Design
An open randomized intervention study was performed comparing an intervention group provided with the clinical pharmacist discharge service, with a control group provided with regular care by doctors and nurses.
The study protocol was approved by a medical ethics committee (Medische-Ethische Toetsing Onderzoek Patiënten en Proefpersonen, Tilburg, The Netherlands).
Setting and study population
The study was conducted at the department of cardiology of a teaching hospital in Tilburg, the Netherlands between May 2007 and July 2008.
Eligible patients were adults (aged over 18 years) admitted with a diagnosis of heart failure and prescribed five or more medicines (from any class) at discharge. We excluded patients living in a nursing home or unable to give informed consent, due to mental incapacity or terminal illness.
All patients who provided written informed consent were randomised using a random number table, to receive intervention or regular care.
Regular care
Patients in the control group received regular care, consisting of verbal and written information about their drug therapy from a nurse at hospital discharge. The discharge prescription was made by the physician and given to the patient to hand over to the GP.
Intervention
First of all, the intervention consisted of a clinical pharmacist identifying potential prescription errors in the discharge medication and discussing them with the cardiologist. This resulted in the final discharge medication. Furthermore, patients in the intervention group received verbal and written information about (side)effects of, and changes in, their in hospital drug therapy from a clinical pharmacist upon hospital discharge. In addition to this, the clinical pharmacist made a discharge medication list which contained additional information related to dose adjustments and discontinued medication. After it had been approved by the physician, the discharge medication list was faxed to the community pharmacy and given as written information to the patient with the instruction to hand it over to the GP.
All patients (both regular care and intervention) collected medication at their community pharmacy and received usual routine management by their cardiologist after discharge. This included an outpatient visit within 6 weeks after hospital discharge and an additional visit to the heart failure nurse if necessary.
Data collection
The following patient characteristics have been collected: age, sex, education (primary school or higher education), living situation (single or cohabitating), chronic co-morbidity (Chronic Disease Score, CDS [14]), routine check ups at the heart failure unit before admission, length of admission, number of medicines at discharge, living conditions after discharge (i.e. living in a nursing home, in a residential home for elderly people or at home with additional care), patient or pharmacy control over medication (i.e. patient is in control or the patient receives a “week box”, prefilled by the pharmacy; a week box contains all the medication for a week arranged by day and hour of intake) and New York Heart Association (NYHA) class upon discharge.
In addition, medication was classified by ATC code and the “source” of each drug, i.e. start or discontinuation during admission, dose adjustment or preadmission medication, was noted.
During the first follow up consultation after discharge with the cardiologist or heart failure nurse at the clinic, an estimate of adherence was made with the “Brief Medication Questionnaire—Regimen Screen” (BMQ), a validated tool for screening for adherence consisting of seven questions. This tool requires patients to list all medication taken in the past week and subsequently for each medicine listed four questions about the use of the medicine are asked, as well as three general questions about medication use. For each item patients received a score of “1” if their initial or spontaneous report indicated potential non-adherence with the current regimen for the target medication (i.e. when the specific question was answered with ‘yes’) and a score of “0” if this reports indicated no non-adherence (i.e. when the question was answered with ‘no’). The maximum score is 7 and a score of 1 or higher is an indication for potential non-adherence. Table 1 shows the questions of the BMQ [15].
Table 1 Questions of the BMQ-Regimen Screen [15]
In addition, the patient’s medication was checked for discrepancies and for prescription errors. Discrepancies were discussed with the patient and the cardiologist or heart failure nurse. A discrepancy was defined as a deviation in medication use compared to the medication on the discharge prescription. Discrepancies were classified as: re-start of discontinued medication, discontinuation of prescribed discharge medication, use of higher or lower dose, more or less frequent use than prescribed and incorrect time of taking medication.
A prescription error was defined as an error which occurs in the process of prescribing medication, namely dosing errors, dosage form errors, contra-indications, drug–drug interactions and double-medication. All prescription errors identified by the clinical pharmacist and agreed upon by the cardiologist were collected.
The clinical relevance of the discrepancy or prescription error was assessed by making use of the NCC MERP-index [16]. This index categorizes medication errors (class A-I), using an algorithm, see Table 2 (briefly, class A: no error, class B, C and D: error, but no harm, class E, F, G and H: error and harm, class I: error and death). Discrepancies and prescription errors in class E or higher (i.e. errors resulting in harm) are considered as clinically relevant. Three pharmacists and a cardiologist assessed the clinical relevance; for those discrepancies they disagreed on they met to reach consensus.
End points
The primary end point in this study is the total sum of the percentage of prescription errors and discrepancies after hospital discharge. The estimate of adherence as indicated by the BMQ was chosen as a secondary end point. Patients with a score of ≥1 were considered to be potentially nonadherent [15].
Data analysis
The program PS sample size (version 2.1.31) was used to determine the sample size [17]. The sample size was calculated at 62 patients per group based on α = 0.05, a power of 0.8, an estimated frequency of the end point in the control group of 30% [3, 4, 11, 18, 19] and an expected reduction to 10% [10, 20].
All data were processed in Microsoft Access 2003 and analysed with SPSS version 16.0.
The average and standard deviation were determined for continuous variables and the percentage was calculated for categorical variables. The differences between the intervention and the control group were analysed by the two sample t test for continuous variables and by the Chi-square test for categorical variables. A P value of ≤0.05 was considered to be significant.
For analysis of the primary and secondary end point the relative risk (RR) was calculated with a confidence interval (CI) of 95%. For the primary endpoint this was performed both on the medication level (number of medications as denominator) and on the patient level (% of patients with one or more discrepancy or prescription errors), and for the secondary endpoint on the patient level.