Abstract 19221: Developing a Clinical Predication Model for 1-Year Health Status Outcomes in Peripheral Arterial Disease

    loading  Checking for direct PDF access through Ovid


Introduction: Patients with peripheral arterial disease (PAD) have different therapeutic options, but it is unknown how their personal characteristics and different treatments might impact their health status. The first step in creating personalized estimates for patients is to model 1-year health status outcomes.

Methods: The prospective PORTRAIT study enrolled 1275 patients with new, or an exacerbation of PAD symptoms from 16 specialty clinics in the US, the Netherlands, and Australia. Disease-specific health status was measured with the Peripheral Artery Questionnaire (PAQ) at presentation, prior to treatment, and at 1-year. Information about patient factors was derived from interviews and medical records. We considered 50 variables and weighted them using a random forest algorithm. A total of 10 variables were retained (all >5% increase in mean square error) in a linear regression model predicting 1-year PAQ summary scores (model 1). To examine the incremental value of 12 additional clinical variables (e.g. ABI), a second model was built.

Results: Model 1 explained 29% of the variance in 1-year health status. Better baseline health status, being white, having new-onset symptoms (vs. worsening), non-US location, and active working status were factors associated with better outcomes at 1 year (Figure 1a). Depressive symptoms, bilateral disease, and sleep apnea were linked with worse outcomes. Country interactions for age, and avoidance of care due to cost were present. Adding 12 clinically relevant variables minimally improved the explained variance (31%) (Figure 1b). No significant interactions with the variables and PAD treatment modality were present in either model.

Conclusion: A prediction model for patients’ 1-year PAD-specific health status was derived from 10 patient characteristics and should be further tested as a support for shared decision-making following validation in distinct PAD cohorts.

Related Topics

    loading  Loading Related Articles