Identifying Poor Adherence in Outpatients with Bipolar Disorder: A Comparison of Different Measures

Objective Unlike schizophrenia, comparisons of different methods of estimating inadequate adherence in bipolar disorder (BD) are scarce. This study compared four methods of identifying inadequate adherence among outpatients with BD. Materials and Methods Two self-reports, the Morisky Medication Adherence Questionnaire (MAQ) and the Drug Attitude Inventory (DAI-10), clinician ratings employing the Compliance Rating Scale (CRS), mood-stabilizer levels, and clinic-based pill counts were compared at intake in 106 outpatients with BD and after 6 months of follow-up (n = 75). Statistical Analysis Rates of nonadherence were determined for each method. The ability to detect inadequate adherence was based on sensitivity, specificity, positive and negative predictive values (PPV and NPV), and positive and negative likelihood ratios (LR positive and LR negative). Correlation coefficients and Cohen's kappa values were used to determine the agreement between measures. Correlation coefficients were also used to evaluate the determinants of inadequate adherence Results The MAQ and the DAI-10 (cut-off score of two) yielded higher rates of nonadherence (35–47%) than the other methods. They were better at detecting adherence (specificity, 34–42%; PPV, 40–44%; and LR negative, 0.70–0.96) than other measures and had moderate ability to identify nonadherence compared with them (sensitivity, 63–73%; NPV, 54–70%; and LR positive, 1.02–1.16). They were associated with several established predictors of nonadherence. The MAQ and DAI-10 scores and the MAQ and CRS scores were modestly correlated. Multivariate analysis showed that 20% of the variance in the DAI-10 scores was explained by the MAQ scores. Despite their low yield, serum levels had a high sensitivity (88%) and higher accuracy (55%) in identifying inadequate adherence. CRS ratings and pill counts had high sensitivity but low specificity to detect inadequate adherence. Conclusion Self-reports appeared to be the most efficient method of ascertaining inadequate adherence among outpatients with BD. However, since none of the measures were adequate by themselves, a combination of different measures is more likely to maximize the chances of identifying inadequate adherence among these patients.