Murray 2007.

Methods	Randomized controlled trial
Participants	The study location was the general medicine and cardiology practices of Wishard Health Services, Indianapolis, Indiana, USA 122 participants were randomized to the intervention group and 192 participants were randomized to the control group The inclusion criteria were clinically stable patients from general internal medicine practices, a cardiology clinic, and Wishard Memorial Hospital (at discharge) to participate in the study. Patients were eligible if they were 50 years of age or older; planned to receive all of their care, including prescribed medications, at Wishard Health Services; had a diagnosis of heart failure confirmed by their primary care physician; regularly used at least 1 cardiovascular medication for heart failure (angiotensin‐converting enzyme (ACE) inhibitor or angiotensin‐receptor blocker, beta‐adrenergic antagonist, diuretic, digoxin, or aldosterone antagonist); were not using or were not planning to use a medication container adherence aid (for example, a pill box); had access to a working telephone; and could hear within the range of normal conversation The exclusion criteria were patients with dementia
Interventions	Intervention: PHARMACIST INTERVENTION A pharmacist delivered the intervention by using a protocol that included a baseline medication history of all prescription and over‐the‐counter drugs and dietary supplements taken by patients, which patients brought with them to the baseline interview, and the results of an assessment of patient medication knowledge and skills. The pharmacist dispensed enough of the patient's medications to last approximately 2 months. The intervention lasted 9 months. When medications were dispensed, the pharmacist provided patient‐centered verbal instructions and written materials about the medications. Each medication category had an icon associated with it. The same icon appeared on the container label and lid and on the written patient instructions. The pharmacist monitored patients' medication use, health care encounters, body weight, and other relevant data by using a study database. Information about patients was communicated as needed to clinic nurses and primary care physicians by face‐to‐face visits, telephone, paging (physician only), and e‐mail (physician only). An interdisciplinary team of investigators trained the intervention pharmacist. The intervention pharmacist also studied guidelines for treating heart failure, key concepts in the pharmaceutical care of older adults, communication techniques, and the pharmacotherapy of the cardiovascular drugs for heart failure Control: USUAL CARE Control groups received usual care. Usual care participants were aware of the purpose of the study. They received their prescription services from the same pharmacy as the intervention participants, but the pharmacists who attended the usual care participants were not trained by the interdisciplinary team
Outcomes	The measures of adherence were Medication Event Monitoring System (MEMS), prescription records, and self reported adherence. MEMS is a prescription container lid and all prescribed cardiovascular medications were dispensed with MEMS lids that recorded the time and date of each opening and closing onto a digital chip. Data retrieved from the lids were used to compute taking adherence and scheduling adherence. Taking adherence is the percentage of prescribed medication taken and measures deviation from the physician's prescription. Scheduling adherence measures the day‐to‐day deviation in the timing of administration. A medication prescribed for once‐daily administration would need to be administered within 24 hours of the previous dose, whereas medications prescribed for twice‐daily administration would need to be administered within 12 hours of the previous dose. Scheduling adherence measures the reliability or consistency of dosing over time. Refill adherence as the medication possession ratio by using prescription records from the Regenstrief Medical Record System. Results for 1 year, incorporating the 9‐month intervention period and the 3‐month postintervention period was computed. Patients came for refills at approximately 2‐month intervals. Considering the carry‐over effect of the intervention that stopped at 9 months, the intervention effect on medication supplies would extend to 11 to 12 months. Hence, the 12‐month period adequately reflects the effect of the intervention. Self reported adherence was determined for the previous month at baseline and 9 months by using validated questionnaires. Using these questionnaire scores, a composite score of self reported adherence was computed The patient outcomes were exacerbations by using hospital admission data from emergency department visits. We extracted data for heart failure–specific, all cardiovascular, and all‐cause reasons for emergency visits and hospitalizations; these were adjudicated by a registered nurse abstractor who used a previously validated method, from the Regenstrief Medical Record System Secondary outcomes included health‐related quality of life. We analyzed disease‐specific quality of life by using the Chronic Heart Failure Questionnaire, which performs well in the clinical setting of our study. This validated questionnaire has 4 dimensions: fatigue, dyspnea, emotion, and mastery. We averaged the scores on each scale, ranging from 1 (worst function) to 7 (best function), across items within each dimension
Notes	―
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A pseudo‐random number generator was used. "A trained interviewer conducted a baseline interview at enrollment. Interviewers were blinded to patients' study status and played no role in the delivery of the intervention. Interviewers contacted a centralized data manager at the end of each interview to determine the patient's study assignment, which was otherwise concealed. We randomly assigned patients, without blocking or stratification, to receive the pharmacy intervention or usual care by using a univariate discrete distribution from the IMSL Fortran Library's subroutine RNGDA pseudorandom number generator (Absoft Corp., Rochester Hills, Michigan) (16). We randomly assigned more patients to the usual care group so that this group could also be a prospective cohort for studying risk factors associated with the clinical deterioration of heart failure. Of the 314 patients included in the study, 229 were recruited from the general internal medicine practices, 15 from the cardiology clinic, and 70 on discharge from the Wishard Memorial Hospital. The numbers of patients assigned to the intervention and usual care groups did not differ by recruitment site" (pg 715)
Allocation concealment (selection bias)	Low risk	Allocation concealment was noted. "Interviewers contacted a centralized data manager at the end of each interview to determine the patient's study assignment, which was otherwise concealed" (pg 715)
Selective reporting (reporting bias)	Low risk	Protocol is available. All pre‐specified outcomes are reported on
Other bias	Low risk	The study seems to be free of other types of bias
Blinding of outcome assessment (detection bias) Adherence measure	Low risk	(PRIMARY) MEMS ‐ Outcome unlikely influenced by outcome assessor. "Interviewers were blinded to patients' study status and played no role in the delivery of the intervention" (pg 717)
Blinding of outcome assessment (detection bias) Patient outcome	Low risk	(PRIMARY) HOSPITAL VISIT RECORDS ‐ Lack of staff blinding unlikely to affect this outcome
Blinding of participants (performance bias) Adherence measure	High risk	(PRIMARY) MEMS ‐ There is no information on blinding. Any blinding could have been broken due to the nature of the intervention
Blinding of participants (performance bias) Patient outcome	Low risk	(PRIMARY) HOSPITAL VISIT RECORDS ‐ Lack of patient blinding unlikely to affect this outcome
Blinding of personnel (performance bias) Adherence measure	Low risk	(PRIMARY) MEMS ‐ This is an objective measure of outcome
Blinding of personnel (performance bias) Patient outcome	Low risk	(PRIMARY) HOSPITAL VISIT RECORDS ‐ Lack of staff blinding unlikely to affect this outcome
Incomplete outcome data (attrition bias) Adherence measure	Unclear risk	(PRIMARY) MEMS ‐ it is unclear whether the reasons for dropouts are related to the intervention; the number of dropouts are not balanced
Incomplete outcome data (attrition bias) Patient outcome	Unclear risk	(PRIMARY) HOSPITAL VISIT RECORDS ‐ it is unclear whether the reasons for dropouts are related to the intervention; the number of dropouts are not balanced