Portal de Periódicos da CAPES

Medicine in Australia: Balancing Employment and Life (MABEL)

2011; Wiley; Volume: 44; Issue: 1 Linguagem: Inglês

10.1111/j.1467-8462.2010.00627.x

1467-8462

Wenda Yan, Terence Chai Cheng, Anthony Scott, Catherine Joyce, John H. Humphreys, Guyonne Kalb, Anne Leahy,

Primary Care and Health Outcomes

Addressing the shortages of the health workforce is an important policy issue in many countries (World Health Organization (WHO) 2006). In Australia, problems such as an overall shortage of health workers as well as their maldistribution by specialty, sector, profession and geographical area have been documented (Australian Health Workforce Advisory Committee (AHWAC) 2004; Australian Medical Workforce Advisory Committee (AMWAC) 2004; Productivity Commission 2005). Shortages can have profound effects on aspects of the health system's performance, including costs, access and quality of care. These problems are set against a background of an increasing proportion of female medical graduates (now over 50 per cent) and trends in falling hours of work among both male and female doctors. Understanding the decisions made by doctors in terms of where they locate, which specialty they choose, which sector they work in, and their drivers for hours worked and labour force participation is crucial for being able to design policies to change behaviour and improve efficiency and equity in the health care sector. There are no national datasets in Australia that contain the range of data required to fully understand these decisions. Research using robust longitudinal data is necessary as the basis for generating evidence of direct relevance in developing more effective medical workforce policies. The Medicine in Australia: Balancing Employment and Life (MABEL) survey is a longitudinal panel survey of Australian doctors launched in 2008. The aim of the MABEL survey is to investigate factors influencing workforce participation, labour supply, specialty choice and mobility of doctors. The MABEL survey is advised by its national Policy Reference Group1 to ensure the survey is informed by current policy issues and priorities, and to assist in the translation of findings into the policy context. The data from Wave 1 of the MABEL survey (2008), released on 1 June 2010, comprise a cohort of 10 498 doctors working in clinical practice in Australia, which represented more than 19 per cent of the underlying population (Joyce et al. 2010). Studies based on data from Wave 1 present initial findings on a variety of policy-relevant issues (for example, Cheng et al. 2010; McGrail et al. 2010; Sivey et al. 2010; Joyce et al. forthcoming)2. The data from Wave 2 (2009), recently released, comprise a sample of 10 304 doctors, including 8180 doctors from the baseline cohort in Wave 1. The aim of this article is to describe the methods of the MABEL survey, summarise response rates and response bias of both Wave 1 and Wave 2 surveys, and compare the two waves in terms of attrition. The population of interest in the MABEL survey is doctors providing clinical medical services in Australia. The sampling frame of the MABEL survey is the Australian Medical Publishing Company's (AMPCo) Medical Directory. The construction of the MABEL sample in the first two survey waves is summarised in Figure 1. During Wave 1 in 2008, 54 750 doctors who made up the population of doctors working in clinical practice in Australia were invited to participate. A total of 10 498 doctors undertaking clinical practice responded, an overall response rate of 19.36 per cent (Joyce et al. 2010). These doctors formed the baseline cohort in the subsequent waves. The Construction of the MABEL Survey Sample For each wave, a top-up sample of doctors not previously invited to participate is included to address the cross-sectional under-coverage of doctors who are new to the medical workforce of the year. Each top-up sample comprises all doctors new to the AMPCo database between each wave; mainly new medical graduates, international medical graduates working in Australia for the first time, and doctors who rejoined the medical workforce after a period of temporary leave (for example, maternity leave or moving back from overseas). The sample in Wave 2 (2009) comprises 15 871 doctors: 10 251 continuing doctors who responded in Wave 1 and a top-up sample of 5620 new doctors. A total of 247 doctors from the baseline cohort in Wave 1 could not be sent a survey in Wave 2 based on updated information provided by AMPCo that they were either not in clinical practice or not contactable.3 According to the AMPCo database as of May 2009, the total population of doctors providing clinical medical services in Australia was 57 565 and thus the Wave 2 sample accounted for 27.6 per cent of the population. Of the 15 871 doctors, 975 (505 new doctors and 470 returning doctors) were surveyed as part of the Wave 2 pilot survey and were included in the Wave 2 MABEL cohort because the survey content changed little as a result of the pilot. Overall, 10 304 doctors responded in Wave 2, with response rates of 79.8 per cent for returning doctors and 37.8 per cent for new doctors. At least four annual waves of the survey are planned; that is, until 2011. From Wave 2 onwards, the initial baseline cohort of doctors who responded in Wave 1, as well as respondents from the top-up samples included in previous waves, will be included in the sample to be surveyed. Wave 2 of the MABEL survey was administered in the same manner as Wave 1 (Joyce et al. 2010). A pilot survey was conducted in April 2009 prior to the main wave. Invitation letters to participate in the main wave of the survey were distributed by mail through AMPCo in late June 2009. The invitation package contained a personalised cover letter, a copy of the questionnaire, an explanatory statement of MABEL survey information, a form to request another version of the survey and a reply-paid envelope. After the first mail-out, three personalised reminders were sent to non-respondents, each of which included a copy of the questionnaire. In the invitation letter, doctors were given a choice to complete the included paper version of the questionnaire or an online version through a secure website using a username and password provided in the invitation letter. To maximise the response rates of general practitioners located in rural and remote areas, a payment of A$100 that was not conditional on response was included with the invitation letter to these doctors. In the MABEL survey we distinguish four broad groups of doctors: general practitioners (primary care practitioners, henceforth GPs); medical specialists; specialists-in-training (vocational trainees or specialist registrars); and hospital non-specialists (including doctors in their early postgraduate years and other hospital doctors not qualified as specialists). For both waves of the survey, specific versions of the questionnaire were designed for each of the four doctor groups, although the majority of questions was common across doctor types. In Wave 2, two extra versions of the questionnaire were created for each doctor type: one for returning doctors who had responded in Wave 1, and another for the top-up sample of new doctors who were not in the medical workforce in 2008. Details of the Wave 1 and Wave 2 questionnaires are available on the MABEL website ( ). Given the longitudinal design of the survey, most of the core questions are repeated each year. Selected questions, such as age, gender and medical school of graduation, are included in the questionnaires for new participants, but are excluded from the returning doctor questionnaires. Key areas covered in the questionnaires are job satisfaction, attitudes to work and intention to quit or change working hours; characteristics of work settings (for example, hospital, private practice); workload (for example, hours worked, on-call arrangements); finances (for example, annual personal earnings, income sources, superannuation); geography; family circumstances (for example, partner, children); and personal characteristics (for example, specialty, qualifications). Several questions included in the Wave 1 questionnaire were modified in Wave 2. For example, questions relating to GPs’ on-call work were expanded to distinguish between on-call ratios for week nights and weekends and practice or hospital work. The categories of doctors’ work settings were modified to include laboratory (specialist survey) and government department, agency and defence force (GP and specialist surveys), while the category ‘Deputising service or after-hours clinics’ was removed from the GP survey. A number of changes were also made in the written instructions and guidance appearing in parentheses after some questions, particularly for hours worked and earnings. A new section at the start of the Wave 2 questionnaire, to be repeated in future waves, asks about doctors’ current working status in terms of whether they are in clinical practice or not (for example, permanently retired, working overseas or on maternity leave). Other questions included in the questionnaire may not be repeated every year. For instance, discrete choice experiments (DCEs) were included in the Wave 1 questionnaire for all doctors to examine preferences and tradeoffs for different types of jobs. The DCEs present a number of paired scenarios describing different job packages and participants are asked, of each pair, which job they prefer. The job packages differ according to a number of predefined job characteristics that include earnings, work sector, hours worked, opportunities for education and training, and characteristics of the work environment. In Wave 2, a new set of DCEs about the impact of different incentive schemes on retention in rural and remote areas was included for GPs practising in those areas. The Wave 3 (2010) questionnaire includes new questions on locus of control and occupational violence and aggression. A new feature of Wave 2 of the MABEL survey was a set of questions aimed at measuring the personality traits of doctors using the 15-item ‘Big Five’ factor model (John and Srivastava 1999). The five broad factors of personality traits are Extraversion, Agreeableness, Conscientiousness, Neuroticism and Openness to Experience. There is a growing literature that shows the important role personality traits play in influencing life and job satisfaction, and how these traits affect workforce participation, location choice and specialty choice of doctors (Lievens et al. 2002; Eley, Young and Shrapnel 2008; Heineck and Anger 2008). These questions have been used in both the German Socio-Economic Panel (GSOEP) and the British Household Panel Survey (BHPS). In addition to the personality trait question, in Wave 2 we also included a 10-point scale measurement of doctors’ life satisfaction (happiness). For both new questions, we examined the Household, Income and Labour Dynamics in Australia (HILDA) survey data for professionals to ascertain that there is enough variation in the responses from an educationally homogenous group of respondents. The set of personality questions will be included in the survey questionnaire for new participants (top-up sample) in future waves. The response rates for Wave 2 of the MABEL survey are shown in Table 1. A summary of the response rates for Wave 1 is also provided at the bottom of the table and more details are shown in Joyce et al. 2010. Of the 15 871 doctors who were invited to participate in Wave 2, 124 (0.78 per cent) declined to participate and 696 (4.4 per cent) of mailed questionnaires were classified as ‘return to sender’. The overall contact rate was 95.6 per cent. The response rate for returning doctors was 79.8 per cent (8180) and for new doctors the response rate was 37.8 per cent (2124). Overall, 10 304 (64.9 per cent) doctors responded. The response rates varied across different doctor types. Among returning doctors, the highest response rate was for specialists (82.5 per cent), followed by GPs (81.2 per cent), specialists-in-training (76.6 per cent) and hospital non-specialists (68.1 per cent). A similar pattern was observed for new doctors. The contact rates among new doctors were lowest for specialists-in-training (80.1 per cent) and hospital non-specialists (88.3 per cent). This is consistent with the rates recorded in Wave 1 and reflects the higher mobility of these doctor groups. Overall, 27.5 per cent of responding doctors chose to complete the survey questionnaire online: the breakdown by doctor type is 20.6 per cent of GPs, 25.1 per cent of specialists, 39.5 per cent of hospital non-specialists and 40.5 per cent of specialists-in-training. These outcomes are lower than those in Wave 1: 25.4 per cent for GPs, 27.6 per cent for specialists, 47.6 per cent for hospital non-specialists and 38.1 per cent for specialists-in-training. Younger doctors (hospital non-specialists and specialists-in-training4) are more likely to complete the survey online. The response rate for the 309 doctors who were practising in rural and remote areas and who received a monetary incentive was 87 per cent. Of the new and returning doctors who were sent a survey and responded (10 304), 521 (5.1 per cent) reported that they were not undertaking clinical practice in Australia at the time of the survey. Of the 10 498 doctors who responded to Wave 1, 483 (4.6 per cent)5 were no longer in clinical practice in Australia during Wave 2. Of the 5620 new doctors invited to participate in Wave 2, 115 (2.0 per cent) were not in clinical practice. The current status of respondents who were not undertaking clinical practice in Wave 2 is described in Table 2. The majority of these doctors was on maternity leave, doing medical work that is non-clinical in Australia, on extended leave or permanently retired. A number of doctors provided multiple responses regarding their status (for example, maternity leave and home duties/child care; enrolled as a student and doing medical but non-clinical work). Table 3 compares the characteristics of respondents in Wave 1 and Wave 2 with the AMPCo population of doctors in Australia in 2008 and 2009, respectively, thus providing information about the cross-sectional representativeness of respondents in both waves. Specialists are over-represented by 5 and 3 percentage points in Wave 1 and 2, respectively, while GPs are under-represented by 4 percentage points in both waves. Females are over-represented by 6 and 8 percentage points in Waves 1 and 2, respectively. There is a slight over-representation of doctors in all age groups until age 60 in Wave 1. In Wave 2, there is an over-representation of doctors under the age of 40 and an under-representation for those aged 60 and over. Doctors practising in rural and remote areas are also over-represented, which was expected given the incentive payments. Comparing the cross-sectional representativeness of Wave 2 with that of Wave 1, it is apparent that Wave 2 is less representative of the 2009 population than Wave 1 was of the 2008 population. The inclusion of the younger top-up sample including a greater relative number of female doctors contributes to this, as does the pattern of attrition between the waves, which is discussed in the next section. Using information on age, gender, doctor type and geographic location of the populations provided by AMPCo in 2008 and 2009, cross-sectional and longitudinal sampling weights are calculated and included in both Wave 1 and Wave 2 MABEL datasets: they can be used to adjust for response and attrition biases. The details of the sampling weights calculation are available in the MABEL user manual. Attrition occurs when the members in a longitudinal survey drop out over the course of the survey. While certain forms of attrition are unavoidable, for example, when there are changes to the underlying population of interest as a result of death or when subjects are no longer within the scope of interest, the attrition that is systematically related to outcome variables of interest can potentially create problems for statistical analyses and result in serious biases (Fitzgerald, Gottschalk and Moffitt 1998). Between the first two waves of the MABEL survey, the adjusted attrition rate, which is defined as the proportion of respondents who were sent a survey in Wave 1 that did not respond in Wave 2, is 20.2 per cent; inclusion of the 247 doctors from Wave 1 who could not be sent a survey increases the attrition rate (unadjusted) to 22 per cent.6 There are few other longitudinal surveys of doctors to compare this rate against. The US Community Tracking Study asked more than 10 000 physicians to participate in a second round of interviews two years after the first round, and obtained a response rate of 77.2 per cent, only slightly lower than that of the MABEL survey after one year (Potter, Sinclair and Williams 2001). The MABEL attrition rate is higher than attrition reported in large household panel surveys of the general population using interviews (13.2 per cent for HILDA and 12.4 per cent for BHPS; Watson and Wooden 2004). We analysed the MABEL attrition rate and how it varies by the observable characteristics of doctors in Wave 1. We further examined how these characteristics are associated with the probability of non-response using a logistic regression. These results are shown in Table 4. Columns 2 and 3 present the number of doctors who attrite and the attrition rates, respectively, and Column 4 reports the estimates of odds ratios and the confidence intervals from the logistic regression. The specification includes basic personal characteristics as well factors thought to influence response rates, including job satisfaction, working hours, attitudes to work and intention to quit. Across the four doctor types, the attrition rate is highest for hospital non-specialists (30.2 per cent), followed by specialists-in-training (24.5 per cent). For the former, the odds of not responding in Wave 2 are 2.26 times higher than is the case for specialists. These rates could be due to the higher mobility of these two doctor groups, which leads to inaccurate contact details in the AMPCo database. Younger doctors have higher attrition rates, but the independent effect of age on the probability of non-response is small and not statistically significant once the doctor type is controlled for. In terms of residency status, temporary and permanent residents are more likely to attrite than Australian citizens. The propensity for attrition is significantly associated with longer working hours, though the relationship is not linear. When compared with those working 20 hours a week or less, the odds of attrition increase for those working 30 to 39 hours a week, fall slightly for those working between 40 and 60 hours a week, and increase thereafter. The probability of attrition is higher for doctors with stressful jobs but is not associated with job satisfaction or intentions to leave either direct patient care or medical work. In terms of geographic factors, doctors practising in Tasmania are more likely to attrite than their counterparts in Victoria. The results also strongly indicate that doctors working in major cities were less likely to respond in Wave 2 than those in regional and remote areas; this may reflect the $100 honorarium paid to doctors in rural and remote areas. MABEL is a unique longitudinal panel survey of doctors that aims to examine the determinants and trends in medical workforce participation, labour supply and mobility. A total of 10 498 doctors form the baseline cohort of Wave 1, while Wave 2 includes 10 304 respondents, of whom 8180 form a panel cohort that have participated in both waves. In Wave 2, 483 (4.6 per cent) doctors who responded in Wave 1 were known not to be in clinical practice in Australia. The response rate was 79.8 per cent for returning doctors and 37.8 per cent for new doctors. The response bias and attrition bias can be adjusted through the use of survey weights; however, unobserved response bias or attrition bias that cannot be corrected by the use of weights may exist. There was some evidence that Wave 2 respondents were less cross-sectionally representative than Wave 1 respondents, and this is likely to be due to the pattern of attrition and to the younger top-up sample of Wave 2 containing a relatively greater number of female doctors. The MABEL data are available for use by researchers. All articles and survey materials from MABEL are available at , including the user manual and information on how to obtain de-identified unit record data.