Medication reviews in the community: results of ... - Wiley Online Library

British Journal of Clinical Pharmacology

DOI:10.1111/j.1365-2125.2004.02220.x

Medication reviews in the community: results of a randomized, controlled effectiveness trial Lene Sorensen,1 Julie A. Stokes,1 David M. Purdie,2 Michael Woodward,3 Rohan Elliott3 & Michael S. Roberts1 1 Theraputics Research Unit, Dept of Medicine, Princess Alexandra Hospital, Brisbane, Australia, 2Aged and Residential Care Services, Austin Health, Heidelberg Repatriation Hospital, Melbourne, Australia, 3Population and Clinical Sciences Division, QIMR, Brisbane, Australia

Aims To examine the effectiveness of a multidisciplinary service model delivering medication review to patients at risk of medication misadventure in the community.

Correspondence Michael S Roberts, Theraputics Research Unit, Department of Medicine, Princess Alexandra Hospital, Ipswich Road, Woolloongabba, QLD4102 Australia

Methods The study was carried out in three Australian states; Queensland, New South Wales and Western Australia, and conducted as a randomized, controlled effectiveness trial with the general practitioner (GP) as the unit of randomization. In total, 92 GPs, 53 pharmacists and 400 patients enrolled in the study. The multidisciplinary service model consisted of GP education, patient home visits, pharmacist medication reviews, primary healthcare team conferences, GP implementation of action plans in consultation with patients, and follow-up surgery visits for monitoring. Effectiveness was assessed using the four clinical value compass domains of (i) functional status, (ii) clinical outcomes, (iii) satisfaction and (iv) costs. The domains of functional status (assessed by the health-related quality of life measure SF-36 subscales) and clinical outcomes (as assessed by adverse drug events (ADEs), number of GP visits, hospital services and severity of illness) were measured at baseline and endpoint. Satisfaction was measured by success in implementation and by par ticipant satisfaction at endpoint, and costs (as assessed using medication and healthcare service costs, less intervention costs) were measured preintervention and during the trial. In addition, process evaluation was conducted for intervention patients, in which problems and recommendations from the medication reviews were described.

Keywords clinical value compass, community pharmacy services, effectiveness trial, home care services, patient care team, randomized controlled trials

Received 5 August 2003 Accepted 7 July 2004

Results The model was successfully implemented with 92% of intervention GPs suggesting that the model had improved the care of participating patients, a view shared by 94% of pharmacists. In addition, positive trends in clinical outcomes (ADEs and severity of illness) and costs (an ongoing trend towards reduction in healthcare service costs) were evident, although the trial was limited to a 6-month intervention time. No differences between intervention and control groups were identified for the healthrelated quality of life domain. The cost–effectiveness ratio for the intervention based on cost savings, reduced adverse events and improved health outcomes was small. The most common problems identified in the medication reviews were potential adverse drug reactions, suboptimal monitoring and adherence/lack of concordance issues. In total, 54.4% of recommendations were enacted, and 23.9% were implemented precisely as recommended in the medication review. Follow-up evaluation showed that 70.9% of actions had a positive outcome, 15.7% no effect and 3.7% had a negative outcome.

Br J Clin Pharmacol

58:6

648–664

648

© 2004 Blackwell Publishing Ltd

Medication reviews in the community

Conclusions Most studies emphasize efficacy and the best achievable clinical outcomes rather than whether an intervention will be effective in practice. The current trial showed that three of the four domains in the clinical value compass showed trends of improvement or were indeed improved in the relatively shor t follow-up period of the trial, suggesting that a service based on this model could achieve similar benefits in practice. A domiciliary medication review programme similar to this model has now been implemented into national Australian practice, where GPs and pharmacists are reimbursed by the Australian government for the provision of these services.

Introduction

Inappropriate medication use in the community and institutional settings has been reported to be a common and continuing problem, particularly in patients taking multiple medications [1]. Inappropriate medication use is often divided into three categories: underuse (e.g. subtherapeutic dosage, insufficient treatment, poor adherence by patient), overuse (e.g. non-adapted doses in a frail elderly patient or in a patient with mild renal insufficiency, poor adherence by patient) or misuse (e.g. failure to recognize potential drug interaction) [2]. Inappropriate medication use may lead to adverse outcomes; for example, medication-related hospital admissions have been reported to range from 5.7% to 16.2% [3–6], with up to 65.5% of the medication-related admissions reported as being possibly preventable [3–6]. In the Australian context, Roughead et al. [7] have reported that medications were associated with 2.4–3.6% of all hospital admissions and 6–7% of emergency admissions, 12% of all admissions to medical wards and 15– 22% of all emergency admissions. They further suggested that between 32% and 69% of drug-related admissions may be preventable. One means of addressing inappropriate medication use is to conduct medication reviews and related interventions as part of a multidisciplinary team process [8]. Our group previously used a randomized, controlled study involving a large number of healthcare professionals and residents to examine whether a similar model improves resident outcomes in Australian nursing homes [9]. Such studies do not appear to have been conducted in the community setting in Australia [10], but studies of pharmacist medication reviews in the community in the USA [11–15] and the UK [16–18] have shown that medication reviews can identify a range of possible medication-related problems and impact positively on medication care. One study used a randomized, controlled trial approach but its generalizability may be limited by all reviews being done by a single university medical centre pharmacist [11]. Other studies

of community pharmacy interventions have shown that interventions during the dispensing process can prevent harm and visits to the general practitioner (GP) and contribute to the prevention of medication-related hospital admissions [19]. In general, most clinical trials have a goal of defining efficacy and rely upon randomized, controlled trial conditions, which demonstrate the significance of healthcare outcomes by maximizing internal validity [20]. In contrast, effectiveness trials seek to show generalizability to practical clinical situations by maximizing external validity. These trials have a focus on ‘real world’ conditions, but may be limited by sponsor requirements. The present trial was one of a number of trials [21, 22] that sought to examine the effectiveness of a multidisciplinary service model delivering medication reviews to patients at risk of medication misadventure in the Australian community. A randomized, controlled design and the clinical value compass [23–25] were used to enable an evaluation of outcomes. Effectiveness in the clinical compass is defined by the domains of health-related quality of life, patient satisfaction, clinical outcomes and costs. Clinical outcomes included identifying and actioning medication problems, changes in patient clinical condition and changes in reported adverse drug events (ADEs). The service model involved the primary healthcare team of various health professionals led by the GP and with community-based, accredited pharmacists conducting medication reviews. A key outcome sought from the trial was whether the process used in this trial and its healthcare outcomes could be effectively translated into the broader community. Methods Trial design

The model for reviewing medication in patients homes (Figure 1) was funded by the Commonwealth Government of Australia and developed by a steering committee consisting of representatives from general (medical) Br J Clin Pharmacol

58:6

649

L. Sorensen et al.

GP TRAINING/CONTINUING EDUCATION • Intricacies of how to “unprescribe” • Quality Use of Medicine (QUM) & ADRs • Involvement of other health professionals • Specialists eg Geriatrician, Physician • Accredited Pharmacist* • Community Nurse

PATIENT at risk Identified by: • GP • Community Nurse • Pharmacist

Health professionals to report risk to GP

GENERAL PRACTITIONER (GP) Makes decision to start review process Patient consultation where GP seeks patient informed consent

• Patient • Medication funders

Home visit by GP

Pharmacist home visit (by accredited* or community pharmacist)

Other • GP Care planning • Refer to specialist • Community nurse home visit

Report visit findings Patient to acknowledge visit receipt

Accredited* Pharmacist Review

Written report to GP • risks & benefits • recommendations

Fund Provider

GP & Pharmacist discuss report via choice of telephone, face-to-face, email, fax

GP Home visit • decides on action based on available information • may need to consult with specialist prescribers

GP implement action plan with patient

Possible implementation options

GP Repeat consultation with patient at 3 months • Reviews changes, follow-up • Clinical audit/quality assurance Feedback to pharmacists on process & outcome

Refer to Community Nurse for monitoring & environmental assessment

If complex, report & refer to geriatrician /specialist physician • diagnosis contribution to polypharmacy • adverse withdrawal events supervision

FLEXIBILITY IN DECISION

GP Action Plan

Report & revised prescription to community & accredited* pharmacist

Figure 1 Intervention model. *An accredited pharmacist was a pharmacist accredited by the Australian Association of Consultant Pharmacy to perform medication reviews (after completion of training and assessment). Accredited pharmacists could be the community pharmacist or an independent practitioner separate from the community pharmacy

650

58:6

Br J Clin Pharmacol


practice, community pharmacy, nursing, patient, government and academic organizations. The committee agreed that the effectiveness of the model should be demonstrated using a randomized, controlled trial with the GP as the unit of randomization. Recruited patients attending GPs in the intervention group were subject to the model, and patients of control group GPs received usual care. The Commonwealth Government of Australia, in funding this effectiveness trial, required that the trial was limited to a single home visit, that the intervention was completed within 6 months, and that patients met predetermined inclusion criteria. The clinical value compass [23–25], with its domains of (i) functional status (health-related quality of life), (ii) clinical outcomes (ADEs, number of GP visits, hospital services and severity of illness), (iii) success in implementation and patient satisfaction and (iv) healthcare service costs), was used to evaluate effectiveness in this trial. Participants

Participants were selected from rural and urban areas in Queensland, northern New South Wales and Western Australia. Patients were enrolled between August 1999 and March 2000, and completed the trial between December 1999 and May 2000. The service model linked patients with their usual GP and pharmacy. GPs were recruited or identified by various sources including articles in GP-targetted media, community pharmacists, community nurses and cold calling by project staff. Community pharmacists were recruited through pharmacist-targetted media, patients, GPs and cold calling by project staff. Patients were identified from a range of sources and enrolled by GPs (Figure 1). Patients were eligible to participate in the trial if they satisfied one or more of the following 10 inclusion criteria: (i) on five or more regular medications; (ii) taking 12 or more doses of medication per day; (iii) suffer from three or more medical conditions; (iv) suspected by GPs to be non-adherent with their medication treatment regimen; (v) on medication(s) with a narrow therapeutic index or requiring therapeutic monitoring; (vi) had significant changes made to their medication regimen in the previous 3 months; (vii) had signs or symptoms suggestive of possible medication-induced problems; (viii) had an inadequate response to medication treatment; (ix) admitted to hospital in the preceding 4 weeks; or (x) at risk in managing their own medications due to language difficulties, dexterity problems or impaired sight. Patients were excluded if they were participating in other clinical trials of other treatment modalities.

Intervention strategy

The intervention model sought to establish effective multidisciplinary teamwork between GPs, pharmacists, nurses and other healthcare professionals with the GPs as team coordinators (Figure 1). The goal of the model was to enable effective intervention for patients at risk of medication misadventure in a community setting by using pre-existing working relationships between healthcare professionals or by initiating collaborative arrangements, where no working relationship had already been established. In the start-up phase, GPs and pharmacists gave written consent to participate in the trial, and all patients consented in writing to trial participation during a visit to the GP surgery. GPs and pharmacists in the intervention group were encouraged to attend educational sessions dealing with prescribing issues (including how to ‘unprescribe’ (Figure 1)). Two education sessions were included in the trial. An initial multisite satellitetransmitted education session for intervention GPs and pharmacists was broadcast on 21 August 1999 and included presentations by a team involving a clinical pharmacologist, a geriatrician, opinion leaders in general practice, pharmacy and nursing and a consumer representative. A copy of the tape was made available to GPs unable to attend. This session was followed by a voluntary workshop based on case studies provided by the research team. The second education session was a video conference on 27 October 1999 in which only intervention GPs participated. Themes included geriatric therapeutics, quality use of medicine, ADEs, how to unprescribe, how to use other members of the primary healthcare team, other therapeutic issues and how to implement the medication reviews. The sessions were video taped and provided as soon as possible to intervention GPs who had been unable to attend. GPs allocated to the control group received the videos at the end of the trial. All participants were provided with oral information (GPs and pharmacist by evaluation team members and patients by GPs) and written information about the trial. The written completed consent form was forwarded on to the evaluation team. The model enabled patients at risk to be identified through a number of sources (Figure 1). In practice, identification was mainly by the local pharmacist and GPs. A key process in the intervention group was establishing effective collaboration in the primary healthcare team (mainly linking GPs with pharmacists) during the study start-up. In order to ensure continuity in patient care, only those pharmacists who were linked to patients of intervention GPs were actively involved in the intervention phase. The predominant process adopted in the intervention was a home Br J Clin Pharmacol

58:6

651

L. Sorensen et al.

visit by a pharmacist. In this case, collection of data for medication reviews was initiated by a structured, written GP referral to the community pharmacist that included patient diagnoses and current medication use. Home visits occurred approximately 2 weeks after enrolment. Such visits identified various medication-related risk factors and other issues that needed to be addressed together with information from the GP in the medication reviews. Whilst we envisaged patient acknowledgement of the visit as evidence of its provision in generalizing the service (Figure 1), we relied on health practitioner invoices in this trial as these were accompanied by data that were used in the trial evaluation. A condition of the trial sponsor was that medication reviews were to be completed only by accredited pharmacists. Accredited pharmacists specialize in medication management (and carry out equivalent functions to clinical pharmacists in many settings) and are defined as such through accreditation by the Australian Association of Consultant Pharmacy (AACP) by a process involving short courses or previous documented training in clinical pharmacy, followed by an open-book case-based examination (see http://www.aacp.com.au/ gettingaccredited/index.html for further information). In the cases where the local pharmacist was not accredited (38%), the local pharmacist could undertake the home visit but the evaluation team helped link in an accredited pharmacist to conduct the medication review. As shown in Figure 1, the accredited pharmacist, most commonly the same person, who had undertaken the home visit, prepared a medication review report using the GP information and home-visit findings. This usually occurred within a week of the home visit. The key requirement for these medication reviews was to raise with the patient’s GP possible medication misadventure risks. Specific guidelines for medication reviews were provided to reviewing pharmacists concerning report format and content. The recommendations from the medication review were forwarded to the GP and within 2 weeks of medication review report preparation, recommendations were discussed at a multidisciplinary conference (between the GP, pharmacist and other professional members of the patient’s healthcare team). The GP developed an action plan based on the outcome of the Primary Health Care Team (PHCT) conference, and implemented the actions in consultation with the patient at a subsequent visit to the surgery. The patients were followed up to monitor the outcomes of the action plan at least 6 weeks later. Remuneration was provided for GP consultations, telephone conversations with the pharmacist and conference participation. Pharmacists 652

58:6

Br J Clin Pharmacol

were remunerated for home visits, medication reviews (including conference participation) and telephone conversations with the GP. The intervention model also allowed a range of implementation options such as GP home visits and referral to other practitioners (Figure 1). In practice, these were rarely exercised, with most interventions involving actions only by the GP and pharmacist. Finally, incentives for the GP to participate were seen to be essential for the continuity of a generalized programme. Support for this programme to be used for GP practice incentive points (e.g. clinical audit and quality assurance) was obtained from the Royal Australian College of General Practice. However, the effectiveness trial was not of sufficient duration fully to refine and develop the programme incentives for GP participation. Data collection

At patient enrolment, demographic and other data were recorded on questionnaires by the patients and their GPs. Data were also collected on questionnaires from the GPs and the patients at the follow-up visits for both groups. Measures used by the GPs to report impressions about the patient’s health included ADEs experienced by the patient in the previous 3 months and the Duke’s Severity of Illness Visual Analogue Scale (DUSOI-A) [26]. The DUSOI-A is a 10-cm visual analogue scale, where 0 indicates low severity of illness and 100 high severity of illness. The patient questionnaire, completed in the GP surgery, included data such as number of hospital admissions, number of non-admission hospital services and number of GP visits. Patient-perceived ADEs were also captured from the responses to the following questions at endpoint: ‘Before participating in this project, have you had any symptoms or health problems, which you think were caused by your medication?’; and ‘Since you became involved in this project, have you had any symptoms or health problems, which you think, were caused by your medication?’ Medication use was also self reported by the patients. Health-related quality of life was measured by the SF-36 [27] with the corresponding physical component score (PCS) and mental component score (MCS) of health-related quality of life calculated using Australian norms. Data for the final domain of satisfaction against expectations was obtained at the follow-up visit from GPs, pharmacists and patients who had participated in the trial. Adoption of recommendations by the GP and the patient was also taken as a proxy of satisfaction [24] with the model. In the intervention group only, process details for


each step of the intervention were recorded. Observations made during home visits were recorded using a structured proforma. All medications found in the home were recorded, and the patients indicated whether or not they were currently taking each medication. Medication review reports were subsequently coded by a clinical pharmacist. The codes for the types of possible medication-related problems identified and the recommendations made to address these were adapted from literature sources [28–30]. Adherence to the medication review guidelines was also rated by the clinical pharmacist with regard to format and content. GPs reported on planned and implemented actions. Data for the cost assessment were obtained from the Commonwealth Department of Health and Ageing and the Commonwealth Department of Veterans’ Affairs. Data were available for 356 patients, 200 control and 156 intervention patients. The databases contain information on prescription medication available under the Pharmaceutical Benefits Scheme (PBS) and the Repatriation Pharmaceutical Benefits Scheme (RPBS), and healthcare services available on the Medical Benefits Scheme (MBS). Patients consented to release of the database information for services between 1 February 1999 and 31 March 2000. Patients also consented to the release of information from the Queensland state government about number of days spent in public hospitals during the same time frame. Cost savings per patient were deduced from the differences in the total sum of medication and healthcare costs between intervention and control groups after adjustment for the number of patients in each group and for baseline differences. Marginal cost benefit per patient was defined as the cost savings per patient assuming no change in patient outcomes as a consequence of the intervention. Sample size

The target sample size was 150 in each group. A clinically significant decrease in DUSOI-A of 6 units was based on the difference between the mean scores in DUSOI-A of 66.7 in a study of a similar elderly population of 1013 subjects and the mean score of 60.8 for those patients (n = 42) who exited the study to enter residential care setting (unpublished data). This study also showed that the DUSOI-A was relatively constant in a subgroup of control patients over approximately a 3-month period (initial 56.7, final 56.1). A change of 6 units in DUSOI-A would be detectable with a power of 80% based on a standard deviation of differences of 20, using a paired t-test with a 5% two-sided significance level and a cluster design effect of 1.67.

Assignment

On receipt of consent forms, GPs were randomized on a 1 : 1 ratio according to a preprepared computergenerated randomization schedule. Statistical analysis

The trial was evaluated using an intention-to-treat methodology. The intervention and control groups were compared on health-related factors at baseline, and in the comparison of endpoint results adjustments were made for any baseline differences. Normally distributed variables (such as DUSOI-A and number of drugs) were summarized using means and standard errors, and comparisons were made between the intervention and control groups using t-tests and linear regression. Dichotomous variables (e.g. ADE) were summarized using proportions and differences between groups were compared using c2 test and logistic regression (to adjust for baseline differences). Variables measured on an ordinal scale (e.g. health status change) were compared between groups, in the cohort completing the trial, using ordinal logistic regression. Number of GP visits followed an approximate Poisson distribution, thus Poisson regression (with over-dispersion) was used to model this variable. Because patients were clustered by GP, and the GP was the unit of randomization, it was necessary to take this clustering into account in the analysis. This was achieved using generalized estimating equations with robust estimation of standard errors and P-values. The statistical package SUDAAN [31] was used to perform all statistical analysis and to take account of patient clustering. Ethical approval

As this was an interstate, multidisciplinary study, ethics approval was sought from, and granted by the following ethics committees: The Princess Alexandra Hospital Research Ethics Committee, the University of Queensland Human Ethics Committee, the Austin and Repatriation Medical Centre Human Research Ethics Committee, the Ethics Committee of the University of Melbourne, the Department of Veterans’ Affairs Ethics Committee and The Royal Australian College of General Practitioners Research and Evaluation Ethics Committee. Results Process and impact evaluation

Altogether 92 GPs, 53 pharmacists and 400 patients enrolled in the trial (Figure 2). Of the GPs enrolling patients (84), 69.0% completed the trial. Overall, 56.3% intervention GPs and 70.5% control GPs completed the Br J Clin Pharmacol

58:6

653

L. Sorensen et al.

Pool* of potentially eligible GPs & pharmacies: ~ 600 GPs ~ 1200 pharmacies

Consent obtained from: 92 GPs 53 pharmacies

Randomisation

Intervention Start Up Phase 48 GPs 32 Pharmacists 177 patients

Excluded GPs: 5/48 did not enrol patients 16/48 withdrew Excluded Patients: 58/177 lost to follow up 6/177 deceased 1/177 did not comply with IC 1/177 left area/changed GP 2/177 did not want to continue 3/177 admitted to hospital at FU

Control Start Up Phase 44 GPs 0 pharmacists 223 patients

Excluded GPs: 3/44 did not enrol patients 10/44 withdrew Excluded Patients: 12/223 lost to follow up 4/223 deceased 7/223 did not comply with IC 3/223 left area/changed GP 1/223 did not want to continue Excluded pharmacists: 21/53 not actively involved due to no link with intervention GP

Home Visit** 152/177 patients (145 forms received)

Medication Review*** 150/177 patients (110 reports received)

Implementation 133 /177 reports received

Follow Up Visit 106 /177 reports received

Follow Up Visit 196/223 reports received

Figure 2 Flow diagram of the progress of the trial. *It was not possible to identify how many potential participants had been reached by the promotional material. **One hundred and forty conducted by pharmacists, 12 by GPs. ***Ninety-three completed by community pharmacists, 57 by noncommunity pharmacists. IC, Inclusion criteria; FU, follow-up. The figure is based on the CONSORT guidelines [42]. Note that 150 out of 177 medication reviews were conducted but only 110 were made available to the study team

654

58:6

Br J Clin Pharmacol


trial. Of the intervention GPs five did not enrol any patients, seven found they were too busy/that the trial was time consuming, five could not be reached or did not want to give any reason for withdrawal, one experienced unexpected commitments and one GP found the trial too awkward. Of the control GPs, three did not enrol any patients, six could not be reached or did not want to give any reason for withdrawal, two found they were too busy/that the trial was time consuming, one had to go overseas unexpectedly and one had to withdraw because only one patient was included – and that patient withdrew. Of the enrolled pharmacists, 21 (39.6%) were not actively involved in the trial, because no intervention GP could be linked to the pharmacist. Of the 400 enrolled patients, 302 (75.5%) completed the trial, 106 (59.9%) intervention and 196 (87.9%) control patients (Figure 2). The reason for the different numbers of enrolled intervention and control patients was that GPs were not asked to enrol a specific number of patients, but a minimum of three patients was suggested. Baseline demographic and health assessment measures were similar for intervention and control participants (Table 1), and the distribution of medication use reported by patients was similar. Whilst the intervention model was designed to allow considerable flexibility in GP training, patient recruit-

ment and in intervention (Figure 1), the actual implementation strategies chosen by the healthcare practitioners were limited. Options such as referral to a specialist, use of a community nurse or implementing GP care planning (marked as a dotted square in Figure 1) were not taken up. The majority of GPs chose to implement the basic model by managing their own patients with input from pharmacists as a part of the home visit model. In the intervention group, an average of 5.5 (range 1– 15) problems were identified per medication review. Potential ADE was the most commonly identified problem in the medication review reports, followed by suboptimal monitoring (Table 2). Consequently, the most frequent recommendations were for laboratory monitoring and changes to prescribed medication (Table 2). An average of 6.8 (range 1–17) recommendations were suggested in each review. The reviewing pharmacists sometimes suggested a range of possible recommendations as options for one identified problem, which explains the larger number of recommendations than problems identified. There was also evidence that the basic model was adopted and followed through with recommendations being acted upon. More than half of the recommendations resulted in an action carried out either by the home

Table 1 Baseline characteristics

GPs

Intervention, n = 48

Control, n = 44

P-value

Gender (female)

10 (20.8%)

12 (27.3%)

0.469

Number of patients/GP

Mean (95% CI) 48.6 (45.7, 51.6) (range 32–73) 3.7 (range 0–20)

Mean (95% CI) 45.4 (42.7, 48.1) (range 29–70) 4.9 (range 0–16)

0.114

Patients

Intervention, n = 176

Control, n = 216

P-value, clustered

Age (years)

Gender (% female)

Age (years) Number of drugs/patient reported by patient Number of drugs/patient reported by GP Severity of illness per DUSOI-A

63.1 Mean (95% CI) 72.3 (70.8, 73.8) (range 37–100) 9.1 (8.3–9.8) 9.7 (9.1–10.3) 62.6 (59.2–60.3)

64.4

0.110

0.756

Mean (95% CI) 71.4 (69.8, 73.0) (range 25–99) 8.2 (7.7–8.8) 8.9 (8.3–9.4) 57.1 (53.9–60.3)

0.756 0.220 0.210 0.094

GP, General practitioner; DUSOI-A, Duke’s Severity of Illness Visual Analogue Scale.

Br J Clin Pharmacol

58:6

655

L. Sorensen et al.

Table 2 Problems identified in medication review reports and recommendations of medication changes based on medication reviews and primary healthcare team discussions % of problems N = 602

Problems identified in medication review reports Condition caused by adverse drug reaction Sub-optimal monitoring Label discrepancy/other adherence difficulties, e.g. drug not taken as indicated by label or in patient profile (apparent non-concordance), under or overuse reported by patient, administration aid or device not effective Disorder treated with wrong (or suboptimal) drug Over- or subtherapeutic dosage of correct drug Condition caused by drug–drug, drug–food, drug–laboratory test interaction Other Untreated indication Sub-optimal drug administration Drug use without valid medical indication Poor medication knowledge Medication hoarding or poor storage of medication Problem associated with drug supply No problem stated

Actions suggested following medication reviews and primary health care team discussions Monitoring – laboratory Change drug – substitute one drug for another Other, e.g. resolve non-concordance between patient-reported and GP-recorded drug regimen (4.4% of N); further investigation of potential ADE (1.6% of N) Provision of patient (or carer) education or information Cease drug or trial withdrawal to confirm need for treatment Change dose, dosage interval or frequency Add a drug No recommendation made Monitoring observation and non-lab monitoring Change administration time, route or dose form Rationalize medications held in the home Non-drug therapy suggested Other intervention to improve adherence Refer patient to another professional

16.9% 16.3% 12.8% 9.3% 9.3% 9.1% 7.8% 6.3% 3.8% 3.5% 1.7% 2.5% 0.5% 0.2% % of actions suggested N = 747 21.0% 12.6% 11.9% 10.4% 9.5% 8.7% 6.4% 6.4% 5.0% 3.5% 1.7% 1.5% 0.8% 0.5%

Feedback for 110 patients. Note there are more recommendations than problems, because the reviewing pharmacist sometimes suggested a range of possible interventions.

visitor (6.0%) or by the GP (48.4%). The actions tended to be based on the pharmacist’s recommendations or were triggered by the conference discussions (Table 3). Many (29.3%) of the recommendations not implemented were suggestions for laboratory tests. Also, at least 10 recommendations were not implemented due to patient reluctance. For those actions implemented after the PHCT 656

58:6

Br J Clin Pharmacol

conference, most were carried out successfully, with a positive outcome for 212 (70.7%) enacted recommendations (Table 4). A negative outcome was recorded for only 11 (3.7%) enacted recommendations. The majority of the medication review reports were unexceptional in format and content. Of the 110 medication review reports examined, 13 (11.8%) were noted as having a particularly good layout and three (2.7%)


Table 3 General practitioner enactment of recommendations based on medication reviews, primary healthcare team discussions at followup data collection Inclusion of recommendations*

Frequency

Percent

Recommendations enacted Enacted as recommended in medication review report Recommendation in action plan/follow-up but not in medication review report (initiated as a part of the conference process) Recommendation from medication review report modified then enacted Medication review report included action taken at home visit, e.g. patient or carer education

307 135 54

54.4 23.9 9.6

84 34

14.9 6.0

Recommendations not enacted Other reasons, e.g. other alternative chosen Recommendations were not implemented due to patient reluctance

232 222 10

41.1 39.3 1.8

Not applicable, e.g. no recommendation made Total

25 564

4.4 100.0

*Feedback for 80 patients (who had 183 recommendations made in their medication reviews).

Table 4 Outcomes of implemented actions reported by general practitioner (GP) at follow-up Outcome of action

Frequency

Positive Action taken leading to improvement in symptoms, function or disease Action taken on drug use with no recorded negative or positive clinical outcomes Action taken but reports of no change in symptoms, function, disease or drug use Information provided to the patient to prevent possible dangers

212 66 51 2 93 47

70.9 22.1 17.1 0.7 31.1 15.7

4 5 25 13 11

1.3 1.7 8.4 4.3 3.7

8 3

2.7 1.0

29

9.7

299

100.0

Neutral Outcome not mentioned Action unclear Monitoring – initiated – no results/effects reported or specified Monitoring – results show normal or within therapeutic range Negative Development of side-effects/worsening in symptoms, function or disease; previous treatment re-started Drug change trialled but re-started without an explanation Actions planned but outcome unknown Total

Percent

Feedback for 71 patients, no records for 30 patients = 183 recommendations; 265 recommendations did not apply because they were not enacted, such as local pharmacy to take action (e.g. introduce Dose Administration Aids), referral to specialist, no recommendation made or information given to pharmacist by GP (Table 3).

were found to be poor. Similarly for content, seven (6.4%) reports were rated as remarkably good and 11 (10.0%) as poor. The content of one (0.9%) report was not rated as deficits were felt to be related to poor home visit data collection, where the home visit was completed by a different pharmacist.

Effectiveness evaluation

Functional status Health-related quality of life was measured by using the SF-36 subscales and the physical and mental component scores. The SF-36 was measured at baseline and at the end of the trial, and there were no Br J Clin Pharmacol

58:6

657

L. Sorensen et al.

differences in the scores between the intervention and control patients for the PCS and the MCS at baseline {PCS, P = 0.17, intervention, mean = 30.20 [95% confidence interval (CI) 27.87, 32.53], control, 32.35 (30.49, 34.21); MCS, P = 0.42, intervention, 51.22 (49.12, 53.32), control, 50.15 (48.62, 51.68)} or at endpoint after adjusting for baseline differences in PCS and MCS [PCS, P = 0.94, intervention, mean =31.04 (29.04, 33.04), control, 30.49 (29.08, 31.90); MCS, P = 0.11, intervention, 48.67 (46.86, 50.49), control, 50.69 (49.06, 52.32)], for the cohort with data at both time points. Clinical outcomes The mean of DUSOI-A was reduced by 4.92 for intervention patients and by 1.34 in the control patients. Although the larger reduction of DUSOI-A for intervention patients was not statistically significant, the trend suggests that the intervention may

have had a positive effect on severity of illness (Figure 3a). After adjusting for baseline differences, intervention patients were less likely (although not statistically significant) to report medication-related ADEs than controls at the end of the trial (Figure 3b). A similar trend was found for ADEs reported by GPs. GPs also reported that the percentage of patients experiencing ADEs fell from 36.9% to 9.3% at trial end for intervention patients while almost no change was reported for the control patients (34.9% at baseline and 34.0% at end). The most common type of ADE reported by patients was of a gastrointestinal nature (12 patients before and nine during the trial). As an aside, one patient with an ADE was advised to initiate walking exercises to improve his bowel function and thereby reduce laxative use. Unfortunately, the patient was so enthusiastic that he strained

Figure 3 Health outcomes for (A) Duke’s severity of illness (DUSOI-A). Control (n = 186) ( ), intervention (n = 100) ( ), and (B) patient-reported adverse drug events. Control (n = 180) ( ), intervention (n = 94) ( )

Duke Severity of Illness score

A 75 65 55 45 35 25 15 5

–5 –15 Baseline mean Punclustered Pclustered

Adjusted endpoint mean 0.129 0.249

0.027 0.129

Adjusted mean difference end-baseline 0.129 0.249

B

Percentage of group

25 20 15 10 5 0 ADE reported at Baseline Punclustered 0.027 Pclustered 0.058

658

58:6

ADE reported at Endpoint 0.959 0.970

Br J Clin Pharmacol

ADE at baseline not reported at end 0.137 0.138

ADE at end not reported at baseline 0.218 0.358


his back while walking vigorously, and consequently pain symptoms were aggravated! There were no differences between intervention and control patients in baseline and endpoint measures for number of hospital admissions, number of nonadmission hospital services and number of GP visits. No apparent differences were found in the cumulative number of bed-days between the intervention and control groups. Satisfaction The view of the majority (92%) of intervention GPs was that the model had improved the care of the participating patients; 94% of participating pharmacists found the model useful. From the patient viewpoint, most reported benefiting from participation in the trial – only three (2.9%) of the intervention patients (5.8% of the control patients) felt they had not benefited from participation in the trial. Of the three intervention patients not reporting benefit, one was a nurse, who felt that changes in her medications would have occurred irrespective of the intervention. The nurse also had high expectations of the home visit and was disappointed that the pharmacist appeared to collect medication information only at the home visit. It is possible that one of the other two patients did not have any room for improvement since the patient reported no medication change during the intervention. Cost domain While the cumulative cost of medication increased at a similar rate for intervention and control patients during the trial period, with a trend of a slightly higher rate for intervention patients, the cumulative cost of health-related services increased more for the control patients after initiation of the intervention (Figure 4a,b). After adjusting for differences in cumulative costs vs. time (medication plus medical service costs) up to the time of patient enrolment, the cumulative cost/patient over the 8 months from enrolment was AUS$5730 (£2234) for the control group and AUS$5401 (£2105) for the intervention group. After subtracting the differences in costs for the trial between intervention and control groups [AUS$275 (£107) per intervention patient], the net cost saving per intervention patient (marginal cost benefit) was AUS$54 (~ £19) per patient relative to controls. Incremental cost–effectiveness ratio in reducing ADEs and in improving DUSOI-A for the groups were AUS$69 (~ £24) and AUS$65 (~ £23), respectively. Discussion Development of the intervention model

Considerable time was spent by members of the stakeholder groups in developing a model that would trans-

late into practice and could be generalized across the nation. The present national home medicine review programme implemented by the Australian Government has a number of features that are very similar to the model described in Figure 1. The intervention component of this model also has a number of features related to the perceived contribution to home medicine reviews by pharmacists which are also present in other models trialled at the same time as that described in this study [21, 22]. Whilst GPs were keen for the inclusion of arms in which a GP could make a home visit, undertake care planning, or refer on to nurses or other specialists (Figure 1), the actioning of these arms proved not to occur to any real extent when the study was implemented. A key outcome in the development of the model was the need for the research study to facilitate effective collaboration between most of the GPs and pharmacists as a primary healthcare team. In order to make the intervention translatable, remuneration for GPs and pharmacists was based on the Australian Medicare schedule of payments for practitioner services and was related to patient contact time. In the present study, GPs were given payments of AUS$38.55 (£15.00) for the initial consultation, AUS$20 (£7.79) for discussion with pharmacist (if the conversation lasted for more than 10 min a larger reimbursement was provided), AUS$120 (£47) for development of action plan and consultation with patient, and AUS$21.30 (£8.30) for the follow-up patient consultation 3 months later. Pharmacists were paid AUS$95 (£37) for the home visit and medication review, and AUS$20 (£7.79) for discussions with the GP, the amount being governed by the time taken (< 30 min, otherwise a larger reimbursement was provided). In the implemented Australian Home Medicine Review programme, GPs are paid AUS$126.10 (£49.11) and pharmacists AUS$154 (£60) for their relative contributions. Effectiveness of intervention

The design of many clinical trials is concerned with demonstrating definitive outcomes for a particular intervention. These trials minimize variability by having strict subject selection, by defining rigorous and selective intervention protocols and by using a limited number of highly trained practitioners. Such studies do not always lend themselves to being translated into the wider community, where there is variability in both the patient population and in the execution of an intervention by practitioners in the field. Classical randomized controlled trial (RCT) designs generally cannot accommodate the multiple, flexible and community-driven strategies typically used in practice, and usually RCTs Br J Clin Pharmacol

58:6

659

L. Sorensen et al.

Figure 4 Cumulative cost per person (AUS$) for (A) medication and (B) medical services. 1, Start of patient enrolments; 2, start of home visits; 3, start of medication reviews; 4, Start of Primary Health Care Team (PHCT) conferences; 5, start of follow-up visits

Cumulative cost/person (AUS$)

A 2500

2000

1500

1000

500

0 31 28 31 30 31 30 Jan Feb Mar Apr May Jun 1999

31 31 30 31 30 31 31 29 31 Jul Aug Sep Oct Nov Dec Jan Feb Mar 2000

B

Cumulative cost/person (AUS$)

4000 3500 3000 2500 2000 1500 1000 1 2 3

500 0 31 28 31 30 31 30 Jan Feb Mar Apr May Jun 1999

4

5

31 31 30 31 30 31 31 29 31 Jul Aug Sep Oct Nov Dec Jan Feb Mar 2000 Date

do not consider process evaluation elements [32]. Evaluation of effectiveness also needs to be broad based and appropriate. For this reason, we chose the clinical value compass [23, 25] with its four equally important domains [33] as the suite of instruments to assess outcomes in this effectiveness trial. The first key outcome from this trial was that a range of healthcare practitioners could conduct the intervention with minimal input from the research team in a timely and effective manner. The second was demonstrating that a range of issues related to quality use of medicines such as inappropriate prescribing, poor adherence, need for monitoring, consumer lack of understanding of how to use medicines, and potential ADEs needed to be addressed. Our Report to the Australian Government and the other Australian studies using a similar programme have described these issues 660

58:6

Br J Clin Pharmacol

in more detail [21, 22]. The third outcome is that statistical significance was not demonstrated in any domain. In three of the clinical value compass domains (clinical outcomes, satisfaction and costs), there was evidence of positive results and, in the first (functional status) there was no apparent change over the time course of the intervention. Greatest evidence of effectiveness occurred in the satisfaction domain. This domain consists of evidence of satisfaction with service delivery [24], as evidenced by people adopting the protocol, following the model and taking consequent actions, as well as people expressing satisfaction with the process and its outcomes [23]. GPs took actions on more than half of the recommendations, which is in line with literature values of acceptance of pharmacists’ recommendations varying from 33% [14] to 98% [34]. It should, however, be kept


in mind that some of the withdrawing GPs could have withdrawn due to dissatisfaction with the intervention. Further, participants gave a very high satisfaction rate about all aspects of the trial in the debriefing at the end of the trial. In the clinical outcomes domain, two outcomes have emerged in the intervention relative to the control group over the short time of the trial: a reduction in disease severity as measured by the DUSOI-A, and a reduction in frequency of adverse events reported by both patients and GPs, although not statistically significant. Given that potential ADEs were the most frequently identified problem in the medication reviews and that drug substitution was commonly recommended, the relative reduction in ADE rates may be attributable, in part, to the intervention process, but it may also be due to a patient acquiescence bias. Other components in the clinical value compass clinical outcomes domain, such as mortality and medication treatment complications [23], showed no differences between the intervention and control groups at the end of the trial. In the domain of costs, there is evidence of a reduction in healthcare costs per person as a consequence of the intervention, albeit with a trend to a slight increase in medication costs (Figure 4). An alternative interpretation is that the cumulative costs of the control group continued to rise at about the same rate as before the study commenced, while the costs for the intervention group may have declined slightly (although it appears that this group had a less steep slope even before the study commenced). Adjustment of the cumulative costs for these initial differences resulted in a very marginal cost saving. Whilst it is possible that the increase in medication cost was a result of the intervention, i.e. initiation and substitution of medications based on medication reviews, the timing for the difference (December to January) corresponds to usual seasonal variation in drug expenditure associated with changes in the level of patient copayment for medications in the Australian healthcare system. The lower rate of healthcare service cost for intervention patients may be an immediate benefit of the intervention in spite of the fact that healthcare services, e.g. pathology tests and corresponding costs, were included as suggestions in the medication review reports. The data sources used in this work have a range of limitations and some aspects of those associated with medication data for the Australian Pharmaceutical Benefit Scheme (e.g. capture of only those drugs above the copayment rate for PBS data) have been discussed elsewhere [35]. A mean cost saving of AUS$54 (~ £19) per patient

over an 18-month period was evident in the trial. Importantly, the divergence between intervention and control costs was continuing at the end of the trial, suggesting that an even greater saving may have been realized had the follow-up time been longer. Our group had previously seen a similar divergence in cumulative drug use in nursing home patients, with intervention patients using 14.8% less medication and costing AUS$64 (~ £25) less than control patients over 1 year [9]. Costs in the clinical value compass should strictly include direct medical and indirect social costs [23], but in this study indirect social costs, such as days lost from work, replacement worker costs, care giver costs, were not measured. The cost-effectiveness for this intervention based on cost savings, reduction in adverse events as improved clinical outcomes was small [average ratio AUS$67 (~ £23) per patient]. Sensitivity analysis [36] suggests that the cost savings and intervention costs are of a similar order of magnitude. Benefits are most likely to be observed over a longer time period than the 6 months to which this analysis was limited. Cowper [37] appears to be the first study in which cost-effectiveness in a randomized controlled trial has been assessed in the community after a clinical pharmacy intervention. However, that study was limited to effectiveness in terms of a medication appropriateness index, which is a measure of drug prescribing rather than measured outcomes, as assessed in this study. A commentary on a number of more recent Australian studies [21, 22] has been presented elsewhere [38]. In the functional status domain, the finding of no differences in the SF-36 physical and mental scores between the intervention and control patients is consistent with outcomes for a similar type of intervention in ambulatory war veterans [39]. Malone et al. [39] concluded that clinical pharmacists had no significant impact on health-related quality of life as measured by the SF-36 for veterans at high risk of medication-related problems. The SF-36 covers all of the recommended aspects of the clinical value compass functional status domain [23] except health risk status. The medication review and multidisciplinary conference process did lead to a trend of reduction in health risk status as indicated by a relative improvement in DUSOI-A. Intervention model limitations

One value of an effectiveness trial such as this is to identify issues that may improve the translation of the model into practice. The considerable flexibility in training, patient recruitment and in intervention provided in planning the implementation model (Figure 1) was not Br J Clin Pharmacol

58:6

661

L. Sorensen et al.

used when the model was implemented. It is possible that implementation strategies such as referral to a specialist, use of a community nurse or implementing GP care planning (marked as a dotted square in Figure 1) were not taken up because of issues in autonomy, costs and time, respectively. It was also possible for the GP to use other options as a part of the action plan, such as referral to specialists, in addition to using the information provided by the pharmacist and discussed at the conference, which might have led to a loss of control of the patient care by the GP. The lack of uptake of community nursing option may have been due to the limited funding for this option in the trial budget – or the GP might have been under the impression that the service was already functioning optimally as a part of the usual care of patients. The GP care planning option may have been excluded by GPs due to lack of time, since the GP would have to invest some time in this option. Finally, the model chosen by GPs may also have been influenced by the trial focus of this particular option. Suboptimal monitoring was the second most frequent problem identified in the medication reviews. One explanation could be that some laboratory monitoring recommendations were redundant where the GP had already carried out the laboratory tests but the information was not included on the referral form to the pharmacist. It is possible if the pharmacist had known of these results on referral, monitoring may have been recommended less often; however, other studies have suggested that GP records and possibly performance of laboratory tests can be quite limited [22, 40]. To improve the process, we suggested that laboratory test results be included in the referral form when the model was translated into practice. In the majority of the cases only the GP and the local pharmacist participated in conferences, but bringing together these health professionals at PHCT conferences led to actions in addition to those initially recommended by pharmacists’ recommendations. This is a benefit from such discussions, since these issues might not have otherwise been uncovered. Further benefit might have arisen with the inclusion of a broader range of healthcare professionals (including specialist doctors) involved in the care of a given patient, as the planned model allowed. The inclusion of the patients’ specialist doctors may have been beneficial, for example, when some GPs were not comfortable in changing specialistinitiated medication. Limitations of the effectiveness trial design

The sponsor’s priority of demonstrating effectiveness within a specified time frame precluded the conduct of 662

58:6

Br J Clin Pharmacol

a classical clinical trial. A longer follow-up period may have shown a larger difference in outcome measures and confirmed reported trends. Additional follow-up consultations to monitor the patients’ progress would have enabled more definitive clinical outcome changes to be measured but have clear cost consequences. A doubleblind strategy would also have been preferred so the measurement of DUSOI-A had been blinded, but since the GPs were included in the intervention model, a blinding of DUSOI-A assessment was not possible. Due to time constraints, the provision of an education session nationally by satellite and the nature of data collection, GPs were randomized before enrolling patients, thus potentially creating selection bias. While Table 1 suggests no significance differences between the intervention and control patient groups at baseline in this study, although there was a nonsignificant difference between the groups for ADEs, intervention GPs may have selected patients they expected to benefit from the medication review, while control GPs may have selected patients less ill. ‘Real world’ randomized trials are well known to be subject to treatment selection bias [41] and this limitation of conventional randomized controlled trials was one we faced in conducting the present trial. Indeed, in the ‘real world’, a GP would be expected to select patients most likely to benefit from such a service. An alternative interpretation of the results is that the intervention process was unacceptable to nearly half of the GPs originally enrolled (and thus an even higher proportion of GPs overall) and resulted in a reduction in the baseline-adjusted proportion of patients reporting ADEs but no significant differences in health-related quality of life, Duke’s severity of illness scale, healthcare expenditure or utilization of medical services or drugs. Collecting complete datasets from both the intervention and control groups was a limitation of this study. Consistent with a lesser evaluation burden for the control group, more control patients than intervention patients completed the trial and more complete datasets from patients and GPs were available for control group patients. It should be emphasized that drop-outs due to evaluation burden would not exist when the model is put into practice. GPs were contacted to explore the reasons for lack of response, but in spite of several attempts it was in many cases not possible to reach the GPs, and the reasons suggested for lower response rates for intervention participants were not fully explored. It is possible that some of the withdrawing intervention GPs, who could not be reached, were not in favour of the intervention.


Only parts of the questionnaires had been tested for validity and reliability, e.g. the SF-36. Time constraints did not allow for a thorough test for validity and reliability of the entire questionnaire; however, all documents were reviewed by a number of experts including those on the steering committee, who supported the content validity of the data collection tools. The quality of the medication review reports was evaluated because good communication of identified problems and suggested recommendations is important for optimal collaboration in the primary healthcare team. Most of the medication review reports were rated as neither good nor poor, but there were a number of poor reports among the medication reviews. Importantly, this variability is likely to reflect national practice. Given the emphasis of having the trial conducted by community-based practitioners (to give patients the widest access to such a service), such an outcome was not unexpected and is being addressed by a range of quality programmes presently being implemented nationally. Nelson et al. [23], in describing the clinical value compass, emphasize that it is ‘prudent to start measurement work with a small number of key measures for which it is possible to collect reasonably accurate information at a cost that’s affordable’ and suggest 4–12 carefully chosen outcome and cost measures should be sufficient initially. The decision in this trial to limit the number of outcome variables measured did make undertaking and completing the trial achievable as well as allowing us to estimate its effectiveness. It is recognized that, in principle, a cost-effectiveness evaluation normally includes all actual costs (as distinct from that defined remuneration agreed to by stakeholders to undertake this study) as well as opportunity and social costs. Inclusion of both of the latter costs would have greatly added to the complexity of the analysis and, in addition, would only be partial as full costs were not known. Further, the collection of such data was beyond the funding available to conduct this study. As a consequence, the health economic analysis should be recognized as being limited by not being able to include these data. Conclusion

The medication review model was successfully trialled with participants reporting high levels of satisfaction. Even in the relatively short period of follow-up, there was some evidence of medication misadventure risk reduction and a trend towards positive patient outcomes. By omitting the evaluation burden, the model would be achievable in practice. Indeed, a domiciliary medication review programme similar to this model has now been

implemented into national practice, where GPs and pharmacists are reimbursed by the Australian government for the provision of these services (see http:// www.health.gov.au/epc/dmmr.htm). We thank all staff and patients involved in the conduct of the trial. We are grateful to Robert Peck for encouraging us to undertake this study, Nirmala Pandeya for her assistance with the statistical analysis and to Elaine Beller for her statistical review of the manuscript. The authors also thank Dr George Parkerson (Duke University, USA), for his helpful comments on DUSOI-A. The project was sponsored by the Australian Commonwealth Department of Health and Ageing. M.S.R. is supported by the Australian National Health & Medical Research Council. The study complied with the current laws of Australia.

References 1 Stuck AE, Beers MH, Steiner A, Aronow HU, Rubenstein LZ, Beck JC. Inappropriate medication use in community-residing older persons. Arch Intern Med 1994; 154: 2195–200. 2 Salzman C. Medication compliance in the elderly. J Clin Psychiatry 1995; 56 (Suppl. 1): 18–22. 3 Dartnell JG, Anderson RP, Chohan V et al. Hospitalisation for adverse events related to drug therapy: incidence, avoidability and costs. Med J Aust 1996; 164: 659–62. 4 Hallas J, Harvald B, Gram LF et al. Drug related hospital admissions: the role of definitions and intensity of data collection, and the possibility of prevention. J Intern Med 1990; 228: 83– 90. 5 Nelson KM, Talbert RL. Drug-related hospital admissions. Pharmacotherapy 1996; 16: 701–7. 6 Raschetti R, Morgutti M, Menniti-Ippolito F et al. Suspected adverse drug events requiring emergency department visits or hospital admissions. Eur J Clin Pharmacol 1999; 54: 959–63. 7 Roughead EE, Gilbert AL, Primrose JG, Sansom LN. Drug-related hospital admissions: a review of Australian studies published 1988–1996. Med J Aust 1998; 168: 405–8. 8 Hanlon JT, Schmader KE, Ruby CM, Weinberger M. Suboptimal prescribing in older inpatients and outpatients. J Am Geriatr Soc 2001; 49: 200–9. 9 Roberts MS, Stokes JA, King MA et al. Outcomes of a randomised controlled trial of a clinical pharmacy intervention in 52 nursing homes. Br J Clin Pharmacol 2001; 51: 257–65. 10 Roughead EE, Semple SJ, Gilbert AL. Quality use of medicines in aged-care facilities in Australia. Drugs Aging 2003; 20: 643– 53. 11 Hanlon JT, Weinberger M, Samsa GP et al. A randomized, controlled trial of a clinical pharmacist intervention to improve inappropriate prescribing in elderly outpatients with polypharmacy. Am J Med 1996; 100: 428–37. Br J Clin Pharmacol

58:6

663

L. Sorensen et al.

12 Hunter KA, Florio ER, Langberg RG. Pharmaceutical care for homedwelling elderly persons: a determination of need and program description. Gerontologist 1996; 36: 543–8. 13 Hsia Der E, Rubenstein LZ, Choy GS. The benefits of in-home pharmacy evaluation for older persons. J Am Geriatr Soc 1997; 45: 211–14. 14 Triller DM, Hamilton RA, Briceland LL, Waite NM, Audette CM, Furman CA. Home care pharmacy: extending clinical pharmacy services beyond infusion therapy. Am J Health Syst Pharm 2000; 57: 1326–31. 15 Chrischilles EA, Helling DK, Aschoff CR. Effect of clinical pharmacy services on the quality of family practice physician prescribing and medication costs. DICP 1989; 23: 417–21. 16 Hawksworth GM, Chrystyn H. Clinical pharmacy domiciliary service an extended role of the community pharmacist. J Appl Ther 1997; 1: 325–9. 17 Lowe CJ, Raynor DK, Purvis J, Farrin A, Hudson J. Effects of a medicine review and education programme for older people in general practice. Br J Clin Pharmacol 2000; 50: 172–5. 18 Nathan A, Goodyer L, Lovejoy A, Rashid A. ‘Brown bag’ medication reviews as a means of optimizing patients’ use of medication and of identifying potential clinical problems. Fam Pract 1999; 16: 278–82. 19 Hawksworth GM, Corlett AJ, Wright DJ, Chrystyn H. Clinical pharmacy interventions by community pharmacists during the dispensing process. Br J Clin Pharmacol 1999; 47: 695–700. 20 UKATT Research Team. United Kingdom Alcohol Treatment Trial (UKATT): hypotheses, design and methods. Alcohol Alcohol 2001; 36: 11–21. 21 Krass I, Smith C. Impact of medication regimen reviews performed by community pharmacists for ambulatory patients through liaison with general medical practitioners. Int J Pharm Pract 2000; 8: 111–20. 22 Gilbert AL, Roughead EE, Beilby J, Mott K, Barratt JD. Collaborative medication management services: improving patient care. Med J Aust 2002; 177: 189–92. 23 Nelson EC, Mohr JJ, Batalden PB, Plume SK. Improving health care, Part 1: The clinical value compass. Jt Comm J Qual Improv 1996; 22: 243–58. 24 Reuber H, Blair A. Developing a clinical value compass to monitor urology outcomes at the Toronto East General Hospital. Healthc Manage Forum 2000; 13: 53–6. 25 Weinstein JN, Brown PW, Hanscom B, Walsh T, Nelson EC. Designing an ambulatory clinical practice for outcomes improvement: from vision to reality – the Spine Center at Dartmouth-Hitchcock, year one. Qual Manag Health Care 2000; 8: 1–20. 26 Parkerson GR Jr, Broadhead WE, Tse CK. The Duke Severity of Illness Checklist (DUSOI) for measurement of severity and comorbidity. J Clin Epidemiol 1993; 46: 379–93.

664

58:6

Br J Clin Pharmacol

27 Ware J, Snow K, Kosinski M, Gandek B. SF-36 Health Survey. Manual and Interpretation Guide. Boston: Nimrod Press, 1993. 28 Ostrom JR, Hammarlund ER, Christensen DB, Plein JB, Kethley AJ. Medication usage in an elderly population. Med Care 1985; 23: 157–64. 29 Bootman JL, Harrison DL, Cox E. The health care cost of drugrelated morbidity and mortality in nursing facilities. Arch Intern Med 1997; 157: 2089–96. 30 Strand LM, Morley PC, Cipolle RJ, Ramsey R, Lamsam GD. Drugrelated problems: their structure and function. DICP 1990; 24: 1093–7. 31 Research Triangle Institute. SUDAAN User’s Manual, Release 8.0. Research Triangle Park, NC: Resarch Triangle Institute, 2001. 32 Rychetnik L, Frommer M, Hawe P, Shiell A. Criteria for evaluating evidence on public health interventions. J Epidemiol Community Health 2002; 56: 119–27. 33 Woodward M, Beer J, Mackay S, Gray L. The Clinical Value Compass: achieving benchmarking and quality improvement in aged care. Aust J Ageing 2000; 19: 11–13. 34 Haxby DG, Weart CW, Goodman BW Jr. Family practice physicians’ perceptions of the usefulness of drug therapy recommendations from clinical pharmacists. Am J Hosp Pharm 1988; 45: 824– 7. 35 King MA, Purdie DM, Roberts MS. Matching of Australian prescription claims with manually collected medication data for nursing home residents: implications for prescriber feedback, drug utilisation studies and selection of prescription claims database. J Clin Epidemiol 2001; 54: 202–9. 36 Salkeld G, Cumming RG, O’Neill E, Thomas M, Szonyi G, Westbury C. The cost effectiveness of a home hazard reduction program to reduce falls among older persons. Aust NZ J Public-Health 2000; 24: 265–71. 37 Cowper PA, Weinberger M, Hanlon JT et al. The cost-effectiveness of a clinical pharmacist intervention among elderly outpatients. Pharmacotherapy 1998; 18: 327–32. 38 Roberts MS. Effective pharmacist involvement in the health care team improves patient outcomes. J Pharm Pract Res 2002; 32: 171–4. 39 Malone DC, Carter BL, Billups SJ et al. Can clinical pharmacists affect SF-36 scores in veterans at high risk for medication-related problems? Med Care 2001; 39: 113–22. 40 Read RW, Krska J. Targeted medication review: patients in the community with chronic pain. Int J Pharm Pract 1998; 6: 216– 22. 41 Simon G, Wagner E, Vonkorff M. Cost-effectiveness comparisons using ‘real world’ randomized trials: the case of new antidepressant drugs. J Clin Epidemiol 1995; 48: 363–73. 42 Elbourne DR, Campbell MK. Extending the CONSORT statement to cluster randomized trials: for discussion. Statist Med 2001; 20: 489–96.