A preliminary predictive model for wheelchair use 4~months after receipt

Objectives: To explore whether use of wheeled mobility devices 4 months after receipt can be predicted based on the type of information available at the time of initial wheelchair prescription/provision across diverse settings and providers. Design: Longitudinal cohort study. Setting: One private teaching hospital and one Department of Veterans Affairs Medical Center. Participants: 86 consecutive, community-dwelling, cognitively intact persons with a wide range of diagnoses newly prescribed a wheeled mobility device. Interventions: Provision of a manual or motorized wheelchair, or scooter. Measurements: The dependent variable was self-reported device use (any versus none) 4 months after initial prescription; independent variables were self-reported medical conditions and impairment, activity limitations, personal and environmental contextual factors. Results: More than half (56.8%) of the subjects were still using their new wheeled mobility device 4 months after receipt. In bivariate analysis, persons still using the device were significantly (p< 0.05) more likely to report that they had experienced confusion in the prior 6~months, that their doctor had told them they had mental problems such as depression/and or anxiety, or that amputation was the reason they were prescribed a wheelchair. They were less likely to report being hospitalized in the previous 6 months. In the final multivariable model, based on stepwise regression, a history of recent confusion or prior amputation significantly increased the odds of continued device use (Odds Ratio [OR] 6.8 and 14.4 respectively) while recent hospitalization or prior fracture significantly decreased the odds (OR 0.15 and 0.30 respectively). An Index predicting wheelchair use, using weights for these 4 variables based on the multivariable regression, was significantly related to device use in 100 bootstrap replicate samples, with an average beta=1.03, OR=2.80, p< 0.001. Depending on the specific combination of variables, the sensitivity of the Index ranged from 0.07 to 0.94 for predicting wheelchair use and the specificity of the Index ranged from 0.27 to 0.99. Conclusions: Although preliminary, the study shows that it is possible to provide evidence-based prediction for short-term use of a new wheeled mobility device, based on information likely to be available at the outset. © 2012-IOS Press and the authors. All rights reserved.

Duke Scholars

Author Carl F. Pieper Biostatistics & Bioinformatics, Division of Biostatistics

Author Helen Marie Hoenig Medicine, Geriatrics and Palliative Care