Study selection

The search results and review process are presented in figure 1, using the Preferred Reporting Items for Systematic Review and Meta-analyses (PRISMA) flow chart. The search performed on Medline (n=3395), Embase (n=1913) and Emerald Management (n=454) yielded a total of 3926 articles after the removal of 1849 duplicates. To test inter-rater reliability, we used Cohen’s kappa and found high levels of agreement between the paired reviewers for the 3% double review, K=0.78 (p<0.0001) for pair 1 and K=0.88 (p<0.0001) for pair 2. The remaining screening of title and abstract resulted in 113 studies eligible for full-text review. The full-text review stage leads to the inclusion of 22 quantitative, qualitative or mixed method studies in the penultimate pool. After an additional exclusion criterion was added, requiring that included studies provide data to allow a comparison of medical and non-medical leadership, eight more studies were eliminated. The literature search identified five literature reviews on topics associated with medical leadership.8 14 28–30 The references of these five reviews were searched for additional studies that met inclusion criteria. Two additional studies were identified as a result of this process, resulting in a final inclusion of 16 studies.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Search and review strategy (PRISMA flow diagram).

Study characteristics

The characteristics of the 16 studies that met inclusion criteria are presented in online supplementary table 1. Fourteen studies conducted quantitative analysis: ten studies analysed questionnaire survey data7 31–39; one study (Jiang et al33) also provided processes of care data, and another provided hospital performance data (Saleh et al38). Two studies analysed US Statewide Health Planning and Development data from California,40 41 one study analysed data from the AMADEUS database and hospital and insurance documents42 and another study analysed UK hospital trust data.43 One study conducted a qualitative analysis of interview and observation data,44 and another conducted a mixed method analysis,45 which combined findings from a review of hospital documents with qualitative analysis of interviews with hospital Chief Executive Officers (CEOs) and board members. Seven studies were from the USA; the remainder were from Finland (2), Germany (2), the UK (1) Ireland (1), Norway (1), Lebanon (1) and Australia (1).

Results of individual studies

Six studies reported on the performance of hospital boards.33 40–43 45 Veronesi et al43 examined the impact of clinicians appointed to the boards of 102 English NHS hospital trusts on quality of hospital care provided from 2006 to 2009. Composition of boards was determined from hospital trust annual reports. Compliance with health and well-being, clinical effectiveness, safety and patient focus, and ease and equity of access care standards was obtained from the UK Healthcare Commission and Dr Foster (a commercial provider of healthcare benchmarking data). A greater percentage of doctors on boards was associated with a better-quality rating of service providers. Trusts achieving a four rating had an average of 15.01% of directors with a medical background, whereas in trusts achieving only a one rating, 11.09% board directors were doctors. This finding was confirmed in relation to lower morbidity rates and tests to exclude the possibility of reverse causality, whereby doctors joined the boards of better performing trusts. No equivalent association was found for clinical professions such as nurses and other allied health professions.

De Andrade41 investigated whether having board members with medical expertise in 281 USA hospitals affected the levels of uncompensated care provided. A quantitative analysis of data from the California Office of Statewide Health Planning and Development between 1997 and 2010 found that doctors’ board membership was not related to provision of uncompensated care, except when the hospital’s ownership status was taken into account. Relative to non-profit and public hospitals, for-profit hospitals provided more uncompensated care, the higher the percentage of doctors on the board. For an average for-profit board size of 10 members, substituting one member by a doctor increased the amount of uncompensated care provided by 19%.

On the theme of boards, Prybil45 sought to determine whether high-performing and mid-range hospitals differ in board structures, processes and practices. High-performance hospitals included at least three of the Solucient Center for Healthcare Improvement’s ‘100 Top Hospitals’ in 1999–2003. A mixed method analysis of hospital documents and interviews with hospital CEOs and board members was conducted for seven matched pairs of USA hospitals, finding that doctors form a larger component of the boards of high-performing hospitals (30.3%) than of mid-range hospitals (20.8%). Doctors comprised 25% or more of the boards’ voting members in five of the seven high-performing hospitals but only one mid-range hospital.

Jiang et al33 examined whether differences exist in quality performance of 490 USA hospitals in relation to adoption of particular practices in board oversight of quality. Data consisted of (1) a survey of 562 hospital CEOs on board practices, (2) process of care measures for three clinical conditions (heart attack, heart failure, pneumonia) and (3) outcomes measures, consisting of risk-adjusted mortality rates, for the same three conditions. Sixty per cent of participating hospitals had a Chief Medical Officer or Vice President of Medical Affairs on the committee; this resulted in significantly higher process of care scores (85.3% vs 81.0%, p<0.05) and lower risk-adjusted mortality rates (5.6% vs 7.3%, p<0.05) than hospitals that did not have a Chief Medical Officer or Vice President of Medical Affairs as a committee member. Eighty-three per cent of participating hospitals had medical staff on the committee; this resulted in significantly higher process of care scores (84.2% vs 80.9%, p<0.05) but no difference in risk-adjusted mortality rates. Sixty-three per cent of participating hospitals had a clinical board member on the committee; this resulted in no difference in process of care scores but significantly lower risk-adjusted mortality rates (5.7% vs 7.2%, p<0.05).

Bai and Krishnan40 examined whether hospitals without medical participation on their boards of directors delivered lower quality of care in 142 non-profit hospitals in the USA. Quantitative data were obtained from the US Hospital Quality Alliance and the California Office of Statewide Health Planning and Development. The study found that boards without medical members were associated with a decrease of three to five percentage points in quality of care for heart failure, pneumonia and surgery infection prevention.

Moving to Germany, Kuntz et al42 examined differences in the financial performance of hospitals with regard to ownership by studying the size and composition of supervisory boards in 175 hospital companies operating 246 hospitals (14% of all German acute care hospitals in 2009). The study reported on a quantitative analysis of hospital financial performance data (from the AMADEUS database) and information on hospital and board characteristics (from business and quality reports, hospital websites and health insurers). Data were obtained from all participants in 2009 and from a subsample of 163 hospital companies in 2010. Financial performance was based on four measures: return on assets (ROA), earnings before interest and tax margin, total profit margin and net income. Doctors comprised, on average, 11.7% board members. Financial performance and board size and composition depended on ownership (p<0.01 for ROA and p<0.001 for the other four performance measures). An increase in board size and greater political participation were negatively associated with all five tested measures of financial performance, an increase in nurse and economist participation was negatively associated with financial performance and no associations were found for clerical participation. An increase in physician participation, however, was positively associated with a 5% increase in ROA (p=0.061).

Two studies reported on doctors’ involvement in strategic decision-making.37 38 Parayitam et al37 examined the self-reported outcomes of decisions when physician executives were involved in strategic decision-making processes in 109 USA hospitals. Hypotheses were that increased numbers of doctors involved in strategic decision-making teams would be associated with better decisions, greater understanding of the rationale of decisions and more commitment to decisions. The sample of 114 CEOs and 254 strategic decision makers (executive officers, director of human resources, chief technical offices, chiefs of staff, personnel involved in facilities, maintenance) completed a survey reporting their decision quality, understanding and commitment. Structural equation modelling of the data suggested that the ratio of doctors was positively correlated with decision understanding, commitment and quality.

Saleh et al38 explored the use of strategic planning processes in 79 (56.4%) Lebanese hospitals and investigated its association with financial performance. Hypotheses included that the level of physician involvement in the strategic planning process is positively associated with hospital performance. Quantitative analysis of survey data on hospital-reported participation in strategic planning processes and hospital performance data from the Lebanese Ministry of Public Health (occupancy rate and revenue per bed) found that there was no association between the level of physician involvement in the strategic planning process and hospital outcomes; generally, physician involvement was low (4.1 out of a possible score of 7).

Two studies reported on medical leadership of organisations.7 32 In the USA, Colla et al7 measured how doctor-led organisations compared with other ACOs in terms of structure, size and care provided and explored the degree of doctor engagement in managing ACOs. ACOs are groups of providers that are jointly responsible for caring for a nominated population of patients. Fifty-one per cent of the 173 ACOs in the study self-identified as doctor-led, 33% self-identified as jointly led by hospital and doctor and the remainder were led by hospitals or other entities. Doctor-led ACOs were found to be more likely to have advanced IT capabilities and better outpatient care than non-doctor-led ACOs and were more likely to measure and report financial performance at practice and clinicians levels; however, this finding was confounded by the fact that doctor-led ACOs are less likely to include hospitals and more likely to include physician groups. Performance of ACOs was not assessed; therefore, it was not possible to determine whether performance was related to whether leaders were doctors or non-medical managers.

Goodall32 examined the relationship between hospital performance and whether the CEO was a doctor or a non-medical manager. Hospital performance was determined by media-generated league tables, produced by US News and World Report’s ‘Best Hospitals’ in 2009. Three hundred healthcare executives from three specialties (cancer, digestive disorders, heart and heart surgery) were surveyed in the top 100 USA hospitals. Positive association was found between physician CEOs and hospital performance for all three hospital specialties (p<0.001). While higher-performing hospitals were associated with physician CEOs, causation was not able to be determined (eg, higher-performing hospitals may just prefer to have doctors as leaders).

Five studies reported on medical leadership within organisations.31 34 35 39 44 Konu and Viitanen34 investigated the incidence of shared leadership among 433 middle-level managers (eg, chief doctors, nursing directors) in social service and healthcare in Finland. Quantitative analysis of survey data on leadership practices found that shared leadership practices were more common among managers without a medical background.

Another study sought to determine whether evaluations on the impact of knowledge sources affecting their decision-making differ depending on the manager’s professional background, activity sector, gender, management experience or age, among 404 middle-level social and healthcare managers in a Finnish hospital (Simonen et al39). Quantitative analysis of survey data revealed that managers who were also doctors more strongly perceived that their decision-making was influenced by their own professional experience, journals and scientific research within their own professional field and nationwide interaction within their own profession. Differences were found between doctor managers and nurse managers with respect to organisation documents and publications, which carried more weight in nurse managers’ decision-making. No differences were found between managers of different professional backgrounds regarding other knowledge sources.

In other work, Agarwal et al31 sought to determine whether there is a positive association between Management Practices Score (MPS) and the level of clinical education of managers in 42 Australian acute care public hospitals. The MPS for each hospital was calculated from interview responses of 116 managers to a survey of 21 hospital management practices across multiple dimensions. The study found no association between the number of doctors in each hospital and the MPS (p=0.779). The coefficient on the level of skills and education within hospitals, however, was positive and significant (p=0.06), indicating that hospitals with a higher proportion of clinically qualified and skilled managers perform significantly better in management practices.

Spehar et al44 investigated how clinicians’ professional background influences their transition into the managerial role and identity as clinical managers in four public hospitals and two health trusts in Norway. Interviews were conducted with 30 clinician managers (13 doctors), 20 of these were also observed in management and staff meetings during the day. Qualitative analysis of interview and observation data revealed that doctors experienced difficulties in reconciling the clinical and management role and used clinical work to gain legitimacy and respect from medical colleagues. In contrast, nurses experienced a faster and more positive transition into the manager role and were more fully engaged in the managerial aspects of the role.

Moving countries again, Kuntz et al35 assessed the relationship between the amount of medical involvement in leadership and staff-to-patient ratios. The study was conducted in Germany, where hospitals are managed by an executive leadership team consisting of a commercial director, an MD and a nursing director. The study was controlled for the size and case mix of the 604 participating hospitals, whether they were public or private, the degree to which they were rural and whether the doctors were salaried or contracted. High staff-to-patient ratios for both nurses and doctors are associated with better hospital performance.46 47 The study found a relationship between full-time MD or heavily involved part-time MD and a higher staff-to-patient ratio. Full-time MDs significantly improved the staff-to-patient ratios for both doctors and nurses (doctors: 1.96, p<0.01; nurses: 4.44, p<0.01), whereas part-time MDs only improved the staff-to-patient ratios for doctors (eg, an increase of part-time involvement from 15% to 25% resulted in an increase of 2.49 doctors per thousand inpatients).

Staying in Europe, O’Keefe36 measured and compared risk aversion in 788 Irish clinicians, clinical managers, non-clinical managers and non-clinical public representatives, in terms of willingness to discharge a patient from the emergency department. The study found no significant difference between clinicians and clinician managers but found significant differences between clinicians and non-clinicians (including between clinical and non-clinical managers), with the non-clinical participants being more risk averse. Limitations included the following: (1) there was a large variation in risk tolerance, even between clinicians; (2) it was a single study in a single country; and (3) clinicians are likely to be more familiar both with actual events depicted in the scenarios and also with the process of making treatment choices that may result in death. The study did not provide indication of an objectively appropriate level of risk, but the authors suggested that the clinicians had a more pragmatic approach to decision-making.

Risk of bias

While studies were generally well designed and executed, generalisability and implications, and usefulness of the findings, was low (table 2).26 There was a risk of bias across studies, in that the majority (12) studies collected self-reported data on aspects of medical management, for example, surveys or interviews, rather than objective data. Many of the studies did not report ethics approval, study funding or authors’ conflicts of interests.

Study synthesis

Extracted data were cross-checked for accuracy and coded by four researchers (ZL, KL LT and RC-W). One researcher (RC-W) completed a narrative synthesis of the coded data, by linking the data to form five inter-related themes: impact of medical leadership on outcomes; doctors on boards; contribution of qualifications and experience; the leader as an individual or part of a team; and doctors transitioning into the medical leadership role. Four broad categories of articles were found in the literature search: (1) individual perspectives on medical leadership (the majority of excluded articles), frequently based on personal experience; (2) empirical research based on self-reported data about medical leadership (surveys, interviews, focus groups); (3) objective empirical research on the role or characteristics of healthcare leaders (observations, other data) and (4) objective empirical research on the relationship between medical leaders and outcomes (hospital performance data, patient outcomes).