Comparative Effectiveness & Technology Assessment Using Medical Records to Interpret Claims-Based Approaches to Comparative Effectiveness Studies of Antidepressants

Research Objective: Comparative effectiveness studies that rely on health insurance claims are difficult to interpret because they lack important clinical information, and could therefore result in biased inferences. The objective of this study was to provide clinical information that would assist in interpreting the results of a comparative effectiveness study of selective serotonin reuptake inhibitors. Study Design: Retrospective analysis of medical and pharmacy insurance claims with supplemental chart review. Abstracted data elements included the chief complaint and reason for the visit, diagnoses, current medications, signs and symptoms of depressive disorder based on DSM-IV criteria, the provider’s assessment and disposition, side-effects, and adherence to antidepressant medications. Population Studied: The claims analysis included 43,921 patients in six health plans across the US with new episodes of antidepressant treatment with escitalopram or an alternative SSRI (citalopram, fluoxetine, or paroxetine) between July 1, 2002 and July 1, 2005. From this universe of patients, we selected a subset for chart abstraction. Our final sample included 457 patients initiating treatment with escitalopram (183) or alternative SSRIs (254), including 169 patients who discontinued treatment two months after initiating treatment (escitalopram 68, other SSRIs 101) and 288 patients who continued treatment at two months (escitalopram 115, other SSRIs 153). Principle Findings: Overall, 30.2% of the sample had four or more symptoms recorded, consistent with DSM-IV criteria for Major Depressive Disorder, while 38.2% had two or three symptoms, consistent with criteria for dysthymia or minor depression. Having more symptoms was significantly associated with continuing treatment at 2 months (?2 =89.9, df=3, p<0.0001). While few patients had complete remission, about two-thirds of patients with a recorded clinician assessment improved within three weeks of starting treatment. Improvement based on clinician assessment (?2 =13.3, df=2, p<0.001) or symptom counts as recorded at follow up (?2 =117.6, df=3, p<0.0001) was significantly associated with continuing treatment. In the claims analysis compared with patients taking other SSRIs, patients on escitalopram were more likely to continue treatment at 2 months ((OR = 1.30; 95% C.I. 1.24 – 1.35; p < 0.01). In the chart analysis escitalopram patients who continued treatment at two months were more likely to have improved based on their physician’s assessment (?2 =8.1, df=3, p<0.05) compared to patients who started on other SSRIs. Few side effects were recorded for any patients. Conclusion: Our analysis of pharmacy claims suggests that patients initiating treatment on escitalopram were more likely to continue treatment at two months. The results of the chart analysis suggest that the claims-based study may have underestimated escitalopram’s clinical effectiveness because it failed to take into account possible differences in clinical response associated with adherence behavior. Implications for Policy, Practice or Delivery: While limited by small sample size and missing data not recorded in the charts, our results suggest that lack of clinical information resulted in a conservative estimate of the relative effectiveness of escitalopram compared to alternative antidepressants. Such possible bias should be taken into account when interpreting the results of claims-based comparative effectiveness studies. Funding: Forest Research Institute