Effect of outpatient pharmacists' non‐dispensing roles on patient outcomes and prescribing patterns

Nancy Nkansah; Olga Mostovetsky; Christine Yu; Tami Chheng; Johnny Beney; Christine M Bond; Lisa Bero

doi:10.1002/14651858.CD000336.pub2

. 2010 Jul 7;2010(7):CD000336. doi: 10.1002/14651858.CD000336.pub2

Effect of outpatient pharmacists' non‐dispensing roles on patient outcomes and prescribing patterns

Nancy Nkansah ¹, Olga Mostovetsky ², Christine Yu ², Tami Chheng ², Johnny Beney ³, Christine M Bond ⁴, Lisa Bero ^5,^✉

Editor: Cochrane Effective Practice and Organisation of Care Group

PMCID: PMC7087444 PMID: 20614422

Abstract

Background

The roles of pharmacists in patient care have expanded from the traditional tasks of dispensing medications and providing basic medication counseling to working with other health professionals and the public. Multiple reviews have evaluated the impact of pharmacist‐provided patient care on health‐related outcomes. Prior reviews have primarily focused on in‐patient settings. This systematic review focuses on services provided by outpatient pharmacists in community or ambulatory care settings. This is an update of the Cochrane review published in 2000.

Objectives

To examine the effect of outpatient pharmacists' non‐dispensing roles on patient and health professional outcomes.

Search methods

This review has been split into two phases. For Phase I, we searched the Cochrane Effective Practice and Organisation of Care (EPOC) Group Specialised Register (January 1966 through March 2007). For Phase II, we searched MEDLINE/EMBASE (January 1966 through March 2008). The Phase I results are reported in this review; Phase II will be summarized in the next update.

Selection criteria

Randomized controlled trials comparing 1. Pharmacist services targeted at patients versus services delivered by other health professionals; 2. Pharmacist services targeted at patients versus the delivery of no comparable service; 3. Pharmacist services targeted at health professionals versus services delivered by other health professionals; 4. Pharmacist services targeted at health professionals versus the delivery of no comparable service.

Data collection and analysis

Two authors independently reviewed studies for inclusion, extracted data, and assessed risk of bias of included studies.

Main results

Forty‐three studies were included; 36 studies were pharmacist interventions targeting patients and seven studies were pharmacist interventions targeting health professionals. For comparison 1, the only included study showed a significant improvement in systolic blood pressure for patients receiving medication management from a pharmacist compared to usual care from a physician. For comparison 2, in the five studies evaluating process of care outcomes, pharmacist services reduced the incidence of therapeutic duplication and decreased the total number of medications prescribed. Twenty‐nine of 36 studies reported clinical and humanistic outcomes. Pharmacist interventions resulted in improvement in most clinical outcomes, although these improvements were not always statistically significant. Eight studies reported patient quality of life outcomes; three studies showed improvement in at least three subdomains. For comparison 3, no studies were identified meeting the inclusion criteria. For comparison 4, two of seven studies demonstrated a clear statistically significant improvement in prescribing patterns.

Authors' conclusions

Only one included study compared pharmacist services with other health professional services, hence we are unable to draw conclusions regarding comparisons 1 and 3. Most included studies supported the role of pharmacists in medication/therapeutic management, patient counseling, and providing health professional education with the goal of improving patient process of care and clinical outcomes, and of educational outreach visits on physician prescribing patterns. There was great heterogeneity in the types of outcomes measured across all studies. Therefore a standardized approach to measure and report clinical, humanistic, and process outcomes for future randomized controlled studies evaluating the impact of outpatient pharmacists is needed. Heterogeneity in study comparison groups, outcomes, and measures makes it challenging to make generalised statements regarding the impact of pharmacists in specific settings, disease states, and patient populations.

Plain language summary

Non‐traditional roles of outpatient pharmacists.

The role of pharmacists in the community includes more than dispensing medications. It involves identifying, preventing, and resolving drug‐related problems, as well as encouraging proper use of medications and general health promotion and education.

This review found forty‐three studies which evaluated non‐traditional roles of pharmacists. In general, the data included in this review supported the roles of pharmacists in patient counseling, therapeutic management, and providing health professional education with the goal of improving patient process of care and clinical outcomes. Non‐traditional roles of outpatient pharmacists improves health care outcomes. The data show that educational outreach visits may impact physician prescribing patterns.

Background

In the past three decades, the roles of pharmacists in patient care have expanded from the traditional tasks of dispensing medications to working with other health professionals and the public. Multiple systematic reviews and meta‐analyses have evaluated the impact of pharmacist‐provided patient care on health‐related outcomes (Machado 2007a; Machado 2007b). It is important to conduct systematic reviews in this area because both the results and quality of the original studies vary. Thus, a rigorous review enables us to assess the best available evidence on the effects of pharmacist interventions.

The systematic reviews conducted thus far have focused on care rendered in specific practice settings (for example, ambulatory care, community pharmacy, acute care, long‐term/intermediate care) (Beney 2000; Blenkinsopp 2003; Christensen 2006; Horn 2006; Kaboli 2006; Kane 2003; Royal 2006; Singhal 1999; Tully 2000; Van Wijk 2005; Westerlund 2006), to specific patient populations (for example, geriatric, pediatric) (Hanlon 2004; Holland 2008; Rollason 2003; Sanghera 2006; van Eijken 2003), and in specified therapeutic areas (for example, anticoagulation, antibiotic utilization, asthma, diabetes, depression, heart failure, hypertension, immunizations, mental health, Parkinson's disease, tobacco cessation) (Dent 2007; Donovan 2006; Finley 2003; Hogue 2006; Holland 2005; Jenkins 1996; Lindenmeyer 2006; Machado 2007a; Machado 2007b; Manley 2002; McLean 2005; Ponniah 2007; Simonson 2007; von Gunten 2007). A few reviews have been conducted to evaluate the impact of pharmacist‐provided care on specific health outcome criteria (for example, humanistic) (Pickard 1999; Pickard 2006; Schumock 1996; Schumock 2003). Although there is some overlap in the focus of previous reviews, there are also gaps in the types of interventions assessed (for example, pharmacist‐care provided to socio‐economically, ethnically, or linguistically diverse patient populations or patients with low health literacy). To our knowledge, there are no comprehensive systematic reviews thoroughly evaluating randomized controlled trials studying the impact of pharmacist‐provided care in outpatient practice settings.

Because the impact of pharmacist‐provided services in the hospital setting has been well‐studied, this systematic review focused on services provided by outpatient pharmacists in community or ambulatory care settings. This review encompassed all outpatient pharmacist services targeted toward patients and health professionals, as well as all types of clinical disease states and health care process measures. This was an update to the previous Cochrane systematic review (Beney 2000) that incorporated the studies that have been published since 2000 as well as studies not included in the original review.

Objectives

The objective of this review was to examine the effect of outpatient pharmacists' roles on patient and health professional outcomes. Relevant health professional outcomes or healthcare practice measures included changes in prescribing patterns (for example, appropriateness of or prescribing, therapeutic duplication) and disease control (for example, disease‐specific test ordering). Relevant patient outcomes included changes in clinical disease markers (for example, blood pressure) and humanistic quality of life outcomes.

We examined the following main hypotheses:

1. Does the delivery of patient‐targeted services by pharmacists improve patient or health professional outcomes compared to the delivery of the same services by other health professionals?

2. Does the delivery of patient‐targeted services by pharmacists improve patient or health professional outcomes compared to the delivery of no comparable services?

3. Does the delivery of health professional‐targeted services by pharmacists improve patient or health professional outcomes compared to the delivery of the same services by other health professionals?

4. Does the delivery of health professional‐targeted services by pharmacists improve patient or health professional outcomes compared to the delivery of no comparable services?

To test the above hypotheses, we examined the following comparisons.

1. Pharmacist services targeted at patients versus services delivered by other health professionals.

2. Pharmacist services targeted at patients versus the delivery of no comparable services.

3. Pharmacist services targeted at health professionals versus services delivered by other health professionals.

4. Pharmacist services targeted at health professionals versus the delivery of no comparable services.

Methods

Criteria for considering studies for this review

Types of studies

Study designs that meet Effective Practice and Organization of Care Group (EPOC) inclusion criteria are randomized controlled trial (RCT), controlled clinical trial (CCT), controlled before and after study (CBA) and interrupted time series (ITS). In this area of research, it has historically been challenging to identify a substantial number of RCTs in the literature. In the original review and 2000 update, all study designs mentioned above were included. Due to the substantial increase in the number of published RCTs studying the effect of pharmacists' interventions on patient and health professional outcomes, we limited the current update to RCT study designs.

We included

RCTs randomizing:   patients;   pharmacists;   practices (pharmacies or medical clinics); or   geographical areas.

Types of participants

The participants for all comparative studies we included in this review were pharmacists (or pharmacies) who deliver services in outpatient settings other than, or in addition to, drug compounding and dispensing. We excluded studies involving services to patients in hospitals or skilled nursing facilities. We included studies of pharmacists delivering services to outpatients in a clinic attached to a hospital or a day hospital.

Types of interventions

The types of interventions we included were any services delivered by pharmacists other than drug compounding and dispensing. When available we collected additional data on the content of each intervention including recipients, format, source, timing, setting, and cost.

Types of outcome measures

We included studies only if 1) reported primary outcomes were objective with respect to measurement of health care process measures or patient outcomes and 2) relevant and interpretable data were presented. We therefore excluded subjective outcomes (for example, self‐reporting of symptoms, medication knowledge, satisfaction with pharmacist services) or outcomes for which reporting was incomplete (for example no numerical values reported, no baseline data provided). To minimize reporting bias, we excluded outcomes that were not primary. For studies that did not explicitly report which outcomes were primary, we included all objective and relevant outcomes.

We excluded adherence outcomes because there is another Cochrane review that assessed interventions to improve adherence (Haynes 2008). We also excluded resource‐utilization and cost outcomes because these endpoints were recently assessed in another systematic review (Perez 2009).

Search methods for identification of studies

When the original review was performed, there were few randomized controlled trials evaluating non‐dispensing roles of outpatient pharmacists. Studies were identified by electronically searching the EPOC Specialised Register, MEDLINE, EMBASE, PHARMLINE and International Pharmaceutical Abstracts from January 1,1966 through December, 1995. Professional librarians were consulted to advise on a broad search strategy for each database. In MEDLINE, broad searches using the MeSH headings 'pharmacy' and 'pharmacist' and each of the following publication types 'randomized controlled trial', 'controlled clinical trial', 'comparative study', 'follow up study', 'prospective study', and 'evaluation study' were performed.

The following journals were hand searched: American Journal of Hospital Pharmacy (1985 through 1995), International Journal of Pharmacy Practice Research (1987 through 1995), Journal of Social and Administrative Pharmacy (1987 through 1995), Scanner (a pharmacy abstract journal) (1987 through1995), and The Pharmaceutical Journal (1960 through1997). The Pharmacy Practice Research Literature Index (1984 through1994) compiled by Peter Abel and published by the UK Pharmacy Practice Research Resource Centre, University of Manchester, England, was also searched.

The reference lists of trials identified for the review, as well as other review articles on the extended roles of pharmacists, were checked. Non‐English language publications, if found, were to be included in the review.

An attempt was made to identify unpublished studies and works in progress by searching, for 1990 through 1995, the published abstracts of the annual meetings of the American Society of Hospital Pharmacists, Health Service and Pharmacy Practice Research Conference (UK), Pharmacy Practice Research Sessions of the Royal Pharmaceutical Society of Great Britain Annual Conference, and proceedings of the UK Clinical Research Association.

For the 2000 update, relevant studies were located by searching the EPOC Specialised Register, electronically searching MEDLINE, and ongoing handsearching of the International Journal of Pharmacy Practice Research and The Pharmaceutical Journal.

Given the significant increase in publications in this area over the past several years, we split the search for this update. We will complete this update in two phases. Phase I (the current update) consists of studies identified in prior versions of this review and studies identified in the EPOC Specialised Register search (January 1966 through March 2007). Phase II (in progress) will include studies identified in prior versions of this review, the Phase I update, and studies identified through a MEDLINE and EMBASE (January 1966 through March 2008) search. Specific search criteria are included in Appendices.

Data collection and analysis

Selection of studies

Two review authors independently selected the trials to be included in the review. We resolved disagreements by discussion of the articles by at least two of the authors of the review.

Data extraction and management

We collected data using the EPOC Data Extraction Checklist. To streamline the data collection process, we built an online database on the Quesgen platform using the Data Extraction Checklist questionnaire. Two review authors independently extracted data for each study with a focus on outcomes and characteristics aimed at reducing bias. We discussed and reconciled differences in coding.

Assessment of risk of bias in included studies

Two review authors independently assessed the risk of bias of all studies eligible for the review using the EPOC Data Extraction Checklist. We adjudicated discrepancies by discussion of the studies. We assessed allocation concealment, blinding, follow‐up of patients or health professionals (when applicable), baseline measurement, reliability of outcome measures, and protection against contamination. For included studies, the risk of bias characteristics are described in the Characteristics of included studies table. We identified studies with unit of analysis errors. No pooled data included unit of analysis errors.

Measures of treatment effect

We reported results for baseline (pre‐intervention) and end‐of‐study (post‐intervention) periods if available (see 'Outcomes Table' under Data and analyses). Where possible, we calculated pre‐post intervention differences for each outcome for control and intervention groups, and the difference of pre‐post intervention change between study groups (result interval). In all cases, we reported a more favorable outcome in the intervention group as a positive finding (that is where changes from baseline are in the intended direction) and vice versa as a negative finding. For quality of life outcomes, we did not report raw data for each quality of life domain; instead we listed each domain measured under the 'primary outcomes' column in the 'Outcomes Table' (under Data and analyses) and indicated in the 'significance' column which domains were significantly improved in intervention versus control groups during the course of the study. All outcomes included in this review are listed under the Characteristics of included studies and Data and analyses tables.

Assessment of heterogeneity

Among the included studies, there was great heterogeneity in comparison groups, intervention type, outcomes assessed, duration of intervention, length of follow‐up, and measurement used for outcomes.

We attempted to perform a meta‐analysis by subgrouping studies based on clinical disease state and outcome type. Unfortunately, there were insufficient data across the 43 included trials to perform subgroup analyses on all disease states.

There was a high degree of heterogeneity in the types of outcomes measured for each disease state. For example, in the four studies assessing disease control in patients with depression, one study used Brief Inventory of Depressive Symptoms (BIDs), Beck Depression Inventory (BDI), and Work and Social Disability Scale (WSDS) (Finley 2003), one study used BDI (Rickles 2005) and two studies used the self‐rating Hopkins Symptom Checklist (SCL) (Brook 2003b; Capoccia 2004). Due to the different outcome measures and measurement units, we were unable to pool these outcomes into one analysis. The same issue was present in studies targeting patients with asthma, chronic obstructive pulmonary disease (COPD), heart failure, hyperlipidemia and anticoagulation therapy. In these cases, we were unable to perform a meta‐analysis due to the reasons described above, as well as the small number of studies performed with these disease states. We present data separately for each of these studies.

Subgroup analysis and investigation of heterogeneity

For studies measuring blood pressure and glycosylated hemoglobin (HbA1C), we collected enough data points to potentially perform a pooled analysis. Although these groups of studies were comparable in terms of disease state studied and outcomes assessed, there was variability in intervention type and length of follow‐up. To minimize heterogeneity in these pooled analyses, we included in the meta‐analysis only studies with similar disease state, intervention type, and length of study.

Performing a pooled analysis for continuous outcomes requires pre‐ and post‐ means and standard deviations for outcome measures for both control and intervention groups. Reporting of standard deviations was incomplete; only three of the seven studies measuring systolic and diastolic blood pressures and one of the five studies measuring HbA1C reported standard deviations.

We considered two methods to yield a standard deviation for data pooling purposes: 1) calculating a standard deviation from a P value and 2) imputation (using the standard deviation reported in other studies included in the analysis). Standard deviations derived from P values resulted in a high degree of study heterogeneity (I² > 80%). Imputation had the least effect on study heterogeneity (I² = 0). Given these observations, we chose the imputation method.

Results

Description of studies

The prior update to this review (Beney 2000) identified 25 studies that met inclusion criteria. Six of the 25 included studies were pre‐post designs, controlled by a separate site (Cody 1998; Lai 1998; Peterson 1995; Peterson 1997; Schaffner 1983; Tamai 1987), two were quasi randomized controlled trials (Erickson 1997; McKenney 1973), and the remainder were randomized controlled trials.

In Phase I of this update, we identified 107 publications that met our search criteria. Of these, 64 were excluded from the final analysis (see Characteristics of excluded studies and Excluded studies). All included studies were randomized controlled trials. One study was a before‐and‐after pragmatic randomized controlled trial (Hall 2001).

Characteristics of interventions

For study details see the Characteristics of included studies table.

Of the 43 included studies, seven studied pharmacist interventions targeted at health professionals (Diwan 1995; Freemantle 2002; Hall 2001; Ilett 2000; Stergachis 1987; Turner 2000; Watson 2001) and 36 reported on pharmacist interventions targeted at patients. In 11 of the 36 studies targeted at patients, the pharmacist intervention also targeted health professionals (Borenstein 2003; Choe 2005; Gattis 1999; Hanlon 1996; Jackson 2004; Mehos 2000; Sadik 2005; Schneider 1982; Sookaneknun 2004; Taylor 2003; Tsuyuki 2002). In most of these studies, pharmacists provided: a) oral or written recommendations to physicians regarding therapy modifications or resolution of medication‐related problems and b) multiple follow‐up visits with patients spanning several months (range: 1 month to 12 months). All but one of the included studies compared pharmacist services targeted at patients or health professionals versus provision of no comparable services (or usual care). One study (Hawkins 1979) compared pharmacist services with services provided by other health professionals. Eight of the 43 studies were randomized by clinical practice or region, with the remainder randomizing by individual patient or health professional.

In all seven studies targeted at health professionals, pharmacists conducted educational outreach visits at physician practices to promote guideline‐based prescribing for certain medication classes including antibiotics (Ilett 2000) and nonsteroidal anti‐inflammatory drugs (NSAIDs) (Freemantle 2002; Stergachis 1987; Watson 2001), and for certain disease states including Helicobacter pylori infection (Hall 2001), heart failure (Freemantle 2002; Turner 2000), and cardiovascular disease (Diwan 1995). Overall, pharmacists conducted one or two visits lasting 10 to 15 minutes within a study period ranging from 3 months to 24 months. Educational outreach visits are the focus of a Cochrane review (O'Brien 2007). This review evaluated all but two (Stergachis 1987; Turner 2000) of the seven studies identified above.

In 8 of 36 studies targeted at patients, the main focus of the pharmacist intervention was patient education (Barbanel 2003; Brook 2003a; Gonzalez‐Martin 2003; Goodyer 1995; Paulos 2005; Rickles 2005; Sarkadi 2004; Van Veldhuizen 1995). One study evaluated the effect of home blood pressure monitoring on blood pressure control with the pharmacist providing telephone follow‐up to assess blood pressures and response to therapy (Mehos 2000). In the rest of the patient‐targeted studies, pharmacist interventions were complex and commonly involved pharmaceutical therapy management consisting of pharmaceutical therapy optimization, monitoring of disease control and adverse drug reactions, identification of drug‐drug interactions, compliance assessment, and patient education. Twenty‐two studies took place in outpatient medical clinics, ten studies took place in community pharmacies (Barbanel 2003; Brook 2003a; Cody 1998; Park 1996; Paulos 2005; Rickles 2005; Sarkadi 2004; Sookaneknun 2004; Tsuyuki 2002; Weinberger 2002), one study took place at a home care agency (Meredith 2002), and two studies involved hospital pharmacists following recently discharged patients at home (Jackson 2004; Peterson 2004). The duration of the intervention ranged from 14 to 120 minutes with 1 to 22 intervention events conducted over the study period of 6 weeks to 23 months. Post‐intervention follow‐up was performed in two trials to assess duration of intervention effect after the studies were completed (Odegard 2005; Sarkadi 2004).

In most patient‐targeted studies, controls were 'usual care' groups in which patients continued to receive standard care from primary care health professionals; the usual care differed from the service provided by the pharmacist to the intervention group. In three of the seven studies targeting health professionals, control groups received no intervention (Diwan 1995; Ilett 2000; Turner 2000). In two of the other health professional‐targeted studies, control groups received a non‐pharmacist intervention. In one study, the control group received a non‐targeted intervention (Freemantle 2002), and in the other study, the control group received mailed practice guidelines, but not the educational outreach visit by the pharmacist (Hall 2001). Two studies had more than one control group (Watson 2001; Weinberger 2002). In the first study, which targeted health professionals to study the effect of an intervention composed of mailed practice guidelines and education outreach visits by the pharmacist, the first control group received no intervention while the second control group received mailed practice guidelines (Watson 2001). In the second trial, which targeted patients with asthma and COPD, the first control group received usual care while the second control group received home peak flow monitors but not follow‐up by the pharmacist (Weinberger 2002).

Characteristics of health professionals delivering the intervention

In all studies, interventions were performed by either practicing pharmacists, pharmacy residents, or doctor of pharmacy students. In most studies, 1 to 4 pharmacists performed the intervention, but some studies involved more than 10 pharmacists across multiple practices (Bond 2000; Brook 2003b; Diwan 1995; Freemantle 2002; Malone 2001; Rickles 2005).

Target population

In six of seven studies targeted at health professionals, participants were selected based on location. Two studies selected participants from general practices within one or more health authorities (Freemantle 2002; Hall 2001) and four studies selected participants within a specific region (Diwan 1995; Ilett 2000; Stergachis 1987; Turner 2000). In one of seven studies, participating practices were selected based on their use of a specific computer system (Watson 2001).

Of the 36 studies targeting patients, 27 studies selected participants based on the clinical disease state; some studies included patients from more than one disease state. The following clinical disease states were represented across the included studies: asthma (Barbanel 2003; Gonzalez‐Martin 2003; Weinberger 2002), COPD (Solomon 1998; Weinberger 2002), depression (Brook 2003a; Capoccia 2004; Finley 2003, Rickles 2005), diabetes (Choe 2005; Clifford 2005; Hawkins 1979; Jaber 1996; Odegard 2005; Sarkadi 2004; Van Veldhuizen 1995), heart failure (Gattis 1999; Goodyer 1995), hyperlipidemia (Bogden 1997; Paulos 2005; Peterson 2004; Tsuyuki 2002) and hypertension (Borenstein 2003; Hawkins 1979; Mehos 2000; Okamoto 2001; Park 1996; Schneider 1982; Solomon 1998; Sookaneknun 2004). Additionally, five studies selected participants based on characteristics other than the clinical disease state; these studies focused on patients with high risk of medication related problems (Malone 2001; Taylor 2003), home care patients (Meredith 2002), patients with repeat prescriptions (Bond 2000), and patients on warfarin therapy (Jackson 2004).

The number of participants ranged from 21 to 6000 patients and 17 to 112 health professionals. Nine studies included fewer than 50 participants, 14 studies had between 50 and 100 participants, 12 studies had between 101 and 500 participants and eight studies had more than 500 participants. One study targeted pediatric patients (Gonzalez‐Martin 2003) and the rest of the studies targeted adults, with nine studies focusing on elderly patients 65 years of age and older.

Risk of bias in included studies

Characteristics aimed at reducing bias are listed in the 'risk of bias' table under each study table in the Characteristics of included studies section. See Figure 1 and Figure 2 for graphic representations of the data presented below.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

There were no major differences in the risk of bias of studies targeted at patients versus studies targeted at health professionals. Three of 43 studies had no risk of bias (Malone 2001; Meredith 2002; Peterson 2004). Only 15 of 43 studies adequately concealed allocation. Adequate follow‐up of patients or health professionals (depending on target subject of study) was done in 27 of 43 studies. Baseline measures of primary outcomes were performed and were similar between intervention versus control groups in 27 of 43 studies, and protection against contamination was adequate in 12 of 43 studies. Because we only included objective primary outcomes in our review, most studies (41) were coded as having reliable outcomes and blinded assessment of outcomes.

Four studies had a unit of analysis mismatch. Of the four studies, three did not correct for clustering in the study analyses (Freemantle 2002; Turner 2000; Weinberger 2002). In two of these studies, the unit of allocation was practice while the unit of analysis was patient (Turner 2000; Weinberger 2002) and in the third study, unit of allocation was health authority while unit of analysis was practice (Freemantle 2002). One study corrected for clustering in the analysis (Bond 2000).

Effects of interventions

All included outcomes are listed under the Characteristics of included studies and Data and analyses sections.

Comparison 1. Pharmacist services targeted at patients versus services delivered by other health professionals

For detailed descriptions of outcomes see Analysis 1.1.

1.1. Analysis.

Comparison 1 Pharmacist services targeted at patients versus services delivered by other health professionals, Outcome 1 Outcomes Table: 2009 Review.

Outcomes Table: 2009 Review
Study	Primary outcomes	Pre‐intervention   (intervention vs control group)	Post‐intervention   (intervention vs control group)	Change due   to intervention (intervention vs   control group)	Result interval   (ΔI ‐ ΔC)	Significance	Notes
Hawkins 1979	PATIENT OUTCOMES   1. Mean systolic blood pressure (mmHg) 2. Mean diastolic blood pressure (mmHg) 3. Mean fasting blood sugar (mg/dl)	1. 145 vs 143 2. 86 vs 86 3. 192 vs 182	1. 147 vs 141 2. 84 vs 84 3. 184 vs 189	1. 2 vs ‐2 2. ‐2 vs ‐2 3. ‐8 vs 7	1. 4 2. 0 3. 15	1. p</= 0.001 2. not sig 3. not sig

Outcomes Table: 2009 Review
Study	Primary outcomes	Pre‐intervention   (I vs C)	Post‐intervention   (I vs C)	Change due   to intervention   (I vs C)	Result interval   (I ‐ C)	Significance Measure	Notes
Diwan 1995	PROCESS   number of Rx of lipid‐lowering drugs/month	Women: 25.6 vs 22 Men: 20 vs 17	Women: 35.5 vs 21 Men: 28 vs 20.6	Women: 9.9 vs ‐1† Men: 8 vs 3.6†	Women: 10.9† Men: 4.4†	n/a* Stat sig change in # of Rxs among women	*exact value not reported   †calculated from reported data
Freemantle 2002	PROCESS   increase in # of pts treated according to practice guidelines	n/a	n/a	n/a	5.2%   (OR = 1.24)	95% CI 1.7, 8.7   (OR 95% CI 1.07‐1.42)
Hall 2001	PROCESS   1. change in omeprazole use due to intervention   2. change in metronidazole use due to intervention	1. n/a   2. n/a	1. n/a   2. n/a	1. n/a   2. n/a	1. ‐0.02 dose units per year   2. ‐0.005 dose units per year	1. 95% CI ‐0.12, 0.08   2. 95% CI ‐0.025, 0.015
Ilett 2000	PROCESS   Antibiotic prescribing (# all ABX RXs)	5182 vs 6666	7262 vs 9654	‐2080 vs ‐2988	908	p‐value n/a*	*not reported
Stergachis 1987	PROCESS   number of Rx per 1000 enrollees per physician (1. piroxicam, 2. ibuprofen, 3. salicylates)	1. 58 vs 27 2. 167 vs 194 3. 142 vs 114	1. 55 vs 40 2. 183 vs 202 3. 213 vs 125	1. ‐3 vs 13† 2. 16 vs 8† 3. 71 vs 11†	1. ‐16† 2. 8† 3. 60†	1. ns* (p‐value n/a) 2. ns* (p‐value n/a) 3. sig* (p‐value n/a)	*exact value not reported   †calculated from reported data
Turner 2000	PROCESS   1. utilization of ACEIs in patient receiving digoxin and   furosemide for CHF (# of patients)   2. targeted daily dose of ACEIs (# of patients)	1. 71 vs 40   2. 52 vs 29	1. 72 vs 38   2. 59 vs 28	1. 1 vs ‐2†   2. 7 vs ‐1†	1. 3†   2. 8†	1. p‐value NS*   2. p‐value 0.14	*exact value not reported   †calculated from reported data
Watson 2001	prescribing of ibuprofen, diclofenac and naproxen as % of total NSAID prescribing (mean %(SD))   (intervention vs mailed guideline vs no intervention)	78.1(2.6) vs 79.0(4.9) vs 77.0(7.6)	82.7(2.6) vs 81.2(3.7) vs 80.3(7.2)	4.6 vs 2.2 vs 3.3†	1. 2.1 (intervention vs no intervention)   2. 1.6 (intervention vs mailed guideline)	1. 95% CI ‐0.8, 5.0   2. 95% CI ‐1.4, 4.7	†calculated from reported data

Date	Event	Description
16 June 2010	New citation required but conclusions have not changed	New search, criteria for included studies changed to only include RCTs, new authors
16 June 2010	New search has been performed	Reconciled old and new studies
21 August 2008	Amended	Converted to new review format.
18 January 2000	New citation required and conclusions have changed	Substantive amendment

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Outcomes Table: 2009 Review			Other data	No numeric data
2 Systolic Blood Pressure (mmHg)	4	734	Mean Difference (IV, Random, 95% CI)	‐6.32 [‐8.80, ‐3.83]
3 Diastolic Blood Pressure (mmHg)	4	734	Mean Difference (IV, Random, 95% CI)	‐3.12 [‐4.57, ‐1.67]
4 Decrease in HbA1C (%)	2	260	Mean Difference (IV, Random, 95% CI)	‐0.75 [‐1.41, ‐0.09]

Methods	RCT (randomized by patient)
Participants	community pharmacy in Tower Hamlets, east London (United Kingdom)   patients with asthma   patients ‐ 24 (12 intervention group, 12 control group)   health professional (delivering intervention) ‐ 1   practice ‐ 1   no unit of analysis error
Interventions	targeted towards PATIENTS   pharmacists reviewed inhaler technique, provided personal education on a variety of asthma‐related topics and followed up with patients with weekly telephone calls vs usual care   length of intervention ‐ 45 to 60 min initial education session and weekly telephone calls   number of interventions ‐ 12 during 3 months
Outcomes	PROCESS not measured   PATIENT improvement in asthma symptoms based on North of England asthma symptom scale
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Patients were randomized using sealed envelopes to intervention or control groups
Blinding?   All outcomes	Low risk	Outcome variables are objective
Follow‐up of professionals?	Low risk	Not applicable
Follow‐up of patients?	Low risk	There were follow‐up data on all 12 in the intervention group and 11 in the control group (1 person moved away).
Baseline measurement?	Low risk	"baseline scores were similar in the intervention and control groups"
Reliable outcome measures?	Low risk	objective validated instrument ‐ North of England asthma symptom scale
Protection against contamination?	Unclear risk	Did not explicitly mention

Methods	RCT (by medical practice)
Participants	university affiliated setting   medical practices in Grampian, United Kingdom   patients on repeat medications   patients ‐ 3074 (1614 intervention, 1460 control)   health professional (delivering intervention) ‐ 62   practice ‐ 19   unit of analysis mismatch corrected (randomized by practice, analyzed by patient‐‐‐analysis accounted for clustering effect)
Interventions	targeted towards PATIENTS   pharmacist dispensed repeat prescriptions following a protocol to check whether items were required, or patients were experiencing side‐effects or drug interactions vs usual care   length of the intervention ‐ not clear   number of interventions ‐ 12 during 12 months
Outcomes	PROCESS   prescribing   costs (excluded)   PATIENT   death rate   adverse drug reactions hospital admissions
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	Random number tables were used
Blinding?   All outcomes	Low risk	Objective outcome assessed
Follow‐up of professionals?	Low risk	Not applicable
Follow‐up of patients?	High risk	Of 1614 patients recruited to intervention group, complete data were available for 905 patients
Baseline measurement?	Unclear risk	Baseline measures not reported
Reliable outcome measures?	Low risk	Objective outcome assessed
Protection against contamination?	Low risk	Allocation was by practice

Methods	RCT (by patient) + CBA (by geographic area)   Similar control site: NOT CLEAR
Participants	community pharmacies of the Kaiser Permanente   patients ‐ 6000 in the RCT / 4600 in the CBA   pharmacies ‐ 9 in the RCT / 101 in the CBA   no unit of analysis error for the RCT   unit of analysis error for the CBA
Interventions	targeted toward PATIENTS   comparison of three models   Control model: usual care before 1992 in California   California state model (1992) which requires outpatient pharmacist to counsel all patients who receive new or changed prescription about direction for use, the importance of compliance, proper storage, and relevant precautions and warnings.   Kaiser Permanente (KP) model that focuses on a more comprehensive pharmacist consultation and other elements of pharmaceutical care on selected high‐risk patients.   length of study ‐ 23 months
Outcomes	PROCESS not measured   PATIENT Quality of life (SF 36)
Notes	see McCombs 1998 for design
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Not explicitly described; appears to have been performed via a central randomized scheme/computer system
Blinding?   All outcomes	Low risk	Objective outcome assessed; nature of study design appears to have ensured blinding ‐ patients were randomly assigned to separate pharmacies with a more comprehensive consultation model
Follow‐up of professionals?	Low risk	Not applicable
Follow‐up of patients?	Unclear risk	Not explicitly described
Baseline measurement?	Unclear risk	Not explicitly described
Reliable outcome measures?	Unclear risk	Patients self‐completed and mailed in quality of life surveys
Protection against contamination?	High risk	NOT DONE for the RCT / DONE FOR the CBA

Methods	RCT (randomized by health center)
Participants	primary care health centers in Sweden, excluding health centers in counties with a university hospital, department of general medicine, or extensive activities for the prevention of cardiovascular disease or both.   education provided by pharmacists focused on treating patients with cardiovascular disease.   patients ‐ 1308   practices ‐ 134 health centers   professional ‐ 1   no unit of analysis error (the unit of analysis was the health center)
Interventions	targeted towards HEALTH PROFESSIONALS   pharmacist vs. no intervention   pharmacist conducted academic group detailing sessions at health centers   length of intervention ‐ four 30 minute detailing sessions over a 5 month period   product related
Outcomes	PROCESS   number of prescriptions of lipid‐lowering drugs/month   PATIENT   not measured
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	Coin toss
Blinding?   All outcomes	Low risk	Objective outcome assessed
Follow‐up of professionals?	Low risk	Data collected from 87% of health centers
Follow‐up of patients?	Low risk	Data were missing on 17 of 1308 patients
Baseline measurement?	Unclear risk	Not explicitly described; some differences existed in the number of baseline prescriptions between control and intervention groups. May not affect outcome
Reliable outcome measures?	Low risk	Objective outcome assessed
Protection against contamination?	Low risk	Detailing occurred in specific health center

Methods	RCT (randomized by health authority)
Participants	outpatient general practices in United Kingdom National Health Service health authorities   professional (delivering intervention) ‐ 12   practice ‐ 75   unit of analysis error (randomized by health authority, analyzed by practice)
Interventions	targeted towards HEALTH PROFESSIONALS   pharmacist performed educational outreach visits on two of four practice guideline topics ((1) use of aspirin as antiplatelet therapy, (2) use of ACEIs in heart failure, (3) use of NSAIDs in pain due to osteoarthritis, (4) choice of antidepressant for depression) vs non‐targeted guideline control   length of the intervention ‐ not clear   number of interventions ‐ 2; 12 months pre‐intervention and 12 months post‐intervention periods
Outcomes	PROCESS   # of patients treated in accordance with guideline recommendations
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	health authorities allocated in pairs through a central random process
Blinding?   All outcomes	Low risk	objective measures
Follow‐up of professionals?	Low risk	6 out of 69 practices refused follow‐up
Follow‐up of patients?	Low risk	not applicable
Baseline measurement?	Unclear risk	baseline adherence to recommendations not reported for intervention vs control groups
Reliable outcome measures?	Low risk	objective outcome measure
Protection against contamination?	Low risk	randomized by health authority

Methods	RCT (randomized by practice) ‐ before and after pragmatic randomized controlled trial
Participants	outpatient general practice clinics in a single health authority district, Enlgand   professional (delivering intervention) ‐ 1   practice ‐79 (38 intervention, 38 control)   no unit of analysis error (general practices the unit of analysis)
Interventions	targeted towards HEALTH PROFESSIONALS   pharmacist conducted educational outreach visits to promote undertaking of H. pylori eradication using mailed consensus guidelines vs mailed consensus guidelines alone   length of the intervention ‐ not clear   number of interventions ‐ 1 between 12 months pre‐intervention and 12 months post‐intervention periods
Outcomes	PROCESS   increase in omeprazole prescribing in accordance with consensus guidelines   increase in metronidazole prescribing in accordance with consensus guidelines
Notes	21 practices from randomized 38 received intervention
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	"computer generated random number list"
Blinding?   All outcomes	Unclear risk	Objective outcome assessed
Follow‐up of professionals?	High risk	3 of 19 intervention practices (that allowed a visit) allowed an audit
Follow‐up of patients?	Low risk	Not applicable
Baseline measurement?	Unclear risk	Not explicitly described
Reliable outcome measures?	Low risk	Objective outcome assessed
Protection against contamination?	Low risk	General practices unit of randomization

Methods	RCT (randomized by health professional)
Participants	outpatient clinics in Australia   patients with antibiotic prescriptions   physicians (receiving intervention) ‐ 112 (56 intervention, 56 control)   health professional (delivering intervention) ‐ not clear   practice ‐ unclear (multiple practices in Osborne Division of General Practice in the Perth Western Australia)   no unit of analysis error
Interventions	targeted towards HEALTH PROFESSIONALS   pharmacist provided academic detailing, consisting of in‐person visit and paper chart, to physicians on best practice guidelines for antibiotic use vs no intervention   length of the intervention ‐ 10 to 15 min   number of interventions ‐ 1 during 3 months
Outcomes	PROCESS Number of prescriptions for all antibiotics   PATIENT none measured
Notes	Guidelines for antibiotic prescribing were developed by an expert panel and were in line with published Australian therapeutic guidelines for antibiotics
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Not explicitly described
Blinding?   All outcomes	Low risk	Objective outcome assessed
Follow‐up of professionals?	Low risk	Outcome data were available for all 112 physicians included in the study
Follow‐up of patients?	Low risk	Not applicable
Baseline measurement?	Low risk	"had similar demographic profile"
Reliable outcome measures?	Low risk	Objective outcome assessed
Protection against contamination?	Unclear risk	Communication between physicians was possible

Outcomes Table: 2009 Review
Study	Primary outcomes	Pre‐intervention   (intervention vs control group)	Post‐intervention   (intervention vs control group)	Change due   to intervention (intervention vs   control group)	Result interval   (ΔI ‐ ΔC)	Significance	Notes
Barbanel 2003	PATIENT   Asthma symptom score ‐ North of England asthma scale	26.3+/‐4.8 vs 27.8+/‐3.7	20.3+/‐4.2 vs 28.1+/‐3.6	‐6.0 vs 0.3	7	p‐value < 0.001   95% CI 4.40‐9.50
Bogden 1997	PATIENT 1. Total cholesterol‐‐Men (mg/dl) 2. Total cholesterol‐Women (mg/dl)	n/a	n/a	1. ‐57 vs ‐26 2. ‐37 vs ‐9	1. 31 2. 28	1. not significant 2. p<0.05
Bond 2000	PROCESS   # items prescribed (median quartile) PATIENT   1. Death rate (% patients)   2. Adverse drug reactions (% patients)	PROCESS   not reported PATIENT   1. n/a   2. n/a	PROCESS   2 vs 3 PATIENT   1. 3.6 vs 3.8   2. 8.3 vs 6.7	PROCESS   n/a PATIENT   1. n/a   2. n/a	PROCESS   n/a PATIENT   1. n/a   2. n/a	PROCESS   p‐value 0.0001 PATIENT   1. p‐value NS*   2. p‐value 0.259	*exact value not reported
Borenstein 2003	PATIENT   1. Systolic blood pressure (mmHg)   2. Diastolic blood pressure (mmHg)   3. % pts achieving blood pressure control	1. 162 vs 156   2. 92 vs 90   3. n/a	1. 140 vs 145   2. 85 vs 82   3. 60 vs 43 %	1. ‐22 vs ‐11   2. ‐7 vs ‐8   3. n/a	1. 11 †   2. 1 †   3. 17%†	1. p‐value <0.01   2. p‐value <0.01   3. p‐value 0.02	†calculated from reported data
Brook 2003a	PATIENT   SCL ‐ general anxiety subscale*	3.1 vs 2.8	1.8 vs 1.8	‐1.3 vs ‐1.0†	0.3†	p‐value 0.4	*intention‐   to‐treat analysis, group mean imputation (GMI)   †calculated from reported data
Capoccia 2004	PATIENT   SCL‐20 score at 12 mo (mean(SD))	1.83(0.10) vs 1.75(0.10)	0.75 vs 0.6*	‐1.08 vs ‐1.15†	‐0.07†	p‐value 0.92	*data estimated from graph   †calculated from reported data
Choe 2005	PATIENT   Hemoglobin A1c (%) (mean(SD))	10.1(1.8) vs. 10.2(1.7)*	8.0(1.4) vs. 9.3(2.1)	‐2.1(2.5) vs ‐0.9(2)	1.2†	p‐value 0.03	†calculated from reported data   *SIG lower in control group at baseline (p‐value 0.046)
Clifford 2005	PATIENT   Hemoglobin A1c (%)	7.5 vs. 7.1	7 vs. 7.1†	‐0.5 vs 0	0.5†	p‐value 0.002*	†calculated from reported data
Cody 1998	PATIENT OUTCOME Quality of Life SF36	n/a	n/a	n/a	n/a	change not statistically significant
Finley 2003	PATIENT   1. BIDS score (mean(SD))   2. % pts with ≥50% reduction in BIDS score   3. % pts with remission (BIDS score < 9)   4. % patients with reduction in WSDS score	1. 18.7(5.8) vs 18.3(5.8)   2. n/a   3. n/a   4. n/a	1. 12.1 vs 9.4†   2. 40.7 vs 54.1   3. 55.6 vs 58.3   4. 56 vs 67	1. ‐6.6(7.3) vs ‐8.9(8.3)   2. n/a   3. n/a   4. n/a	1. ‐2.3†   2. n/a   3. n/a   4. n/a	1. p‐value 0.23   2. p‐value 0.27   3. p‐value 0.36   4. p‐value 0.36	†calculated from reported data
Gattis 1999	PATIENT   All cause mortality and heart failure events at 6 mo (# of events)	n/a	4 vs 16 (OR 0.22)	n/a		p‐value 0.005   (OR 95% CI 0.06, 0.63)
Gonzalez‐Martin 2003	PATIENT   1. PAQLQ score (mean(SD)):   a) emotions domain   b) activities domain   c) symptoms domain 2. Spirometry testing (mean(SD)):   a) FVC   b) FEV1	1.   a) 5.2(0.4) vs 5.2(0.4)‡   b) 3.8(0.3) vs 4.0(0.3)‡   c) 4.1(0.5) vs 4.6(0.4)‡ 2.   a) 3.08(0.97) vs 2.66(0.19)   b) 2.41(0.76) vs 2.34(0.22)	1.   a) 6.5 vs 5.3   b) 6 vs 4.1   c) 6 vs 4.8 2.   a) 3.13(1.14) vs 2.85(0.29)   b) 2.48(0.89) vs 2.51(0.27)	1.   a) 1.3 vs 0.1   b) 2.2 vs 0.1   c) 1.9 vs 0.2 2.   a) 0.07 vs 0.19†   b) 0.07 vs 0.17†	1.   a) 1.2   b) 2.1   c) 1.7 2.   a) ‐0.12†   b) ‐0.1†	1.   a) p‐value < 0.001   b) p‐value < 0.001   c) p‐value < 0.002 2.   a) p‐value NS*   b) p‐value NS*	*exact p‐value not provided   †calculated from reported data   ‡ data extrapolated from a graph
Goodyer 1995	PATIENT OUTCOMES     Exercise test (1. distance in 6 min and 2. distance till breathless)	1. 138 vs 145 2. 85 vs 91	1. 159 vs 123 2. 111 vs 71	1. 21 vs ‐22 † 2. 26 vs ‐20†	1. 43† 2. 46†	p<0.001 (unclear for which outcome)	†calculated from reported data
Hanlon 1996	PATIENT   SF 36 (Health Related Quality of Life) SF‐36 domains: physical functioning, social functioning, physical role function, emotional role function, mental health, energy, pain, general health perception	1. n/a	1. n/a	1. n/a	1. n/a	1. p=0.99 (NS all domains)
Jaber 1996	PATIENT   1. Fasting plasma glucose (mmol/L) 2. Glycated hemoglobin (%) 3.Quality of life	1. 11.1 (4.0) vs 12.7 (4.7) 2. 11.5 (2.9) vs 12.2 (3.5) 3. n/a	1. 8.5 (2.3) vs 11 (3.9) 2. 9.2 (2.1) vs 12.1 (3.7) 3. n/a	1. ‐2.6 vs ‐1.8 2. ‐2.2 vs ‐0.1 3. n/a	1. 0.8† 2. 2.1† 3. n/a	1. p<0.05 (in final fasting plasma glucose) 2. p<0.05 (mean ablsolute change in glycated hemoglobin) 3. no stat sig differences*	*p‐value not reported †calculated from reported data
Jackson 2004	PATIENT   1. % pts with therapeutic INR   2. median INR   3. total bleeding up to 90 days after discharge (% pts)	1. 42 vs 45   2. 2.0 vs 2.2   3. n/a	1. 67 vs 41   2. 2.4 vs 2.1   3. 15 vs 36	1. 25 vs ‐4   2. 0.4 vs ‐0.1   3. n/a	1. 29   2. 0.5   3. n/a	1. p‐value 0.01*   2. p‐value 0.84*   3. p‐value 0.009	*p‐value for I vs C at day 8
Jameson 1995	PROCESS OUTCOMES   1. # of drugs   2. doses of drugs/day	1. 5.6 vs 5.7 2. 9.5 vs 9.9	1. 5.0 vs 6.2 2. 7.9 vs 10.5	1. ‐0.6 vs 0.5 2. ‐1.6 vs 0.6	1. 1.1 2. 2.15	1. p=0.004 2. p=0.007
Malone 2001	PATIENT   HRQOL change over 12 mo ‐ SF‐36 domains: physical functioning, role physical, bodily pain, general health perceptions, vitality, social functioning, role emotional, mental health, change in health	n/a	n/a	n/a	n/a	*p‐value < 0.05	bodily pain and change in health domains were significantly higher in control group at baseline
Mehos 2000	PATIENT 1. Systolic blood pressure (mmHg) (mean(SD))   2. Diastolic blood pressure (mmHg) (mean(SD))   3. Mean arterial blood pressure (mmHg) (mean(SD))	1. 157.9(16.4) vs 153.9(14.6)   2. 91.1(10.8) vs 89.6(9.8)   3. 113.4(8.0) vs 111.0(6.4)	1. 140.8 vs 146.9   2. 80.6 vs 85.8   3. 100.7 vs 106.1	1. ‐17.1 vs ‐7   2. ‐10.5 vs ‐3.8   3. ‐12.7 vs ‐4.9	1. 10.1 †   2. 6.7 †   3. 7.8 †	1.p‐value 0.069   2. p‐value 0.022   3. p‐value 0.01	†calculated from reported data
Meredith 2002	PROCESS ‐ improvement in prescribing   1. any medication use (% pts)   2. therapeutic duplication (% pts)   3. cardiovascular medication use (% pts)   4. psychotropic medication use (% pts)   5. NSAID use (% pts)	1. n/a   2. n/a   3. n/a   4. n/a   5. n/a	1. n/a   2. n/a   3. n/a   4. n/a   5. n/a	1. 50 vs 38   2. 70.8 vs 23.5   3. 55 vs 17.6   4. 40.3 vs 31.6   5. 42.2 vs 52.1	1. 12   2. 47.3   3. 37.4   4. 8.8   5. ‐10	1. p‐value 0.05, 95% CI 0‐24   2. p‐value 0.003, 95% CI 20.2‐74.5   3. p‐value 0.02, 95% CI 9‐65.7   4. p‐value > 0.2, 95% CI ‐9.7‐27.4   5. p‐value > 0.2, 95% CI ‐30.4‐10.5	calculated from reported values
Odegard 2005	PATIENT   1. HbA1C at 6 mo (end of intervention) (%) (mean(SD))   2. HbA1C at 12 mo (6 mo usual care follow‐up) (%)(mean(SD))	1. 10.2(0.8) vs 10.6(1.4)   2. 10.2(0.8) vs 10.6(1.4)	1. 8.7 vs 8.8‡   2. 8.2 vs 8.3‡	1. ‐1.5 vs ‐1.8†   2. ‐2 vs ‐2.3†	1. 0.3†   2. 0.3†	p‐value 0.61	†calculated from reported data   ‡ data extrapolated from a graph
Okamoto 2001	PATIENT Note‐authors did not define units for all outcomes; based on other data reported in study, it is assumed that data is reported as (mean(SD)). 1. Systolic blood pressure (mmHg)   2. Diastolic blood pressure (mmHg)   3. QOL ‐ SF‐36 domains: physical functioning, role‐physical*, bodily pain, general health, vitality, social functioning, role‐emotional, mental health	1. 144.23(18.4) vs 142.91(18)   2. 82.79(11.2) vs 82.13(11.4)   3. n/a	1. 135.10(15.3) vs 141.66(17.90)   2. 77.65(11.2) vs 80.67(10.2)   3. n/a	1. ‐9.13(17.1) vs ‐1.32(15.7)   2. ‐5.14(9.2) vs ‐1.46(10.1)   3. n/a	1. 7.81 †   2. 3.68 †   3. n/a	1. p‐value < 0.001   2. p‐value < 0.001   3. *p‐value SIG	†calculated from reported data
Park 1996	PATIENT OUTCOMES   1. Systolic Blood Pressure (mmHg) 2. Diastolic Blood Pressure (mmHg)   3. Health Status Questionnaire (HSQ)	1. 155 (21.1) vs 147.9 (18.6) 2. 87.9 (9.9) vs 83.3 (8.5) 3. n/a	1. 143.2 (11.5) vs 148.6 (20.1) 2. 83.2 (8) vs 83.7 (10.9) 3. n/a	1. ‐11.8 vs 0.7 2. ‐4.7 vs 0.4 3. n/a	1. 12.5 2. 5.1 3. n/a	1. no p‐value reported 2. no p‐value reported 3. p<0.05 only in energy/fatigue category (no sig in other 7 categories)
Paulos 2005	PATIENT   1. Total cholesterol (mg/dL)   2. Triglycerides (mg/dL)   3. % pts with decrease in total cholesterol   4. % pts with decrease in triglycerides	1. 205.1(44.7) vs 203.2(40.6)   2. 190.7(88.7) vs 163.6(116.4)   3. n/a   4. n/a	1. 178(31.1) vs 199.1(37.6)   2. 140.3(47.6) vs 193.2(108.0)   3. 72.8 vs 33.3   4. 77.3 vs 27.8	1. ‐27.1(41.1) vs ‐1.4(37.2)   2. ‐50.5(80.3) vs 29.6(118.5)   3. n/a   4. n/a	1. 25.7   2. 80.0   3. n/a   4. n/a	1. p‐value 0.0266*   2. p‐value 0.0169*   3. p‐value n/a   4. p‐value n/a	*p‐value for change in intervention group over study period
Peterson 2004	PATIENT   cholesterol levels ‐ mmol/L (mean(SD))	4.8(0.70) vs 4.8(0.9)	4.4(0.6) vs 4.6(0.8)	‐0.4 VS ‐0.2†	0.2†	p‐value 0.24*	*p‐value for I vs C at follow‐up   †calculated from reported data
Rickles 2005	PATIENT   ≥50% improvement in BDI‐II score (# (%) pts)	n/a	21 (75) vs 21 (65.6)	n/a	n/a	p‐value NS*	*exact value not reported
Sadik 2005	PATIENT   1. mean distance walked in two min test @ 12 mo (meters)   2. QOL ‐ SF‐36 domains: physical functioning, role‐physical, bodily pain, general health, vitality, social functioning, role‐emotional, mental health	1. 124 vs 120.8   2. n/a	1. 140.2 vs 117.2   2. n/a	1. 16.2 vs ‐3.6 †   2. n/a	1. 19.8 †   2. n/a	1. p‐value 0.001‡   2. *p‐value SIG	†calculated from reported data   ‡p‐value for I vs C at 12 months
Sarkadi 2004	PATIENT   1. Hemoglobin A1c at 12 months (end of study) (%)   2. Hemoglobin A1c at 24 months (follow‐up) (%)	1. 6.5 vs. 6.5†   2. 6.5 vs. 6.5†	1. 6.3 vs. 6.4†   2. 6.2 vs. 6.6†	1. ‐0.2 vs. ‐0.1†   2. ‐0.3 vs. 0.15†	1. 0.1†   2. 0.4†	1. p‐value NS*‡   2. p‐value 0.023‡	*actual value not provided   †values extrapolated from graph   ‡p‐value for I vs C at the end of study period
Schneider 1982	PATIENT   1. Systolic blood pressure (mmHg)   2. Diastolic blood pressure (mmHg)	1. 169 vs 162   2. 98 vs 97	1. 139 vs 153   2. 89 vs 92	1. ‐20 vs ‐9 †   2. ‐9 vs ‐5 †	1. 11 †   2. 4 †	1. p‐value <0.05*   2. p‐value <0.05*	*p‐value for I vs C at 12 mo   †calculated from reported data
Solomon 1998	PATIENT   1. Systolic Blood Pressure ‐ 1st measurement (mmHg) (mean(SD))   2. Systolic Blood Pressure ‐ 2nd measurement (mmHg) (mean(SD))   3. Diastolic Blood Pressure ‐ 1st measurement (mmHg) (mean(SD))   4. Diastolic Blood Pressure ‐ 2nd measurement (mmHg) (mean(SD))   5. BORG Scale ‐ pre‐challenge score   6. BORG Scale ‐ post‐challenge score	1. 146.7(16.8) vs 146.2(17.0)   2. 144.4(17.2) vs 146.4(16.3)   3. 84.6(13.2) vs 87.0(10.9)   4. 85.0(13.0) vs 86.0(10.4)   5. 2.63(2.33) vs 2.03(1.71)   6. 0.7(0.092) vs 0.68(0.94)	1. 138.5(13.9) vs 144.9(21.3)   2. 138.2(12.9) vs 144.0(20.1)   3. 80.2(9.6) vs 83.2(11.5)   4. 80.6(8.7) vs 83.3(11.5)   5. 2.24(2.46) vs 2.44(2.26)   6. 0.87(1.10) vs 1.24(1.48)	1. ‐8.2 vs ‐1.3   2. ‐6.2 vs ‐2.4   3. ‐4.4 vs ‐3.8   4. ‐4.4 vs ‐2.7   5. ‐0.39 vs 0.41†   6. 0.17 vs 0.56†	1. 6.9 †   2. 3.8 †   3. 0.6 †   4. 1.7 †   5. 0.8†   6. 0.39†	1. p‐value 0.044‡   2. p‐value NS   3. p‐value NS   4. p‐value NS   5. p‐value NS   6. p‐value NS**	†calculated from reported data   ‡p‐value for I vs C at the end of intervetion   p‐value for I vs C at end of intervention   * actual p‐values not reported
Sookaneknun 2004	PATIENT   1. Systolic blood pressure (mmHg) (mean(SD))   2. Diastolic blood pressure (mmHg) (mean(SD))   3. % patients controlled for systolic blood pressure and diastolic blood pressure	1. 144.76(19.69) vs 142.41(19.81)   2. 85.72(13.56) vs   85.86(12.94)   3. 22.9 vs 17.9	1. 121.47(14.90) vs   124.77(17.97)   2.71.55(10.80) vs   74.23(11.87)   3. 66.1 vs 57.3	1. ‐23.29(19.10)   vs ‐18.64(17.67)   2. ‐14.18(11.20) vs ‐11.73(10.08)   3. 43.2 vs 39.4	1. 4.65 †   2. 2.45 †   3. 3.8%	1. p‐value <0.001   2. p‐value <0.001   3. p‐value 0.061	†calculated from reported data
Taylor 2003	PROCESS   1. # of medications at 12 mo (mean(SD))   2. #of inappropriate prescriptions   for all MAI domains PATIENT   1. # (%) pts at goal for hypertension   2. # (%) pts at goal for diabetes   3. # (%) pts at goal for dyslipidemia   4. # (%) pts at goal for anticoagulation   5. QOL ‐ SF‐36 domains: physical functioning,   social functioning, physical role function, emotional role function, mental health, energy, pain, general health perception	PROCESS   1. 6.3(2.2) vs 5.7(1.7)   2. 831 vs 895 PATIENT   1. 3 (12.5) vs 9 (31)   2. 3 (23.1) vs 9 (56.3)   3. 2 (10.5) vs 3 (15.8)   4. 1 (25) vs 3 (50)   5. n/a	PROCESS   1. 4.7(2.0) vs 6.2(2.0)   2. 264 vs 978 PATIENT   1. 22 (91.7) vs 8 (27.6)   2. 13 (100) vs 5 (26.7)   3. 14 (77.8) vs 1 (5.9)   4. 4 (100) vs 1 (16.7)   5. n/a	PROCESS   1. ‐1.6 vs 0.5†   2. ‐567 vs 83† PATIENT   1. 19 (79.2) vs ‐1 (‐3.4)†   2. 10 (76.9) vs ‐4 (‐29.6)†   3. 12 (67.3) vs ‐2 (‐9.9)†   4. 3 (75) vs ‐2 (‐33.3)†   5.n/a	PROCESS   1. 2.1†   2. 650† PATIENT   1. 20 (82.6)†   2. 14 (106.5)†   3. 14 (77.2)†   4. 5 (108.3)†   5. n/a	PROCESS   1. p‐value 0.002*   2. not reported PATIENT   1. p‐value 0.001*   2. p‐value 0.001*   3. p‐value 0.001*   4. p‐value 0.048*   5. p‐value NS (all domains)	†calculated from reported data   *p‐value for I vs C   at 12 mo
Tsuyuki 2002	HEALTH PROFESSIONAL   % patients reaching a composite outcome†	n/a	57 vs. 31%   (odds ratio 3.0)	n/a	n/a	p‐value <0.001 95% CI 2.2‐4.1	†composite score of complete cholesterol panel done by PCP OR new prescription for cholesterol lowering medication OR increase in dosage of cholesterol lowering medication
Van Veldhuizen 1995	PATIENT OUTCOMES   Mean Blood glucose levels (mg/dl) (Group II=1st; Group III=2nd; Control=3rd)	164† vs 168† vs 189†	140† vs 152† vs 158†	‐24† vs ‐16† vs ‐31†	Gp II vs C = 7* GpIII vs C =15*	Pts in intervention groups achieved lower average weekly blood glucose values than control group (p<0.05)	†values extrapolated from graph *calculated
Weinberger 2002	PATIENT (PC vs PFM vs UC)*   1. unadjusted PERFs at 12 mo (% predicted) (mean(SD))   2. overall HRQOL with asthma, COPD	1. 63.8(21.6) vs 61.2(22) vs 60.4(22)   2. n/a	1. 65.5(19.5) vs 64.2(21.5) vs 61.6(22.6)   2. n/a	1. ‐1.7 vs ‐3 vs ‐1.2†   2. n/a	1. 1.3 (PC vs PFM), 0.5 (PC vs UC)†   2. n/a	1. p‐value 0.006 across all groups, 0.28 for PC vs PFM at 12 mo   2. p‐value NS	*PC = pharmaceutical care program   PFM = peak flow meter monitoring program   UC = usual care program   †calculated from reported data

Methods	open‐label RCT (randomized by patient)
Participants	home‐based follow‐up of patients discharged from Royal Hobart acute care teaching hospital in southern Tasmania, Australia   patients ‐ 128 (60 intervention, 68 control)   health professional (delivering intervention) ‐ 1   practice ‐ 1   no unit of analysis error
Interventions	targeted towards PATIENTS   pharmacist conducted home‐visit to test INR and educate patients regarding anticoagulant therapy using printed educational materials   targeted towards HEALTH PROFESSIONALS   pharmacist informed physicians regarding patients' INR, recommended dosage adjustments and implemented therapy changes vs usual care   length of the intervention ‐ 24 min   number of interventions ‐ 4 during 90 days
Outcomes	PATIENT   therapeutic INR as defined by ACCP on day 8 after‐discharge   total, major, and minor bleeding complications within 90 days of discharge
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	Patients were home‐based; allocation was likely adequately concealed
Blinding?   All outcomes	Low risk	Objective outcomes assessed
Follow‐up of professionals?	Low risk	Not applicable
Follow‐up of patients?	Low risk	127 of the 131 patients completed the study
Baseline measurement?	High risk	"The two groups were well matched with regard to baseline demographics. There was a significantly higher incidence of previous myocardial infarction in the intervention group compared to the control group"
Reliable outcome measures?	Low risk	Objective outcomes assessed
Protection against contamination?	Low risk	Patients were home‐based, contamination unlikely

Methods	RCT (randomized by patient) ‐ randomized, controlled, unblinded, mixed experimental design
Participants	community pharmacies within a large managed care organization in Wisconsin, United States   patients presenting with new antidepressant prescriptions   patients ‐ 63 (31 intervention, 32 control)   health professional (delivering intervention) ‐ 14   practice ‐ 8   no unit of analysis error
Interventions	targeted towards PATIENTS   pharmacist provided monthly telephone‐based education on antidepressant use and goal of therapy and monitoring of adverse effects and adherence vs usual care   length of the intervention ‐ 19, 12, and 11 min for first, second, and third phone call, respectively   number of interventions ‐ 3 during 3 months
Outcomes	PATIENT   greater than 50% improvement in depression symptoms measured with Beck Depression Inventory‐II (BDI‐II) instrument
Notes	‐past use of psychiatric medications was different between groups at baseline   ‐study was powered to detect compliance outcomes only
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	assignment sealed in an envelope; envelope not reported as "opaque"
Blinding?   All outcomes	Low risk	Objective outcome assessed
Follow‐up of professionals?	Low risk	not applicable
Follow‐up of patients?	Low risk	60 out of 63 patients completed the study
Baseline measurement?	Low risk	Intervention group was more likely to have a history of psychotropic medication use
Reliable outcome measures?	Low risk	Objective outcome assessed
Protection against contamination?	High risk	Patients were randomized. There was the possibility of communication between subjects within the same household or family

Methods	RCT (randomized by clinic)
Participants	outpatient primary care clinic that is part of a managed care organization in the United States (Washington)   patients receiving NSAIDS and salicylates   patients ‐ not clear   health professionals ‐ 2 pharmacists, 17 physicians   practices ‐ 2   no unit of analysis error
Interventions	targeted towards PATIENTS (minimal) AND HEALTH PROFESSIONALS (primary intervention)     pharmacist in family practice clinic provided educational/drug monitoring services to physicians and counseling to patients vs usual care   length of intervention ‐ 6 months   product related
Outcomes	PROCESS   number of prescriptions   drug ingredient costs   operating cost of clinical pharmacy program   PATIENT   not measured
Notes	This was one of few studies that included an assessment of the cost of the intervention
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Allocation procedure not described explicitly
Blinding?   All outcomes	Low risk	Objective outcome assessed
Follow‐up of professionals?	Low risk	Data were included for all 17 physicians
Follow‐up of patients?	Low risk	Not applicable
Baseline measurement?	Low risk	"There were no significant differences among" demographic characteristics
Reliable outcome measures?	Low risk	Objective outcome assessed
Protection against contamination?	Unclear risk	Intervention part of broader education campaign affecting cases and controls

Methods	RCT (randomized by practice)
Participants	outpatient practices in Avalon Peninsula, Newfoundland, Canada   all physicians in the region were invited to participate in the study; intervention focused on physicians treating CHF (congestive heart failure) patients   patients ‐ not clear   health professional (delivering intervention) ‐ not clear   practice ‐ 72   unit of analysis error (no correction)
Interventions	targeted towards HEALTH PROFESSIONALS   pharmacist provided academic detailing to physicians on use and dosage of ACEIs and ARBs for prevention and management of CHF using Canadian consensus guidelines on management of CHF vs no intervention   length of the intervention ‐ not clear   number of interventions ‐ 1 during 3 months
Outcomes	PROCESS   utilization of ACEIs in patient receiving digoxin and furosemide for CHF   targeted daily dose of ACEIs
Notes	Number of patients allocated to each group not reported   Statistical analysis of results did not account for clustering and unit of analysis error
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Not explicitly described
Blinding?   All outcomes	Low risk	Objective outcome assessed
Follow‐up of professionals?	Low risk	"First and second interviews were successfully completed with 25 and 23 physicians, respectively, in the control group and 32 and 31 physicians in the intervention group"
Follow‐up of patients?	Low risk	Not Applicable
Baseline measurement?	Unclear risk	Not explicitly described
Reliable outcome measures?	Low risk	Objective outcome assessed
Protection against contamination?	Low risk	"All members of a group practice were randomized to the same study group to minimize cross‐contamination"

Methods	RCT (by patient)
Participants	outpatient diabetes center in United States (Indiana)   diabetic patients   patients ‐ 41   health professionals ‐ 1   practices ‐ 1   no unit of analysis error
Interventions	targeted towards PATIENTS   1‐on‐1 interaction with pharmacist plus follow‐up vs group education from pharmacist vs standard education.     length of intervention ‐ 2 months   product related
Outcomes	PROCESS   not measured   PATIENT   knowledge of diabetes   perception/attitudes towards diabetes, therapy, pharmacists   blood glucose levels
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Not explicitly described
Blinding?   All outcomes	Low risk	Objective outcome assessed
Follow‐up of professionals?	Low risk	Not applicable
Follow‐up of patients?	High risk	Based on figures, 32 out of 41 patients completed the study (for blood glucose values)
Baseline measurement?	Unclear risk	Not explicitly described, but there appears to be differences in the numbers presented (statistical significance not reported); data not reflected in a table or figure.
Reliable outcome measures?	Low risk	Objective outcome assessed
Protection against contamination?	Unclear risk	Not explicitly described; subjects recruited from the same Regional Diabetes Center

Methods	cluster RCT (randomized by practice)
Participants	university affiliated teaching clinic   outpatient general practices in Avon, United Kingdom   health professional (delivering intervention) ‐ 3   practice ‐ 20   no unit of analysis error
Interventions	targeted towards HEALTH PROFESSIONALS   pharmacist performed educational outreach visits to promote mailed practice guideline on NSAID use vs mailed practice guideline alone vs no intervention   length of the intervention ‐ less than 10 minutes   number of interventions ‐ 2 during 12 months
Outcomes	PROCESS   change in volume of prescribing for ibuprofen, diclofenac and naproxen as % of total NSAID prescribing
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	Not explicitly described
Blinding?   All outcomes	Low risk	Objective outcome assessed
Follow‐up of professionals?	Low risk	Data are presented for all practices included in the study
Follow‐up of patients?	Low risk	not applicable
Baseline measurement?	Low risk	"There were only slight differences in NSAID prescribing between the three groups at baseline"
Reliable outcome measures?	Low risk	Objective outcome assessed
Protection against contamination?	Low risk	Randomised by practice

Methods	RCT (by practice ‐ 36 drugstores divided into 12 clusters of 3 geographically proximal drugstores)
Participants	CVS community pharmacies in Indianapolis, United States   patients with COPD and asthma   patients ‐ 1113   asthma ‐ 660 (pharmaceutical care program 262, peak flow monitoring control 233, usual care control 165)   COPD ‐ 453 (pharmaceutical care program 185, peak flow monitoring control 130, usual care control 138)   health professional (delivering intervention) ‐ not clear   practice ‐ 36   unit of analysis error (randomized by practice, analyzed by patient)
Interventions	targeted towards PATIENTS   pharmaceutical care (patients received peak flow monitor + instructions for use, written educational materials, and monthly telephone calls from research personnel to collect PEFR results; pharmacist assessed PEFR results and other relevant medical information (medications, refill history, emergency department visits and hospitalizations) and implemented pharmaceutical care activities) vs peak flow monitoring (patients received peak flow monitors and instructions for use and monthly telephone calls from research personnel to collect peak flow PEFR results (results were not seen by the pharmacist)) vs usual care (patients did not receive peak flow monitors but received monthly follow‐up phone calls from research personnel)   length of the intervention ‐ not clear   number of interventions ‐ mean 19.4 in asthma, 22.4 in COPD patients during 12 months
Outcomes	PATIENT   peak flow rate (PEFR) (combined for asthma and COPD patients) at 12 months   HRQOL (health‐related quality of life) for asthma patients at 12 months   HRQOL for COPD patients at 12 months
Notes	No statistical analysis done to adjust for unit of analysis error
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	"Used a randomized number chart"
Blinding?   All outcomes	Low risk	Objective outcome assessed
Follow‐up of professionals?	Low risk	Not applicable
Follow‐up of patients?	Low risk	"Completed interviews with 947 patients (85.1%) at 6 months and 898 patients (80.7%) at 12 months. Completion rates did not differ significantly by disease or study group."
Baseline measurement?	High risk	DONE for asthma patients, NOT DONE for COPD patients; "study groups were comparable at baseline, except for race (asthma/COPD) and PEFR (COPD only)...we controlled for race in all analyses and for baseline PEFR among COPD patients only."
Reliable outcome measures?	Low risk	Objective outcome assessed
Protection against contamination?	Low risk	Sites were randomized in clusters.

Study	Reason for exclusion
Abramowitz 1982	Hospital‐based intervention
Bogden 1998	Not RCT
Bolas 2004	Hospital‐based intervention
Bouvy 2003	Subjective primary outcome
Bozovich 2000	Not RCT
Brook 2003b	Subjective primary outcomes
Brook 2005	Subjective primary outcome
Bucci 2003	Intervention not performed solely by pharmacist ‐ patients followed by a multidisciplinary team including a pharmacist
Charrois 2004	Design paper ‐ no results reported. A search was performed for the completed manuscript. The completed manuscript was found by searching for studies that referenced the original design paper. The completed manuscript is not indexed in PubMed; it is in the Studies awaiting classification section and will be coded in Phase II.
Chisholm 2001	Relevant and interpretable data are not presented
Cowper 1998	Economic data only reported as primary outcomes
Davidson 2000	Not RCT (experimental gp‐not prospectively assigned)
de Maat 2004	Sequential study design
De Tullio 1987a	Subjective primary outcome
De Tullio 1987b	Not RCT
Erickson 1997	Not RCT
Fischer 2002	Resource utilization and costs only reported as primary outcomes
Fornos 2004	Design paper ‐ no results reported. Manuscript for completed study is in Studies awaiting classification section
Forstrom 1990	Costs only reported as primary outcomes
Garnett 1981	Patient self‐reported data
Gourley 1998	Subgroup analysis of Solomon 1998; focused on humanistic outcomes only
Helling 1979	RCT   Subjective outcome measure ‐ patient satisfaction only was measured
Holland 2005	Subjective primary outcomes
Ibrahim 1990	No control group
Jameson 2001	Subjective primary outcome ‐ adverse effect and symptom score self reported by patients
Johnson 1998	Patient self‐reported data
Jones 1991	Costs only reported as primary outcomes
Karki 1988	Hospital‐based intervention
Knoell 1998	Not RCT
Krska 2001	Subjective primary outcome ‐ pharmaceutical care issues identified by pharmacist
Lai 1998	Not RCT
Law 2003	Cost‐savings only reported as primary outcome
Lim 2004	Relevant and interpretable data not reported for primary outcome
Malone 2000	Economic analysis only reported as primary outcome
Malone 2003	Not RCT
McKenney 1973	Not RCT
Murray 2004	Intervention not primarily performed by a pharmacist
Murray 2004a	Design paper ‐ no results presented; Manuscript for completed study is in Studies awaiting classification section.
Peterson 1995	Not RCT
Peterson 1996	Not RCT
Peterson 1997	Not RCT
Powers 1983	Not RCT
Raisch 1990	All three comparison groups included a pharmacist intervention, therefore there was no relevant control group for this review
Rathbun 2005	Subjective primary outcome (medication compliance)
Rodgers 1999	Not RCT
Rogers 1998	Cost‐savings only reported as primary outcomes
Schaffner 1983	Not RCT
Sczupak 1977	Results not interpretable
Sellors 2001	Results not interpretable ‐ no baseline data
Sellors 2003	Results not interpretable ‐ no baseline data
Shaw 2000	Hospital‐based intervention
Shibley 1997	Patients served as their own control
Sidel 1990	Subjective primary outcome
Simpson 2001	Subgroup analysis
Simpson 2004	Subgroup analysis
Smith 1999	Cost‐savings only reported as primary outcome
Soumerai 1986	Cost‐savings only reported as primary outcome
Steele 1989	Cost‐savings only reported as primary outcome
Tamai 1987	Not RCT
Varma 1999	Hospital‐based intervention
Vrijens 2006	Subjective primary outcome (medication compliance)
Wandless 1981	Subjective primary outcome (medication compliance)
Yamada 2005	Not RCT
Zermansky 2001	Primary outcome not relevant ‐ number of changes to prescriptions over one year

PERMALINK

Effect of outpatient pharmacists' non‐dispensing roles on patient outcomes and prescribing patterns

Nancy Nkansah

Olga Mostovetsky

Christine Yu

Tami Chheng

Johnny Beney

Christine M Bond

Lisa Bero

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Plain language summary

Background

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Search methods for identification of studies

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Assessment of heterogeneity

Subgroup analysis and investigation of heterogeneity

Results

Description of studies

Characteristics of interventions

Characteristics of health professionals delivering the intervention

Target population

Risk of bias in included studies

1.

2.

Effects of interventions

Comparison 1. Pharmacist services targeted at patients versus services delivered by other health professionals

1.1. Analysis.

Comparison 2. Pharmacist services targeted at patients versus the delivery of no comparable service

2.1. Analysis.

2.2. Analysis.

2.3. Analysis.

2.4. Analysis.

Comparison 3: Pharmacist services targeted at health professionals versus services delivered by other health professionals

Comparison 4: Pharmacist services targeted at health professionals versus delivery of no comparable service

3.1. Analysis.

Discussion

Authors' conclusions

Implications for practice.

Implications for research.

Feedback

Pharmacist interventions

Summary

Reply

Contributors

What's new

History

Notes

Acknowledgements

Appendices

Appendix 1. EPOC search strategy (Phase I of review)

Appendix 2. MEDLINE/EMBASE search strategy (Phase II of review)

Appendix 3. EMBASE search strategy (Phase II of review)

Data and analyses

Comparison 1. Pharmacist services targeted at patients versus services delivered by other health professionals.

Comparison 2. Pharmacist services targeted at patients versus the delivery of no comparable service.

Comparison 3. Pharmacist services targeted at health professionals versus the delivery of no comparable service.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Barbanel 2003.

Bogden 1997.

Bond 2000.

Borenstein 2003.