Used for many years in market research, one tool with potential to suggest working condition priorities is conjoint, or trade-off, analysis. This constellation of techniques provides the researcher with the ability to elicit individuals’ stated preferences. A common form of conjoint analysis is the discrete choice experiment (DCE), in which respondents (e.g. midwifery students) are presented with a choice of several competing hypothetical job posting scenarios each characterized by a variety of attributes (e.g., salary, housing, offer of a car allowance). Respondents are then asked to select their preferred scenario. The benefit of this model is that of opportunity cost. We know individuals want the best of everything, but when resources are limited, the DCE gives a weighted relevance to distinguish which attributes are the most highly incentivizing [14] to motivate individuals to locate to rural areas. Well established in the use of inferring patients’ preferences, this technique has recently been used to test providers’ preferences as well [15].
In designing the DCE, we selected attributes (motivators) to increase attraction to rural practice based on conversations with the Ghana Ministry of Health and eight focus groups with final-year midwifery students (n = 49) at the two largest Ghanaian midwifery training colleges. The seven attributes identified by content analysis [16] included: salary, study leave, housing, supportive management, infrastructure, transportation and children’s education. The DCE was then designed to estimate the relative value or utility of different work conditions that might incentivize students to locate to rural areas to practice after graduation. The survey consisted of demographic and background questions followed by a series of 11 discrete choice questions. In these questions, students were asked to compare two hypothetical job postings (Figure 1).
Participants were asked to imagine that upon completion of their midwifery training, they were offered two postings in two rural deprived areas by the Ministry of Health. Deprived area was defined as an area that is distant from a big city with few social amenities such as schools, roads, or pipeborne water. Participants were asked to imagine themselves making a real decision between two rural postings and were asked of the two offered, which they felt was better. Further, students were asked to answer whether or not they would accept this posting if it were offered.
Setting and sample
Midwifery education in Ghana is a three-year, post-secondary school diploma program. Since 2003, fourteen midwifery training schools in Ghana have been accredited. Of the ten regions in Ghana, each is home to at least one midwifery education program. There is a national curriculum with the first three semesters focused on general nursing and the final three semesters devoted to midwifery knowledge and skills. Students spend at least one clinical rotation at a rural district hospital.
We chose third-year midwifery students (n = 238) about to graduate and considering employment perspectives for our sample. We used purposive sampling to obtain a wide diversity of experiences and opinions. Students at two of the largest midwifery training schools in Ghana were invited to participate in the study. These two schools combined graduate the largest number of midwifery students per year. This survey was part of a larger collaboration between the University of Michigan, the Kwame Nkrumah University of Science and Technology and the Ghana Ministry of Health. The research was approved by the Ghana Health Service Ethical Review Committee, the Kwame Nkrumah University of Science and Technology Committee on Human Research, Publications and Ethics, the University of Ghana Medical School, and the University of Michigan Ethical Review Board.
Data collection
Informed consent was obtained prior to participation in the DCE. Each computerized survey took approximately 30–45 minutes. Students were given an incentive of 10 Ghana Cedis (approximately 7 US dollars) upon completion of the survey. Students signed in and the names were compared to a class list generated by the head of each college to determine response rate.
Data analysis
Sawtooth Software (Orem, UT) was used to construct, field and score the surveys. Using market simulator software in Sawtooth’s Choice-Based Conjoint with Hierarchical Bayes module, we used individual-level utilities to estimate the proportion of respondents who would prefer specific incentive packages. The software calculates total utilities of the simulated options for each respondent by summing attribute utilities. The respondents were repeatedly sampled to stabilize these preferences. In addition, we added a random error term to the estimates of utilities to correct for any similarities in scenarios. We used Sawtooth’s Choice-Based Conjoint with Hierarchical Bayes statistical program to estimate coefficients for the individual utilities of each attribute level.