A Discrete Choice Experiment on Factors Inuencing Attrition and Motivation of Frontline Health Workers in Remote and Rural Areas in Nigeria

Background The current policy thrust in Nigeria is to ensure that there exists qualied, skilled, and adequate health workforce for the achievement of universal health coverage. To achieve this, evidence combination of strategies/interventions on factors inuencing attraction, retention, and motivation of the health workforce is needed to ensure the equitable distribution of the health workforce across the country. Methods We conducted a discrete choice experiment study to determine the combinations of incentives that may increase the retention of frontline health workers in rural and remote areas in Nigeria. The study was undertaken across rural and urban locations in Bauchi State between August and October 2018 amongst 145 students and practicing health workers. Results Health workers are 14.6 and 14.4 times more likely to take up a rural posting or continue to stay in their present rural posts if there was basic housing and improvement of the quality of the facilities respectively. The preference for rural job location increased by 6.17 times if good schools for children's education were provided. The highest willingness-to-pay was for the provision of basic housing or housing allowance with a high utility of 0.609 followed by improved quality of facility with sucient staff number and type, equipment, drugs, and medical supplies with a utility value of 0.607. Conclusion Improving the working conditions of health workers will support retention in rural health posts. Based on the ndings, we recommend the introduction of housing incentives in rural and remote areas to support the retention of health workers. This should be accompanied by deliberate interventions to improve the quality of the health facilities ensuring adequate and functional equipment and uninterrupted supplies.

The current policy thrust in Nigeria is to ensure that there exists quali ed, skilled, and adequate human resources for health at primary, secondary, and tertiary levels of care towards the achievement of universal health coverage for all Nigerians [5,6]. To achieve this, the government has established a policy framework to generate evidence and implement strategies aimed at ensuring the availability, equitable distribution, and retention of health workers with a focus on the rural areas [5][6][7]. The current situation in Nigeria indicates that health workers are not motivated and this is negatively impacting their performance and e ciency. Several factors, including poor remuneration, inadequate sta ng levels with resultant high workloads, lack of social amenities, poor working environments, inadequate medical supplies and equipment, and delayed salaries and emoluments payment have been documented [8][9][10][11][12][13][14][15][16][17]. Additionally, institutionalized attraction, retention, posting, and motivation mechanisms are absent, especially for primary level facilities often located in rural and remote areas [6,18]. To improve the attraction, retention, and motivation of health workers, investments have to be made in the work environment, performance management and recognition systems, work environment, infrastructure, career development, and sta ng levels with a focus on rural areas are needed [6,9]. Evidence suggests that attraction and retention are informed by a complex interaction of factors and thus bundles/ sets of strategies/interventions are needed (4).
As Nigeria is on the path of improving the health system towards achieving the national health sector goals and universal health coverage, policymakers need contextual evidence to improve human resources for health (HRH) policy and management. Information on factors in uencing attraction, retention, and motivation of the health workforce is key in ensuring the equitable distribution of the health workforce across the nation for quality service delivery, especially at the primary level. There is, however, limited evidence on these factors as well as the combination of strategies/interventions. This informed this study using the discrete choice technique to ascertain the applicable combinations. This technique has been applied in several settings globally with essential evidence generated for policy and planning [11,[19][20][21][22].
The objective of this study was to determine the combinations of incentives that may increase the retention of frontline health workers in rural and remote areas.

Study Setting
The study was conducted in Bauchi State, one of the seven States in North-Eastern Nigeria. The population is mainly rural and agrarian. With a signi cant shortage of health workforce, the State relies on lower-level frontline health workers to deliver health services, especially in the vast rural settings. According to the Nigeria Demographic and Health Survey of 2018, while about 51.6% of pregnant women in Bauchi State had their ante-natal care (ANC) attended to by a skilled provider [23,24] in 2018, only 3% of them were attended to by a physician against the national average of 17.2%. ANCs were mainly attended to by Nurse/Midwives (46%) compared to the national value of 48%. Availability and effective Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js distribution of these frontline health workers are therefore essential in efforts to lift Bauchi State and indeed much of the North-Eastern region from poor healthcare conditions. Despite the 56% ANC attendance, only 21.8% of the women eventually delivered in a health facility while 78% of the deliveries were at home. The unavailability and inadequacy of health workers especially in rural settings have been suggested to be the reason for this trend [10].

Study Design
This study used the design of discrete choice experiment (DCE) studies. The DCE design aims to elicit favored preferences of individuals based on options. The theoretical underpinning is the random utility theory that is based on the assumption that each individual is an independent rational decision-maker, maximizing utility to his or her choices. Our hypothesis for this study speculates that health workers in Bauchi State will maximize their utility by selecting the working condition and location options that give them the maximum utility.
To establish the attributes and levels for the DCE, we employed qualitative methods. Focus group discussions and key informant interviews were conducted to obtain information about the dimensions of the work conditions in primary health facilities that are important to frontline health workers when choosing to take up posting or stay in their rural work locations. This was undertaken across rural and urban locations in Bauchi State between August and October 2018. We conducted ve focus group discussions among students in the School of Nursing and School of Midwifery in the Bauchi State College of Nursing and Midwifery Bauchi and the College of Health Technology Ningi. Key informant interviews were conducted among 16 frontline healthcare providers (nurses, midwives, community health o cers, community health extension workers, and junior community health extension workers). Seventeen policymakers and health managers were also interviewed.
Five attributes were identi ed as recurring from the students and healthcare workers interviewed after analysis of the qualitative ndings. Literature and desk reviews were conducted to determine appropriate levels for each of the four ranked attributes ( Table 1). The levels of attributes were chosen to re ect the range of situations that respondents might expect to experience. We subsequently designed a structured questionnaire to collect information on the socio-demographic information and work experience of respondents. The questionnaire also had a series of options for two hypothetical post options -rural/remote health facility and urban health facility to healthcare workers meeting eligibility criteria. In all, there were 24 different choice sets in the questionnaire with each containing two choices of established attributes and differing levels. Respondents were asked to select either a rural or urban work location choice based on the attributes and levels presented in each choice set.

Data collection and Analysis
The sample size for the DCE was determined based on a formula postulated by Orme [25] as follows: Where c = largest level of any variable; t = Number of choices; a = Number of alternatives.
The largest level of any attribute was three (3), the total number of choice tasks being ve (5), and the number of alternative choices as two (2). The completed 145 questionnaires were entered into CS-Pro and analyzed using STATA 15®. Diagnostic correlation analysis was used to determine the absence of multicollinearity (Orthogonality) which is a basic assumption in tting logit models in a DCE. Further analysis was explored to ensure the level balance and utility balance of attributes.

Ethical Issues
The National Health Research Ethics Committee of the Federal Ministry of Health reviewed the study protocol with ethical approval issued. Approval for data collection was obtained from the Bauchi State Ministry of Health.

Socio-demographic Characteristics
The characteristics of the DCE survey respondents presented in Table 3 show that the proportion of males and females are 46% and 54% respectively. The respondents were youthful with about 64% of the population less than 35 years of age. Those aged 35 to 44 were 21% and the remaining 15% are 45 years or older. More respondents were married (58%) compared to those who were single (41%  Conditional Regression Analysis and Willingness to Pay Table 4 shows the conditional regression analysis measuring factors or attributes that in uences utility as well as determine the willingness-to-pay (WTP) for such attributes for health workers. The WTP was determined by the ratio of the estimated coe cient for each attribute to that of the cost coe cient. The highest willingness-to-pay was for provision of basic housing or housing allowance for which workers Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js were willing to for-go about NGN32,233 (About $90) in monthly wages with a high utility of 0.609 followed by improved quality of facility with su cient staff number and type, equipment, drugs and medical supplies (WTP: NGN31,107; Utility value: 0.607). There was a preference for a job posting in locations with the availability of good schools for children's education rather than one with no good schools available at an increased utility of 0.343 and for which workers were willing to for-go NGN18,171 (About $50) in salary and allowances per month. The least preference for opportunities for career advancement was approval for continuing education was given at an enhanced period of 2 years postemployment rather than the current guideline of 3 years. This had a negative utility of -0.263 and a WTP value of NGN13,894 (about 40%).

Choice Probabilities and Covariates
To determine the change in probability of choosing between posting location alternatives due to a change in the levels of the attributes, a logit probability was tted to the data. Table 5 shows that frontline health workers in Bauchi State were 14.6 and 14.4 times more likely to take up a rural posting or continue to stay in their present rural duty posts against an urban alternative if there was basic housing and improvement of the quality of the facilities respectively. The preference for rural job location increases by 6.17 times if good schools for children's education were provided compared to when such a facility is not available. When offered a salary increment, preference for rural posting increases by 2.72 times; this magnitude of preference remains unchanged regardless of the magnitude of salary increment. We explored the possible covariates that may in uence directly self-reported preference for rural or urban uptake of job posting by tting a multiple logistic regression model. It gave an insight into relevant characteristics and the factors to consider when considering individual health workers for posting to rural locations.  About 145 frontline healthcare providers in rural and urban centers as well as students of training institutions in Bauchi State responded to the DCE questionnaire. There is a near equal proportion of males (46%) and females (54%) respondents which show the usual higher proportion of male health workers especially Nurses and CHEWs in Northern Nigeria. With the respondents of this study being young health workers, there is potential to groom new sets of health workers with appropriate incentives to start life in desired locations and commit to a prolonged stay if the ndings are used to inform policy [26][27][28][29]. The outcome of the study will therefore be well targeted at the appropriate age bracket and demographics who may have longer working years to spend in rural postings if policy changes are considered for implementation from trends shown in this study.
The study suggests that the provision of basic housing is the most essential job attribute with the highest willingness-to-pay for health workers in Bauchi State. Bauchi State is largely rural and agrarian with sparse communities, hamlets, and villages. It is therefore plausible that the availability of housing will be a priority incentive for workers in this setting. Second to housing, quality of facility was identi ed as an important factor increasing utility for a job location choice. The reasons will not be far from the same one adduced for housing need. With the shortage of health workforce and low government expenditure on health infrastructure, health facilities located in remote areas are likely to be short-staffed, poorly equipped, and poorly supplied with materials for patient care. These will form a major disincentive for the retention of health workers. A Zambian study [30]found that technical quality of care, represented by the thoroughness of the examination, avoiding rude staff, and drug availability are the most highly valued attributes in uencing the demand for hospital care among hospital attendees. This report presents client satisfaction opinions and points to the level of motivation of workers to provide adequate and respectful services as well as healthcare management changes that will support improved healthcare services. Even though that the geography and settings may differ, this supports the notion that improving the quality of the facility and providing incentives like housing are likely to improve client satisfaction in these settings. Our ndings were corroborated by DCE studies conducted amongst health care workers which identi ed the provision of basic government housing as having the greatest impact on the probability of choosing a job [19,31]. Also, a study amongst healthcare students identi ed improved facility infrastructure as important attributes to health worker retention [21,26]. The ndings of our study place a low utility on salary despite the literature indicating that an increase in salaries is important [32] and it often causes strike actions in Nigeria's health sector [16]. Studies have also shown that salaries and other factors including those bothering on working conditions are often implied in instances of unrest [11].
The probability of choosing a rural post with the changing levels of attributes in the presence of an urban alternative as shown in the logit model is in accord with the results of the willingness to pay analysis. The high odds of choosing a rural posting with the provision of basic housing or housing allowance and improvement of quality of the facility was demonstrated in other studies conducted within the West African sub-continent [21,26]. These signi cantly demonstrate that the state and functionality of the health facilities are important when health workers are considering offers for job postings, whether in rural or urban locations [9]. They however represent better incentives for rural preference when other rural advantages like low cost of living and opportunities for farming for extra income are considered.
Studies on retention of health workers in Nigeria have suggested that the poor level of infrastructure in facilities as well as accommodation, where available, in rural areas impact negatively on the choice of health workers to take up or stay in postings. Thus, government investment in the provision of accommodation and face-lift of facilities would enhance retention in rural areas. This should be augmented by improvements in working conditions in rural areas in relation to ensuring that appropriate staff and equipment are in place, and investments to ensure improved career advancement [9].
Our study had some limitations. DCEs are known to suffer from limitations of offering hypothetical preferences that may not be correlated with real-life options [20,31]. In generating the job attributes and its dimensions, we ensured that it was informed by the qualitative perspectives of the study participants and a literature review. To ensure that the respondents understood the trading-off options, we thoroughly reviewed the questionnaire and explained the content and implications before their completion. This was done to reduce the likelihood pf the respondents not understanding the implications of the trade-offs to the study ndings.

Conclusion
This study applied the principles of a discrete choice experiment (DCE) to explore important factors responsible for the retention and motivation of frontline health workers to take up posting or continue to stay in rural locations. Of more importance is the establishment of utility values placed on certain job attributes with the willingness to pay demonstrated across attributes and levels. It is now evident that improving the working conditions of frontline health workers in terms of adequate staff strength, good skills mix, equipment, etc as well as improved opportunities for career advancement will support retention in rural health posts. Based on these ndings, we recommend the introduction of housing incentives in rural and remote areas of Bauchi State to support the retention of health workers. This should go alongside deliberate effort and strategy to improve the quality of the health facilities ensuring adequate and functional equipment and uninterrupted supplies. By focusing on the analysis of locally relevant, actionable incentives, generated through the involvement of policy-makers at the design stage, this study provides an example of research directly linked to policy action to address a vitally important issue in global health. The ndings presented in this study provide important evidence for input into policy formulation and guidelines for retaining frontline health workers in Nigeria as efforts are geared towards the achievement of universal health coverage and sustainable development goals.

AVAILABILITY OF DATA AND MATERIALS
The tools and data related to this study are available from the corresponding author on reasonable request.