Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Meta-Regression Analyses to Explain Statistical Heterogeneity in a Systematic Review of Strategies for Guideline Implementation in Primary Health Care

  • Susanne Unverzagt ,

    Affiliation Institute of Medical Epidemiology, Biostatistics and Informatics, University Halle/Wittenberg, Halle (Saale), Germany

  • Frank Peinemann,

    Affiliation Children's Hospital, University of Cologne, Cologne, Germany

  • Matthias Oemler,

    Affiliation Section of General Practice, Institute of Medical Epidemiology, Biostatistics and Informatics, University Halle/Wittenberg, Halle (Saale), Germany

  • Kristin Braun,

    Affiliation Section of General Practice, Institute of Medical Epidemiology, Biostatistics and Informatics, University Halle/Wittenberg, Halle (Saale), Germany

  • Andreas Klement

    Affiliation Section of General Practice, Institute of Medical Epidemiology, Biostatistics and Informatics, University Halle/Wittenberg, Halle (Saale), Germany

Meta-Regression Analyses to Explain Statistical Heterogeneity in a Systematic Review of Strategies for Guideline Implementation in Primary Health Care

  • Susanne Unverzagt, 
  • Frank Peinemann, 
  • Matthias Oemler, 
  • Kristin Braun, 
  • Andreas Klement


This study is an in-depth-analysis to explain statistical heterogeneity in a systematic review of implementation strategies to improve guideline adherence of primary care physicians in the treatment of patients with cardiovascular diseases. The systematic review included randomized controlled trials from a systematic search in MEDLINE, EMBASE, CENTRAL, conference proceedings and registers of ongoing studies. Implementation strategies were shown to be effective with substantial heterogeneity of treatment effects across all investigated strategies. Primary aim of this study was to explain different effects of eligible trials and to identify methodological and clinical effect modifiers. Random effects meta-regression models were used to simultaneously assess the influence of multimodal implementation strategies and effect modifiers on physician adherence. Effect modifiers included the staff responsible for implementation, level of prevention and definition pf the primary outcome, unit of randomization, duration of follow-up and risk of bias. Six clinical and methodological factors were investigated as potential effect modifiers of the efficacy of different implementation strategies on guideline adherence in primary care practices on the basis of information from 75 eligible trials. Five effect modifiers were able to explain a substantial amount of statistical heterogeneity. Physician adherence was improved by 62% (95% confidence interval (95% CI) 29 to 104%) or 29% (95% CI 5 to 60%) in trials where other non-medical professionals or nurses were included in the implementation process. Improvement of physician adherence was more successful in primary and secondary prevention of cardiovascular diseases by around 30% (30%; 95% CI -2 to 71% and 31%; 95% CI 9 to 57%, respectively) compared to tertiary prevention. This study aimed to identify effect modifiers of implementation strategies on physician adherence. Especially the cooperation of different health professionals in primary care practices might increase efficacy and guideline implementation seems to be more difficult in tertiary prevention of cardiovascular diseases.


Cardiovascular diseases (CVDs) are major causes of morbidity and mortality across developed countries [1]. Various guidelines are published that provide information on health care in the prevention and treatment of CVD [2][5]. A time gap between publication and implementation in clinical practice may evolve. In a recent systematic review, we found that using implementation strategies (versus not using implementation strategies) improved the adherence of physicians to guidelines on primary care of patients with CVDs [6]. We also found a considerable variability of treatment effects among the two evaluated groups.

This heterogeneity may be an artifact of methodological factors [7][9]. These factors may include differences in study design features such as predefined primary outcomes, unit of randomization [10], duration of follow-up [11], and risk of bias. High risk of bias may change the magnitude and even the direction of treatment effects [12][16]. On the other hand, heterogenic effects of implementation strategies may result from a wide range of clinical factors such as variations in the patients with CVD in primary care, types and timing in the measurement of adherence, and special characteristics of investigated strategies [7][9], [17], [18]. Implementation strategies may be characterized as unimodal or multimodal or by their distinct quality improvement components (e.g. Shojania 2004 [19]). Strategies to improve physician adherence are complex interventions and an understanding of the effectiveness of these interventions is based on the assessment of what works best in different populations, circumstances and contexts [9], [20], [21]. These conditions may cause variations in treatment effects across different studies included in the review and provide an opportunity to identify clinical factors that may modify the treatment effect on physician adherence and increase scientific understanding [7]. Discordance between recommended and observed behaviour of physicians in the treatment of cardiovascular diseases is influenced by modifiable context-specific barriers as providers' and patients' knowledge of and attitudes towards adherence to health care recommendations and by external factors related to the health system including lack of policy support for chronic care and prevention or limited access to health-care resources [22].

Primary care frequently occurs in small health care organizations with a team of professionals consisting of physicians, nurses, and other professionals in the primary, secondary, and tertiary prevention of CVD with a variety of symptoms and other diseases. The differences among professionals and/or professions involved in the care of patients with different indications and severity of CVD may influence the quality of dissemination and implementation of guidelines [23].

Statistical heterogeneity describes the variability in treatment effect estimates between studies and may arise from methodological or clinical heterogeneity, from other unknown, unrecorded or unreported study characteristics, or may be due to chance [17]. The causes and the extent of heterogeneity should be evaluated as they may compromise the implications of systematic reviews [24]. Statistical heterogeneity between treatment effects estimated by individual studies can be visually assessed in forest plots, tested for statistical significance, and quantified using the percentage of total variation across studies I2 [25], [26], or the between-study variance τ2 [27], [28]. Heterogeneity may be explored by conducting subgroup analyses or weighted meta-regression in complex interventions [29], [30]. Meta-regression analysis can be used for a simultaneous exploration of potential methodological and clinical effect modifiers. One way of dealing with statistical heterogeneity in meta-analyses and meta-regression analyses is to incorporate a term to account for it in a random-effects model.

This study is a workup of a previously published systematic review [6]. The main aim is to explain the heterogeneity of findings from the original review to identify possible methodological and clinical heterogeneity factors that may have influenced estimated treatment effects of guideline implementation strategies on physician adherence in primary care of patients with CVD.


Study design

The design of this systematic review and the efficacy results of different implementation strategies in comparison to usual care have been previously published [6]. In brief, searches included Medline, Embase, the Cochrane Library, references of included studies, conference proceedings, register of ongoing trials and references of all included studies published between 1990 and 2012. The review considered all randomized studies that investigated guideline dissemination and their implementation into treatment of patients with CVD in primary care practices. Trials had to report guideline adherence of physicians over a minimum period of 3 months after initiating the implementation strategies. We extracted information on design, indication of patients according to the International Classification of Diseases-10 (ICD-10), implementation strategies and outcomes from all eligible trials into standardized data extraction tables. Implementation strategies were categorized as provider reminder systems, provider education, facilitated relay to data, audit and feedback, promotion of self-management, patient education, patient reminder, and organizational change [31]. The internal validity of trials was judged according to the Cochrane Collaboration risk of bias tool [10] with extension to cluster randomized trials (c-RCTs) [32][34] as high, unclear or low in six specific domains including bias in random sequence generation, allocation concealment, blinded outcome assessment, documentation of incomplete outcome data and selective reporting. Furthermore, baseline comparability between treatment groups and the use of adjustment methods to cope with potential imbalances in both cluster and individual characteristics were summarized as other sources of bias. All steps were done by two independent authors, disagreements were resolved by discussion until consensus was obtained. All treatment effect measures for the primary endpoint (physician adherence) were presented as odds ratios (OR) with 95% confidence intervals (CI) and recoded so that an OR higher than 1 indicates a beneficial effect with higher physician adherence in the experimental implementation strategy. Multiple endpoints were summarized using the mean of logarithmic ORs and ORs were recalculated from relative risks [10], standardized mean difference with standard deviation [35], and absolute frequencies of physician adherence in different groups. Results of c-RCTs without hierarchical modelling were corrected with the reported intra-cluster correlation coefficient and the mean number of patients per cluster [10]. Effect sizes were interpreted in categories of small to describe effect sizes ≤20%, moderate to describe effect sizes>20 and ≤50%, and large to describe effect sizes of >50% increase of physician adherence in comparison to passive implementation.

Investigation of heterogeneity

In our first publication we explored treatment effects in subgroups of unimodal interventions, graphically displayed them in forest plots, and quantified the remaining heterogeneity of treatment effects using the I2 and τ2 values [6]. Due to the high heterogeneity between included trials we used the random-effects model for meta-analysis of the relevant comparisons of implementation strategies to usual care. This study adds a simultaneous assessment of the influence of multimodal implementation strategies and different effect modifiers on physician adherence in six random effects meta-regression models. We separately added six single sources of heterogeneity to binary variables describing the components of multimodal implementation strategies and investigated their influence on the treatment effect. Sources of heterogeneity included the staff responsible for implementation, level of prevention, definition of the primary outcome, unit of randomization, duration of follow-up and risk of bias domains.

We quantified the influence of all investigated sources of heterogeneity with ratios of odds ratios (ROR) comparing OR of studies with different values of the sources of heterogeneity (e.g. studies in secondary prevention with those in tertiary prevention of CVD). ROR with 95% confidence intervals (95%CI) not containing the null value (ROR = 1) will be interpreted as significant. Heterogeneity was measured using the τ2 statistics which estimates the between-trial variability. The amount of heterogeneity explained by different effect modifiers was described by the relative reduction of τ2. All modifiers were investigated as binary traits. Categorical traits (staff responsible for implementation, type of prevention) were recoded into dummy variables. We conducted the statistical analyses using RevMan5 for the systematic review and SAS 9.2 (PROC MIXED statement) for this study.


Meta-analyses of treatment effects on physician adherence revealed considerable heterogeneity in all subgroups of unimodal interventions [6]. In this study, altogether 75 trials were pooled in 84 comparisons between unimodal or multimodal active and passive implementation strategies (Table S1). Of these comparisons, 13 indicated a negative treatment effect with an OR<1, 17 a small effect size, 22 a moderate effect size, and 32 a large effect size of physician adherence to guidelines.

In the majority of trials the implementation strategies were directed at physicians (70 trials). Physicians were supported by nurses in 19 trials or other non-medical professionals such as pharmacists (12 trials), (study)-assistants (3 trials), health workers (2 trials), (peer-) supervisors, praxis managers, or other specialists (1 trial). In four trials, strategies were exclusively implemented by specialized nurses and in one trial by a team of nurses, Asian link (health-) workers, and community diabetes specialists. Most trials concentrated on patients in the secondary prevention of CVD, 10 trials additionally included patients in primary prevention, and 6 trials included patients in tertiary prevention. Only 6 trials were limited to primary prevention and 12 trials to tertiary prevention. The most successful strategies based on patient education and organizational change regularly included nurses (46 and 41% of trials) or other professionals (31 and 71% of trials) in the implementation process (Table 1).

Table 1. Investigated sources of heterogeneity in 75 trials.

Approximately half of the trials were pre-planned with physician adherence as the primary outcome (36 trials) and the other half of trials reported adherence as the secondary outcome or parameter describing the process of care (39 trials). Units of analyses were individual patients in 23 RCTs and practices in 52 c-RCTs. Implementation strategies that were directed to the staff of general practices, such as provider reminder systems, audit and feedback, provider education, and organizational change, were frequently investigated in c-RCTs with physician adherence as the primary outcome. On the other hand, trials on strategies directed to patients such as patient education, promotion of self-management, and patient reminders predominantly concentrated on patient-related outcomes and reported physician adherence to the implementation process mostly as a secondary outcome. Follow-up periods between 3 and 36 months were used to investigate the efficacy of the intervention with a median length of follow-up of 12 months. Of the 75 trials, 61 trials had follow-up periods of between 3 and 12 months and 14 had longer follow-up periods of up to 36 months (Table 1).

Of the 75 trials, 48 reported the method of randomization in the text. The treatment allocation of clusters or patients was described as concealed in 63 trials. Physician adherence was assessed on objective criteria (such as number of medications, or by external monitors) and/or blinded in 63 trials. In 61 trials, the analyses were done by intention-to-treat, both at the individual and at the cluster levels. Total numbers of dropouts were low (<10%) and their causes were given by group. Primary endpoints were pre-specified in sample-size calculations and were adequately reported in 47 trials. Other sources of bias were evident in 28 trials that made no use of appropriate methods for the adjustment of treatment effects on physician adherence to cope with potential imbalances in cluster and individual characteristics. Summarizing these results, only 15 trials had low risk of bias in all the investigated domains.

Association between heterogeneity factors and estimated treatment effects

We calculated relative frequencies of negative, small, moderate, and large treatment effects depending on subgroups with special clinical and methodological characteristics. These characteristics were correlated with special implementation strategies, as shown in Table 2.

Table 2. Association between trial characteristics and treatment effect in 84 comparisons.

In general, the inclusion of other non-medical professionals seems to be most successful in improving physician adherence to guidelines. Large treatment effects are indicated in 57% of comparisons where non-medical professionals were included in the implementation process, compared to 38% if exclusively physicians were included, and 42% if nurses were included. Furthermore, large treatment effects are more frequently found in trials on primary or secondary prevention (48 or 40% of trials) compared to tertiary prevention (24%) of CVD.

Comparisons of trials investigating the process of implementation with physician adherence as a primary outcome less frequently showed a negative direction of effect (12% vs. 19%) and more frequently a moderate or large effect size with OR>1.2 (71% vs. 58%). Finally, RCTs were more frequently successful with a moderate or large effect size compared to c-RCTs (79 vs. 58%), and studies with longer follow-up periods of at least 12 months more frequently showed a large effect size (50% vs. 35%) compared to shorter trials.

Table 3 summarizes the ratios of odds ratios (ROR) for the influence of all investigated clinical and methodological heterogeneity factors on treatment effect. Between-trial variability and the relative reduction of between-trial variability by single effect modifiers compared to the variability of the original model describe the reduction of statistical heterogeneity.

Table 3. Association of different effect modifiers on treatment effect.

These results showed that the treatment effect varied depending on clinical heterogeneity factors. The receiver (i.e., person/profession responsible for implementation) of the implementation seems especially to influence the efficiency of implementation strategies on physician adherence. The inclusion of the receiver of the implementation into the meta-regression reduced between-trial variability (τ2) by 27%. Physician adherence was improved by 62% (ROR 1.62; 95% CI 1.29 to 2.04) in trials where other non-medical professionals were included in the process of implementation, and expanding the role of nurses in the curative process was successful in improving physician adherence by 29% (ROR 1.29; 95% CI 1.05 to 1.60).

Inclusion of level of prevention as a clinical effect modifier reduced τ2 by 11%. Improvement of physician adherence was most successful in the treatment of patients in primary and secondary prevention of CVD (ROR 1.30; 95% CI 0.98 to 1.71 and 1.31; 95% CI 1.09 to 1.57) compared to patients in tertiary prevention.

However, methodological issues seem to have a smaller influence on the investigated association. We found an association between the definition of the primary endpoint as a quantitative measure of the main objective of a trial and the estimated implementation effect. The inclusion of this effect modifier was able to reduce τ2 by 10%. The implementation effect on the primary outcome in process optimization studies with the primary endpoint of physician adherence was increased by 38% (ROR 1.38; 95% CI 1.12 to 1.70) compared to outcome optimization studies where physician adherence was investigated as a secondary or process of care parameter. Moreover, we also found an increased implementation effect in RCTs compared to c-RCTs by 28% (ROR 1.28; 95% CI 1.02 to 1.60) and with longer duration of follow-up (ROR 1.38; 95% CI 1.03 to 1.83), with only small reductions of τ2 by 2 and 8%, respectively.

Inclusion of six risk of bias domains reduced between-trial variability by 22%, but no single component was associated with a significant overestimation of implementation effect. We found a tendency to an overestimation of implementation effects in trials with potential bias by selective reporting of the primary outcome by 58% (ROR 1.58; 95% CI 0.96 to 2.60).


This study is based on an analysis of 75 trials and on eight classes of implementation strategies to improve physician's guideline adherence. It investigates the influence of six possible effect modifiers on estimated implementation effects and remaining statistical heterogeneity. These influences are quantified by RORs and the relative change of between trial variability. Our investigations revealed a substantial reduction of statistical heterogeneity explained by five of the investigated effect modifiers.

We found that clinical effect modifiers such as the cooperation of physicians with non-medical health professionals, the setting of the primary care of patients in early prevention of CVD, and the duration of implementation were especially associated with the improvement of physicians' adherence to guidelines. We have found a considerable reduction of statistical heterogeneity by these factors of 27, 11, and 8%, respectively. Furthermore, the inclusion of methodological effect modifiers as different sources of bias or the definition of the primary endpoint was able to reduce statistical heterogeneity by 22 and 8%, respectively.

A considerable amount of statistical heterogeneity is explained by organizational structures in the primary care practices. Improvement of adherence could be achieved if physicians accepted support from non-medical health professionals such as pharmacists, health workers, qualified nurses, or nurse practitioners in improving their professional and organizational performance. Such cooperation could take place within team or (smaller) teamlet structures in single practices or networks of care [36]. Teamwork among different professions and/or professionals aimed at implementing guidelines explains the different estimated treatment effects on the adherence of primary care physicians. Our results are in line with the conclusions of Grol and Grimshaw (2003) [37] and Unverzagt et al. (2013) [6] which suggest that the whole primary care team (or network) is important for the implementation success.

We have further showed that different levels of prevention may cause heterogeneous treatment effects with greater improvements in physician adherence in the primary care of patients in the early prevention of CVD. Guidelines are more difficult to implement in tertiary prevention, where patients frequently suffer from complex co-morbidities and guidance on interventions and information on risks of specific interventions is missed [38]. This observation reflects the problem how far results from RCT can be generalized due to selection in patients included in RCTs and narrow inclusion/exclusion criteria [39]. Especially patients with complex comorbidities should be well represented in RCTs and guidelines should provide guidance for the often complex need of these patients including information on risks of specific interventions [38].

Moreover, our explorations revealed artificial sources of variation resulting from the inclusion of two different types of study with different main aims, although both types reported physician adherence. Physician adherence summarizes the degree of conformity between knowledge, cognition, and action in a primary physician center with evidence-based recommendations from guidelines on different aspects of quality of care [40].

The primary aim of the first type of study is to improve the care process by implementing guidelines and the primary aim of the second type is to improve the health outcomes of patients. A “change” to a more evidence-based treatment aimed at improving health outcomes is a stepwise complex process in which several individual and organizational barriers have to be removed and intermediate outcomes (such as adherence) and final outcomes (such as health outcomes) may be improved. First, different strategies should be used to implement evidence-based guidelines in the care process and to enhance the adherence of physicians. Secondly, patients must be adherent to evidence-based recommendations, and finally, the health outcomes for patients might change. The benefit they receive from guideline-oriented treatment by reduced hospital admissions and prolonged survival has been shown in several studies [41][43]. In trials designed to improve the implementation of guidelines, physician adherence was regularly chosen as a primary endpoint and measured by multiple indicators, which were summarized to adherence scores or described by different frequencies.

The second type of study concentrates on the improvement of health outcomes of patients and describes physician adherence as a step towards improving mortality, morbidity, health-related quality of life, or surrogate parameters such as the achievement of targets for blood-pressure, cholesterol concentrations, physical activity, body mass index, smoking cessation or reduced smoking, or the reduction of cardiovascular risk score. Some statistical heterogeneity of treatment effects can be explained by these different types of studies.

In addition, we stated an increased treatment effect in studies with longer follow-up periods where implementation strategies were used over a longer period with a potentially better chance to improve providers knowledge, to integrate guidelines recommendation into organizational structures and processes and to overcome as well as negative staff attitudes and beliefs and time and resource constraints in primary care centers.

Furthermore, we identified an influence of risk of bias on the variability of treatment effect and were able to reduce statistical heterogeneity, but we were not able to identify one single source of bias causing biased treatment effects. However, different additional factors show divergent influences on the process of implementation and health outcomes and may influence both the process of implementation and the efficacy of the recommended intervention for the patient. These factors include concerns about the quality of guidelines such as the quality of evidence on which they are based, lack of agreement, differences in strength of recommendations, practicability of guidelines and recommended interventions, and the benefit/harm ratio of the intervention [44], [45]. Finally, financial constraints and organizational structures (e.g., health systems) may modify the process of care [46].

Strengths and limitations

We are aware that our research may have some limitations. The conclusions of our systematic review [6] were limited by a large amount of imprecision and inconsistency of reported results [24], [47], [48]. Unexplained heterogeneity was caused by different methodological and clinical effect modifiers and some of them were investigated in this study. Other, not investigated potential effect modifiers are differences in health care systems, stability, attitudes and resource constraints of health care teams, types of guideline-recommended interventions, patient decisions and treatment. This study contributes an improved insight into the processes and elements of successful change and shows the environment in which implementation strategies work best, but all these ideas are exploratory and should be considered as hypotheses for evaluation in future studies [17]. Developing of precise definitions of effect modifiers for complex definitions of participants, intervention and outcomes was problematic because our priori definitions were influenced from our growing knowledge on existing evidence. To reduce the dependence on the results of data syntheses, we discussed these potential effect modifiers and tried to find the best classifications on the basis of diverse extracted clinical and methodological characteristics before we started data synthesis. Moreover, we are assured that the investigated heterogeneity factors are unlikely to influence the pathway from implementation strategy to an improved physician adherence and patient outcome independently and may reflect associations with other correlated but not investigated factors [25]. Furthermore, it is possible that especially in studies with inclusion of patients in early and late prevention relationships across trials may not be the same as relationships for patients within trials [29].


We recommend a careful discussion of the pathway from the intervention to the outcomes and pre-definition of potential effect modifiers at an early time point when conducting a systematic review. This seems to be especially important in a review of complex interventions with a broad range of participants, interventions, comparators, and definitions of outcomes. Our work has explained some statistical heterogeneity by clinical and methodological effect modifiers. These effect modifiers cause variability in estimated treatment effects, and taking them into account reduced statistical heterogeneity.

The results of this study provide some evidence that the incorporation of different health professionals in the practice can change professional and organizational performance in a primary care practice. This study properly investigated the role of implementation strategies in the treatment of patients in primary and secondary prevention of CVD and provides some evidence of promising results. However, the implementation of guidelines into tertiary prevention of CVD in general practice requires improved guidance for this patient group. Finally, we propose choosing the process parameter of physician adherence as the primary outcome parameter only in cases where a theoretical model explains the route from the intervention to the anticipated health outcomes for patients. Nevertheless, the health outcomes for patients should always be measured and reported additionally to process parameters (e.g., physician adherence) to ensure the association between the process parameter and improved health outcomes and to improve understanding of the pathway from implementation over physician adherence to improved health outcomes.

Supporting Information

Table S1.

Included studies and comparisons with treatment effect, implementation strategies in the intervention group, risk of bias and effect modifiers. Abbreviations: lnOR: logarithmic Odds Ratio, SElnOR: standard error of logarithmic Odds Ratio.



We thank Roland Prondzinsky for his constructive comments and providing clinical background.

Author Contributions

Conceived and designed the experiments: SU. Performed the experiments: SU. Analyzed the data: SU. Contributed reagents/materials/analysis tools: SU. Wrote the paper: SU FP MO KB AK.


  1. 1. Murray CJL, Lopez AD (1996) Evidence-based health policy - lessons from the global burden of disease study. Science 274: 740–743.
  2. 2. AHA (2014) ACC/AHA Joint Guidelines. Dallas: American Heart Association.
  3. 3. CDC (2014) Prevention. Guidelines and Recommendations. Atlanta: Centers for Disease Control and Prevention.
  4. 4. ESC (2014) ESC Clinical Practice Guidelines section. Sophia Antipolis: European Society of Cardiology.
  5. 5. WHO (2014) Prevention of cardiovascular disease (CVDs). Geneva: World Health Organization.
  6. 6. Unverzagt S, Oemler M, Braun K, Klement A (2013) Strategies for guideline implementation in primary care focusing on patients with cardiovascular disease: a systematic review. Fam Pract 31: 247–266.
  7. 7. Thompson SG (1994) Why sources of heterogeneity in meta-analysis should be investigated. BMJ 309: 1351–1355.
  8. 8. Glasziou PP, Sanders SL (2002) Investigating causes of heterogeneity in systematic reviews. Stat Med 21: 1503–1511.
  9. 9. Pigott T, Shepperd S (2013) Identifying, documenting, and examining heterogeneity in systematic reviews of complex interventions. J Clin Epidemiol 66: 1244–1250.
  10. 10. Cochrane Collaboration (2011) Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 [updated March 2011]; Higgins JPT, Green S, editors. Oxford: The Cochrane Collaboration.
  11. 11. Rosén M (2009) The aprotinin saga and the risks of conducting meta-analyses on small randomised controlled trials - a critique of a Cochrane review. BMC Health Serv Res 9: 34.
  12. 12. Schulz KF, Chalmers I, Hayes RJ, Altman DG (1995) Empirical evidence of bias. Dimensions of methodological quality associated with estimates of treatment effects in controlled trials. JAMA 273: 408–412.
  13. 13. Moher D, Pham B, Jones A, Cook DJ, Jadad AR, et al. (1998) Does quality of reports of randomised trials affect estimates of intervention efficacy reported in meta-analyses? Lancet 352: 609–613.
  14. 14. Balk EM, Bonis PA, Moskowitz H, Schmid CH, Ioannidis JP, et al. (2002) Correlation of quality measures with estimates of treatment effect in meta-analyses of randomized controlled trials. JAMA 287: 2973–2982.
  15. 15. Wood L, Egger M, Gluud LL, Schulz KF, Juni P, et al. (2008) Empirical evidence of bias in treatment effect estimates in controlled trials with different interventions and outcomes: meta-epidemiological study. BMJ 336: 601–605.
  16. 16. Egger M, Juni P, Bartlett C, Holenstein F, Sterne J (2003) How important are comprehensive literature searches and the assessment of trial quality in systematic reviews? Empirical study. Health Technol Assess 7: 1–76.
  17. 17. Gagnier JJ MH, Altman DG, Berlin J, Chang S, McCulloch P, et al. (2013) Consensus-based recommendations for investigating clinical heterogeneity in systematic reviews. BMC Med Res Methodol 13
  18. 18. Higgins J, Thompson S, Deeks J, Altman D (2002) Statistical heterogeneity in systematic reviews of clinical trials: a critical appraisal of guidelines and practice. J Health Serv Res Policy 7: 51–61.
  19. 19. Shojania KG, McDonald KM, Wachter RM, Owens DK (2004) Closing Closing The Quality Gap: A Critical Analysis of Quality Improvement Strategies, Volume 1—Series Overview and Methodology. Technical Review 9 (Contract No. 290-02-0017 to the Stanford University–UCSF Evidence-based Practices Center). AHRQ Publication No. 04-0051-1. Rockville: Agency for Healthcare Research and Quality.
  20. 20. Petticrew M, Anderson L, Elder R, Grimshaw J, Hopkins D, et al. (2013) Complex interventions and their implications for systematic reviews: a pragmatic approach. J Clin Epidemiol 66: 1209–1214.
  21. 21. Petticrew M, Rehfuess E, Noyes J, Higgins JP, Mayhew A, et al. (2013) Synthesizing evidence on complex interventions: how meta-analytical, qualitative, and mixed-method approaches can contribute. J Clin Epidemiol 66: 1230–1243.
  22. 22. Nieuwlaat R, Schwalm JD, Khatib R, Yusuf S (2013) Why are we failing to implement effective therapies in cardiovascular disease? Eur Heart J 34: 1262–1269.
  23. 23. Wensing M, van der Weijden T, Grol R (1998) Implementing guidelines and innovations in general practice: which interventions are effective? Br J Gen Pract 48: 991–997.
  24. 24. Guyatt GH, Oxman AD, Kunz R, Woodcock J, Brozek J, et al. (2011) GRADE guidelines: 7. Rating the quality of evidence–inconsistency. J Clin Epidemiol 64: 1294–1302.
  25. 25. Higgins JP, Thompson SG (2002) Quantifying heterogeneity in a meta-analysis. Stat Med 21: 1539–1558.
  26. 26. Higgins JP, Thompson SG, Deeks JJ, Altman DG (2003) Measuring inconsistency in meta-analyses. BMJ 327: 557–560.
  27. 27. DerSimonian R, Laird N (1986) Meta-analysis in clinical trials. Control Clin Trials 7: 177–188.
  28. 28. Rucker G, Schwarzer G, Carpenter JR, Schumacher M (2008) Undue reliance on I(2) in assessing heterogeneity may mislead. BMC Med Res Methodol 8: 79.
  29. 29. Thompson SG, Higgins JP (2002) How should meta-regression analyses be undertaken and interpreted? Stat Med 21: 1559–1573.
  30. 30. Bower P, Gilbody S, Richards D, Fletcher J, Sutton A (2006) Collaborative care for depression in primary care. Making sense of a complex intervention: systematic review and meta-regression. Br J Psychiatry 189: 484–493.
  31. 31. AHRQ (2004) Closing the Quality Gap: A Critical Analysis of Quality Improvement Strategies (Vol. 1: Series Overview and Methodology); Shojania KG, McDonald KM, Wachter RM, Owens DK, editors. Rockville: Agency for Healthcare Research and Quality.
  32. 32. Puffer S, Torgerson D, Watson J (2003) Evidence for risk of bias in cluster randomised trials: review of recent trials published in three general medical journals. BMJ 327: 785–789.
  33. 33. Giraudeau B, Ravaud P (2009) Preventing bias in cluster randomised trials. PLoS Med 6: e1000065.
  34. 34. Campbell MK, Piaggio G, Elbourne DR, Altman DG Group C (2012) Consort 2010 statement: extension to cluster randomised trials. BMJ 345: e5661.
  35. 35. Borenstein M, Hedges LV, Higgins JPT, Rothstein HR (2009) Introduction to meta-analysis. Chichester: John Wiley & Sons.
  36. 36. Chen EH, Thom DH, Hessler DM, Phengrasamy L, Hammer H, et al. (2010) Using the Teamlet Model to Improve Chronic Care in an Academic Primary Care Practice. J Gen Intern Med 25: 610–614.
  37. 37. Grol R, Grimshaw J (2003) From best evidence to best practice: effective implementation of change in patients' care. Lancet 362: 1225–1230.
  38. 38. Francke AL, Smit MC, de Veer AJ, Mistiaen P (2008) Factors influencing the implementation of clinical guidelines for health care professionals: a systematic meta-review. BMC Med Inform Decis Mak 8: 38.
  39. 39. Bailey KR (1994) Generalizing the results of randomized clinical trials. Control Clin Trials 15: 15–23.
  40. 40. Hasenbein U, Wallesch CW (2007) [What is “adherence to guidelines”? Theoretical and methodological considerations on a new concept for health system research and quality management]. Gesundheitswesen 69: 427–437.
  41. 41. Kuepper-Nybelen J, Hellmich M, Abbas S, Ihle P, Griebenow R, et al. (2012) Association of long-term adherence to evidence-based combination drug therapy after acute myocardial infarction with all-cause mortality. A prospective cohort study based on claims data. Eur J Clin Pharmacol 68: 1451–1460.
  42. 42. Frankenstein L, Remppis A, Fluegel A, Doesch A, Katus HA, et al. (2010) The association between long-term longitudinal trends in guideline adherence and mortality in relation to age and sex. Eur J Heart Fail 12: 574–580.
  43. 43. Komajda M, Lapuerta P, Hermans N, Gonzalez-Juanatey JR, van Veldhuisen DJ, et al. (2005) Adherence to guidelines is a predictor of outcome in chronic heart failure: the MAHLER survey. Eur Heart J 26: 1653–1659.
  44. 44. Ploeg J, Davies B, Edwards N, Gifford W, Miller PE (2007) Factors influencing best-practice guideline implementation: lessons learned from administrators, nursing staff, and project leaders. Worldviews Evid Based Nurs 4: 210–219.
  45. 45. Prior M, Guerin M, Grimmer-Somers K (2008) The effectiveness of clinical guideline implementation strategies–a synthesis of systematic review findings. J Eval Clin Pract 14: 888–897.
  46. 46. SIGN (2008) SIGN 50: A guideline developer's handbook. Edinburgh: Scottish Intercollegiate Guidelines Network.
  47. 47. Balshem H, Helfand M, Schunemann HJ, Oxman AD, Kunz R, et al. (2011) GRADE guidelines: 3. Rating the quality of evidence. J Clin Epidemiol 64: 401–406.
  48. 48. Guyatt GH, Oxman AD, Kunz R, Brozek J, Alonso-Coello P, et al. (2011) GRADE guidelines 6. Rating the quality of evidence–imprecision. J Clin Epidemiol 64: 1283–1293.
  49. 49. Beaulieu MD, Brophy J, Jacques A, Blais R, Battista R, et al. (2004) Drug treatment of stable angina pectoris and mass dissemination of therapeutic guidelines: a randomized controlled trial. QJM 97: 21–31.
  50. 50. Bebb C, Kendrick D, Coupland C, Madeley R, Stewart J, et al. (2007) A cluster randomised controlled trial of the effect of a treatment algorithm for hypertension in patients with type 2 diabetes. Br J Gen Pract 57: 136–143.
  51. 51. Casebeer LL, Klapow JC, Centor RM, Stafford MA, Renkl LA, et al. (1999) An intervention to increase physicians' use of adherence-enhancing strategies in managing hypercholesterolemic patients. Acad Med 74: 1334–1339.
  52. 52. Cobos A, Vilaseca J, Asenjo C, Pedro-Botet J, Sanchez E, et al. (2005) Cost effectiveness of a clinical decision support system based on the recommendations of the European Society of Cardiology and other societies for the management of hypercholesterolemia: Report of a cluster-randomized trial. Dis Manag Health Out 13: 421–432.
  53. 53. Eccles M, McColl E, Steen N, Rousseau N, Grimshaw J, et al. (2002) Effect of computerised evidence based guidelines on management of asthma and angina in adults in primary care: Cluster randomised controlled trial. BMJ 325: 941–944.
  54. 54. Frances CD, Alperin P, Adler JS, Grady D (2001) Does a fixed physician reminder system improve the care of patients with coronary artery disease? A randomized controlled trial. West J Med 175: 165–166.
  55. 55. Fretheim A, Oxman AD, Havelsrud K, Treweek S, Kristoffersen DT, et al. (2006) Rational prescribing in primary care (RaPP): A cluster randomized trial of a tailored intervention. PLoS Med 3: 783–791.
  56. 56. Frijling BD, Lobo CM, Hulscher ME, Akkermans RP, van Drenth BB, et al. (2003) Intensive support to improve clinical decision making in cardiovascular care: a randomised controlled trial in general practice. Qual Saf Health Care 12: 181–187.
  57. 57. Gill JM, Chen YX, Glutting JJ, Diamond JJ, Lieberman MI (2009) Impact of decision support in electronic medical records on lipid management in primary care. Popul Health Manag 12: 221–226.
  58. 58. Gilutz H, Novack L, Shvartzman P, Zelingher J, Bonneh DY, et al. (2009) Computerized community cholesterol control (4C): Meeting the challenge of secondary prevention. Isr Med Assoc J 11: 23–29.
  59. 59. Goff DC Jr, Gu L, Cantley LK, Sheedy DJ, Cohen SJ (2003) Quality of care for secondary prevention for patients with coronary heart disease: results of the Hastening the Effective Application of Research through Technology (HEART) trial. Am Heart J 146: 1045–1051.
  60. 60. Heidenreich PA, Gholami P, Sahay A, Massie B, Goldstein MK (2007) Clinical reminders attached to echocardiography reports of patients with reduced left ventricular ejection fraction increase use of beta-blockers: A randomized trial. Circulation 115: 2829–2834.
  61. 61. Hicks LS, Sequist TD, Ayanian JZ, Shaykevich S, Fairchild DG, et al. (2008) Impact of computerized decision support on blood pressure management and control: a randomized controlled trial. J Gen Intern Med 23: 429–441.
  62. 62. Hung CS, Lin JW, Hwang JJ, Tsai RY, Li AT (2008) Using paper chart based clinical reminders to improve guideline adherence to lipid management. J Eval Clin Pract 14: 861–866.
  63. 63. Lester WT, Grant RW, Octo BG, Chueh HC (2006) Randomized controlled trial of an informatics-based intervention to increase statin prescription for secondary prevention of coronary disease. J Gen Intern Med 21: 22–29.
  64. 64. Montgomery AA, Fahey T, Peters TJ, MacIntosh C, Sharp DJ (2000) Evaluation of computer based clinical decision support system and risk chart for management of hypertension in primary care: randomised controlled trial. BMJ 320: 686–690.
  65. 65. Murray MD, Harris LE, Overhage JM, Zhou XH, Eckert GJ, et al. (2004) Failure of computerized treatment suggestions to improve health outcomes of outpatients with uncomplicated hypertension: results of a randomized controlled trial. Pharmacotherapy 24: 324–337.
  66. 66. Rossi RA, Every NR (1997) A computerized intervention to decrease the use of calcium channel blockers in hypertension. J Gen Intern Med 12: 672–678.
  67. 67. Santschi V, Rodondi N, Bugnon O, Burnier M (2008) Impact of electronic monitoring of drug adherence on blood pressure control in primary care: a cluster 12-month randomised controlled study. Eur J Intern Med 19: 427–434.
  68. 68. Sequist TD, Gandhi TK, Karson AS, Fiskio JM, Bugbee D, et al. (2005) A randomized trial of electronic clinical reminders to improve quality of care for diabetes and coronary artery disease. J Am Med Inform Assoc 12: 431–437.
  69. 69. Tierney WM, Overhage JM, Murray MD, Harris LE, Zhou XH, et al. (2003) Effects of computerized guidelines for managing heart disease in primary care. J Gen Intern Med 18: 967–976.
  70. 70. Van Wyk JT, van Wijk MAM, Sturkenboom MC, Mooseveld M, Moormann PW, et al. (2008) Electronic Alerts Versus On-Demand Decision Support to Improve Dyslipidemia Treatment: A Cluster Randomized Controlled Trial. Circulation 117: 371–378.
  71. 71. Villeneuve J, Genest J, Blais L, Vanier MC, Lamarre D, et al. (2010) A cluster randomized controlled Trial to Evaluate an Ambulatory primary care Management program for patients with dyslipidemia: the TEAM study. CMAJ 182: 447–455.
  72. 72. Benner JS, Erhardt L, Flammer M, Moller RA, Rajicic N, et al. (2008) A novel programme to evaluate and communicate 10-year risk of CHD reduces predicted risk and improves patients' modifiable risk factor profile. Int J Clin Pract 62: 1484–1498.
  73. 73. Bonds DE, Hogan PE, Bertoni AG, Chen H, Clinch CR, et al. (2009) A multifaceted intervention to improve blood pressure control: The Guideline Adherence for Heart Health (GLAD) study. Am Heart J 157: 278–284.
  74. 74. Fiscella K, Volpe E, Winters P, Brown M, Idris A, et al. (2010) A novel approach to quality improvement in a safety-net practice: concurrent peer review visits. Journal of the National Medical Association 102: 1231–1236.
  75. 75. Kogan JR, Reynolds EE, Shea JA (2003) Effectiveness of report cards based on chart audits of residents' adherence to practice guidelines on practice performance: a randomized controlled trial. Teach Learn Med 15: 25–30.
  76. 76. Luders S, Schrader J, Schmieder RE, Smolka W, Wegscheider K, et al. (2010) Improvement of hypertension management by structured physician education and feedback system: cluster randomized trial. Eur J Cardiovasc Prev Rehabil 17: 271–279.
  77. 77. Mitchell E, Sullivan F, Watt G, Grimshaw JM, Donnan PT (2004) Using electronic patient records to inform strategic decision making in primary care. Stud Health Technol Inform 107: 2–61.
  78. 78. Sondergaard J, Hansen DG, Aarslev P, Munck AP (2006) A multifaceted intervention according to the Audit Project Odense method improved secondary prevention of ischemic heart disease: A randomised controlled trial. Fam Pract 23: 198–202.
  79. 79. Verstappen WHJM, van der Weijden T, Sijbrandij J, Smeele I, Hermsen J, et al. (2003) Effect of a Practice-Based Strategy on Test Ordering Performance of Primary Care Physicians: A Randomized Trial. JAMA 289: 2407–2412.
  80. 80. Amado GE, Pujol RE, Pacheco HV, Borras JM Group A (2011) Knowledge and adherence to antihypertensive therapy in primary care: results of a randomized trial. Gaceta Sanitaria 25: 62–67.
  81. 81. Anguita Sánchez M, Jiménez-Navarro M, Crespo M, Alonso-Pulpón L, de Teresa E, et al. (2010) Using the Teamlet Model to Improve Chronic Care in an Academic Primary Care Practice. Revista espaňola de cardiologia 63: 677–685.
  82. 82. De Muylder R, Tonglet R, Nackers F, Boland B (2005) Randomised evaluation of a specific training of general practitioners in cardiovascular prevention. Acta Cardiol 60: 199–205.
  83. 83. Feder G, Griffiths C, Eldridge S, Spence M (1999) Effect of postal prompts to patients and general practitioners on the quality of primary care after a coronary event (POST): Randomised controlled trial. BMJ 318: 1522–1526.
  84. 84. Feldman RD, Zou GY, Vandervoort MK, Wong CJ, Nelson SA, et al. (2009) A simplified approach to the treatment of uncomplicated hypertension: a cluster randomized, controlled trial. Hypertension 53: 646–653.
  85. 85. Goldberg HI, Wagner EH, Fihn SD, Martin DP, Horowitz CR, et al. (1998) A randomized controlled trial of CQI teams and academic detailing: can they alter compliance with guidelines? Jt Comm J Qual Improv 24: 130–142.
  86. 86. Jafar TH, Hatcher J, Poulter N, Islam M, Hashmi S, et al. (2009) Community-based interventions to promote blood pressure control in a developing country: a cluster randomized trial. Ann Intern Med 151: 593–601.
  87. 87. Kasje WN, Denig P, Stewart RE, de Graeff PA, Haaijer-Ruskamp FM (2006) An educational programme for peer review groups to improve treatment of chronic heart failure and diabetes mellitus type 2 in general practice. J Eval Clin Pract 12: 613–621.
  88. 88. Kiessling A, Henriksson P (2002) Efficacy of case method learning in general practice for secondary prevention in patients with coronary artery disease: Randomised controlled study. BMJ 325: 877–880.
  89. 89. Langham J, Tucker H, Sloan D, Pettifer J, Thom S, et al. (2002) Secondary prevention of cardiovascular disease: A randomised trial of training in information management, evidence-based medicine, both or neither: The PIER trial. Br J Gen Pract 52: 818–824.
  90. 90. Levine DA, Funkhouser EM, Houston TK, Gerald JK, Johnson-Roe N, et al. (2011) Improving care after myocardial infarction using a 2-year internet-delivered intervention: the Department of Veterans Affairs myocardial infarction-plus cluster-randomized trial. Arch Intern Med 171: 1910–1917.
  91. 91. Martensson J, Stromberg A, Dahlstrom U, Karlsson JE, Fridlund B (2005) Patients with heart failure in primary health care: effects of a nurse-led intervention on health-related quality of life and depression. Eur J Heart Fail 7: 393–403.
  92. 92. McAlister FA, Fradette M, Majumdar SR, Williams R, Graham M, et al. (2009) The enhancing secondary prevention in coronary artery disease trial. CMAJ 181: 897–904.
  93. 93. McBride P, Underbakke G, Plane MB, Massoth K, Brown RL, et al. (2000) Improving prevention systems in primary care practices: the Health Education and Research Trial (HEART). J Fam Pract 49: 115–125.
  94. 94. Mehler PS, Krantz MJ, Lundgren RA, Estacio RO, MacKenzie TD, et al. (2005) Bridging the quality gap in diabetic hyperlipidemia: A practice-based intervention. Am J Med 118: 1414.e1413–1414.e1419.
  95. 95. Simon SR, Majumdar SR, Prosser LA, Salem-Schatz S, Warner C, et al. (2005) Group versus individual academic detailing to improve the use of antihypertensive medications in primary care: a cluster-randomized controlled trial. Am J Med 118: 521–528.
  96. 96. Zillich AJ, Ackermann RT, Stump TE, Ambuehl RJ, Downs SM, et al. (2008) An evaluation of educational outreach to improve evidence-based prescribing in Medicaid: A cautionary tale. J Eval Clin Pract 14: 854–860.
  97. 97. Holbrook A, Pullenayegum E, Thabane L, Troyan S, Foster G, et al. (2011) Shared electronic vascular risk decision support in primary care: Computerization of Medical Practices for the Enhancement of Therapeutic Effectiveness (COMPETE III) randomized trial. Arch Intern Med 171: 1736–1744.
  98. 98. Lapointe F, Lepage S, Larriv‚e L, Maheux P (2006) Surveillance and treatment of dyslipidemia in the post-infarct patient: can a nurse-led management approach make a difference? Can J Cardiol 22: 761–767.
  99. 99. Montgomery AA, Fahey T, Peters TJ (2003) A factorial randomised controlled trial of decision analysis and an information video plus leaflet for newly diagnosed hypertensive patients. Br J Gen Pract 53: 446–453.
  100. 100. Muhlhauser I, Sawicki PT, Didjurgeit U, Jurgens V, Trampisch HJ, et al. (1993) Evaluation of a structured treatment and teaching programme on hypertension in general practice. Clin Exp Hypertens 15: 125–142.
  101. 101. Munoz MA, Vila J, Cabanero M, Rebato C, Subirana I, et al. (2007) Efficacy of an intensive prevention program in coronary patients in primary care, a randomised clinical trial. Int J Cardiol 118: 312–320.
  102. 102. Oddone EZ, Weinberger M, Giobbie-Hurder A, Landsman P, Henderson W (1999) Enhanced access to primary care for patients with congestive heart failure. Veterans Affairs Cooperative Study Group on Primary Care and Hospital Readmission. Eff Clin Pract 2: 201–209.
  103. 103. O′Hare JP, Raymond NT, Mughal S, Dodd L, Hanif W, et al. (2004) Evaluation of delivery of enhanced diabetes care to patients of South Asian ethnicity: The United Kingdom Asian Diabetes Study (UKADS). Diabet Med 21: 1357–1365.
  104. 104. Saunders LD, Irwig LM, Gear JS, Ramushu DL (1991) A randomized controlled trial of compliance improving strategies in Soweto hypertensives. Med Care 29: 669–678.
  105. 105. Wood DA, Kotseva K, Connolly S, Jennings C, Mead A, et al. (2008) Nurse-coordinated multidisciplinary, family-based cardiovascular disease prevention programme (EUROACTION) for patients with coronary heart disease and asymptomatic individuals at high risk of cardiovascular disease: a paired, cluster-randomised controlled trial. Lancet 371: 1999–2012.
  106. 106. Grover SA, Lowensteyn I, Joseph L, Kaouache M, Marchand S, et al. (2007) Patient knowledge of coronary risk profile improves the effectiveness of dyslipidemia therapy: the CHECK-UP study: a randomized controlled trial. Arch Intern Med 167: 2296–2303.
  107. 107. McManus RJ, Mant J, Roalfe A, Oakes RA, Bryan S, et al. (2005) Targets and self monitoring in hypertension: randomised controlled trial and cost effectiveness analysis. BMJ 331: 493.
  108. 108. McManus RJ, Mant J, Bray EP, Holder R, Jones MI, et al. (2010) Telemonitoring and self-management in the control of hypertension (TASMINH2): a randomised controlled trial. Lancet 376: 163–172.
  109. 109. Parati G, Omboni S, Albini F, Piantoni L, Giuliano A, et al. (2009) Home blood pressure telemonitoring improves hypertension control in general practice. The TeleBPCare study. J Hypertens 27: 198–203.
  110. 110. Baig AA, Mangione CM, Sorrell-Thompson AL, Miranda JM (2010) A randomized community-based intervention trial comparing faith community nurse referrals to telephone-assisted physician appointments for health fair participants with elevated blood pressure. Journal of General Internal Medicine 25: 701–709.
  111. 111. Bogden PE, Koontz LM, Williamson P, Abbott RD (1997) The physician and pharmacist team: An effective approach to cholesterol reduction. J Gen Intern Med 12: 158–164.
  112. 112. Community Pharmacy Medicines Management Projekt Evaluation Team (2007) The MEDMAN study: A randomized controlled trial of community pharmacy-led medicines management for patients with coronary heart disease. Fam Pract 24: 189–200.
  113. 113. Bond CM, Fish A, Porteous TH, Reid JP, Scott A, et al. (2007) A randomised controlled trial of the effects of note-based medication review by community pharmacists on prescribing of cardiovascular drugs in general practice. Int J Pharm Pract 15: 39–46.
  114. 114. Borenstein JE, Graber G, Saltiel E, Wallace J, Ryu S, et al. (2003) Physician-pharmacist comanagement of hypertension: a randomized, comparative trial. Pharmacotherapy 23: 209–216.
  115. 115. Carter BL, Ardery G, Dawson JD, James PA, Bergus GR, et al. (2009) Physician and pharmacist collaboration to improve blood pressure control. Arch Intern Med 169: 1996–2002.
  116. 116. Fihn SD, Bucher JB, McDonell M, Diehr P, Rumsfeld JS, et al. (2011) Collaborative care intervention for stable ischemic heart disease. Arch Intern Med 171: 1471–1479.
  117. 117. Hunt JS, Siemienczuk J, Pape G, Rozenfeld Y, MacKay J, et al. (2008) A randomized controlled trial of team-based care: impact of physician-pharmacist collaboration on uncontrolled hypertension. J Gen Intern Med 23: 1966–1972.
  118. 118. Jolly K, Bradley F, Sharp S, Smith H, Thompson S, et al. (1999) Randomised controlled trial of follow up care in general practice of patients with myocardial infarction and angina. Final results of the Southampton heart integrated care project (SHIP). BMJ 318: 706–711.
  119. 119. Kaczorowski J, Chambers LW, Dolovich L, Paterson JM, Karwalajtys T, et al. (2011) Improving cardiovascular health at population level: 39 community cluster randomised trial of Cardiovascular Health Awareness Program (CHAP). BMJ d442
  120. 120. Katon WJ, Lin EHB, Korff MV, Ciechanowski P, Ludman EJ, et al. (2010) Collaborative care for patients with depression and chronic illnesses. N Engl J Med 363: 2611–2620.
  121. 121. Lobo CM, Frijling BD, Hulscher ME, Bernsen RM, Braspenning JC, et al. (2002) Improving quality of organizing cardiovascular preventive care in general practice by outreach visitors: a randomized controlled trial. Prev Med 35: 422–429.
  122. 122. McInnes GT, McGhee SM (1995) Delivery of care for hypertension. J Hum Hypertens 9: 429–433.
  123. 123. Simpson SH, Majumdar SR, Tsuyuki RT, Lewanczuk RZ, Spooner R, et al. (2011) Effect of adding pharmacists to primary care teams on blood pressure control in patients with type 2 diabetes: A randomized controlled trial. Diabetes Care 34: 20–26.