Which incentive package will retain regionalized health personnel in Burkina Faso: A discrete choice experiment

Background: The lack of motivation of health workers to practice in rural areas remains a crucial problem for decision-makers, as it deprives the majority of access to health care. To solve the problem, many countries have implemented health worker retention strategies. However, the development of such strategies requires an understanding of the preferences of health workers. The objective of the study was to identify a package for attracting and retaining health workers in underserved areas. Methods: A cross sectional study was conducted in three health regions of Burkina Faso in 2012. A discrete choice experiment was used to investigate preferences for incentive packages among health workers recruited under the regionalized policy. In-depth interviews and focus group discussions with health workers currently working in the East and Sahel regions and policy makers, and a literature review on attraction and retention in low income countries, were performed to identify the attributes and levels. These attributes were: the regionalized recruitment policy, health insurance, work equipment, housing, and specific incentive compensation. The final design resulted in 16 choice sets. A multinomial logistic regression was used to determine the influence of socio-demographic characteristics on choice of a given option. A probit logistic regression model was then used to analyze the effect of these difference variables on choice, to identify the incentive package best suited to health workers. In total, questionnaires were administered to 315 regional health workers. Results: For all participants, choice of package was strongly influenced by length of commitment under the policy and provision of housing. Sex, number of years in profession, and location also influenced the choice of package. Women are twice more likely to choose a package with free housing and the cancellation of the policy. Conclusion: It is important that governments consider health worker preferences in crafting policies to address attraction and retention in underserved areas. In addition, the methodology of discrete choice experiment has been particularly useful, not only for better understanding the factors explaining the reluctance of health workers to work in underserved areas, but also to provide practical advice to the government, to improve its retention policy.

This is a cross-sectional study of health personnel working in the public health sector. Data was collected by six data collectors in February 2012. The sample size required to conduct a discrete choice experiment varies widely [20]. However, 50 appears to be the minimum size required [9]. The sample was selected using a stratified approach. Three health regions were selected purposively. As such, the Sahel, East, and Boucle du Mouhoun regions were selected. These hard-to-reach regions are often characterized by high maternal mortality ratios, a low ratio of personnel to primary health care centre, and a reputation of being difficult according to health professionals. In addition, these health regions have the highest number of health workers recruited under the regionalized recruitment policy. Each health region is comprised of a number of health districts and a regional hospital. Two health districts were selected in each region based on the number of health officers identified through the census. The two districts with the highest number of officers were selected. In total, six health districts were selected. Each health district covers a number of health centres including several primary health centres and one district hospital. All health centres in the district were selected. In each centre, the questionnaire was administered to all health workers recruited through the regionalization program except those who were absent for sickness or on leave. In total, 315 regional health workers were included. DCE is a method that has been used in health economics since the 1990s [16]. It was used in health status development studies, determination of patient preferences, and identification of decision-making criteria in medical or political interventions [16]. Currently, it is increasingly being used to determine the preferences of health professionals to motivate them to practice in rural or underserved areas [17]. DCE is an evaluation method that consists of proposing to individuals different hypothetical scenarios typical of the good or service to be evaluated. Faced with a series of hypothetical choices, the individual assigns a level of utility to each scenario presented and chooses the one that would give him maximum utility. This method is founded on the random utility model. According to Lancaster’s consumer theory [18], the satisfaction that an individual (n) derives from the consumption of a good is explained by the combination of good characteristics. When faced with several choices or alternatives, the individual n will choose the alternative that provides the highest individual benefit or utility. The individual n will choose the alternative i from a series of choices C, comprised of several other alternatives j, if and only if the alternative i procures maximum utility U (Uin = Vin +Ԑin): Uin > Ujn , ⇔ (Vin + Ԑin) > (Vjn + Ԑjn) ⇔ (Vin – Vjn ) > (Ԑjn- Ԑin) Given that utility cannot be observed, it is, however, possible to collect the preferences of individuals from among the different alternatives proposed. It is assumed that individuals choose an alternative based on the utility it provides. Consequently, the probability that the individual n will choose the alternative i from among all (j) is expressed as: Pn (i/C) = Pn [(Vin – Vjn ) > (Ԑjn- Ԑin)], ⇔ Pn (i/C) = Pn [Ԑjn< Ԑin + (Vin – Vjn )], This theory of utility requires a multiple attribute approach that consists of breaking down the examined good into its different components or attributes, to which levels (or states) they are likely to take. A monetary attribute is added, which takes into account the consumer’s budget constraints in the selection process. Breaking down the good into attributes and identifying the levels of attributes results in a process of fractional generation of experiences that consists of combining the attributes to get several scenarios. Each scenario reflects a specific state of the studied good [17]. In this study, we first conducted a qualitative study among health professionals in the target population to identify the different attributes [19]. In-depth interviews and focus group discussions with health workers currently working in the East and Sahel regions and policy makers, and a review of international literature on attraction and retention in low income countries were performed. This study allowed us to identify a total of five attributes, one of which is financial and four of which are non-financial. These attributes are: the regionalized recruitment policy, health insurance, work equipment, housing, and specific incentive compensation. The attribute values or levels were chosen to be realistic (Table ​(Table11). Final DCE attributes and levels for health workers under regionalised recruitment policy All motivation allowance figures presented in CFA Francs (FCFA); 1 euro = 655 FCFA The experience construction step enabled the creation of a design that combines the different attribute levels, to identify all possible incentive packages that include health workers’ aspirations and preferences. For the DCE, a labelled choice design with two choices in each choice set was used. The experimental plan was made up of all possible combinations, taking into account the number of attributes and levels of each attribute. The regionalized recruitment policy attribute had four levels and all other attributes had three levels. This specification resulted in a design with 324 (41 x 34) possible combinations. As the possible number of combinations was too high to reasonably conduct a study, a fractional or full factorial experimental plan was generated. The Hahn and Shapiro catalogue was used to select combinations for an orthogonal main effects design, and to organize the selected profiles into the most D-efficient choice design, given our design parameters. The final design was resulted in a 16 choice sets. Respondents were then asked to select their preferred option from each of 16 choice sets (15 random and 1 fixed) or could decide to stay in their current job. They were asked to give the reason for the opt-out option and for how long the chosen option could motivate them in their current position. The survey questionnaire included three sections. In the first section, an information notice on the study, and the meaning of each attribute and levels was presented. In the second section, information was collected on respondents’ socio demographic and professional characteristics including age, sex, ethnicity, religion, marital status, number of children, current professional title and qualifications, and years worked in the health sector. The third section presented the DCE choice sets. The full questionnaire was pre-tested among health professionals in the health district of Kaya. Minor corrections were made after the pre-test and before data collection. The questionnaire was administered to participants by six data collectors. Each participant was asked to place him/herself in a hypothetical situation, the implementation of a new policy with an incentive package, before presenting the different choice sets. The data entry was performed using CSPro software. The data was analyzed using Stata 12 software. A descriptive analysis allowed us to select the options that recorded the highest percentages of agreement by region and by professional category; these options were then selected to conduct multinomial logistic regression to determine the influence of socio-demographic characteristics on the choice of a given option. We also calculated the difference between the level of attributes according to each alternative, and this was done based on dummy variables. A probit logistic regression model was then used to analyze the effect of these difference variables on the choice of an option in order to identify the incentive package best suited to health workers. This study is a sub-component of a larger study on a recruitment policy for human resources for health in Burkina Faso, for which ethical clearance was obtained from the Ethics Committee for Health Research in Burkina Faso. Respondents participated on a voluntary basis and could withdraw from the study at any time. Informed consent was obtained from all participants and signed consent forms were obtained prior to the interviews.

N/A

• Dima Health