Several factors play a role in determining the numbers of health care workers and skill mix of a particular health system, including resource availability, regulatory environment, culture and customs. The extent to which any one of these factors influences the typical mix of health care occupations in different countries remains unknown. A recent review of the literature pointed to rapidly growing interest in examining the roles and mix in medical and nursing occupations. In terms of the published literature, few studies were found offering a cross-national perspective [5]. The most frequently used bases for comparing international health care resources are health care expenditures, measured either as a fraction of gross domestic product or on a per capita basis. Assessments of non-monetary resources, such as medical equipment or health personnel, are less widespread; working the latter into international comparisons of health care resources has been taken up only slowly [6].

The availability of quantitative, methodologically sound analyses of the stock and mix of health care occupations across different settings and health systems could be an important catalyst towards better understanding labour issues in health care and identifying appropriate solutions for human resources for health (HRH) planning and management. Despite the importance of an evidence base for policy decisions, information on the health workforce tends to be fragmented, and the sources that can potentially produce statistics relevant to this issue are often underused in health research. Although many general data sources can be exploited as tools for conducting HRH assessments – including population censuses, routine administrative records and sample surveys – their potential for defining the scope of the health system including human resources has generally not been met [7]. For example, censuses can provide valuable information on the occupational distribution and other sociodemographic characteristics of the population of economically active age; however, they tend to contain limited information on other indicators of labour force activities. Administrative records can provide data on licensing and regulations of health occupations in some countries, but often only with emphasis on the public sector. While each type of data source tenders at least some analytical potential, household-based employment and income surveys offer the advantage of providing nationally representative information on many aspects of labour force participation.

Data on health occupations were available for 18 countries with surveys conducted between 1989 and 1997. Twelve of the countries were characterized with developed market economies: Austria, Canada, Denmark, Finland, France, Germany, Netherlands, Norway, Spain, Switzerland, United Kingdom and the United States of America. Six were countries with economies in transition: Czech Republic, Hungary, Poland, Russian Federation, Slovakia and Slovenia. Moreover, for 11 countries two or more surveys were accessible that allowed identification of health occupations, enabling us to conduct time-trend analyses. (The LIS project had compiled surveys for nine other countries that were not used here because the occupational data did not enable differentiation of the health workforce.)

The standardization of classification of health occupations was facilitated through the International Labour Office's latest revision, in 1988, of the International Standard Classification of Occupations (ISCO-88). This internationally comparable classification pools occupational titles into a hierarchical four-digit system, which can be aggregated to progressively broader groups, representing a value set describing the different tasks and duties of jobs [10]. Within ISCO-88, occupations are essentially organized according to two dimensions: skill level and skill specialization [11]. The former refers to the complexity of skills required for the job (but not necessarily the way the skills were acquired). Skill specialization is related more to areas of knowledge required, such as subject matter, services produced or equipment used. Different user areas may have different degrees of interest in the various elements, so classification structures may vary nationally. Many national statistical agencies participating in the LIS project mapped their occupational classifications to ISCO-88 for data dissemination. Otherwise, where possible, the project provided ISCO-88 classification codes by reconciling national classifications through standardized mapping techniques of occupational status scales (for example, techniques cited in [12]).

Among the 10 major ISCO-88 occupational groups, two were of interest here: group 2 "professionals" (generally well-trained workers in jobs that normally require a university or advanced-level degree for recruitment) and group 3 "technicians and associate professionals" (generally requiring skills at a non-university educational qualification level). Identification of the health workforce is possible when the classification is coded to a degree of detail that minimally corresponds to the three-digit level, and preferably to the four-digit level for distinction of practitioner specializations. The professional major group includes physicians, nursing and midwifery professionals and other health professionals, such as dentists, pharmacists and veterinarians (see Table ​Table1).1). Classified as associate professionals are modern health associate professionals (except in nursing), nursing and midwifery associate professionals and traditional medicine practitioners. The former encompass medical assistants, dental assistants, pharmaceutical assistants, opticians, veterinary assistants, physiotherapists, sanitarians and others. Traditional medicine practitioners include herbalists and faith healers.

We performed basic analyses on characteristics of the health workforce where occupational data were standardized at the three-digit ISCO-88 level or equivalent. Further in-depth analyses were conducted where identification of health occupations was possible at the four-digit ISCO-88 level or equivalent. In either instance, occupational categories were aggregated to reflect national classifications or sample size limitations for some surveys. It should be noted that despite efforts to standardize, the definition of certain categories of health occupations may have varied across surveys; for example, in some cases the classification of nurses and midwives did not distinguish between professionals and associate professionals. The precision of mappings to ISCO-88 would have largely depended on the level of detail in the national classifications. Moreover, while certain related occupations aside from medical and nursing practitioners are identifiable at the four-digit ISCO-88 classification level – in particular, medical equipment operators (code 3133), health and safety inspectors (code 3152), and institution-based personal care workers (code 5132) – they were excluded from the present analysis to maintain comparability with data where the selection of occupations was possible only at the three-digit level or equivalent.

The surveys' sampling designs and sizes were not homogeneous: while most sampling frames drew on stratified random selections of private households, some datasets were based on income tax or other administrative records of government agencies. Although several different data types may have been available in many countries, only selected surveys were retained by the LIS project based in part on comparability of information on income sources or other labour market indicators. Non-response rates, for the entire interview or per item, varied and were treated differently across countries [9]. No attempt was made in the present analysis to further adjust the data for coverage or completeness of information.

To monitor the relative allocation of human resources to the health system, all samples were limited to the population of economically active age (15 years and over) declaring an occupation. The numbers in the samples with health-related occupations ranged from 60 to 12,248 (see Table ​Table2).2). Our study includes profiles of the health workforce by selected sociodemographic and labour force characteristics, including sex, age, migration status, education, income and industry. Such information can offer valuable insight into specific aspects of HRH as an input to assessing health systems performance [7].

Standardization of indicators was ensured to the extent the available data permitted. In terms of migration of health workers, an audit of human resources can show movement between localities (e.g. rural to urban), between sectors (e.g. public to private), or between countries. Relying on information on immigration status available from the LIS datasets, we defined migrants as non-native born. Education was assessed by university-level attainment versus secondary schooling at most, as a gauge of the skill distribution of health care personnel. Depending on the source, this was captured by either the individual's highest level of general education or vocational training, total length of education, or age when the highest level was obtained. The indicator for income – information of value when discussing countries' health care financing options – was measured through either net or gross occupational wages. Workforce industry encapsulated the economic activity of the main job establishment. However, the classification of health services only sometimes distinguished the various health care activities, such as hospitals or practitioners' clinics versus veterinary or pharmaceutical services.

Gender issues were emphasized as being important not only for assessing equity in human resources, but also for health services planning. Studies have shown that increased participation of women in the medical field may be accompanied by differences in working patterns; in particular, female physicians are likely to work fewer hours than their male counterparts [13,14], and to present different styles of care provision that may be reflected in the levels of patient participation [15].

Our analysis included descriptions of the health workforce for a number of demographic and socioeconomic indicators. Notably, an examination of the sex distribution revealed a health labour market characterized by a large presence of women. Further assessment of data from five countries helped to reveal the extent to which women and men may have equal opportunities in career choice. The evidence suggested that gender inequity in human resources for health remains an important shortcoming of many health systems. Large imbalances were seen in terms of occupational segregation and wage gap. It is difficult to establish the causal links of such imbalances, as the influences may be dynamic and multidirectional, related to both demand and supply factors. But it is important to point out that the gender inequity observed in the health field may be even more pronounced in some respects than for workers in other fields: there were greater wage gaps in some countries compared to other occupations in physical and engineering sciences or in teaching. Assuming that monetary incentives are important in labour participation decisions, such results suggest that recruitment and retention in health occupations, especially nursing and midwifery, might suffer in comparison with other non-health occupations that propose better earnings for a similar skill level.

The household-based survey data on labour force activities and occupational wages available through the LIS project presented both advantages and constraints in conducting international comparisons. Because of the discontinuous nature of the data collection procedures, performed by the various national statistical agencies with differing targets and interests, LIS undertakes variable harmonization to facilitate public use. In assessing the existing data, we found certain strengths and weaknesses. In many ways our analyses were guided as much by the character of the data as by the base required for formulating policy decisions.

The first challenge lay in defining health occupations themselves. In this article, the terms "health occupations" and "health workforce" were used interchangeably. Occupations were generally classified according to ISCO-88, which broadly groups occupations according to skill levels and specializations. We focused on selected professionals and associate professionals with a health care-related specialization. Since the survey data were nationally representative, they included those practising medicine and nursing in both public and private institutions, as well as those in administrative, research and industry positions. For example, findings from the Netherlands, United Kingdom and United States revealed that some 7% to 9% of physicians and 8% to 18% of nursing and midwifery professionals were engaged outside of the health services industry. Persons with health qualifications but not working in the public health sector are often excluded from national registries, for instance. On the other hand, our analyses did not consider those employed in the health system with non-health occupational backgrounds, such as managers, accountants, equipment operators, drivers and other support staff. Again for the same three countries, an estimated 36% to 48% of workers in health services had a vocation other than the selected health occupations.

Even when occupations were classified according to ISCO-88, comparability issues arose. In some cases, discrepancies were evident between mappings of national classifications with the international standard, particularly with regard to the nursing and midwifery specializations. In addition, the results did not enable distinguishing the different types of health activities: clinical, research and public health interventions; preventive and curative personal care; health systems planning and management; etc. Moreover, health-related specializations under ISCO-88 place veterinary occupations in the same minor (three-digit) groupings as human care occupations. WHO is currently collaborating with ILO and other interested parties to refine the descriptions for some categories of health and personal care occupations, in order to facilitate analyses of human resources for health. Such issues may grow increasingly important, because use of or mapping to ISCO-88 is expected to become more widespread across countries [11].

Other constraints included the sometimes small sample sizes of surveys (from which, for example, census data do not suffer) as well as occasional definitional inconsistencies among the selected indicators. Availability and comparability of certain types of information were dependent on the source. Education was a variable that posed problems, as the categories and details varied across countries. While in a few cases the variable captured the level of attainment, in others it was years of schooling and in yet others, age of completion. Some discrepancies might be related to the structure of national educational systems. To the extent possible, information was recoded for the present analysis to approximate education at the tertiary level. We recommend the use in future data collection and processing of a cross-nationally comparable instrument for definitions of levels and fields of education, such as the International Standard Classification of Education [21].

Migration of skilled health workers has become a cause for global concern, as mass emigration of health professionals from less-developed countries can put great pressure on the health systems and workers remaining [22]. We found that, for instance, health professionals – and especially physicians – were over-represented among the foreign-born compared to the total labour force in the United Kingdom and the United States. But migration was an area covered somewhat inadequately in the surveys. There was generally no information on period of international migration, other aspects of mobility (such as rural exodus), or time-trends in the countries under observation. Moreover, data on emigration were notably lacking. WHO advocates better cooperation between the many agencies supporting processes for strengthening national health systems, with equitable geographical distribution as one of the core policy areas. Human resource constraints have been identified as a significant barrier to ensuring sustainability of health systems in many developing countries and to scaling up interventions on major health problems among the poor (see, for example, [23,24]). The promise of higher salaries and other incentives in one country may be met by outflows of labour from other countries, creating or exacerbating personnel shortages. Assessing the determinants and impacts of international migration across countries at different stages of development remains an important research domain.