This study examines the relationship between clinician support of patient activation (CS-PAM) and three sets of outcomes: clinician behaviors that support patient self-management, clinician behaviors in support of behavior change, and actual changes in patient activation measure scores (PAM) among a PCP’s patient panel. Thus we present two types of analyses: Our primary analysis—exploring the relationship between CS-PAM and clinician behaviors—is cross-sectional; our supplementary analysis examines the relationship between CS-PAM and changes in patient activation.
The institutional review boards of the University of Oregon, University of Minnesota and George Washington University approved the study procedures. The study was conducted in collaboration with the Fairview Health System in the summer of 2013. This health system includes 44 primary care clinics; PCPs from these clinics completed an online survey. The online survey was based on qualitative findings from a preliminary study exploring provider perceptions about a new compensation model implemented within the Fairview system . Survey items included items about quality of care, productivity, and strategies providers may use to influence patient health behavior .
For some analyses, we used data extracted from Fairview’s electronic health record to compute descriptives of PCPs’ patient panels. Additionally, to compute the average change in activation for a PCP’s panel of patients, we used a panel of 10,957 patients who had two PAM scores in consecutive years between 2010 and 2012. For this portion of the analysis, we only included those PCPs who had at least 20 patients on their panels with two PAM scores.
Clinician Support for patient activation
We measured providers’ beliefs about the patients’ role in their own care using the CS-PAM (Appendix). This 13-item scale was modified from the Patient Activation Measure, which asks patients how much they agree or disagree with statements related to how they manage their health. In the CS-PAM, clinicians respond to how important to them as clinicians are each patient skill, areas of knowledge, or beliefs. For example, clinicians were asked, “How important is it to you that your patients with chronic conditions can follow through on medical treatments they need to do at home?” PCPs answer using a four-point scale in which 1 = not important and 4 = extremely important. Scores are translated into a 1–100 scale, based upon Rasch methods , with higher scores indicating more positive beliefs about the importance of patient knowledge and involvement in his/her care. The original version of this measure (14-items) was used among a sample of clinicians from the US, UK, and The Netherlands [6, 15]. The CS-PAM demonstrated high reliability in this sample (Cronbach’s alpha α = .97).
Clinicians’engagement in chronic illness management support behaviors
The first set of 7 items—adapted from the Patient Assessment for Chronic Illness Care (PACIC)—were designed to assess the degree to which clinicians engage in partnership-building behaviors with patients around self-management . An example of an original item from the PACIC is – “…when I received care from my chronic conditions, I was: asked for my ideas when we made a treatment plan.” The items where then adapted to be relevant for providers; for example, one item asks: “Over that last month when you treated patients with chronic conditions, how often did you make sure patients were involved in setting the agenda for the visit? (1 = almost never, 4 = almost always).
Using specific strategies to support patient behavior change
The second set of 8 items was derived from qualitative interviews with Fairview clinicians and is based on their descriptions of their strategies for supporting their patients’ behavior changes. For example, “When a patient is not making needed behavioral changes to meet quality metrics, how often do you … have frank and sometimes difficult conversations with a patient about his or her behaviors?” (1 = never, 5 = very often). These 8-items and the 7-items adapted from the PACIC were all treated as individual items.
The final dependent variable was the average amount of change in patient activation score a PCPs’ patient panel had over 1 year. The “patient panel” is the collective term that refers to the patients seen by a provider over 1 year. Given that Fairview routinely collects patient activation data on patients, this latter outcome measure was derived from data from the electronic medical records; thus, the patient activation scores did not come from the same data source as the clinician survey.
We examined the relationships between provider characteristics and CS-PAM score, and providers’ patient panel characteristics (percent women, average age, average risk score, average mean income for ZIP Code), and CS-PAM score. None of the patient panel characteristics were related to the CS-PAM, and for the sake of parsimony, we did not include the patient panel characteristics in further analysis. To examine the relationship between CS-PAM scores and the dependent variables (Clinician engagement in chronic illness management support behaviors and using specific strategies to support patient behavior change), we initially conducted bivariate analyses. Specifically, we examined what percentage of PCPs in each CS-PAM tercile “almost always” or “very often” (depending on the set of items) engaged in the specific behavior. For the third dependent variable, we examined the relationship between the CS-PAM scores and the providers’ mean patient panel change in PAM scores.
We then developed multivariate regression analyses to examine the relationships between CS-PAM and the dependent variables, controlling for the following provider characteristics: gender, provider age, years of work at Fairview, and type of provider. For the first two sets of dependent variables, which were treated as dichotomous, we developed logistic regression models, and for the last dependent variable, we used ordinary least squares regression.