Improving the Detection of Medication Nonadherence

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Medication non-adherence is a significant clinical problem in the management of hypertension, with rates ranging from 30-80% and an average of 50%. There is no 'gold standard' for measuring non-adherence. Self- reports are advantageous because they can be administered in any setting, take little time to complete, can provide immediate feedback at the point of care, cost little to administer and analyze, and can detect the specific reasons for non-adherence, thereby identifying areas for intervention. Nonetheless, self-reports are criticized for yielding low non-adherence rates that are 10-20% lower than rates obtained by other methods. These limitations are due, in part, to measurement issues related to latent variable models that have been largely ignored. The first measurement issue is the conflation of causal and effect indicator models. In an effect indicator model, responses to items on a measure are reflected (influenced) by the underlying latent variable (in this case, adherence). In a causal indicator model, responses to items on a measure give rise to the latent variable. Both types of indicators are important for detecting non-adherence; effect indicators reflect the extent to which patients are nonadherent (e.g., how often doses are missed), whereas causal indicators assess specific reasons for non-adherence (e.g., experiencing side effects). Accordingly, our first specific aim is to use both types of indicators to develop and validate a two-step method for assessing medication non-adherence. The first, brief questionnaire will assess the presence and extent of non-adherence. The second, longer questionnaire will assess the reasons for non-adherence. In Study 1, the measures will be administered twice, 3 to 5 days apart, to 200 patients with a diagnosis of hypertension (HTN) taking at least one blood pressure (BP) medication for at least 3 months. Intraclass correlations will provide evidence of the stability (reliability) of the measures. The association between the newly developed measures and BP and other measures (e.g., pharmacy refills, social desirability) will provide evidence of construct validity. The second measurement issue that has been ignored is that non-adherence is analyzed cross-sectionally, which assumes that it is trait-like (i.e., stable over time). Although some people may take their medications consistently, others may not. Accordingly, the second specific aim is to use longitudinal data analytic methods to determine the extent to which adherence is trait-like versus state-like. In Study 2, the two measures developed in Study 1 will be administered by telephone four times at 2-week intervals to 250 patients with HTN taking at least one BP medication for at least 3 months. Mixture distribution latent state-trait analyses will be conducted to determine the number of latent classes (subgroups) of medication takers, the size of each class, and the probability that each person belongs to each class. Personality variables (e.g., conscientiousness) will be assessed at baseline to determine whether membership in each latent class can be predicted. PUBLIC HEALTH RELEVANCE: Treatment decisions are based on estimates of medication non-adherence. These estimates would be most informative if they included how often doses are missed, reasons for missed doses, and whether missing doses is an occasional or ongoing problem. The goals of the proposed research are to develop new questionnaires to assess these aspects of non-adherence. The new questionnaires will help researchers and clinicians assess medication non-adherence more accurately and meaningfully, leading to improved tailored interventions to decrease medication non-adherence.