Search

Summary: We consider a Markov structure for partially unobserved time-varying compliance classes in the Imbens-Rubin (1997, The Annals of Statistics 25, 305-327) compliance model framework. The context is a longitudinal randomized intervention study where subjects are randomized once at baseline, outcomes and patient adherence are measured at multiple follow-ups, and patient adherence to their randomized treatment could vary over time. We propose a nested latent compliance class model where we use time-invariant subject-specific compliance principal strata to summarize longitudinal trends of subject-specific time-varying compliance patterns. The principal strata are formed using Markov models that relate current compliance behavior to compliance history. Treatment effects are estimated as intent-to-treat effects within the compliance principal strata.

Objective: Our purpose was to determine the personal characteristics associated with different patterns of adherence to the antidepressant citalopram in a primary care trial of depression management., Method: The study sample consisted of 228 adults aged 60 years and older recruited from primary care settings and who participated in a depression intervention. The intervention consisted of services of trained care managers, who offered recommendations to physicians following a clinical algorithm and helped patients with treatment adherence. Adherence to the antidepressant citalopram was measured using pill counts. We employed the latent class model to classify patients according to the pattern of adherence to citalopram over time. We examined the association of sociodemographic characteristics, depression status, cognitive status, and medical comorbidity with the resulting classes of adherence., Results: The latent class model generated three classes of adherence: known to be adherent, unknown adherence, and known to be nonadherent. Participants who were white were more likely to be in the known to be adherent class than in the known to be nonadherent class (odds ratio (OR) = 10.38, 95% confidence interval (CI) [3.47, 31.12]). Married participants were less likely to be in the unknown adherence class than the known to be nonadherent class (OR = 0.28, 95% CI [0.09, 0.85]). No associations between age, gender, education level, depression status, cognitive status, or medical comorbidity and classes of adherence were found., Conclusions: We found stronger relationships between ethnicity and marital status with patterns of adherence to citalopram than we did other personal characteristics. Identification of a subgroup of patients at particularly high risk of nonadherence is important for the development of adherence interventions.