Self-reported medication nonadherence predicts cholesterol levels over time.

OBJECTIVE: Self-report measures of medication nonadherence are frequently adapted to new clinical populations without evidence of validity. We evaluated the predictive validity of a medication nonadherence measure previously validated in patients with hypertension among patients taking cholesterol-reducing medications. METHOD: This secondary analysis involves data from a randomized trial (VA HSR&D IIR 08-297) conducted at the Durham Veterans Affairs Medical Center. At baseline, 6-months, and 12-months, serum cholesterol was obtained and participants (n = 236) completed a 3-item measure of extent of nonadherence to cholesterol-reducing medications. Two cross-lagged panel models with covariates, in addition to growth curve analysis, were used to examine the predictive utility of self-reported nonadherence on concurrent and future cholesterol levels, while accounting for potential reverse-causation. RESULTS: Extent of nonadherence items produced reliable scores across time and fit a single-factor model (CFI = 0.99). Nonadherence, and changes in nonadherence, moderately predicted future cholesterol values, and changes in cholesterol values (7 of 9 longitudinal associations were significant at p < .05; B's ranged from 0.16 to 0.35). Evidence for reverse associations was weaker (3 of 9 longitudinal associations were significant at p < .05; B's ranged from 0.16 to 0.36). CONCLUSION: Analyses support the predictive validity of this medication nonadherence measure over the competing reverse-causation hypothesis.

Duke Scholars

Author Daniel Blalock Psychiatry & Behavioral Sciences, Behavioral Medicine & Neur ...

Author Leah L Zullig Population Health Sciences

Author Hayden Barry Bosworth Population Health Sciences