Association Between Adherence to Diuretic Therapy and Health Care Utilization in Patients With Heart Failure

Michelle A. Chui, Pharm.D., Ph.D., Melissa Deer, B.S., Susan J. Bennett, D.N.S., Wanzhu Tu, Ph.D., Stacey Oury, B.S., D. Craig Brater, M.D., Michael D. Murray, Pharm.D., M.P.H.


Pharmacotherapy. 2003;23(3) 

In This Article


This prospective study measured adherence to diuretic therapy in patients with heart failure and determined the relationship between adherence and health care utilization for 6 months. All aspects of the protocol were reviewed and approved by the institutional review board and the medicine and specialty clinics review process at Indiana University Purdue University Indianapolis. Written informed consent was provided by all enrolled patients.

Inpatient and outpatient care are prospectively monitored by the computerized Regenstrief medical record system,[9] which captures and stores data on clinical events, diagnostic tests, and treatment for all patients.[10] For this study, data on hospital admission and emergency department visits were extracted from the system.

Adults (aged ≥ 18 yrs) with a medical diagnosis of heart failure, confirmed by preestablished echocardiogram criteria, and an active prescription for a diuretic were identified by screening automated medical, pharmacy, and administrative records. They were sent a letter explaining the study and informing them that the investigator would meet them at their next scheduled clinic appointment to discuss their participation. Participants spoke English, had access to a working telephone, had no conversational hearing impediments, and were alert and oriented as determined by the Pfeiffer short portable mental status questionnaire administered to ensure the patients' ability to answer questions appropriately.[11] The maximum score for the questionnaire is 10, one point for each correctly answered question. Examples of questions are, "What is your telephone number?" and "Who is president of the United States?" Patients with a score of 6 or higher were eligible to be enrolled.

On enrollment, patients received a 6-month supply of their prescribed diuretic at no cost to them, in a bottle with an electronic monitor lid (Medication Event Monitoring System V [MEMS V]; formerly known as eDEM, APREX Corp., San Diego, CA), a valid measure of adherence, to quantify adherence to therapy.[12,13] The lid contains microelectronic circuitry that records the date and time it is removed and replaced. Over time, the monitor provides a temporal distribution of lid openings and closures. Data stored in the lid are downloaded using a manufacturer-supplied communicator and software on a personal computer.

Patients were informed about the purpose of the study and shown how to open and close the electronic monitors. In several cases, the 6-month supply of diuretic required patients to come to the pharmacy for prescription refills. Patients were contacted and met at the pharmacy to ensure that they would continue to have enough drug for the entire study period. Those who had a systematic method of adhering to their regimens, such as a weekly planner box, were excluded if they believed participating in the study would jeopardize their established routines.

Over the 6 months, the primary investigator met patients at the clinic or pharmacy for physician visits, laboratory or diagnostic work, or prescription refills. At those times the investigator confirmed patients' participation and verified the prescribed drug and dosage. At the end of the study, patients returned the monitors and were paid $10 for participating.

Two aspects of adherence were measured. Taking adherence was considered the proportion of prescribed bottle openings captured by the MEMS V lids (within 1 day before or after the prescribed day). Scheduled adherence was defined as the proportion of prescribed openings that occurred within a specific time interval (determined by preceding dose). We prespecified time intervals to represent adherence: for dosing once/day the interval was ± 6 hours from the average time of ingestion, for dosing twice/day the interval was 3 hours, and for dosing 3 times/day it was 2 hours. Taking adherence was thus a measure of the proportion of prescribed doses taken during the 6-month monitoring period, and scheduled adherence represents the proportion of observed doses that were consistently timed.

Health care utilization was defined as the number of hospital admissions and emergency department visits during the study period. These were stratified three ways: utilization for all diagnoses, for heart failure, and for all cardiovascular diagnoses based on a validated method.[2] This method involved adjudication by three physicians of reasons for admission of 88 patients with heart failure. Agreement with the hospital's assigned International Classification of Diseases, ninth revision (ICD-9) code and agreement among physicians on the admitting diagnoses were good ( ≥ 0.78). To determine health status, patients were given an ambulatory care group code that identified the number of comorbidities and their severity as related to resource utilization.[14] Ambulatory care groups were stratified into four groups: low, moderate, high, and very high resource utilization. We found that 88% of patients were categorized as moderate, high, and very high resource utilization, suggesting that virtually all of them had several comorbidities.

The primary interest of the data analysis was to examine the effects of the two adherence measures on subjects' utilization of the health care system. Toward this end, we used log-linear regression models for count data to describe the numbers and types of hospitalizations that occurred over the 6 months. In addition to the adherence, other relevant patient information, such as age, sex, race, years of education, household income indicator, and New York Heart Association (NYHA) classification, were included in the model as covariates. To reduce the number of parameters to be estimated, we dichotomized income into a binary variable (1 = not enough to make ends meet; 0 = enough to made ends meet or living comfortably). We have found this way of measuring income to be accurate among our indigent patients, whereas dollar income is elusive. In fitting the models, stepwise model-selection procedures were used to identify important covariates to be included in the models. The analysis was carried out using SAS procedure GENMOD (Generalized Linear Models; SAS Institute, Cary, NC). An level of 0.05 was used in all statistical inferences.

In addition to the primary analysis, we examined nonrespondent related issues. Descriptive statistics were reported on patient demographics, adherence measures, and health care utilization. Correlation of the taking adherence and scheduling adherence was calculated by a Pearson correlation coefficient. Linear regression models were used to examine effects of various patient characteristics on the two adherence measures.