Institutional members access full text with Ovid®

Variations in Patients Adherence to Medical Recommendations: A Quantitative Review of 50 Years of Research

DiMatteo, M. Robin PhD

Original Article

Background: The literature on patient adherence to treatment includes hundreds of empirical studies. A comprehensive examination of the findings requires the organization and quantification that is possible with meta-analysis.

Objectives: The goals of this research are retrieval, compilation, and averaging of adherence rates in all published empirical studies from 1948 to 1998; assessment of variation according to sample characteristics, time period of publication, measurement method, disease, and regimen; and examination of the effects on adherence of patient demographic characteristics.

Methods: We calculated a meta-analysis of 569 studies reporting adherence to medical treatment prescribed by a nonpsychiatrist physician, and 164 studies providing correlations between adherence and patients’ age, gender, education, and income/socioeconomic status; group comparison and multiple regression analysis of moderators.

Results: The average nonadherence rate is 24.8%. Controlling for intercorrelations among moderator variables, adherence is significantly higher in more recent and smaller studies and in those involving medication regimens and adult samples. The use of physical tests and self-report have respectively significant and borderline negative effects on the level of adherence, and disease severity and use of the medical record have no significant effects. Adherence is highest in HIV disease, arthritis, gastrointestinal disorders, and cancer, and lowest in pulmonary disease, diabetes, and sleep. Demographic effects on adherence are small and moderated by sample, regimen, and measurement variables.

Conclusions: This review offers insights into the literature on patient adherence, providing direction for future research. A focus on reliability and validity of adherence measurement and systematic study of substantive and methodologic moderators are recommended for future research on patient adherence.

Research on patient adherence (compliance) grew rapidly over the past 50 years, as chronic disease became more prevalent and treatment more dependent on patient self-management. 1–4 Patients are adherent when they do what their health professionals recommend. In some views, nonadherence wastes resources and causes preventable morbidity and mortality and the loss of healthcare dollars and productivity. 5 In other views, nonadherence represents rational choice as patients attempt to maintain their personal identity, achieve their goals, and preserve their quality of life. 6–11 Data suggest that the difference in health outcomes between high and low adherence is 26%, and that the adherence–outcome relationship varies with the regimens, measurements, and diseases studied. 12

Many theoretical models (eg, Health Belief Model, Theory of Planned Behavior, and Transtheoretical Model) focus on understanding, predicting, and improving adherence. 13,14 Their common components involve health professional–patient communication, patients’ cognitive and social processes (eg, beliefs, norms), and patients’ resources (eg, financial, psychologic, and social support). The empirical literature on adherence is large but not well understood, even with elegant conceptual frameworks. 14 Studies vary widely in methodologies, and operational definitions of adherence are as varied as the diseases, regimens, and patients examined. Both measurement and context differences produce wide variations in adherence estimates, correlates, and outcomes. 12,15,16

This article examines “variations in patients’ compliance …”17,18 as a function of methodologic and contextual factors in adherence research. Meta-analysis is used as a methodology for integrating research findings with statistical analysis of results from many individual studies. 19 Metaanalysis offers a rigorous alternative to narrative discussions of the empirical literature 20 and allows examination of trends in substantive and measurement issues 21–24 The goals of the present research are: 1) to retrieve, compile, and average the adherence rates reported in all published empirical studies of adherence (and compliance) over 50 years (1948–1998); 2) to assess variations in adherence according to contextual and measurement variables (across time periods, sample characteristics, measurement strategies, disease conditions, and treatment regimens), providing norms for subgroups and their interactions; and 3) to examine the role of patients’ demographic characteristics (age, gender, education, and income) in adherence. Based on commonalities in existing adherence models, particularly as they relate to patients’ identity, illness self-image, and the meaning of treatment in attempts to preserve, maintain, or restore a preferred (pre-illness) self-identity, 8,9,25,26 it is hypothesized that adherence could be more difficult to achieve: 1) in the context of regimens that are more pervasive (eg, behavioral) rather than circumscribed (eg, medication), 2) when suffering is less (ie, when illness is less serious and the consequences of nonadherence might seem lower), and 3) when resources such as income and education are wanting. Demographic variables (patient age and gender) are not expected to have consistent effects on adherence, which is predicted to vary consistently according to the subjectivity of its measurement.

From the Department of Psychology, University of California, Riverside, California.

Reprints: M. Robin DiMatteo, PhD, Department of Psychology, University of California, Riverside, CA 92521. E-mail:

© 2004 Lippincott Williams & Wilkins, Inc.