Students’ intentions to practice primary care are associated with their motives to become doctors: a longitudinal study

Background Medical schools can contribute to the insufficient primary care physician workforce by influencing students’ career preferences. Primary care career choice evolves between matriculation and graduation and is influenced by several individual and contextual factors. This study explored the longitudinal dynamics of primary care career intentions and the association of students’ motives for becoming doctors with these intentions in a cohort of undergraduate medical students followed over a four-year period. Methods The sample consisted of medical students from two classes recruited into a cohort study during their first academic year, and who completed a yearly survey over a four-year period from their third (end of pre-clinical curriculum) to their sixth (before graduation) academic year. Main outcome measures were students’ motives for becoming doctors (ten motives rated on a 6-point scale) and career intentions (categorized into primary care, non-primary care, and undecided). Population-level flows of career intentions were investigated descriptively. Changes in the rating of motives over time were analyzed using Wilcoxon tests. Two generalized linear mixed models were used to estimate which motives were associated with primary care career intentions. Results The sample included 217 students (60% females). Career intentions mainly evolved during clinical training, with smaller changes at the end of pre-clinical training. The proportion of students intending to practice primary care increased over time from 12.8% (year 3) to 24% (year 6). Caring for patients was the most highly rated motive for becoming a doctor. The importance of the motives cure diseases, saving lives, and vocation decreased over time. Primary care career intentions were positively associated with the motives altruism and private practice, and negatively associated with the motives prestige, academic interest and cure diseases. Conclusion Our study indicates that career intentions are not fixed and change mainly during clinical training, supporting the influence of clinical experiences on career-related choices. The impact of students’ motives on primary care career choice suggests strategies to increase the attractivity of this career, such as reinforcing students’ altruistic values and increasing the academic recognition of primary care. Supplementary Information The online version contains supplementary material available at 10.1186/s12909-021-03091-y.

would ignore the dependence between observations of the same individual. The classical approach for modelling longitudinal data of this type is to rely on generalized linear mixed models (GLMM, see below) [1], which can model the difference between groups (individuals in our case) by specifying random effects that are common to all observations of a group. Given the data provided, and as the estimation of GLMM relies on the approximation of intractable integrals and is known to be unreliable with complex random structure, we considered a simple random structure with a random intercept per individual. As the logistic regression model only allows to distinguish two classes, we first compared primary care committed students to students intending to practice a non-2 primary care specialty (model 1). We then compared students intending to practice primary care to those who were undecided (model 2).
To ensure a reliable estimation of the parameters, we considered the Event Per Variable (EPV) rule of thumb (which specifies to have at least 10 observations per variable) [2] and eliminated three motives from the analysis: reward because of a higher number of missing data, mission because of its high inter-correlation with vocation (>0.6), and care for patients because of its very low variance over study years.

Interpretation and predictive power
The interpretation of our models does not differ from that of a standard logistic regression for the fixed effect. The sign of the estimated coefficients (positive or negative) indicates the direction of the effect of the variable on the probability of being assigned to the class of interest (i.e., the intention to practice primary care). In our study, a positive sign of the estimated coefficient suggests that a positive increment of the associated variable will increase the probability of being interested in a primary care career. The magnitude of the estimated coefficients is directly related to the scale of the variable. Thus, the magnitude of the estimated coefficients of two variables can only be meaningfully compared when the two variables are on the same scale. In our study, this means that the coefficients relative to the "motives" variables can be compared to each other, but that the coefficient related to the "gender" and "age" variables cannot be compared to the others, for example.
For a complete discussion of the interpretation of estimated coefficients in GLMM, we recommend [3]. 3 The estimated variance of the random intercepts gives insights on the importance of effects that are not accounted for in our model with respect to the observed variables. We obtained a measure of the importance of the fixed effects in the modeling of the probability of primary care career intention by performing a leaveone-out cross-validation procedure (LOOCV), considering only the estimated fixed effects. We obtained a LOOCV accuracy of 78.14% (model 1) and 64.68% (model 2), suggesting that both models provide adequate predictions.

Illustration of the relationship between motives to become doctors and the probability to be interested in a primary care career
To illustrate the impact of the importance of the rating of the different motives for becoming a doctor on the probability of primary care career intentions, we created a Webapp that interested readers may access directly. By allowing to combine different ratings of motives, this app in fact visualizes the effect of the different magnitudes of the estimated coefficients of the "motives" variables (explained above) on the probability of indicating an intention to practice primary care in the future (π).