Physicians’ intention to leave direct patient care: an integrative review

Background In light of the growing shortage of physicians worldwide, the problem of physicians who intend to leave direct patient care has become more acute, particularly in terms of quality of care and health-care costs. Methods A literature search was carried out following Cooper’s five-stage model for conducting an integrative literature review. Database searches were made in MEDLINE, PsycINFO and Web of Science in May 2014. Results A total of 17 studies from five countries were identified and the study results synthesized. Measures and percentages of physicians’ intention to leave varied between the studies. Variables associated with intention to leave were demographics, with age- and gender-specific findings, family or personal domain, working time and psychosocial working conditions, job-related well-being and other career-related aspects. Gender differences were identified in several risk clusters. Factors such as long working hours and work–family conflict were particularly relevant for female physicians’ intention to leave. Conclusions Health-care managers and policy-makers should take action to improve physicians’ working hours and psychosocial working conditions in order to prevent a high rate of intention to leave and limit the number of physicians actually leaving direct patient care. Further research is needed on gender-specific needs in the workplace, the connection between intention to leave and actually leaving and measures of intention to leave as well as using qualitative methods to gain a deeper understanding and developing validated questionnaires.


Background
Losing physicians from the workforce is a problem in many countries and particularly in those where there is already a physician shortage [1,2]. Research on the physician shortage has found that the causes in many countries are the poor geographical distribution of physicians [3,4], a mismatch between growing medical demands (for example, population ageing, which increases utilization of medical services [5]) and restricted supply (for example, intake rates to medical schools [4,6]) and retaining physicians in patient care until retirement age [4,7,8]. This study focuses on retention. Studies have found that the process of losing physicians from the workforce starts directly after graduation and persists throughout the physician career [9][10][11]. The problem is exacerbated by an increasing proportion of female physicians working part-time during childbearing age and the fact that female physicians are less likely than their male colleagues to have extended working hours [6,8,12]. To prevent actual leaving, reasons for leaving must be better understood. There is a long tradition of review studies focusing mainly on organizational turnover in employees [13], registered nurses [14] and physicians [15]. Less research has been conducted in recent decades on professional or occupational turnover intent, which is the focus of this review. To date, there is little evidence about the risk factors for physicians actually leaving patient care, although one study has shown that intention to leave (ITL) is a strong precursor of actually leaving [2]. Therefore, research on ITL could help identify a high-risk group for early intervention before leaving actually occurs.
As such, we aim to review the literature and summarize factors associated with the intention to leave direct patient care as reported by physicians. This is the first review study on this subject. Special attention is given to potential risk factors for the development of ITL and factors that can potentially be modified, such as working conditions and gender-specific workplace needs.

Review design
An integrative review was conducted according to the steps in the five-stage method recommended by Cooper [16]: formulation of the research problem, data collection, data evaluation, data analysis and presentation of the results.

Search strategy and study selection
The first author conducted the database search. Relevant literature was extracted using a search of the following databases: MEDLINE, PsycINFO and Web of Science. Search terms were combinations of "intent(ion)(s) to" and "leave/withdraw/quit", "leaving intention(s)", combinations of "leaving" and "medicine/clinical care/patient care/the medical profession" and combinations of "quitting/withdrawing" and "intention(s)". All search terms with more than 100 search results were restricted by AND (physician* OR doctor*). The search identified publications with at least one search term in the title or abstract. The earliest publication found was from 1992. As such, it was not necessary to limit the search period to select only recent publications on physician ITL direct patient care. The search covered all articles published prior to May 2014 in English or German. The references within the identified publications were also considered in order to find relevant literature not identified by the database search.

Study selection criteria
Studies were included in the review if the original research was presented, the study population comprised postgraduate physicians, the study focus was the intention to leave direct patient care and the language was English or German. All studies that focused on turnover intentions such as intention to leave the current practice [17] or intentions to quit the current employment [18] were excluded from the review since there would be no net loss of physicians in the workforce in these cases.

Search outcome
The first database search identified 539 hits (104 MEDLINE, 254 PsycINFO and 181 Web of Science), and snowballing procedures identified an additional 20 hits ( Figure 1). After screening the titles and abstracts using the study selection criteria, 42 articles remained. Full-text examination of these articles reduced the number of studies included in this review to 17. Decisions to include or exclude studies were made by the first author under the supervision of the second author.

Framework for analysis
Studies were synthesized using the following criteria: author information, sample, country, medical speciality, study design, career phase, research focus, study  instruments, measurement of ITL, proportion of ITL and variables associated with ITL. To allow readers to evaluate the quality of the studies [19] included in the review, we provide additional details in Table 1 (research  question, response rate, sample demographic, sample size) and Table 2 (research instrument). While we identified some weaknesses in the research instruments, which stemmed from the use of non-standard, self-designed questionnaires, we decided to analyse all studies for the sake of a broader literature review.

Study characteristics
The 12 cross-sectional and 5 longitudinal studies were published between 1998 and 2011 ( Table 1). All studies used quantitative methods, and two included open questions on reasons behind ITL. The studies were conducted in different countries: the UK (seven studies), Germany (four), the US (three), France (two) and Finland (one). While most studies (11) investigated a wide range of specialities, 2 focused on emergency physicians, 3 on general practitioners (GPs) or family physicians, and 1 on surgical fields (surgery, gynaecology and obstetrics). All studies had mixed-gender samples, and the proportion of male physicians ranged from 40.8% to 89.9%. Most studies (10) covered all career phases, while 3 studied physicians in the early career phase, 3 studied mid-career physicians and 1 studied the mid-and late-career phases. The sample sizes in the studies reviewed ranged from 296 to 2926 (Table 2).
Measuring ITL and the percentage of physicians intending to leave direct patient care The reported percentage of physicians with ITL varied from 3.2% [20] to 53.7% [21] across the studies. Three measures were used to determine ITL: prospective measures asking physicians for the likelihood of leaving or intention to leave on a prospective time scale or until retirement (nine), retrospective measures asking for thoughts in the preceding 12 months (five) and universal measures without a time frame (three). The most commonly used prospective measure was "likelihood of leaving direct patient care within five years" (four); here, the percentage of physicians with ITL ranged from 11.8% to 22% (scores of 4 or 5 on a five-point Likert scale). The most frequently used retrospective measure was "how often during the past 12 months/ course of the past year have you thought about giving up the profession/clinical practice" (five); here, the percentages ranged from 17.4% to 25.7% (scores of 3, 4 or 5 on a five-point Likert scale). Overall, the prospective percentages of physicians with ITL were slightly smaller than the retrospective percentages.

Variables associated with ITL
Many different variables are associated with ITL. Therefore, similar explanatory factors were grouped under the following general terms: demographics, family or personal domain, working time and psychosocial working conditions, job-related well-being and career-related aspects. a

Demographic variables
The studies reported significant associations of ITL with having children, ethnicity, medical speciality, age and gender. Physicians from ethnic minorities [22] and emergency physicians compared to hospital physicians [23] had a higher ITL. Female physicians working in paediatrics, anaesthesiology and surgery had lower ITL than those in internal medicine, while the same was true for male physicians in surgery and orthopaedics compared to those in internal medicine [24]. Lower ITL was also found in physicians with children under the age of 18 [22,25].
The demographic aspects of age and gender were discussed frequently in the studies. The next two paragraphs summarize these discussions.

Age-specific findings
The impact of age on ITL was examined in eight studies. Three found null association with samples of physicians in their first or second postgraduate year [24] or primarily after the age of 45 [26] or primarily before the age of 45 [23]. Five studies with age ranges from postgraduate to retirement age found mixed results. Two studies found that younger physicians had a higher ITL [21,27]. The relative frequency of ITL at different ages was presented in one study [21], which indicated low ITL for very young physicians, the highest ITL for those aged 33-35 and decreasing ITL for physicians after the age of 50. Conversely, three studies indicated that older physicians had a higher ITL than those under the age of 35 [22,25,28].

Gender-specific findings
Gender-specific analyses were conducted in seven studies. Two UK studies found that female physicians were more likely than their male colleagues to intend to remain in the National Health Service (NHS) [12,29]. Another study showed that male physicians with a partner and children under the age of 16 were less likely to indicate ITL [24].
Further gender-specific findings were identified for many of the explanatory factors. We describe these in the following after the general findings for each factor.

Family or personal domain
Evidence of a connection between work-family conflict and ITL was observed in two studies. High work demands

Cross-sectional, Quantitative
All career phases Clarify the relationships between ITL, overall job satisfaction, domains of job satisfaction and personal and job characteristics by using a structural recursive model  "Likelihood that they would leave direct patient care within the next 5 years" Five-point scale ("none" to "high") 139 (11.8%) rated the likelihood of ITL within next 5 years as "likely" or "high" "Considering leaving medicine but remaining in the UK" and "considering leaving medicine and the UK" were grouped to "considering leaving medicine" Five-point scale ("yes, definitely" to "no, definitely not") 42 of the 1326 physicians analysed (3.2%) were "probably not" or "definitely not" going to continue to practise medicine Intention to leave medicine: "Whether or not they hoped to leave medical practice for another career in the next 5 years" (1 = "yes", 0 = "no") Not specified that interfered with family duties were associated with higher ITL [30]. Of all the reasons behind ITL, 26.1% belonged to the category "family reasons, time for leisure and other interests" and 19.6% belonged to the category "maintaining good health and wanting a healthy retirement" [29]. Sleeping problems were also associated with higher ITL [27]. Gender-specific findings showed that female physicians were more than twice as likely as their male colleagues to cite "family reasons, time for leisure and other interests" as a cause of ITL [29]. ITL was also higher for female hospital physicians with poor sleep quality [28], male emergency physicians and male hospital physicians who worried about making mistakes [23,28].

Working time and psychosocial working conditions
Four studies reported an association between factors related to working time and ITL. Physicians with oncall duties and fewer breaks were more likely to indicate ITL [21]. One major reason for ITL was long working hours [20].
Three studies reported gender-specific findings. Female physicians reported higher ITL with 50 or more working hours each week [23], weekly overtime of 5 to 10 h [24] and long working hours [21]. Conversely, male physicians with fewer weekend duties reported lower ITL [24].
A variety of psychosocial working conditions were associated with ITL. A high degree of job control was associated with lower ITL; this factor also mitigated the positive effects of distress and sleeping problems on ITL [27]. Low quality of teamwork [28] and little or no work-related support from colleagues and superiors [24] increased ITL. Higher ITL was associated with high psychosocial stress (measured by effort-reward imbalance and job strain) [31], non-permanent positions [21] and dissatisfaction with the ability to choose method of working [25]. More than a third of physicians (36.5%) reported that the main reason for ITL was work-related pressure [29], while 35% cited poor working conditions in general [20].
Gender-specific results: ITL was higher for male emergency physicians with a low quality of teamwork [23], male hospital physicians reporting hostile relations with the administration [28] and men with fewer opportunities for postgraduate training [24]. Among women, ITL was higher for emergency physicians reporting monthly or more frequent harassment by superiors [23] and those with a low influence at work [23] and a lack of performance feedback [24].

Job-related well-being
Several studies reported an association between job dissatisfaction and higher ITL [2,22,25,32]. The lack of job satisfaction was also regarded as an important reason behind ITL [20,29]. Another study found that job satisfaction mediated the link between perceived stress and ITL [32].
Four studies investigated mental health aspects. Higher ITL was linked with a high general burnout among emergency physicians [23]; patient-related burnout among hospital physicians [28]; poorer mental health as measured by depression, anxiety and burnout [32]; and high levels of psychosocial distress [27].
Gender-specific findings on high ITL were reported for female emergency physicians with high patientrelated burnout [23] and female hospital physicians with high general burnout [28].

Career-related aspects
Links were reported between career-related aspects such as job opportunities, salary and career satisfaction and ITL. Emergency physicians with job opportunities outside the health-care sector [23] and less salary progression reported higher ITL [26].
Gender-specific results for high ITL were found for male hospital physicians with job opportunities within and outside the health sector, women with a low salary satisfaction [28] and male emergency physicians with a lower career satisfaction [26].

Associations between ITL and actually leaving
Two studies examined ITL and physicians actually leaving medical practice. A national survey of family physicians in England collected data (N = 1,174) on physicians' ITL within the next 5 years [2]. The prevalence of ITL at the baseline survey was 11.8%, and the incidence of leaving direct patient care at the 5-year follow-up survey was 16.5%. The study showed that a higher ITL was associated with a greater likelihood of actually leaving.
In a survey of specialist physicians (N = 968) in California, data were collected on projected working status in 3 years time [32]. At the baseline, 8.8% of physicians reported ITL, and 10.6% (N = 101) of the initial sample was defined as having left clinical practice after 3 years.

Discussion
In this review, we identified 17 studies of physicians' intention to leave direct patient care from five countries and synthesized the study results. Variables associated with ITL were demographics, with age-and genderspecific findings, family or personal domain, working time and psychosocial working conditions, job-related well-being and career-related aspects. Three kinds of ITL measures were used: prospective, retrospective and universal. The share of physicians reporting ITL varied considerably between studies. The studies reviewed were diverse in terms of their aims, study design and instruments, making direct comparisons challenging.

Critical consideration of the studies
Country-specific effects, such as different incentives and disincentives for leaving, could explain the diverse findings on age and ITL; for example, the studies with a high ITL among young physicians were from Germany [21] and Finland [27], while the studies reporting a high ITL among older physicians were from the UK [22,25] and France [28].
Different measures of ITL were used ( Table 2) and different scale categories were determined for identifying ITL. Studies that determined ITL on a five-point Likert scale with scores of 3 or 4 [32] or on a three-point Likert scale with a score of 2 [21] should be interpreted with caution, as these scores included categories that indicated only a moderate ITL.
In the studies on ITL and actual leaving, the definition for actual leavers was not restricted to permanent leavers during the working-age period [2,33]. The results should be interpreted carefully in the context of these studies; for example, in the US study, most of the actual leavers were likely to have left due to normal retirement, as 9.9% of all physicians in the initial sample were aged 65 or over.

Limitations of the review
Search terms including "turnover" were excluded from the review because of the broad definition of turnover intentions [34] and the problem of separating turnover intention from intention to leave a job and intention to leave the profession. As our aim was to identify risk factors for intention to leave direct patient care, we decided not to include turnover intention in order to obtain rigorous results on the intention to leave direct patient care.
Decisions to include or exclude studies were made by two authors, whereas the database search was conducted by just one author. Therefore, the validity of the initial records included was not verified by a second author.
In two studies [12,29], Davidson investigated the intention to practise in the NHS until normal retirement age. In the UK, it is possible to work as a physician in the NHS or in private practice. This makes it possible for physicians to leave the NHS and switch to private practice. In the study by Davidson [29], the analysis of the reasons behind intending to leave before retirement age makes it clear that changing to private practice was not a reason for considering early retirement. Davidson [12] researched a different cohort of physicians from the same database, but the reasons behind considering early retirement are not known for these physicians. This last study by Davidson [12] was not excluded from the review, as the two cohorts differed by just 3 years. Furthermore, most physicians in the UK tend to work predominantly for the NHS and take only part-time work in private practice, if at all. This highlights that retiring from the NHS early and shifting to private practice may not be an attractive option, as this leaves too little time to amortize the additional investment and operating costs.
This shows that the differences in health-care systems, working environments and physician career systems in the countries investigated could lead to research bias. Further research is needed for a systematic examination of the prevailing conditions in each of these countries to reduce this potential bias. Another limitation concerns the small number of countries (Germany, the UK, France, Finland and the US) included in this review and the absence of low-and middle-income countries. This is partly due to the requirement that articles be in English or German, which introduced language bias.

Recommendations for future research
Unfortunately, we were unable to identify a relevant study that used qualitative methods to conduct an indepth analysis of the personal perspectives of and context behind physicians with ITL [35]. In-depth qualitative analysis could provide an overall picture of reasons behind ITL and lead to the development of a standard and rigorously psychometrically tested instrument for this research topic. Most of the studies reviewed used self-developed instruments.
Secondly, attention should be paid to how ITL is measured. It is not clear which kind of ITL measure has better explanatory power when it comes to linking risk factors and actual leaving. This methodology decision to use a prospective, retrospective or universal ITL measure is an issue for further research.
Thirdly, studies in the review had diverse aims and study designs, resulting in different sets of variables being considered in this analysis. Therefore, no single study included all of the job and personal variables that have been attributed to ITL. To present the results of similar factors together, we grouped the key findings into several categories (see the "Results" section above). In reality, all of these factors are correlated and are difficult to differentiate from each other. According to earlier findings [25], overall job satisfaction combined with job and personal characteristics was better able to predict the outcome than any single variable set. In future, a wide range of all relevant predictors should be considered in ITL analyses.
Finally, gender-specific analyses were conducted in seven studies and revealed that men and women often had different risk factors under similar circumstances; for example, in terms of psychosocial working conditions, higher ITL was most closely associated with a low quality of teamwork for men and a low influence at work for women. This makes it clear that men and women have different workplace needs. Deeper exploration of gender differences is warranted on this topic.

Conclusions
This integrative review provides an overview of postgraduate physicians' intentions to leave direct patient care. Early intervention on risk factors for ITL could prevent physicians from actually leaving. Based on the findings of this current review, starting points for intervention could focus on work-family conflict [29,30], especially for women [29]. Factors related to working time also appear to be important and, again, particularly for women [21,23,24]. Health-care managers and policymakers should consider working time and work-family conflict as priorities when it comes to retaining physicians. These issues become even more critical as female physicians form an increasingly large share of the physician workforce [4][5][6]. There is also much evidence to indicate that job stress caused by adverse psychosocial working conditions increases ITL. Health-care managers, hospital administrators and policy-makers could address this ITL driver by improving psychosocial working conditions for the medical profession. The success of such interventions has been observed with long-term effectiveness for a wide range of psychosocial work factors and well-being among health-care professionals [36].