Managing uncertainty in forecasting health workforce demand using the Robust Workforce Planning Framework: the example of midwives in Belgium

In Belgium, the training and practice of midwives are regulated by the Law of 10 May 2015 on the practice of health professions. Although the minimum content requirements for midwifery training are described in this law, each linguistic community may define the duration of the training needed to acquire the required skills, i.e. 3 years in the Flemish Community and 4 years in the French Community.

The law distinguishes between activities performed by midwives with complete autonomy and those that require medical supervision. Autonomous midwifery activities are listed in the law and include pregnancy diagnosis, follow-up of low-risk pregnancies (maternal and child risk assessment, birth preparation, and parent education), eutocic deliveries (including amniotomy, episiotomy, perineal suturing), postnatal care, care of healthy newborns, preventive measures, and emergency procedures. They also have the right to prescribe a limited number of drugs listed in the law. The management of fertility problems, high-risk pregnancies, high-risk deliveries, and newborns with life-threatening conditions requires medical supervision. In addition, the law describes procedures that are explicitly prohibited for midwives, namely: artificial dilation of the cervix; use of forceps and vacuum; administration of anaesthesia (except local anaesthesia for performing or suturing episiotomy); and inducing abortion. Except in emergencies, midwives are also prohibited from performing the following procedures: internal version, breech extraction, manual removal of the placenta and manual exploration of the uterus.

In Belgium, patients are free to choose their care provider and the setting of care. Antenatal care can be provided in a variety of settings, including hospitals, private practices and other centres. The majority of women give birth in hospital while the number of outpatient deliveries (at home, in a birth centre, or in a one-day hospitalisation) has remained fairly stable over time, accounting for about one per cent of the total number of deliveries [1‐3]. Early postnatal care is usually provided in hospital, while postnatal care can be provided at home or in other settings (hospital, private practice, etc.).

In 2019, 12 088 midwives were licensed to practise [4], of whom 57% worked in the healthcare sector (7 175 FTEs) [5]. With more than 9 out of 10 births taking place in hospital [1‐3], the majority of midwifery activity took place in hospitals. As a result, less than 10% of the midwives worked in an outpatient setting [5].

In Belgium, the planning of healthcare professionals is a responsibility shared by the federal state and the federated entities [6].

The Planning Commission for Medical Supply, under the authority of the Federal Minister of Health, is responsible for monitoring human resources for health (HRH) and proposing the regulation of the workforce through a system of federal quotas. Recently, besides this commission, two commissions have been set up in the federated entities to monitor HRH in their territories. The authorities of the federated entities regulate the candidates for training in order to meet the federal quotas.

The Planning Commission for Medical Supply is supported by the Planning Unit for the Supply of the Healthcare Professions. This Unit quantifies and forecasts the supply of health professionals, including midwives, on the basis of a stock and flow model [7]. Stock refers to the available supply of midwives expressed in terms of headcount and Full Time equivalents (FTE) working in the health sector. Flows are made up of inflows (new entrants to the labour market through graduations, immigration, or return to the profession) and outflows (resignations, dismissals, retirements, or deaths). The forecasting model estimates the future evolution of the workforce using historic trends. Based on demographic trends, the output of the model is the midwife-to-population ratio weighted by the consumption of care [8]. The weighting is done by multiplying the “gross” population projections by a consumption rate, based on the reimbursement of midwifery fees.

In 2019, the Planning Unit charged the KCE (Belgian Health Care Knowledge Centre) to develop additional forecasting scenarios to complement the standard historic trend approach. The mission consisted of developing different scenarios for the management of pregnancy and childbirth to estimate their impact on the future demand for midwives, over a 25-year period.

This paper illustrates both the development and the quantification of alternative forecasting scenarios for the demand for midwives in Belgium.

This Framework was created by the Centre for Workforce Intelligence, which was active between 2010 and 2016 to support the planning for future health workforce needs in the UK [9].

The purpose of the Framework is to focus on the uncertainty that arises in the context of long-term forecasting of healthcare supply and demand. To capture the complexity of factors influencing the workforce supply, the organisation of health care, and the demand for health services, the Robust Workforce Planning Framework consists of a four-step approach [9]:

The Framework requires a high level of stakeholder involvement and therefore uses participatory methods such as surveys and workshops. It can be complemented by targeted expert review and text mining [10].

The first two steps of the Framework were carried out in five workshops with 31 invited participants (see Fig. 1). Workforce modelling and policy analysis were then carried out by the Planning Unit. These steps are further summarised in the discussion section.

The horizon scanning was carried out in two phases. First, a preparatory online survey of stakeholders was conducted in French and Dutch. Stakeholders from midwifery professional associations, midwifery schools, the hospital sector, public authorities and administrations, patient associations, and individual professionals (midwives and gynaecologists) were interviewed to identify significant factors to be added to the original forecasting model. The variables included in the original forecasting model, and the description of the participants can be found in Additional file 1. The online survey (see Additional file 1) was sent to 122 participants, with a response rate of 75%. Participants were invited via a call in both French and Dutch. Participants could choose their language before starting the survey. The survey presented 50 factors derived from a preparatory exercise [11] and grouped into 14 domains [12] (see survey in Additional file 1). For each domain, participants were asked to identify factors to add to the initial list.

Results of the survey were then discussed in an initial workshop to assess the relevance and the potential impact of the proposed factors on the future workforce (see Fig. 1 and Additional file 1: Material Sect. S2.3).

All workshops were led by an external team (ShiftN) with no prior connection to the participants.

From the preparatory online survey, 101 potential factors were suggested for inclusion in the midwifery workforce forecasting model. During the first workshop (n = 31 participants), 77 relevant factors that were not yet included in the existing model developed by the Planning Unit, were selected. Detailed results on the retained factors and their potential impact on the future workforce as assessed by the participants are presented in Additional file 1 (see Table 2).

Factors identified during the horizon scanning were merged into 12 clusters (see Additional file 1), placed on an impact/uncertainty matrix by the workshop participants (Fig. 3). Ten clusters were positioned above the median line of impact meaning that participants in Workshop 2 (n= 24) considered these clusters likely to have a high impact on future midwifery practice. Two other clusters, positioned below the median line of impact, were judged to have a lower impact and were not further considered further in the development of the conceptual framework for narrative scenarios.

The upper right quadrant contains the 5 clusters representing critical uncertainties while the upper left quadrant contains the 5 clusters representing driving forces. Critical uncertainties and driving forces are likely to shape the future of midwifery with varying degrees of probability. It is still possible to simplify the picture by distinguishing 2 underlying meta-clusters, namely the rationale for health care provision (clusters 1–4) and the nature of the networking (clusters 5–7). Contextual factors are independent of the organisation of care but will have a more or less strong impact on supply and demand (clusters 8–10).

The framework for the scenarios can be illustrated in Fig. 4.

The first (horizontal) continuum depicts the rationale for care provision, with one extreme representing care provided mainly by midwives (primary care or ‘midwifery-led care’) and the opposite representing care provided mainly by medical specialists (secondary care or ‘specialist-led care’). The second (vertical) continuum illustrates the nature of the networking of perinatal providers around pregnant women and future parents from hospital-centred networks (one extreme) to outpatient networks (the opposite extreme).

The combination of these continua creates a matrix with four quadrants within which all possible alternative scenarios for the future can be positioned alongside a baseline scenario. The baseline scenario takes into account current trends towards a reduction in postnatal length of stay in maternity units and contextual factors (demographic and epidemiological trends, socioeconomic vulnerability and budgetary constraints).

Workshop participants agreed that 3 plausible alternative scenarios were sufficient to predict likely futures. Scenario A is the closest to the current organisation in which the gynaecologist plays a central role, facilitating a close collaboration between the hospital and outpatient networks. In the other two scenarios, care would be provided primarily by midwives, either in a hospital-centred care network (scenario B) or in an ambulatory-centred care network (scenario C). The scenarios are summarised in Fig. 4.

The scenarios illustrate a gradual shift: (1) the transition from the baseline scenario to scenario A was characterised by a greater involvement of midwives in the antenatal care of low-risk pregnancies; (2) the difference between scenario A and B was a shift from a ‘gynaecologist-led-care’ model to a ‘midwifery-led care’ model; and (3) the difference between scenario B and C was the transition from hospital to outpatient care (see Fig. 4).

During the elicitation process, 34 professionals, mostly midwives (n = 22), gave their input to elicit the parameters of the model, i.e. they were allowed to assign a value to each parameter of the model (see Fig. 2). Their estimates were discussed with 15 midwives in order to reach a consensus on the value assigned to some parameters.

Table 1 shows the values of the parameters that are common to all scenarios, while Table 2 reports the assumptions specific to each scenario. The penultimate column indicates the source of the data.

During the elicitation process, 94% of the participants estimated that 80% of the midwifery services were provided in hospitals and 20% in outpatient settings (Table 3). Using data on the workforce distribution within hospitals, we estimated that 60% of hospital midwifery services were devoted to postnatal care and 40% to other activities. In outpatient settings, the relative proportions of midwifery services were calculated using the annual number of sessions and the duration of midwifery sessions (Table 1).

Given these assumptions, the expected variation in consumption is shown for the different scenarios in Table 4.

Between 2016 and 2026, the baseline scenario predicts an increase of 7.1% in the demand for midwifery activities, mainly driven by increasing activities in outpatient postnatal care activities (+ 8.6%), offsetting a 4.6% decrease in inpatient postnatal care activities (due to shorter hospital stays).

Over the same period, the implementation of Scenario A would lead to an increase of 11.4% in midwifery activities. This increase is mainly related to childbirth preparation and parenting education (+ 5.1%) and to a lesser extent to prenatal activities (+ 1.3%).

The transition from Scenario A to Scenario B would have a very small increase related to antenatal and postnatal care (+ 0.3 point of percentage for each activity between the two scenarios).

However, the transition from Scenario B to Scenario C would involve a major organisational change from a hospital-centred model to an outpatient-centred model. This would be reflected in a significant reduction in the length of stay in the maternity ward and an increase in the use of outpatient postnatal services. Scenario C show a 17.4% increase in demand for midwifery activities between 2016 and 2026. This is mainly due to a very strong increase in outpatient postnatal activities (+ 21.6%), partly compensated by a decrease in inpatient postnatal activities (− 12.0%).

The scenario originally developed by the Planning Unit to forecast the midwifery workforce did not initially consider possible, likely, or plausible potential deviations in demand from historical demographic trends [8]. The Robust Workforce Planning Framework led us to develop alternative scenarios to better capture the future demand for midwives, taking into account epidemiological, organisational, technological, economic, environmental, political, social, and ethical influences on future workforce development. Depending on the scenario considered, we estimated an increase of between 11 and 17% in the demand for midwives in 2026 compared with 2016. However, the robustness of the projections depends on several factors such as the quality and completeness of the administrative data, the reliability of the elicitation, the appropriateness of the proxies used, the likelihood of the forecast assumptions and the time horizon of the forecast.

As administrative data are not collected for the purpose of HRH planning, proxies are often used to capture the necessary parameters. For example, care consumption has been used as a proxy for care needs. However, it only represents the expressed demand and does not provide information on unmet need or overuse. In addition, some of the data needed to quantify the scenarios are not available in the administrative databases. Therefore, this was compensated by an elicitation process organised in two steps: (1) a survey sent to 74 stakeholders with a response rate of 47% (including 3 gynaecologists, 22 midwives and 9 professionals from institutions dealing with mother and child care or HRH planning); and (2) a workshop to discuss the survey results with 14 midwives and 1 gynaecologist. The elicitation process was based entirely on the opinions of the experts interviewed. The quality of the results, therefore, depends on the ability of the experts consulted to anticipate the future.

The determination of the allocation key by the elicitation step is a critical feature of the model. Variation in estimation could significantly alter the results obtained. However, an independent check by the midwifery working group of the Planning Unit on their own databases confirms the reported values [8]. A regular monitoring of all elicited parameters would be necessary to ensure the accuracy of the estimates or to integrate deviations. In the future, a prospective collection of missing data could be envisaged to increase the reliability of the estimated parameters.

In Belgium, the Planning Commission produces forecasts with a 25-year horizon. However, the estimation of the future value of some parameters is based on short(er) time series, which means that the degree of uncertainty increases with time. Some authors therefore recommend limiting the forecast period to 10 years [21].

The scenarios described by the stakeholders and experts in the field favoured a greater involvement of midwives in maternity care, from the antenatal period to the postnatal period, especially for pregnant women with a low risk of complications. This greater involvement would be accompanied by varying degrees of practice autonomy, depending on the scenario. The likelihood of each scenario depends on the time horizon envisaged and many other factors, such as the political will and incentives to reorganise antenatal care and their preference for a particular scenario, the propensity of the system to ensure a more intensive and rapid shift to outpatient care, and access to specialists reserved for high-risk clinical situations. In the near future (5 to 10 years), scenario A is indeed the most likely. However, if we consider a longer time horizon (20 to 25 years), the relevance of the other two scenarios becomes equally likely. In fact, they follow an international trend towards more patient-centred care provided by first-line professionals (midwives and general practitioners) [22]. Such changes usually take several years to implement in daily practice and require adequate training of the professionals concerned. Four years after this study, the Federal Minister of Health's policy note (2023) shows a willingness to invest in the training of midwives and to improve multidisciplinary coordination in the outpatient care of pregnant women and women in labour.

The generation of scenarios is essentially based on the collection and analysis of proposals from midwifery experts in Belgium. It cannot be excluded that a group composed of different experts would have led to different scenarios. The low participation of gynaecologists in the process (despite repeated invitations) and the decrease in the number of participants in the workshops (29, 24, 19, 18 and 15, respectively, for workshops 1 to 5) probably influenced the forms of care proposed for the future. Therefore, this type of exercise needs to be monitored regularly to ensure that the pathways identified for the future remain plausible in the light of legislative, contextual, epidemiological, social, and scientific developments. Forecasts of the HRH workforce, based on a similar methodology and involving stakeholder groups of comparable size, have been developed in the Netherlands [23], The United Kingdom [24, 25], Portugal [26] and Australia [27‐29], of which three models concerned midwives. In the Netherlands [23], the training needs of midwives were predicted in 3 steps, combining interviews with 9 experts, 2 workshops with 21 participants to elaborate scenarios and a quantification process based on statistical data and elicitation. Four scenarios were developed on the basis of a theoretical framework presented in 2 dimensions: (1) a continuum from no to a greater shift of tasks from gynaecologists and general practitioners to midwives; and (2) a continuum from no to full collaboration between care providers in primary and specialised care. The more conservative scenario showed a 21% reduction in the number of midwives to be trained over 10 years. The opposite scenario (i.e. greater delegation of tasks to midwives, coupled with the removal of the barrier between primary and specialist care) predicted a 31% increase in the number of midwives to be trained. The UK [25] and Australia [28] focused their scenarios on the supply side of the forecasting model whereas the demand forecast was only based on projected birth rates. The reflection carried out in this study goes beyond the question of HRH alone and raises the question of the future organisation of mother and child care. This strategic positioning depends on the political choices made.

Finally, subjecting the various scenarios to the critical opinion of decision-makers and users of the health system could confirm or undermine their plausibility and, if necessary, make some corrections. Other factors could be taken into account, such as the accessibility of care, the risk of two-tier medicine when moving from one scenario to another, or the quality of care for low-risk pregnancies (caesarean section rates, etc.) [30].

The Planning Unit used the results of this study to enrich the debate in the Planning Commission. Over a 10-year horizon, the density of midwives (expressed in FTE weighted by the population’s consumption of care by age category—see above) could increase from 13 to 37%, depending on the scenario considered [8].

In addition, the methodology described here is currently being used by the Planning Unit to forecast nurses [31].