Analysis of longitudinal adherence data: Hitting the target with consistency.

Typical descriptions of partial adherence summarize to what extent a patient follows a prescribed treatment plan. Another aspect of adherence relates to how consistently a patient behaves. We distinguish these two notions as "adherence target" and "adherence consistency." Most research on adherence focuses on the former concept, but the latter can be a useful idea in assessing patients' compliance. Quantifying these two notions goes beyond typical summaries of center and spread, because they may be related to each other. That is, the mean is often related to the standard deviation via the variance function. Our approach defines a parameter that measures consistency separate from the adherence target. For exponential dispersion or quasi-likelihood families, the parameter corresponds to the dispersion parameter. Assessing adherence consistency to medical treatment requires longitudinal data, which can be collected with new technology. Two examples from pediatric medicine demonstrate that detangling the notion of consistency from the adherence target is useful.