The desire and capability of Australian general practitioners to change their working hours

The Australian Institute of Health and Welfare (AIHW) estimated that the number of primary care practitioners in 2009 was 25 707, a significant increase from 20 616 in 1999.1 While general practitioner numbers have risen, the number of hours worked by a typical GP has fallen. The AIHW Medical Labour Force Survey estimated that, over the same period, the mean weekly hours worked fell from 45.6 to 42.2. In the context of an ageing population, the demand for GPs is likely to continue to rise over time. Recent workforce policy established a number of new medical school places to increase the supply of medical graduates. However, the effects on the availability of GP services will be reduced if working hours continue to decline. Based on existing patterns of demand growth and supply, Health Workforce Australia has predicted a shortfall in doctors in 2025, with an increased reliance on poorly coordinated skilled migration to meet workforce requirements.2 The issue of GP labour is particularly acute in rural areas. In 2010, the National Primary Health Care Strategic Framework stated that the number of GPs per 100 000 population varies from under 60 in very remote Australia to almost 200 in major cities.3

The workload of a GP is likely to be a function of a number of factors. Preferences for reducing the numbers of hours worked, or for changing the type of work undertaken, are likely to depend on a range of characteristics of the doctor, such as age, job satisfaction, and family circumstances. Job satisfaction of Australian doctors is correlated with good support networks, realistic expectations, self-assessed health and income, among other things.4 These findings are largely corroborated in the international literature.5 In the United States, it has been demonstrated that dissatisfaction with a medical career was a strong predictor of reducing hours per week below 20, and of retiring altogether.6

Thus, a better understanding of doctors’ work-participation intentions and how they change their labour supply is a sound basis for developing workforce policy. We aim to investigate two linked issues in the determination of GP working hours. First, by using a 2010 cross-section of an ongoing longitudinal study, we consider the stated preference for changing working hours, to identify the factors influencing GPs’ preference for reducing working hours. Second, using the longitudinal nature of the data, we explore whether doctors were able to change their working hours in line with their previously stated preference. The separation of the issue into two components is valuable, as it helps to identify whether changing working hours is being driven by doctor preferences or by external factors that help or impede doctors to reduce their time commitment to care delivery.

The data sources for our analysis were the two most recent waves (Waves 3 and 4, conducted in 2010 and 2011 respectively) of the Medicine in Australia: Balancing Employment and Life (MABEL) survey.7 It has been employed to explore, among other things, the determinants of doctors’ earnings, work–life balance, and junior doctors’ preference for specialty.8-10 The Wave 1 (2008) panel of this survey consisted of a census of 54 750 doctors listed in the Medical Directory of Australia.11 Of those invited to participate, 10 498 responses were collected. The survey has been administered in three further years, allowing longitudinal exploration of the data. In subsequent waves, those who had responded to previous waves and doctors new to the Directory were invited to participate. The total number of invitations in Waves 3 and 4 were 16 327 and 15 967, respectively, yielding 9949 and 9773 responses. These consisted of both GPs and hospital-based doctors, of whom 30%–40% were GPs.

Three modifications were made to the dataset before analysis. Annual net personal income was converted into quintiles (defined as < $60 000, $60 000–$84 999, $85 000–$119 999, $120 000–$155 999 and ≥ $156 000). A sixth category for those who did not disclose income was included. Second, family circumstances were considered using a composite variable that included both whether the GP had children and whether they had a partner. The composite variable took on a value representing each of the four possible combinations of these two variables. Third, the number of hours currently worked was grouped into categories of < 20, 20–39.5, 40–59.5 and ≥ 60 hours per week.

The determinants of desire to reduce workload used Wave 3 data collected in 2010. The dependent variable was whether or not the doctor stated a desire to reduce hours. As only 3% of the sample stated a willingness to increase their hours, we combined those responses with those who did not want to change. The second part of the analysis focused on those GPs who stated a desire to reduce their hours in Wave 3, and their subsequent change in hours between Waves 3 and 4. We selected a criterion of reducing weekly working hours by 5 or more.

The independent variables were age (linear and quadratic); sex; self-assessed health; family circumstance; work satisfaction; weekly working hours; whether the GP was required to do on-call work; the business relationship the GP had with their main practice (principal or partner, associate, salaried, contracted, locum, other); whether the practice was metropolitan, inner regional or other (outer regional, remote, very remote); income quintiles; and measures of density of GPs (defined as number of GPs per 1000 population) and composition of patients (the proportion of patients aged < 5 years or > 65 years). In addition to running a pooled analysis with both male and female GPs, separate analyses were run for each sex.

We used ordinary logistic regression using Stata version 12.1 (StataCorp), using weights included in the MABEL survey to ensure representativeness of the sample to the Australian GP population. Sensitivity analyses to test the robustness of the findings to the decision to limit to the most recent waves of data (ie, Waves 3 and 4) and to the use of a 5-hour cut-off to denote a reduction in hours resulted in similar results (available on request).

Of the 3377 GPs from Wave 3 included in the analysis, 43% stated a desire to reduce hours. The regression results are presented in Box 1. Age was a strong positive predictor of desire to reduce hours (P < 0.01), although the negative effect for age squared suggests that the increase slows and even reverses in higher age groups. Female doctors were more likely to want to reduce hours (odds ratio [OR], 1.36; 95% CI, 1.11–1.67; P < 0.01). Those who reported excellent health were less likely to want to reduce hours than those reporting good health (OR, 0.59; 95% CI, 0.45–0.76; P < 0.01). Those who were satisfied (OR, 0.50; 95% CI, 0.31–0.79; P < 0.01) or very satisfied (OR, 0.17; 95% CI, 0.11–0.28; P < 0.01) with their work were less likely to want to reduce hours relative to those who were neither satisfied nor dissatisfied. Working hours strongly predicted the desire to reduce hours; relative to a GP working 20–39.5 hours per week, a GP working ≥ 60 hours per week was over six times more likely to want to reduce hours. Partners were most likely to want to reduce their working hours. Associates, contracted GPs and locums were all less likely to want to reduce their hours (all P < 0.01) compared with partners. There was a weak income effect — doctors in the higher quintiles were more likely to state a desire to reduce their hours, although statistical significance was only achieved for the difference between those in the first and third quintiles (ie, doctors earning < $60 000 per year compared with those earning $85 000–$119 999 per year).

When the analysis was run for each sex separately (results available on request), the notable age effect was driven by male GPs. That is, the association between age and stating a preference to reduce working hours was very strong in males (P < 0.01 for both the linear and quadratic terms), but not in female GPs. Middle-aged men were most likely to state a desire to reduce working hours, with a maximum desire occurring at 56 years of age.

Two thousand seven hundred and twenty GPs had complete working-hours data for both Waves 3 and 4, 1177 of whom had stated a desire to reduce their hours in Wave 3. Of these, 316 (26.8%) successfully reduced their weekly hours by more than 5. Of the 861 doctors who were unable to reduce their hours despite stating that preference, 647 (75.1%) continued to state a desire to reduce working hours in Wave 4, suggesting that their failure to reduce hours was not due to changing preferences, but to external factors preventing a reduction in working hours. The regression results are reported in Box 2. The ability to reduce hours was predicted by three factors: age, sex and the number of hours currently worked. Regarding age, there was a U-shaped relationship suggesting that middle-aged doctors (around 50 years of age) are the least likely to achieve desired reduction in hours. Female doctors were more likely to successfully reduce hours (OR, 1.91; 95% CI, 1.32–2.75; P < 0.01). Doctors working longer hours were more likely to successfully reduce hours. For doctors working 40–59.5 hours per week, the OR was 1.97 (95% CI, 1.30–3.00; P < 0.01) and for those working ≥ 60 hours per week, the OR rose to 7.71 (95% CI, 4.47-13.31; P < 0.01).

One potential issue with this type of analysis is the generalisability of the result to the Australian GP population. The population weights supplied in the MABEL survey ensure generalisability in terms of age, sex and location only. However, the MABEL survey represents a relatively large proportion of the Australian GP population, reducing the risk of volunteer bias. Additionally, the GP characteristics over which the MABEL survey is population-weighted represent those for which we are likely to have whole-population data. Our results also show that most of the GPs who wish to reduce their working hours fail (or are unable) to do so within the next year. The key predictors of reducing hours are age, sex and current working hours. Middle-aged doctors are least likely to reduce their hours. Female GPs are more likely to reduce hours if they want to do so. These results are robust to other factors that might explain these relationships (eg, practice ownership, family circumstance and income). This may be due to different motivations. Family constraints are more pressing for younger GPs and female GPs, while older male GPs are contemplating moving to retirement gradually. The variables that matter in predicting GPs’ success in reducing hours are either working hours themselves or demographic characteristics that cannot be changed through policy levers. The implication of this is that policy designed to retain labour supply in the health workforce should focus on the predictors of the desire to reduce hours (Box 1).

There is an increasing emphasis on the development of policy based on research evidence, but to date there has been little insight into doctors’ changing work patterns. Our results show that improved GP job satisfaction is associated with less desire to reduce working hours. This implies that further investigation of policies that can improve GP job satisfaction may well ensure improvements in the doctor supply, which are needed to meet the increasing demand for primary care into the future. Meeting the needs of the population for primary health care and supporting those providing this care represents an important balance that must be made, and an area of ongoing challenge and opportunity.