Improving PACT Coordination Across Settings and Services (QUE 15-276)

Background: High-risk Veterans are defined as individuals who are at increased risk for poor clinical outcomes and higher use of unplanned health services relative to their non-high-risk counterparts. These Veterans typically have multiple chronic health problems and are vulnerable to gaps in care due to impaired physical, psychological, and/or social functioning. Despite efforts to integrate care through VA's Patient Aligned Care Teams (PACT) in primary care, deficits in care coordination persist. In VA, most high-risk Veterans are managed in primary care rather than a specialty service. PACT was expected to improve care coordination by creating the care manager role for the PACT teamlet nurse. However, there have been significant challenges in implementing the care manager role as intended. Many of the care coordination challenges involve the "medical neighborhood" outside of PACT.

To improve the quality of care coordination in outpatient care and also develop better methods for spreading innovations, the Coordination Toolkit and Coaching project was funded by VA's Quality Enhancement Research Initiative (QUERI). This project develops and pilots an online toolkit and distance-based coaching process, and then compares the effectiveness of the toolkit alone to the combination of the toolkit plus distance coaching for improving VA patients' experience of care. Both toolkit and combined toolkit/coaching strategies have been used individually in VA quality improvement initiatives, and each strategy has been compared individually to other alternatives. However, to the investigators' knowledge, these strategies have not formally been compared head-to-head.

Additional Outcome Information: The project's primary outcome is a measure of patient experience, the Health System Hassles Scale. This 16-item scale asks patients questions such as whether their medications are being refilled on time, whether they were given information about why they were referred to a specialist, whether there has been poor communication between different doctors or clinics, or whether there have been disagreements between doctors about the patient's diagnosis or the best treatment for the patient.

Sample Size Calculations: The sample size calculation for this study is based on a simple presumption of a difference-in-differences analysis (across the two time points) for the comparison of the two implementation strategies. The primary outcome is the Health System Hassles Scale. The investigators assume 12 clinics in the study (6 per study group), which will be viewed as clusters in order to evaluate the sample size. Since the number of patients per cluster may vary, the investigators assume a coefficient of variation of cluster sizes of about 0.9. With an effect size of 0.3 standard deviations (which is considered to be a small to medium effect size in Cohen's terminology) for the difference-in-difference analysis and an intra-cluster correlation of 0.023 (based on preliminary evaluation of prior data), then with 80% power and two-sided 5% significance level, 149 patients per clinic are needed for a total of 1788 patients (evenly divided between the two groups).

Statistical Analysis Plan: The primary endpoint of the Health System Hassles Scale will be compared between the two implementation groups (toolkit and combined toolkit/coaching) using a difference-in-differences (between the two time points: baseline and 12 months) analysis adjusted for the clustering by clinic. This analysis will be performed initially with a general linear model using the between time point difference as the dependent variable and study group as the independent variable, with clinic as the clustering variable (and, thus, using an appropriately chosen variance-covariance matrix). A further adjustment model may incorporate appropriate covariates including patient-level factors, such as gender, age, and use of non-VA care.