Measurement Matters - Ascertaining Response to Depression Treatment in Primary Care

Abstract

Introduction/Background: Depression is a common behavioral health problem that has been linked to elevated disability, morbidity, mortality and medical spending. Although there are numerous evidence-based treatments for depression in various settings, the accuracy and reliability of symptom measurement remain elusive, making it difficult to ascertain clinical progress. There are inherent flaws in the longitudinal use of validated symptom scales, such as the Patient Health Questionairre-9 (PHQ-9), and the literature is divided on how to interpret changes in patients’ scores over time. Different metrics have been used to denote depression “response” or “remission,” but they have had minimal empirical validation. Further, it remains unclear how changes in rating scale scores impact patient-centered outcomes (PCOs), such as quality of life and social connectivity. Methods: This investigation was conducted in two phases. The first was via a secondary analysis of data from the Improving Mood–Promoting Access to Collaborative Treatment (IMPACT) trial, an 1800-participant RCT of Collaborative Care for the treatment of depression in primary care. In IMPACT, all participants were surveyed and interviewed with a variety of evidence-based instruments at baseline and periodically. PHQ-9 scores were also tracked for the treatment group (n=906). Baseline and follow-up surveys for all participants included, among other instruments, the Health Information National Trends Survey (HINTS), the Strengths Self-Efficacy Scale (SSES), a self-reported Quality of Life measure and the Sheehan Disability Scale (SDS). For the treatment group only, we calculated response rates and analyzed the average associated change in general health, social functioning, quality of life and degree of disability (with simple linear regression significance testing) across nine PHQ-9 depression “response” and “remission” metrics and three intervals of time (3, 6 and 12 months). Pair-wise tests for equivalence in mean differences in outcome for all combinations of single-component PHQ-9 depression “response” and “remission” metrics were also conducted using ANOVA with specified linear contrasts. The second phase was analyzing data collected by the University of Washington’s Advancing Integrated Mental Health Solutions (AIMS) center from 42 health care organizations serving more than 11,000 patients in real-world collaborative care implementations nationwide. Using the same nine PHQ-9 “response” and “remission” metrics, we ranked all 42 health care organizations with direct sample means and a multilevel logistic regression model with empirical Bayes predictions. For both methods, correlations between all pair-wise rank orders tests for significance were conducted using the Spearman's rank-order correlation coefficient. Results: Using PHQ-9 data from IMPACT, choice of metric had a substantial impact on depression treatment response rates, with 3- 6- and 12-month rates ranging from 32.3-76.7%, 37.9-82.2% and 42.1-84.9%, respectively. Further, PCOs such as general health, social functioning, quality of life and degree of disability change all demonstrated statistically significant improvement with depression response or remission. Three metrics were top-performing with respect to associated PCO differences- (1) 50% decrease from baseline, (2)  50% decrease from baseline AND score of <10 and (3) Score <5 (remission). However, when combinations of the five single-component metrics were tested for equivalence, most pair-wise comparisons were found to be statistically insignificant. Remission (score < 5) was more consistently different from the single-component response metrics. Using real-world AIMS data, organization-level rankings differed across depression response and remission metrics, though most were highly correlated. Of all metrics, remission was least positively correlated with the others. Conclusions: Our findings demonstrate that choice of depression response metric substantially impacts observed response rates and that PCOs tend to improve with depression. However, our comparative associations between depression response/remission metrics and patient-centered outcomes were mixed overall and we are consequently unable to suggest an optimal metric. Organization-level rankings for depression response/remission vary depending on choice of metric, but their rankings are highly correlated. Given the increasing global focus on population health and measurement-based care, future research should prioritize the determination of an optimal, pragmatic metric for depression outcome ascertainment.