To achieve the United Nations Millennium Development Goals (MDGs) for health, improving access to key interventions such as anti-retroviral therapy, immunizations, and tuberculosis and malaria treatment, are top priorities for most health systems [1–3]. However, in addition to financial resources for commodities [4] scaling up requires a well-functioning health system and adequate workforce capable of delivering interventions at a large scale [5–9]. However, the 2004 Joint Learning Initiative report on Human Resources for Health and others have concluded that shortage and mal-distribution of health personnel, with few in rural areas, undermine scaling up efforts, particularly in low-income countries [10–13]. In sub-Saharan Africa it has been argued that nearly 1 million additional health workers are needed to realize MDG health outcomes [14].

In Kenya understaffing in rural primary care facilities (dispensaries and health centers) is particularly acute. Staffing norms for dispensaries are 2 to 5 nurses, but 9% have no nurse at all and 48% only one. Similarly, health centers should have 8 to 12 nurses, with the larger health centers expected to offer short-term residential basic obstetric care, but 48% have less than three [15] (physicians are not an expected part of the workforce at either of these rural facility types). The most obvious solution to address the rural and general nursing workforce shortage is a major increase in recruitment, as there are many qualified and unemployed nurses in Kenya [1]. In an attempt to address staff shortages quickly, donor partners have supported recruitment of just over 3000 nurses from 2005 to date, hired on short-term contracts of 2 to 3 years duration, and posted to public sector facilities in underserved rural areas with high HIV prevalence [16]. However, major shortages remain.

Posts in rural areas are often considered less desirable. Urban settings have superior infrastructure and services [3, 17, 18] while rural posts are often associated with high workloads and poorly aligned incentive systems (for example lower housing allowances) that disadvantage rural staff. Once they are posted in rural areas, some health workers make repeated demands to be reallocated to urban postings (Chief Nursing Officer, personal communication).

There are two main cadres of nurses in Kenya: registered nurses (RNs) and enrolled community nurses (ECNs). RNs and ECNs undergo a 3.5 year and 2.5 year pre-service training respectively. The majority of Kenya’s nursing colleges now train predominantly RNs (90% of trainees) while previously the majority of trainees were ECNs. Rural posts in health centers and dispensaries are currently mainly filled by ECNs while posts in hospitals are mainly filled by RNs. As of 2008, entry-level RNs and ECNs received monthly basic salaries of KES 22 519 (USD 289) and KES 11 518 (USD 148) respectively converted at a rate of (1 USD = KES 78.19) [19]. ECNs have the option to upgrade to RN status either through a 1.5 year on-campus program or a 2 year distance learning program, which many are currently choosing to do, given the salary differentials and the limited career progression for ECNs. A minority of RNs have a bachelor’s degree but this group is not considered in this paper.

The debate in Kenya over how to ensure adequate numbers of nurses in rural areas to provide and supervise essential community and primary care services is poorly informed by evidence. This study, focusing on RN trainees near graduation, set out to explore the views of final year nursing students on rural and urban areas, nursing posts in rural areas, and strategies to recruit and retain nurses. Results are presented separately for those completing pre-service training and those upgrading from ECN status, as variation in characteristics and experience between the two groups is expected to affect their attitudes.

Between September 2008 and January 2009 data were collected from 345 RN students in their final year of training at four Medical Training Colleges (MTCs), using a structured self-administered questionnaire (SAQ), and focus group discussions (FGDs). MTCs were eligible for the study if they had at least 30 pre-service and 30 upgrading students scheduled to complete training in 2008. Of a total of eight eligible MTCs, four were purposively selected (Nairobi, Murang’a, Meru and Kakamega). Nairobi MTC is located in the capital city, while Murang’a, Meru and Kakamega are 87 km, 238 km and 355 km from Nairobi respectively, representing a range of settings across four provinces. All students who intended to sit their final exams in the second half of 2008 were invited to participate.

The SAQ data covered respondents’ backgrounds, training plus sets of likert scale statements, designed to assess their attitudes towards living and working in rural areas. The terms “rural” and “urban” are open to varied interpretation [20]. We advised students completing the SAQ to consider rural areas as those in remote places, far from major cities and towns and with poor infrastructure and limited recreational facilities, though they may also have incorporated their own individual views of these terms. Their interpretations were explored further in the FGDs. The development of the likert scale statements drew on those reported in the literature [21], adapted to improve applicability to the local context through discussions with locally qualified nursing staff and researchers. Statements were adapted to ensure that they covered all key issues raised by local stakeholders, that all issues were relevant to the Kenyan nursing profession, and that the language was easily understandable in the local context. Students were asked to respond to each statement on a 6 point scale of 1 (strongly disagree), 2 (moderately disagree), 3 (somewhat disagree), 4 (somewhat agree), 5 (moderately agree) and 6 (strongly agree). To reduce response bias, 18% of the statements were phrased negatively, and randomly distributed within question sets. The SAQ tool was pre-tested with students from two MTCs not included in the study, including a group discussion of the meanings of the questions and statements.

Two FGDs were carried out in each MTC, one each for pre-service and upgrading students. 6-8 participants were randomly selected for each group from all those completing the SAQ. Each group included both male and female students. The discussions covered trainees’ experiences, attitudes towards rural areas, rural postings, and nursing in general, and potential interventions to improve health worker distribution. Discussions lasted between 45 and 90 minutes, and were digitally recorded, supplemented by note taking. Draft topic guides were piloted with nursing students from urban and rural MTCs.

The voluntary nature of participation was emphasized, and informed written consent was obtained from all participants. No MTC tutors were present during data collection.

Data from the SAQ were double-entered and verified. Simple descriptive analysis was used to compare the characteristics of pre-service and upgrading students, and to calculate the mean response (and 95% confidence interval) for each likert scale statement. As each likert scale statement was rated on a scale from 1 to 6 we considered an upper confidence interval less than 3.5 as an indication of general disagreement with the statement, and a lower confidence interval above 3.5 as an indication of general agreement. Principal components analysis (PCA), a data reduction technique that enables one to identify from a large set of variables those that contain information common to all, was performed across all statements on rural/urban perceptions to determine how the statements grouped together under the broad constructs of ‘perception of life in a rural area’ and ‘perception of working in rural areas’. Additional PCAs were then run within these two constructs, the first component for each construct was retained and PCA derived scores calculated for each construct. These scores were used as dependant variables in separate linear regressions to assess the influence of a range of trainee characteristics on attitudes to rural life and work. All analyses were conducted in STATA version 10.1 (Stata, http://www.stata.com).

FGD recordings and field notes were reviewed for clarity, transcribed, uploaded into the qualitative analysis software, NVivo7, and subjected to content analysis. This involved development of a coding “tree” or thematic framework. A draft coding tree was developed from the FGD topic guide, tested by two researchers separately on about 30% of all data collected, and refined using themes emerging from transcripts. Information under each code was then compiled and tabulated to obtain a clearer picture of the issues arising from the data, and to compare views across different groups of participants.

Results are presented on respondent characteristics, their perceptions of life and posts in rural areas, and their perceptions of strategies to improve rural recruitment and retention, drawing from both the SAQ and FGD data.

Inadequate human resources have been identified as a key constraint to scale up of health interventions [7, 8]. Mal-distribution of health workers continues to be a global challenge with high concentration of health workers in urban areas while rural areas are understaffed [22]. This situation threatens equitable delivery of health services to people living in rural areas who are often less educated, poorer and with a higher disease burden. There has been a general neglect of health policies which focus on strengthening human resources for health, reflected in poor budget allocations and limited time within the national health agenda devoted to this topic in both developed and developing countries [5, 20]. This paper provides evidence from Kenya about the challenges of recruiting and retaining nurses in rural posts, and the potential of a range of strategies to address this.

We studied responses from both pre-service and upgrading students. The latter were older, more likely to be married, to have children, and to have been born in a rural area than pre-service trainees. More upgraders reported that nursing was their first career choice. This may have reflected a survivor bias as upgraders interviewed had already worked in the profession for several years, or the declining popularity of the nursing over time, reflecting the reduced attractiveness of the profession (e.g. due to relative pay, prestige and support), or the wider range of career options available in more recent years. However, responses from the two groups in the SAQ and FGDs were broadly similar; the only significant differences in likert scale responses were that pre-service students were more positive about the potential impact of greater responsibility and improved housing on rural recruitment. Multivariable analysis showed that when other factors were controlled for, being an upgrading student had a significantly negative impact on perceptions of rural posts. The reasons for this are unclear, but could reflect their greater experience of the realities of rural posts, or a preference for urban areas making it more likely that an ECN will choose to upgrade. Kenya is planning to phase out ECNs, an approach already adopted in other African countries [23]. There is concern that as a result Kenya will have a more qualified nursing workforce which is more marketable locally and internationally, and less willing to work in rural areas. However, the data do not allow a direct comparison of RN and ECN preferences as all respondents were close to becoming RN graduates at the time of data collection.

Across both groups likert scale responses indicated generally negative perceptions of nursing posts in rural areas, although stated preferences were rarely strong. During FGDs trainees raised both positive and negative aspects of rural life. On the positive side they mentioned lower costs of living and more autonomy at work. On the negative side they cited poor infrastructure, inadequate education facilities and opportunities, higher workloads, and inadequate supplies and supervision, factors also highlighted in other studies [22]. Parents working in rural areas may also be forced to separate from their children who attend schools in urban areas, an issue also highlighted in South Africa[24]. Particular concern was expressed about working in communities dominated by other tribes, reflecting Kenya’s recent election-related violence.

Quantitative analysis showed that attitudes to living in rural areas were significantly positively affected by studying at the MTC furthest from Nairobi, and attitudes to working in rural areas were positively affected by being older. One might expect students born in a rural area to have more positive attitudes towards living and working in such locations. For example a review related to doctors found strong evidence that students with a rural origin are more likely to practice in a rural setting [20]. However, in this study, rural birth did not have a significant impact on attitudes once other factors where controlled for. However, such conclusions should be considered tentative given the relatively low Cronbach’s alpha associated with the PCA derived constructs and the limited explanatory power of the regression models.

Students expressed mixed views on the relative merits of working in private and public sectors. While pre-service students perceived commercial and NGO facilities to pay the highest salaries, upgraders emphasised that the FBO organisations had the lowest salaries despite their high workload. The fact that the FBO segment of the private sector is located predominately in rural areas, while the commercial segment is predominately in urban areas, may be a further factor influencing rural/urban job choices.

Quantitative and qualitative data indicated that students believed a number of strategies could work to improve rural recruitment and retention, with particular emphasis on the introduction of substantial rural allowances. Many studies point to the value of financial incentives to enhance rural recruitment and retention [25], although it has been argued that multidimensional programmes are more successful than those relying on financial incentives alone [20]. The implementation of such strategies requires clear definitions of rural areas to ensure equity and fairness in distribution of allowances [26], and the allocation of adequate funds to ensure sustainability [27].

Pre-service trainees perceived that increased responsibility would motivate nurses, a factor also noted in other studies [22]. FGD data also indicated that trainees thought more supportive supervision would encourage nurses to practice in rural areas. Other studies have highlighted the important role that managers can play in job satisfaction [23, 28].While there is some evidence that improved infrastructure and supportive supervision improve retention [29, 30], it is worth noting that infrastructural problems themselves limit the amount and frequency of supervisory visits to remote areas. Views of compulsory rural service were positive but not strongly supportive; other studies have found that the evidence that this has a positive long-term impact on rural-urban mismatch is weak [20].

Qualitative work revealed, importantly, that in a context of political instability, interventions aimed at improving recruitment and retention in rural areas may fail to meet their objectives. As a consequence of recent post election violence, trainees expressed concern about accepting rural posts where ethnic tensions may affect their personal security, highlighting that for rural recruitment strategies to be attractive to nurses, their basic safety needs must be assured [24]. The ability to select the location of a rural post was also highlighted as likely to encourage application during FGDs and reflected in likert scale results. Interestingly, the recently introduced short term, donor funded contracts that target critical human resource gaps in underserved areas were regarded as less attractive than standard government posts due to their lack of pension plans and job security, despite a stated policy that such posts be absorbed by the government at the end of the contract. However, applications for such posts are plentiful (Chief Nursing Officer, personal communication), probably reflecting both the strong supply of nursing graduates and the lack of alternative opportunities in Kenya.

A number of limitations should be highlighted. First for logistical reasons it was not possible to select a nationally representative sample of trainees. In addition, data were collected within a year of the post election violence which may have modified nurse’s perceptions of rural postings. Thirdly, the term “rural” is difficult to define [20] and its interpretation is likely to have varied across interviewees. For example, some may have considered district towns as urban, while others may have classified them as rural. However, the characteristics students identified for rural areas during FGDs were similar to those that we proposed during quantitative data collection. Other terms used in the SAQ may also have been open to varied interpretations such as “lifestyle”, “motivation” “stressful” and “decent housing”. The list of potential interventions assessed was not comprehensive; others identified as potentially useful in other studies include conducting training placements in rural settings, providing scholarships with an enforceable rural service agreement, accelerated promotion, and increased recognition by the employer or community [20, 22, 27].

Finally, the findings are based on the stated opinions of trainees and may not translate into actual career decisions. It is notable that while many strategies to improve recruitment and retention have been proposed in the literature [31], few studies have investigated their effectiveness. Further research is therefore necessary to investigate the extent to which the intentions of qualifying nurse graduates translate into career moves, and the effectiveness of interventions to influence recruitment and retention patterns over time. In developed countries, retrospective studies have been conducted to document the career paths of health workers [32] but similar databases are not available in Kenya, while results from the developed world are unlikely to be applicable to developing countries which have very different health labor markets and health workforce distribution [33].

The issue of workforce shortage and mal-distribution is complex and not unique to the nursing cadre or to Kenya. Poor infrastructure, limited training opportunities, high workloads, inadequate supplies and supervision, undisclosed job locations for public sector jobs, and most recently political instability all continue to be barriers to successful rural recruitment and retention. Interestingly we found no suggestion that those born in or with experience working in rural areas are more willing to seek rural employment. While donor funded short-term contracts have increased recruitment in recent years, it is possible that their impact will be compromised by their unpopularity among nurses due to their lack of pension plans and job security. The most popular proposed policy intervention among respondents was the provision of additional financial incentives for rural posting, though these may be more effective if implemented as part of a multi-dimensional package. Such a package would require collaboration between economic and health policy-makers to earmark funding to not only secure salaries but also improve working conditions. It should also be accompanied by investment in information systems capable of monitoring its impact with rigor.