Guidelines in professions allied to medicine

Abstract

Background

Clinical practice guidelines aim to reduce inappropriate variations in practice and to promote the delivery of evidence‐based health care.

Objectives

To identify and assess the effects of studies of the introduction of clinical practice guidelines in nursing (including health visiting), midwifery and other professions allied to medicine.

Search methods

We searched the Cochrane Effective Practice and Organisation of Care Group specialised register, MEDLINE (1975 to 1996), EMBASE, Cinahl and Sigle to 1996, the NHS Economic Evaluations Database (1994 to 1996), DHSS‐Data (1983 to 1996), the Database of Abstracts of Reviews of Effectiveness (1994 to 1996) and reference lists of articles. We also hand searched the journal Quality in Health Care, made personal contact with content experts and contacted libraries identified by an expert panel.

Selection criteria

Randomised trials, controlled before‐and‐after studies and interrupted time series analyses of the introduction of interventions comparing 1. Clinical guidelines plus dissemination and/or implementation strategies versus no guidelines; 2. Guidelines plus dissemination and/or implementation strategies versus guidelines plus alternative dissemination and/or implementation strategies; and 3. (post hoc) Guidelines used by professions allied to medicine versus standard physician care. The participants were nurses, midwives and other professions allied to medicine.

Data collection and analysis

Two reviewers independently extracted data and assessed study quality.

Main results

Eighteen studies were included involving more than 467 health care professionals. The reporting of study methods was inadequate for all studies. In all but one study, nurses were the targeted professional group; one study was aimed solely at dieticians. The various behaviours targeted included the management of hypertension, low back pain and hyperlipidaemia. Nine studies were identified for comparison 1. Three out of five studies observed improvements in at least some processes of care and six out of eight studies observed improvements in outcomes of care. Only one study included a formal economic evaluation, with equivocal findings. Three studies were identified for comparison 2 but it was difficult to draw firm conclusions because of poor methods. Six studies were identified for comparison 3 (post hoc). These studies generally supported the hypothesis that there was no difference between care given by nurses using clinical guidelines and standard physician care.

Authors' conclusions

There is some evidence that guideline‐driven care is effective in changing the process and outcome of care provided by professions allied to medicine. However, caution is needed in generalising findings to other professions and settings.

Plain language summary

Professions allied to medicine, such as nurses, midwives and dietiticians, can effectively incorporate clinical guidelines to improve patient care

The issuing of clinical guidelines to nurses, midwives, dieticians and other health‐care professionals allied to medicine may reduce variations in practice and improve patient care. This review found that, despite limited research, there is some evidence that guidelines can improve care and that professional roles can be substituted effectively, for instance a nurse can perform the function of a physician in certain circumstances. Such interventions offer the possibility of reduced costs but further research is needed in all areas of this topic.

Background

Clinical guidelines are 'systematically developed statements to assist practitioner decisions about appropriate health care for specific clinical circumstances' (IOM 1990). Guidelines are used to reduce inappropriate variations in practice and to promote the delivery of high‐quality evidence‐based health care (CRAG 1993). However, most of the emphasis in the literature has been on guidelines solely or principally in the medical domain; for example, Grimshaw and Russell (Grimshaw 1993) undertook a systematic review of the effects of introducing clinical guidelines for medical staff. A preliminary search of the literature in relation to clinical practice guidelines in nursing, midwifery and professions allied to medicine has revealed systematically developed guidelines in community pharmacy (Bond 1995) and dietetics (Flanel 1995). Early examples have also been found where medical protocols were implemented by nurses for dysuria, frequent urination and vaginal discharge (Greenfield 1975a) and for low back pain (Greenfield 1975b). In both studies, protocols were evaluated using patient outcome measures, rather than compliance with the protocols by health professionals.

To date, there has been no systematic review of the effectiveness of dissemination and implementation strategies for guidelines targeting nurses, midwives and professions allied to medicine. In the USA, guideline dissemination appears to be via 'massive distribution' to professional and lay groups (Colling 1994), despite evidence that dissemination by publication or direct mailing are the least successful strategies (eg Grol 1992). It cannot be assumed that those strategies evaluated for medicine will transfer to other groups of health care professionals that have a different educational preparation and management structure.

Objectives

1. To identify rigorous evaluations of the introduction of clinical practice guidelines in nursing (including health visiting), midwifery and other professions allied to medicine. Both hospital and community sectors were included.

2. To determine the effectiveness and efficiency of introducing clinical practice guidelines targeting nursing, midwifery and professions allied to medicine to promote improved professional practice and patient outcomes.

Comparison groups were as follows:

Comparison 1: guidelines plus dissemination and/or implementation strategy versus no guidelines;

Comparison 2: guidelines plus dissemination and/or implementation strategy A versus guidelines plus dissemination and/or implementation strategy B;

During the review, a subset of studies was identified which examined the ability of clinical guidelines to enable role substitution (in all cases substituting nurses for doctors). These studies evaluate an organisational intervention and have a different objective to studies evaluating the effectiveness of introducing guidelines on improving patient outcomes. However, as these studies comprised rigorous evaluations of guidelines by nurses, they were included as a post‐hoc subset in our review.

Comparison 3: guidelines (used by professions allied to medicine) versus no guidelines (standard physician care), ie skill substitution studies.

Methods

Criteria for considering studies for this review

Types of studies

We included randomised controlled trials (RCTs), interrupted time series (ITS) studies and controlled before‐and‐after (CBA) studies . (See EDITORIAL INFORMATION under GROUP DETAILS for METHODS USED IN REVIEWS for complete definitions of study designs.)

Types of participants

The following professions were included in the review: nursing, midwifery, health visiting, chiropody, speech and language therapy, physiotherapy, occupational therapy, dietetics, clinical psychology, pharmacy and radiography. For brevity, all professions included in the review are subsequently referred to as 'professions allied to medicine'.

On detailed reading and data extraction of a number of papers which had initially met our inclusion criteria, some problems were identified. These included the inability to separate the effects of guidelines on the behaviour of professionals allied to medicine from the impact on the behaviour of other health professionals. A number of papers were found where guidelines were aimed at a multi‐disciplinary team. We decided that if processes or outcomes for professions allied to medicine were not evaluated separately, 50% or more recipients of the guideline had to be professionals allied to medicine for the paper to be included. Papers that initially appeared to meet our inclusion criteria but which were subsequently excluded for these reasons are recorded in the excluded studies table.

Types of interventions

We adopted the Institute of Medicine's (IOM 1990) definition of clinical guidelines as 'systematically developed statements to assist practitioner decisions about appropriate health care for specific clinical circumstances'. In recognition of the variety of terms used for guidelines, we did not insist that the term 'guideline' was explicitly used, as long as the intervention in question implicitly met the definition. Synonyms for clinical guidelines included 'protocol', 'standard' and 'clinical pathway'. We were interested in all types of guideline, not simply those based on evidence. 
 
 To be eligible for inclusion, a study had to be a rigorous evaluation of the introduction of a clinical guideline on the behaviour of professionals allied to medicine and/or on patient outcomes. We included studies that evaluated an entire or identified component of a guideline aimed at these professions and that specified which part(s) of the guideline should be implemented by the profession(s) in question. We considered not only those studies on guidelines developed by the target profession for their own use, but also those where the guideline was developed and/or disseminated by another professional group but where the targeted behaviour was that of professionals allied to medicine. We included all types of dissemination and implementation strategies. (See EDITORIAL INFORMATION under GROUP DETAILS for METHODS USED IN REVIEWS for details of the EPOC intervention taxonomy.)

'Skill substitution' studies, evaluating the ability of guidelines to facilitate role substitution, were included.

Studies evaluating discharge planning or case management protocols were included only if they included a list of actions for patient groups, rather than for individual patients.

Types of outcome measures

Studies were included if they reported objective assessments of the process or outcome of care provided by professionals allied to medicine, for example on patient assessment or on the provision of appropriate interventions. Studies that examined staff satisfaction and knowledge were included only if they also examined the behaviour of professionals allied to medicine or patient outcome. Studies measuring economic outcomes (ie anything about the costs of guideline implementation, especially costs versus benefits) were included.

Search methods for identification of studies

Search strategies were developed using the methodological component of the EPOC search strategy (see EDITORIAL INFORMATION under GROUP DETAILS for METHODS USED IN REVIEWS) and expanding the section on clinical guidelines. Search terms included: 'practice guideline', 'health planning guidelines', 'care map' and 'professional standard'.

A cut‐off date of 1975 was imposed because MEDLINE (the only database used in this review going back prior to 1975) does not contain abstracts before this year. Because of the nature of the review, searching on title only would have resulted in retrieving potentially hundreds of papers not meeting our inclusion criteria; the likely yield was judged by the research team not to merit the resources and expenditure required by the task.

Electronic searches

Searches of electronic databases including: MEDLINE (1975‐1996); CINAHL (1982‐1996); PsycLIT (1975‐1996); EmBASE (1980‐1996); NHS Economic Evaluations Database (1994‐1996); DARE (Database of Abstracts of Reviews of Effectiveness; 1994‐1996); DHSS‐DATA (1983‐1996); SIGLE (1980‐1996); and the National Research Register (1996 version);

Searching other resources

• hand searching of the EPOC register of studies: this involved visiting EPOC editorial base and scanning copies of papers;

• scanning reference lists of papers identified;

• hand searching of the journal Quality in Health Care, identified as likely to contain relevant papers, but not indexed on the electronic databases;

• personal contact with content area experts: a panel of experts was asked to identify sources of information on guidelines in their profession (eg information on databases, key journals and conferences);

• requests for assistance were posted on internet discussion lists including evidence‐based‐health@mailbase.ac.uk, nursenet@listserv and physio@mailbase.ac.uk. We asked for unpublished studies which may be relevant and for details of other useful contacts. No included studies were identified from this source;

• contacting libraries of appropriate professional bodies: we made contact with libraries identified by an expert panel (see acknowledgements) and requested information on library holdings related to guideline evaluation. No library produced information specific enough to inform our review.

Data collection and analysis

Each relevant study was assessed for inclusion in the review independently by two reviewers. Data extraction was undertaken by two reviewers working independently. Reviewers did not always work in the same pairs. Discrepancies were discussed and resolved by the data extractors; those that could not be resolved easily were referred to the remaining authors.

Where separate papers reported on different aspects of the same study (for example, one paper describing the methodology and one the findings, or one exploring the impact of guidelines on professional behaviour and another the effect on patient outcome), they were treated as one study and only one data extraction template was completed. Where separate papers reported on a pilot study and a main study of the impact of the same guideline, we extracted both separately only if they reported on a separate sample of patients; otherwise, we extracted data from the paper reporting the main study only.

Time constraints precluded following up data missing from papers by contacting authors; this will be undertaken prior to the next revision of the review.

As the heterogeneity of clinical areas, design of studies, source and format of interventions, outcomes measured and participating health professionals was substantial, a meta‐analysis would have had little practical meaning. We therefore opted to report the effects on the process and outcome of care in the same way that they were reported in the original papers. Where reported analyses were inadequate, for example where the original authors did not report the statistical significance of their findings, or where a more powerful analysis of time‐series data was possible (eg test for trend), we re‐analysed the data presented in the source papers.

Results

Description of studies

We identified 18 studies which met our inclusion criteria (see included studies table). A further 13 were initially believed to meet the criteria but were rejected at the data extraction phase for various reasons, the most common being fundamental methodological flaws or the failure to separate the impact of the guideline on professions allied to medicine from that on other professions (see excluded studies table).

All but one of the identified studies evaluated the introduction of guidelines targeting nurses. In five of these studies, physicians were also targeted (Herman 1994; McDonald 1980; Jewell 1988; Zeler 1992; French 1989). The study by Shaffer and Wexler (Shaffer 1995) evaluated the introduction of a multi‐disciplinary guideline targeting a lipid clinic team comprising nurses, pharmacists, psychologists and dieticians. The study by Franz and colleagues (Franz 1995) evaluated the introduction of guidelines for dieticians.

Targeted behaviours were nutrition therapy for non‐insulin‐dependent diabetics (Franz 1995), infection control (French 1989), management of labour (Frigoletto 1995), management of dysuria, frequency and vaginal discharge (Greenfield 1975a), management of low back pain (Greenfield 1975b), management of headache (Greenfield 1976), vaccinations (Herman 1994), management of hypertension (Jewell 1988), referral for X‐ray (Klassen 1993), management of intravenous therapy (Larson 1984), test ordering, referral, recording and changes in treatment (McDonald 1980), reduction in fall rates (Mitchell 1996), discharge planning (Naylor 1990; Naylor 1994), catheter practices (Seto 1991). lipid lowering (Shaffer 1995), identification of battered women (Tilden 1987) and management of post‐operative bleeding after cardiac surgery (Zeler 1992).

Study settings were in‐patient care (nine studies), out‐patient/ambulatory care (seven studies) and emergency room (one study).

Twelve studies were based in the United States of America, two in Australia, two in Hong Kong, one in Canada and one in the United Kingdom.

Risk of bias in included studies

The methodological characteristics of each study are shown in the included studies table. We identified 13 RCTs. In three studies the unit of allocation was the health professional or provider unit (Herman 1994; McDonald 1980; Seto 1991). Two trials (Herman 1994; Seto 1991) had a unit of analysis error: the unit of randomisation was the provider or ward, but the unit of analysis was the patient. Such unit of analysis errors artificially increase the precision of the study and may lead to misleadingly significant findings.

There were ten RCTs where the unit of allocation was the patient (Franz 1995; Frigoletto 1995; Greenfield 1975a; Greenfield 1975b; Greenfield 1976; Jewell 1988; Klassen 1993; Naylor 1990, Naylor 1994; Zeler 1992). There were major differences in methodological quality; the proportion of quality criteria met ranged from five out of seven (Naylor 1990) to one out of seven (Naylor 1994; Jewell 1988). Just over half of the studies (7/13) had good follow up, blindly assessed outcomes and reliable measurement of at least some key outcomes. A minority of studies reported method of randomisation (done in 4/13).

We identified two CBA studies where a second site acted as the control (Larson 1984; Shaffer 1995). In the Larson study (Larson 1984), five units in a single hospital participated; two received the intervention and three served as controls. No data were presented about the comparability of intervention and control sites; baseline measurement showed significant differences on one parameter likely to influence post‐intervention differences (accuracy of recording of last intravenous infusion). At least some outcomes were measured reliably but none were assessed blindly. Protection against contamination appeared adequate. The analysis is almost certainly inappropriate: there is no explanation as to how the dependent variable was calculated and clustering of the data has probably not been taken into account.

In the Shaffer study (Shaffer 1995), two clinics (one intervention, one control) within the same medical centre participated. No data were presented on the similarity of the clinics; patients selected for the study from each clinic were matched according to age and cholesterol level. Outcomes measured were reliable and were assessed blindly. As the lipid clinic team worked only in the intervention clinic, it is likely there was successful protection against contamination between clinics.

Three studies were ITS designs (French 1989; Mitchell 1996; Tilden 1987); quality criteria met ranged from four out of nine (French 1989) to one out of nine (Mitchell 1996). In the French study (French 1989) the intervention appeared independent of other changes during the study period; in the Mitchell study (Mitchell 1996) no details were provided as to whether this was or was not the case. In the Tilden study (Tilden 1987), where the targeted behaviour was identification of battered women, the authors state that media coverage and an initiative by hospital administration (aimed at all staff) to increase the level of detail in patient documentation may have intensified the effect of the intervention. No ITS study had sufficient data points before and after the intervention; only one study (French 1989) did a formal test for trend. All ITS studies were prone to detection bias. In two studies (French 1989; Tilden 1987) data collection methods were identical before and after the intervention. However in the Mitchell study (Mitchell 1996), pre‐intervention data on fall rates were collected from retrospective analysis of fall incident forms, while prospective data were collected by adding a 'falls reporting document' to the incident form. As this appeared to be part of the intervention, this directly influenced data collection. Whether the intervention directly influenced data collection in the two other ITS studies is not clear. Outcomes were not assessed blindly in one study (Mitchell 1996); in the remaining two studies it is not clear whether this was the case. In only one study (Tilden 1987) was information provided about the completeness of the data set: all adult female trauma patients were included. No ITS study measured outcomes reliably.

Effects of interventions

For comparisons 1 and 2 improved processes and outcomes of care were anticipated if the guidelines were effective.

Comparison 1: guidelines plus dissemination and/or implementation strategy versus no guidelines.

Nine studies compared the introduction of guidelines with a 'no guideline' control (Franz 1995; Frigoletto 1995; Naylor 1990; Naylor 1994; Larson 1984; Shaffer 1995; Tilden 1987; French 1989; Mitchell 1996); these studies aimed to improve the quality of care delivered by the target profession(s). Of these, four were RCTs (Franz 1995; Frigoletto 1995; Naylor 1990; Naylor 1994), three were ITS studies (Tilden 1987; French 1989; Mitchell 1996) and two were CBA studies (Larson 1984; Shaffer 1995).

Five studies assessed the effect of guidelines in terms of the process of care. Significant changes in at least some of the assessed processes of care were identified in three studies (Frigoletto 1995; Shaffer 1995; Tilden 1987). For example, Frigoletto's study (Frigoletto 1995) of the management of labour by nurse‐midwives found significant differences in practices relating to the management of labour between the active management protocol and usual care groups. In the active management group, the frequency of vaginal examinations was significantly higher (mean interval 1.6 hours versus 2.5 hours for standard care group), significantly more women received oxytocin to hasten labour (70% versus 56%) and significantly fewer women requested and received epidural anaesthesia (54% versus 64%). The study by Larson observed improvements in process of care, however these are difficult to interpret given the problems with the analysis (see Methodological Quality).

Eight studies measured outcomes of care; statistically significant differences in favour of the intervention group were found for at least some outcomes of care in six studies (Franz 1995; Frigoletto 1995; Naylor 1990; Naylor 1994; Shaffer 1995; French 1989). However, the findings from several of these studies were equivocal, with no significant change being noted in other outcome variables assessed. For some studies, the observed change in what appeared to be the primary outcome measure was non‐significant (Franz 1995; Frigoletto 1995; Naylor 1994; Mitchell 1996).

Only one study (Franz 1995) included a formal economic evaluation. This study assessed the effect of medical nutrition therapy for non‐insulin‐dependent diabetics administered according to practice guidelines with medical nutrition therapy administered in line with standard care. Findings were suggestive of a cost‐efficiency advantage in favour of the control group when one outcome measure (glycated haemoglobin) was considered, but no difference when an alternative outcome measure (fasting plasma glucose level) was examined. The studies described by Naylor (Naylor 1990; Naylor 1994) and Larson and Hargiss (Larson 1984) suggest that direct health care costs are decreased by guideline‐driven care, but these authors do not provide sufficient data to ascertain whether these cost savings are offset by the cost of guideline dissemination and implementation.

Comparison 2: guidelines plus dissemination and/or implementation strategy A versus guidelines plus dissemination and/or implementation strategy B.

Three studies (Herman 1994; McDonald 1980; Seto 1991) explicitly compared two or more dissemination or implementation strategies. All were RCTs. Within these studies, the interventions were heterogeneous, precluding combining the findings from two or more studies. Seto et al (Seto 1991) in their evaluation of opinion leaders and in‐service lectures as methods of disseminating a guideline on urinary catheter care to nurses, compared three groups: one received guidelines, lectures and opinion leaders (A), one received guidelines and opinion leaders (B) and one received guidelines and lectures (C). Performance improved in all three groups, however improvements were greater in groups A and B. There was a unit of analysis error and it is not possible to determine whether these improvements were statistically significant.

Herman et al (Herman 1994) found that a 'nurse protocol' intervention led to positive changes in the process of care (offer of influenza and pneumococcal vaccine). There were three comparison groups: all received guidelines, educational materials and lectures, group two received additional provider and patient education at each clinic visit and group three received an additional protocol (which comprised a flowchart record incorporating the recommendations of the guideline and a redefinition of the nurse's role). They (Herman 1994) found that an active dissemination strategy (involving provider education) had a minor effect on professional behaviour in the group receiving only guidelines, educational materials and lectures. The impact of the guideline on outcomes of care was also assessed: a difference between intervention and control groups was found, but in the opposite direction to that which was expected: the percentage of patients declining immunisation increased in the intervention period to 13.2% in the prevention team group; 5.8% in the patient education group and 5.9% in the control group. However, there was a unit of analysis error in this study, hence it is not possible to say whether these findings are statistically significant.

McDonald (McDonald 1980) compared the effect of patient‐specific computerised prompts at the time of consultation, computerised prompts plus bibliographic citations and standard care on the use and follow up of medications. However, the results for nurses are presented with the two intervention groups combined so it is not possible to compare the two methods of implementation. With the computerised prompts and computerised prompts plus citations groups combined, the intervention did not appear to lead to a significant change in nurses' behaviour; this finding was contrary to that for physicians in the same study. It is possible that this negative finding was due to a lack of power to detect a difference (n=five nurses).

Comparison 3: skill‐substitution studies.

In the remaining six studies (Greenfield 1975a; Greenfield 1975b; Greenfield 1976; Jewell 1988; Klassen 1993; Zeler 1992) the performances of nurses operating in accordance with a guideline were compared with 'standard care', generally provided by a physician. All six studies were RCTs. The aim was skill substitution, generally in the interests of greater efficiency.

In these studies, a finding of no significant difference or of a positive effect in favour of the profession allied to medicine would suggest that the intervention was effective. The findings from all six studies that examined the processes and all five that examined outcomes of care generally support the hypothesis of no difference. Whilst appropriate sample size calculations were not reported in any study, with the exception of one study (Jewell 1988), sample sizes were probably large enough to detect clinically significant differences.

Discussion

Findings from the 18 studies identified provide some evidence that guideline‐driven care can be effective in changing the process and outcome of care provided by professions allied to medicine. Significant improvements in the outcome of care were found in seven out of nine studies comparing the introduction of guidelines to a no guideline control. The three studies comparing two or more dissemination and implementation strategies were compromised by a small sample size (McDonald 1980) or unit of analysis errors (Herman 1994; Seto 1991). As a result it is difficult to draw firm conclusions from these studies. The findings from the six studies that examined the ability of clinical guidelines to enable role substitution generally support the effectiveness of this intervention.

Some caution is needed in interpreting the findings from our review and in generalising these findings to other professions and settings due to the poor methods observed in most studies (see Methodological quality) and the study settings. The majority of the studies were carried out in a single site, focused on only one profession and often involved very small numbers of health professionals (in several cases, only one nurse). The possibility that the findings were specific to the setting or to the individual(s) cannot be ruled out. In some cases, the health professionals who were studied were selected specifically because of their enthusiasm or expertise (Franz 1995); the guidelines may have been less effective in the hands of less committed or less highly qualified or experienced staff. As many of the authors themselves recognise (Frigoletto 1995; Greenfield 1976; Larson 1984; Shaffer 1995), the majority of the studies were threatened by the Hawthorne effect, as health professionals were aware that their performance was being assessed. The timing of the assessment of process and outcome may also have had an impact: the studies we identified did not, in general, provide sufficient information to assess whether any observed changes in performance were sustained after the initial impetus of guideline dissemination.

Authors' conclusions

Implications for practice.

Target professions

All but one study was aimed at nurses. No studies evaluating the effectiveness of guidelines were found for several professions included in the review (eg physiotherapy; occupational therapy). These conclusions, therefore, relate mainly to nursing.

Guideline dissemination and implementation

Findings from this review provide some evidence that guideline‐driven care can be effective in changing the process and outcome of care provided by professions allied to medicine. There is currently insufficient evidence to make firm conclusions about the effectiveness of the different dissemination and implementation strategies. We recommend that future dissemination and implementation activities should be informed by theoretical perspectives (Grol 1992) and by evidence of those strategies shown to be most effective in relation to changing doctors' behaviour.

Implications for research.

Research into guideline development, dissemination and implementation strategies

The studies included in this review provide only limited evidence of the relative merits of alternative dissemination and implementation strategies. More research is required into the relative effectiveness of these strategies in relation to professions allied to medicine. Studies are needed in which two or more strategies for guideline development, dissemination or implementation are examined in the same population.

Study design and reporting

In studying professional behavioural change, the strongest evidence is provided by RCTs in which the units of randomisation and allocation are the provider. This design should be considered as the 'gold standard' in designing future studies of the effectiveness of guidelines. Where it is not feasible to randomise by provider, care should be taken to minimise the risk of contamination. Potential confounding factors and sources of bias should be identified explicitly. Ideally, these studies will include a qualitative component to provide answers about not only whether, but how and why an intervention works.

Reports of study design and results should include sufficient detail to allow a 'quality assessment' of the study. The unit of randomisation should be the same as the unit of analysis to avoid artificially increasing the precision of the study and misleadingly small confidence intervals. The primary 'outcome measures' (ie the processes or outcomes of care expected to change as a result of the intervention) should be clearly identified. The size of difference considered to be 'clinically significant' should be stated a priori and the study should be powered accordingly. Reports of the study should include details of sample size calculations. Where possible, an intention‐to‐treat analysis should be carried out; in any case, the number of units of randomisation lost to follow up should be reported (separately for intervention and control groups).

What's new

Date	Event	Description
12 November 2008	Amended	Minor changes

History

Protocol first published: Issue 3, 1996
 Review first published: Issue 1, 1999

Date	Event	Description
27 October 2008	Amended	Converted to new review format.
24 November 1998	New citation required and conclusions have changed	Substantive amendment

Data and analyses

Comparison 1. Guidelines plus dissemination and/or implementation strategy versus no guidelines.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Results			Other data	No numeric data

1.1. Analysis.

Comparison 1 Guidelines plus dissemination and/or implementation strategy versus no guidelines, Outcome 1 Results.

Results
Study	Outcome Measurement	Results Comparisons	Effect on Process	Effect on Patient	Notes
Franz 1995	6 weeks, 3 months, 6 months post intervention (Quality of Life 6 months)	1. Guidelines 2. Standard care	Changes in therapy: 18% (17/94) vs 11% (9/85) p=0.203	Fasting plasma glucose: 9.1 mmol/L (from 10.2) vs 9.2 mmol/L (from 9.8) n.s. Glycated haemoglobin: 7.4 mmol/L (from 8.3) vs 7.6 mmol/L (from 8.3) n.s. Cholesterol: 5.4 mmol/L (from 5.6) vs 5.5 mmol/L (from 5.7) Statistical significance not stated LDL‐Cholesterol: 3.25 mmol/L (from 3.34) vs 3.51 mmol/L (from 3.50) Statistical significance not stated HDL‐Cholesterol: 1.07 mmol/L (from 1.09) vs 1.07 mmol/L (from 1.14) Statistical significance not stated Triglycerides: 2.40 mmol/L (from 2.57) vs 2.62 mmol/L (from 2.54) Statistical significance not stated Weight: 92.4 kg (from 93.8kg) vs 92.0 kg (from 93.7kg) Statistical significance not stated Body Mass Index: 32.4 kg/m2 (from 32.9 kg/m2) vs 32.4 kg/m2 (from 33.0 kg/m2) Statistical significance not stated Quality of life: Resisting temptation: 69 vs 79 p<0.05 Positive social: 62 vs 76 p<0.01 Exercise: 48 vs 64 p<0.01 Neg. emotions: 62vs 76 p<0.01 Per patient cost per unit of change in outcome (fasting plasma glucose level): $5.84 vs $5.75
French 1989	6 monthly pre and post intervention 4 data points before and 3 after for hospital acquired urinary tract infection	1. Guidelines 2. No intervention		Hospital acquired urinary tract infection (HAUTI) rates (crude): Pre‐intervention: 3.1%, 3.8% 2.8%, 3.0% Post‐intervention: 2.1%, 1.0%, 2.0% p=0.0008 HAUTI rates (adjusted): Pre‐intervention (95% CI): 3.2% (2.0‐4.3), 3.1% (2.3‐4.0), 2.6% (1.8‐3.4), 2.9% (2.1‐3.7) Post‐intervention (95% CI): 2.3% (1.6‐3.0), 1.9% (1.1‐2.6), 2.3% (1.5‐3.1) p=0.032 (significance of difference pre versus post)	Findings on a further outcome measure, hospital acquired infection, are not presented as there were insufficient data points (2) before the intervention
Frigoletto 1995	Immediate	1. Protocol (midwives) 2. Standard care	Frequency of caesarean section: 19.5% (197/1009) vs 19.4% (176/906) Relative risk 1.0 95% CI (0.8‐1.2) Mean frequency of vaginal exam (per hour): 1.6 vs 2.5 Artificial rupture of membranes: 61% vs 51% Relative risk 1.2 95% CI (1.1‐1.3) Rupture within 1 hour of admission: 76% vs 15% Relative risk 17.8 95% CI (13.3‐23.8) Administration of oxytocin: 70% vs 56% 1.3 (1.2‐1.4) Administration of epidural anaesthesia: 54% vs 64% 0.8 (0.8‐0.9)	Rate of caesarean section: 19.5% (197/1009) vs 19.4 (176/906) for all patients 10.9% vs 11.5% for protocol eligible group Relative risk 0.9 95% CI (0.4‐1.9) Duration of labour (protocol‐eligible sub‐group): 6.2 vs 8.9 hours Statistical significance not stated Rate of maternal fever (protocol‐eligible sub‐group): 7% vs 11 % p=0.007 Rate of other maternal outcomes: no significant differences in foetal distress, placental abruption, shoulder dystocia, vaginal lacerations Duration of labour > 12 hours (protocol‐eligible sub‐group): 9% vs 26% (p<0.0001) Infant outcomes: no significant differences in frequency of jaundice, seizures, treatment for sepsis, resuscitation at birth or admission to neonatal ICU
Larson 1984	4 weeks pre intervention; 2 month gap; 6 weeks post intervention	1 Guidelines; outreach visits 2 Standard practice	Percentage of items documented correctly: intervention group; increase from 57% to 79%. Control group; increase from 52.5% to 68.5%; p<0.001	Frequency of bacterial colonisation of infusion devices: intervention group; increase from 12.7% to 19.4% p=0.25; Control group; decrease from 7.1% to 5.9% p=0.90 Patient comfort with IV insertion: not reported in appropriate format
Mitchell 1996	6 months pre intervention, 6 months post intervention	1. Protocol 2. No intervention		Fall rates reduced from 7.8 per 1000 bed days pre intervention to 4.4 post intervention p=0.06	Large monthly variation in rate suggests some covariates (eg frailty scores per month) are needed to explain variation
Naylor 1990	Post discharge 12 weeks post intervention	1. Protocol 2. Standard practice		Mean length of stay (days): 8.2 + 4.9 vs 9.05 + 7.66 n.s. Post‐discharge infection rates: 33.3% (6/20) vs 50% (10/20) Relative risk 0.43 95% CI (0.11‐1.57) Post‐discharge rehospitalisation rates: 16.7% (4 rehospitalisations, 3 patients) vs 64.7% (12 rehospitalisations, 11 patients) p<0.05
Naylor 1994	2 weeks post discharge 12 weeks post discharge 6‐12 weeks post discharge	1. Protocol 2. Standard practice		Results reported for medical & surgical diagnostic‐related group MDRG & SDRG separately. Mean length of initial hospitalisation (days): MDRG: 7.4 + 3.8 vs 7.5 + 5.2 difference (95% CI) 0.1 (‐1.6 to 1.4) SDRG: 15.8 + 9.4 vs 14.8 + 8.3 difference (95% CI) 1.0 (‐2.0 to 4.0) Mean days to rehospitalisation: 45.6 vs 31. p=0.12 Rate of rehospitalisation: MDRG: (15%) 11 vs (16%) 11 difference (95% CI) ‐1% (‐8% to 12%) SDRG: (10%) 7 vs (7%) 5 difference (95% CI) 3% (‐7% to 13%) Mean duration of rehospitalisation: MDRG: 94 vs 100 difference (95% CI) ‐6 (‐83 to 71) SDRG: 52 vs 26 difference (95% CI) 26 (‐8 to 60)
Shaffer 1995	12 months, intervention; 18 months, control	1. Guidelines (nurse, clinical pharmacist, dietitian, clinical psychologist) 2. No intervention (doctors)	Percentage of patients for whom a new drug was prescribed: 85% (51/60) vs 18% (11/60) p<0.001 Number of patients in whom an existing drug was stopped: 22 patients vs 3 patients. p<0.001 Percentage of patients receiving expert dietary counselling: 100% (60/60) vs 42% (25/60) p<0.001	Total cholesterol at follow‐up: 6.08 mmol/L (from 7.65) vs 6.72 mmol/L (from 7.50) p=0.004 LDL‐Cholesterol at follow‐up: 3.78 mmol/L (from 5.20) vs 4.60 mmol/L (from 5.17) p<0.001 HDL‐Cholesterol at follow‐up: ‐1.03 mmol/L (from 1.11) vs ‐ 1.14 mmol/L (from 1.19) n.s. Triglycerides at follow‐up: ‐3.39 mmol/L (from 4.13) vs ‐3.56 mmol/L (from 4.45) n.s. Body Mass Index: 28.3 kg/m2 (from 27.9 kg/m2) vs 28.5 kg/m2 (from 28.8 kg/m2) n.s. Percentage achieving NCEP LDL‐Cholesterol goal (3.36 mmol/L): 44% (19/43) vs 11% (3/28) Relative risk=4.1 (95% CI = 1.4 ‐ 12.7) Percentage on lipid lowering drug: 91% (51/56) vs 35% (21/60) Relative risk=3.75 (95% CI = 2.05 ‐ 6.87)
Tilden 1987	Monthly for 4 months pre and post intervention	1. Guidelines; staff education programme; local consensus processes 2. No intervention	Increase in rate of identification of positive battering from 7 cases (9.72% of total) to 17 cases (22.97% of total) p=0.03 Odds ratio: likelihood of identification 2.74 times greater post‐intervention

Comparison 2. Guidelines plus dissemination and/or implementation. Strategy A versus strategy B.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Results			Other data	No numeric data

2.1. Analysis.

Comparison 2 Guidelines plus dissemination and/or implementation. Strategy A versus strategy B, Outcome 1 Results.

Results
Study	Outcome Measurement	Results Comparisons	Effect on Process	Effect on Patient	Notes
Herman 1994	3 months post intervention	1. Guidelines; educational materials; lectures 2. As 1; provider and patient education provided at each clinic visit 3. As 1; plus protocol	Offer of influenza vaccination: 68% (166/243 ‐ nurse protocol group) vs 50% (122/242 ‐ patient education group) vs 48% (129/271 ‐ control group) Offer of pneumococcal vaccination: 28.3% (89/314) vs 6.5% (19/292) and 5.4% (16/295)	Influenza vaccination: 13.2% (nurse protocol) vs 5.8% (patient education) vs 5.9% (control) Pneumococcal vaccination 6.7% vs 1.4% vs 2.0%	Possible unit of analysis error; levels of significance not reported
McDonald 1980	Processes measured during one 5‐week control period and two 5‐week intervention periods	1. Computer reminders 2. Computer reminders + bibliographic citations 3. Standard practice	Individual provider response rate = number of appropriate responses/total number possible. Average Nurse Response Rate to prompts 30% (I), 25% (C) (n=5 nurses)	Responses to computer prompts measured during 1 5‐week control period and 2 5‐week intervention periods
Seto 1991	2 weeks pre intervention, 5 weeks post intervention	1. Guidelines; local opinion leaders; lecture 2. Guidelines; local opinion leaders 3. Guidelines; lecture	Change in reported practice, 2 weeks before vs 5 weeks after intervention: Opinion leader + lecture (A) 5.63 (50%); Opinion leader only (B) 4.96 (35%); Lecture only (C) 3.29 (38%) Mean change in Group C lower than Groups A and B Percentage of correct catheter practices, 1 month after intervention: Opinion leader + lecture (A): 50% Opinion leader only (B): 35% Lecture only (C): 38%		Possible unit of analysis error; levels of significance not reported

Comparison 3. Guidelines (used by professions allied to medicine) versus no guidelines (standard physician care).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Results			Other data	No numeric data

3.1. Analysis.

Comparison 3 Guidelines (used by professions allied to medicine) versus no guidelines (standard physician care), Outcome 1 Results.

Results
Study	Outcome Measurement	Results Comparisons	Effect on Process	Effect on Patient	Notes
Greenfield 1975a	Within one week and 3 weeks post intervention	1. Protocol (nurses) 2. No intervention (doctors)	Concordance between nurse and physician re: medical history: 139/146 95% CI 95.2% (91.6%‐98.7%) Concordance re: results of physical examination: 137/146 95% CI 93.8% (89.8%‐97.7%) Concordance re: therapy and referral: (99%) 144/146 All referrals appropriate; protocol did not fail to recommend referral where there should have been one Laboratory agreed with nurse in (93%) 54/58 urinalyses 95% CI 93.1% (86.5%‐99.6%)	Alleviation/improvement of symtoms: (97%) 66/68 vs (97%) 63/65 p=1.00 Results of treatment with antibiotics (urine samples sterile): (71%) 10/14 vs (53%) 8/15 p=0.45
Greenfield 1975b	Within 5 weeks and at 4 months post intervention	1. Protocol (nurses) 2. No intervention (doctors)	Number of x‐rays ordered: 21 vs 36 p<0.01 Initial/final diagnosis: no significant differences except in initial diagnosis of 'other' (10.8% vs 1%) p<0.01	Intervention (nurse protocol) patients significantly more satisfied than control (physician) p<0.005 Symptom relief: resolved or improved in 71% (144/203) vs 74.6% (144/193) p=0.43 Complications: no patient managed by nurses alone developed serious complications Return with back problems within 3 months: 27% vs 18% p=0.04
Greenfield 1976	Within 5 weeks and at 4 months post intervention	1. Protocol (nurses) 2. No intervention (doctors)	Final diagnosis: diagnosis of muscle headache: 62.1% (126/203) vs 43.5% (84/193) p<0.001 Prescription of medication: no significant differences except for prescription of minor tranquilisers: 6.4% (13/203) vs 15.5% (30/193) p<0.05	Symptom relief: resolved or improved in 73% (162/222) vs 68.5% (135/197) p=0.33 Patient satisfaction: satisfaction with 6 of 9 elements significantly higher in intervention group; no significant difference for remaining 3 elements; p ranged from <0.02 to <0.00001 for these 6 elements Development of serious sequelae: no instances of serious disease and no hospitalisation related to headache in either group
Jewell 1988	In 12 months after entry to study	1. Protocol (nurses) 2. Protocol (doctors)	Percentage of patients with record of: Smoking habit (once): (100%) 15 vs (84%) 16 patients p=0.32 Urine test (once): (93%) 14 vs (21%) 4 patients p<0.00002 Blood pressure (at each visit) (100%) 90 vs (100%) 103 visits p=1.00 Pulse at each visit: (98%) 88 vs (65%) 67 visits p<0.01 Weight (at each visit): (90%) 81 vs (37%) 38 visits p<0.001	Difference in mean systolic BP: 8 mmHg 95% CI (‐8.7 to 24.7) Difference in mean diastolic BP: 0.4 mmHg 95% CI (‐6.2 to +7.0) Percentage achieving treatment objective of diastolic BP <90 mmHg: 67% vs 63% p>0.5 Frequency of attendance: mean attendance rates for BP control: 6.0 vs 5.7 n.s. Knowledge of medication: drug names recalled: 13 vs 20 p=1.00 drug names not recalled: 12 vs 17 Patient reactions: number reporting satisfaction: 15/17 vs 17/19 p=1.00; 2 in nurse protocol group chose to return to doctor care; only 3 patients in both groups did not wish to be looked after by a nurse
Klassen 1993	Processes measured at baseline and during the intervention period (length of this period not specified)	1. Protocol (nurses) 2. No intervention (doctors)	Percentage of patients having X‐rays ordered: 81.9% vs 87.1% p=0.03 Positive findings as percentage of ordered X‐rays: 40.8% vs 42.6% p=0.21 Missed positive findings as percentage of whole group: 3.2% vs 0% p<0.001 Mean time spent in the emergency department: 3.3 hours vs 3.6 hours p<0.001
Zeler 1992	24 hours post‐operatively	1. Guideline (nurses) 2. No intervention (doctors)	Number of patients in whom coagulation tests performed:13% (14/108) vs 14% (20/146) p=0.4 Mean time to coagulation test: 291 minutes vs 454 minutes p=0.3 Volume of packed cells infused: Presented graphically p<0.01 Volume of total blood and blood products infused: Presented graphically p<0.05	Mean postoperative blood loss: 817 ml vs 976 ml p=0.08 Mean activated partial thromboplastin time: 31.7 seconds vs 47.3 seconds p=0.01 Number requiring emergency sternotomy: 4% (4/108) vs 8% (12/146) p=0.32

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Franz 1995.

Methods	RCT‐Pt Randomisation concealment: NOT DONE Follow up: providers: NOT APPLICABLE patients: NOT DONE Blinded assessment: DONE Baseline: NOT CLEAR Reliable outcomes: DONE Protection against contamination: NOT DONE
Participants	Dieticians (number not specified) in US outpatient clinics Patients: 247 with non‐insulin‐dependent diabetes mellitus Proportion of patients followed up: 72% in total Outpatient care Academic/teaching status: NOT CLEAR
Interventions	1. Guidelines 2. Standard care
Outcomes	Process: Changes in therapy Patient: Fasting plasma glucose Glycosylated haemoglobin Cholesterol Low density lipoprotein cholesterol High density lipoprotein cholesterol Triglycerides Weight Body Mass Index Quality of Life Cost per patient
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

French 1989.

Methods	ITS Protection against secular changes: Intervention independent: NOT CLEAR Sufficient data points: NOT DONE Test for trend DONE Protection against detection bias: Data collection: DONE Intervention unlikely to affect data collection: NOT CLEAR Blinded assessment: NOT CLEAR Completeness of data set: NOT CLEAR Reliable outcomes: NOT CLEAR
Participants	Three infection control nurses in one Hong Kong hospital Patients: ranged from 809 to 1260 in 7 prevalence surveys Inpatient care University based/teaching setting
Interventions	1. Guidelines 2. No intervention
Outcomes	Patient: Rates of hospital acquired urinary tract infection
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Frigoletto 1995.

Methods	RCT‐Pt Randomisation concealment: NOT DONE Follow up: providers: NOT APPLICABLE patients: DONE Blinded assessment: DONE Baseline: NOT CLEAR Reliable outcomes: DONE Protection against contamination: DONE
Participants	Two maternity units in one US hospital Patients: 1017 intervention 917 control Protocol eligible sub‐group: 678 intervention 585 control Proportion of patients followed up: 99% in total Inpatient care Academic status/teaching setting: NOT CLEAR
Interventions	1. Protocol (midwives) 2. Standard care
Outcomes	Process: Frequency of caesarean section Mean frequency of vaginal exam (hours) Artificial rupture of membranes Rupture within one hour of admission Administration of oxytocin Administration of epidural anaesthesia Patient: Rate of caesarean section Duration of labour (protocol eligible sub‐group) Rate of maternal fever (protocol eligible sub‐group) Rate of other maternal outcomes Duration of labour > 12 hours (protocol eligible sub‐group) Infant outcomes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

Greenfield 1975a.

Methods	RCT‐Pt (Cross over trial ‐ pt) Randomisation concealment: NOT CLEAR Follow up: providers: NOT APPLICABLE patients: DONE Blinded assessment: DONE Baseline: DONE Reliable outcomes: NOT CLEAR Protection against contamination: NOT CLEAR
Participants	One nurse and 13 doctors in one US clinic Patients: 151 patients randomised; 146 analysed: 76 intervention 70 control Proportion of patients followed up: 99% in total Outpatient care Academic status/teaching setting: NOT CLEAR
Interventions	1. Protocol (nurses) 2. No intervention (doctors)
Outcomes	Process: Concordance between nurse and physician re: medical history (percentage agreement) Concordance between nurse and physician re: results of physical examination (percentage agreement) Concordance between nurse and physician re: therapy and referral (percentage agreement) Evaluation of laboratory work by nurses, and physicians' records Patient: Alleviation/improvement of symptoms Results of treatment with antibiotics (urine samples sterile)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Greenfield 1975b.

Methods	RCT‐Pt Randomisation concealment: NOT CLEAR Follow up: providers: NOT APPLICABLE patients: NOT DONE Blinded assessment: NOT CLEAR Baseline: DONE Reliable outcomes: NOT CLEAR Protection against contamination: NOT CLEAR
Participants	5 nurses and 32 doctors in one US 'walk‐in' clinic Patients: 592 randomised 419 completed study (of remainder, some found not to have low back pain; others lost to follow‐up); 221 Intervention, 197 Control Proportion of patients followed up: 70% in total Outpatient care Academic status/teaching setting: NOT CLEAR
Interventions	1. Protocol (nurses) 2. No intervention (doctors)
Outcomes	Process: Number of X‐rays ordered Initial diagnosis Final diagnosis Patient: Satisfaction Symptom relief Complications Return with back problems within 3 months
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Greenfield 1976.

Methods	RCT‐Pt Randomisation concealment: NOT CLEAR Follow up: providers: NOT APPLICABLE patients: NOT DONE Blinded assessment: DONE Baseline: NOT CLEAR Reliable outcomes: NOT CLEAR Protection against contamination: NOT DONE
Participants	5 nurses and 32 doctors in one US 'walk‐in' clinic Patient ‐ 203 intervention 193 control Proportion of patients followed up: 72.6% in total Outpatient care Academic status/teaching setting: NOT CLEAR
Interventions	1. Protocol (nurses) 2. No intervention (doctors)
Outcomes	Process: Final diagnosis Prescription of medication Patient: Symptom relief Patient satisfaction Development of serious sequelae
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Herman 1994.

Methods	RCT‐Pr Randomisation concealment: NOT CLEAR Follow up: providers: NOT APPLICABLE patients: NOT DONE Blinded assessment: NOT CLEAR Baseline: DONE Reliable outcomes: DONE Protection against contamination: DONE
Participants	Nurses (number not specified) and 66 doctors in three US ambulatory medical practices Provider: 1 practice in each of 3 arms Proportion of patients followed up: 100% of eligible patients for influenza vaccination group 75% for pneumococcal vaccination group Outpatient care University based/teaching setting
Interventions	1. Guidelines; educational materials; lectures 2. As 1.; provider and patient education provided at each clinic visit 3. As 1.; plus protocol
Outcomes	Process: Offer of influenza vaccination Offer of pneumococcal vaccination Patient: Acceptance of both vaccines
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Jewell 1988.

Methods	RCT‐Pt Randomisation concealment: NOT CLEAR Follow up: providers: NOT APPLICABLE patients: DONE Blinded assessment: NOT CLEAR Baseline: NOT CLEAR Reliable outcomes: NOT CLEAR Protection against contamination: NOT CLEAR
Participants	One nurse and one doctor in one UK general practice Patient ‐ 17 intervention 19 control Proportion of patients followed up: 88% intervention 100% control Community based care Non‐teaching setting
Interventions	1. Protocol (nurses) 2. Protocol (doctors)
Outcomes	Process: Number (%) of patients with record of: smoking habit (once) urine test (once) blood pressure (at each visit) pulse (at each visit) weight (at each visit) Patient: Systolic blood pressure Diastolic blood pressure Frequency of attendance Knowledge of medication Patient reactions
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Klassen 1993.

Methods	RCT‐Pt Randomisation concealment: NOT CLEAR Follow up: providers: NOT APPLICABLE patients DONE Blinded assessment: DONE Baseline: DONE Reliable outcomes: DONE Protection against contamination: NOT CLEAR
Participants	21 triage nurses and doctors (number not specified) in one Canadian emergency room Patient ‐ 491 intervention 494 control Proportion of patients followed up: 98.9% Inpatient care Non‐teaching setting
Interventions	1. Protocol ( nurses) 2. No intervention (doctors)
Outcomes	Process: Percentage of patients having radiographs ordered Positive findings as percentage of ordered X‐rays Missed positive findings as percentage of whole group Mean time spent in the emergency department
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Larson 1984.

Methods	CBA Baseline measurement: NOT DONE Baseline characteristics: NOT CLEAR Blinded assessment: NOT DONE Protection against contamination: DONE Reliable outcomes: DONE Follow up: providers: NOT CLEAR patients: NOT CLEAR
Participants	10 nurses (in intervention group) and doctors (number not specified) in 2 intervention and 3 control wards in one US university‐affiliated hospital Patients: 707 (in total) Proportion of patients followed up: NOT CLEAR Inpatient care Academic/teaching setting: NOT CLEAR
Interventions	1. Guidelines; outreach visits 2. Standard practice
Outcomes	Process: Percentage of items documented correctly Patient: Occurrence of IV‐related phlebitis Frequency of bacterial colonisation of infusion devices Patient comfort with IV insertion
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

McDonald 1980.

Methods	RCT‐Pr (Cross‐over trial) Randomisation concealment: NOT CLEAR Follow up: providers: DONE patients: NOT APPLICABLE Blinded assessment: NOT CLEAR Baseline: DONE Reliable outcomes: DONE Protection against contamination: NOT DONE
Participants	5 nurses and 26 doctors in one US medical outpatient clinic 3‐treatment cross‐over design ‐ all providers experienced all conditions in one of 6 possible different treatment sequences Proportion of patients followed up: NOT APPLICABLE Outpatient care Academic status/teaching setting: NOT CLEAR
Interventions	1. Computer reminders 2. Computer reminders + bibliographic citations 3. Standard practice
Outcomes	Process: Appropriate responses to prompts in areas of test ordering, referral, recording, changes to treatment
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Mitchell 1996.

Methods	ITS Protection against secular changes: Intervention independent: NOT CLEAR Sufficient data points: NOT DONE Test for trend NOT DONE Protection against detection bias: Data collection: NOT DONE Intervention unlikely to affect data collection: NOT DONE Blinded assessment: NOT DONE Completeness of data set: NOT CLEAR Reliable outcomes: NOT CLEAR
Participants	Nurses (number not specified) on one medical ward in an Australian acute care hospital Inpatient care University based/teaching setting
Interventions	1. Protocol 2. No intervention
Outcomes	Patient: Fall rate
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Naylor 1990.

Methods	RCT‐Pt Randomisation concealment: DONE Follow up: providers: NOT APPLICABLE patients: DONE Blinded assessment: DONE Baseline: DONE Reliable outcomes: DONE Protection against contamination: NOT DONE
Participants	2 nurses (in intervention group) in one US hospital Patients: 20 intervention 20 control Proportion of patients followed up: 100% Inpatient care Academic status/teaching setting: NOT CLEAR
Interventions	1. Protocol 2. Standard practice
Outcomes	Patient: Mean length of stay (days) Post‐discharge infection rates Post‐discharge rehospitalization rates
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Naylor 1994.

Methods	RCT‐Pt Randomisation concealment: NOT CLEAR Follow up: providers: NOT APPLICABLE patients: NOT DONE Blinded assessment: NOT CLEAR Baseline: NOT CLEAR Reliable outcomes: NOT CLEAR Protection against contamination: NOT DONE
Participants	2 nurses and doctors (number not specified) in one US hospital Patients: 364 patients randomised; 276 analysed: 140 intervention 136 control Proportion of patients followed up: 76% in total Inpatient care Academic status/teaching setting: NOT CLEAR
Interventions	1. Protocol 2. Standard practice
Outcomes	Patient: Mean length of initial hospitalisation (days) Mean days to rehospitalization Rate of rehospitalization Mean duration of rehospitalization (days)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Seto 1991.

Methods	RCT‐Pr Randomisation concealment: NOT CLEAR Follow up: providers: DONE patients: NOT APPLICABLE Blinded assessment: NOT CLEAR Baseline: DONE Reliable outcomes: NOT CLEAR Protection against contamination: DONE
Participants	220 nurses in one Hong Kong hospital Provider ‐ 2 wards in each of 3 arms Proportion of patients followed up: NOT APPLICABLE Inpatient care University based/teaching setting
Interventions	1. Guidelines; local opinion leaders; lecture 2. Guidelines; local opinion leaders 3. Guidelines; lecture
Outcomes	Process: Change in reported practice Percentage of correct catheter practices
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Shaffer 1995.

Methods	CBA Baseline measurement: DONE Baseline characteristics: NOT CLEAR Blinded assessment: DONE Protection against contamination: DONE Reliable outcomes: DONE Follow up: providers: NOT APPLICABLE patients: DONE
Participants	One lipid clinic team (nurse‐led) and one general medicine clinic team in one US medical centre Provider ‐ 1 intervention 1 control Patients: 60 intervention 60 control Proportion of patients followed up: 93% intervention 88% control Outpatient care Academic status/teaching setting: NOT CLEAR
Interventions	1. Guidelines (nurse, clinical pharmacist, dietitian, clinical psychologist) 2. No intervention (doctors)
Outcomes	Process: Percentage of patients for whom a new drug was prescribed Percentage of patients in whom an existing drug was stopped Percentage of patients receiving expert dietary counselling Patient: Total cholesterol at follow‐up LDL cholesterol at follow‐up HDL cholesterol at follow‐up Triglycerides at follow‐up Body Mass Index Percentage achieving NCEP LDL‐cholesterol goal (3.36mmol/L) Percentage on lipid‐lowering drug
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Tilden 1987.

Methods	ITS Protection against secular changes: Intervention independent: NOT DONE Sufficient data points: NOT DONE Test for trend: NOT DONE Protection against detection bias: Data collection: DONE Intervention unlikely to affect data collection: NOT CLEAR Blinded assessment: NOT CLEAR Completeness of data set: DONE Reliable outcomes: NOT CLEAR
Participants	22 nurses in one US hospital emergency department Inpatient care Non‐teaching setting
Interventions	1. Guidelines; staff education programme; local consensus processes 2. No intervention
Outcomes	Process: Percentage of positive cases identified as battered
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Zeler 1992.

Methods	RCT‐Pt Randomisation concealment: NOT DONE Follow up: providers: NOT APPLICABLE patients: NOT DONE Blinded assessment: DONE Baseline: NOT CLEAR Reliable outcomes: DONE Protection against contamination: NOT DONE
Participants	Nurses and doctors (numbers not specified) working in one Australian cardiothoracic intensive care unit Patients: 159 intervention 163 control Proportion of patients followed up: 67% intervention 91% control Inpatient care Academic status/teaching setting: NOT CLEAR
Interventions	1. Guideline (nurses) 2. No intervention (doctors)
Outcomes	Process: Number of patients in whom coagulation tests performed Mean time to coagulation test Volume of packed cells infused Volume of total blood and blood products infused Patient: Mean postoperative blood loss Mean activated partial thromboplastin time Number requiring emergency sternotomy
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Belcher 1990	Less than 50% of recipients of the protocol professions allied to medicine.
Campbell 1991	Behaviour of professionals allied to medicine is measured only in the experimental group.
Cimino 1994	The control site is not comparable on targeted activity; it is not clear whether the professional allied to medicine is the target or the deliverer of the intervention; the study is not aimed at changing the behaviour of professionals allied to medicine.
Fortner 1985	It is difficult to separate the impact of the guidelines from the rest of a complex intervention.
Fullard 1987	Outcomes measured do not relate to the guideline.
Greenfield 1978	Behaviour of professionals allied to medicine cannot be separated from that of 'physician's assistants'.
Hong 1990	Duplicate publication.
Jaber 1996	This study is not about a guideline.
Khan 1990	The guideline is not aimed at professions allied to medicine, but at 'community health workers'.
Landefeld 1995	It is not possible to separate the effect of protocols from that of other interventions.
McDonald 1984	Less than 50% of recipients of the protocol professionals allied to medicine.
Safran 1995	Behaviour of professionals allied to medicine cannot be separated from that of physicians.
Shrake 1994	Behaviour of professionals allied to medicine cannot be separated from that of physicians.
Wilson 1983	The guideline is not aimed at professions allied to medicine, but at 'pamosists'.

Sources of support

Internal sources

• University of Newcastle upon Tyne, UK.

• University of York, UK.

External sources

• NHS R&D Programme on evaluating methods to promote the implementation of R&D, UK.

Declarations of interest

None known.

Edited (no change to conclusions)