PATIENT ACTIVATION AMONG HIGH-RISK PATIENTS: DOES THE PATIENT ACTIVATION MEASURE PREDICT UTILIZATION?
Embargo until
Date
2014-03-10
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Johns Hopkins University
Abstract
Integrated delivery systems increasingly seek to identify high-risk patient sub-groups where effective interventions can reduce costs. Most predictive models for utilization incorporate un-modifiable factors such as health status, clinical severity, and demographics. In contrast, patient activation – the beliefs, knowledge, and skills that support self-management behavior – is modifiable. We explore whether activation, measured by the Patient Activation Measure (PAM), predicts utilization and whether motivational interviewing (MI) improves patient activation.
We analyzed a retrospective cohort of high-utilizing Medicaid patients enrolled in a MI program who completed PAM surveys between 2009 and 2011. We used proportional hazard models to predict time to emergency department (ED) visits and hospitalizations using PAM, clinical risk, and health status as main predictors. We used generalized estimating equations to model PAM score change based on type of MI intervention. We explored interactions of demographics and patient engagement with key predictors. Our eligible population included 1,676 patients from four medical centers within Kaiser Permanente Northern California (KPNC), a large integrated health care delivery organization. Our study population comprised 1,041 patients (62 percent) who completed ≥1 PAM surveys.
While the relationship between activation with ED visits was insignificant in fully adjusted models, lower activation was associated with higher ED visit risk (HR: 1.40, p<0.01) for patients lacking stable primary care physician relationships. The association of activation with hospitalization was insignificant after adjusting for clinical risk. For the 915 patients with 2+ PAM surveys, the overall MI intervention was associated with an unadjusted first PAM-to-last PAM mean score improvement of a clinically meaningful 4.1-points (p<0.01). Lower activation patients received the most interventions (PAM Level 1: 2.5/month; Level 2: 1.9/month; Level 3: 1.7/month; Level 4: 1.5/month; χ2 p<0.01) and improved the most (adjusted PAM score change for Level 1 versus Level 4: 12.7 points; Level 2 versus Level 4: 11.9; Level 3 versus Level 4: 8.6; all p<0.01).
We infer that PAM may be more useful for targeting interventions for behavior sensitive outcomes such as ED visits. MI seems to be effective in improving activation in this complex patient population.
Description
Keywords
predicting utilization, patient activation, self-management, motivational interviewing