0.80, CI: 0.68 – 0.95, p=0.009) were less likely than other veterans to report one or more PrUs. The fit of the partial proportional odds model was significantly better than the ordinal logistic regression when compared with a likelihood-ratio test (p=0.000). These results suggest how important it is for clinicians to focus on prevention of chronic illnesses in SCI to lower risk of PrUs.
Veterans with SCI&D who smoke, have diabetes, longer durations of injury, and frequent mental distress were more likely to have PrUs, but less likely to have PrUs if they received their care exclusively at the VA or if they had a service connected injury.
This study is an example of the potential usefulness of the generalized logistic regression model for health services research. Additionally, it provides information about risk factors for PrUs, suggesting potential areas such as smoking prevention/cessation or diabetes management for future intervention efforts.