This mixed-methods study used qualitative and quantitative research approaches sequentially. The study population included medical students, interns, residents, and practicing physicians from 3 participating osteopathic medical schools, associated partner residency hospitals, and a regional association of osteopathic physicians. Institutional review board (IRB) approval was obtained for the study from the Ohio University's IRB and from the IRB of each participating institution, as applicable.
Qualitative data were collected through 1 on 1 interviews and focus groups to create the survey. We solicited participants via email and word of mouth. The sample was a self-selected convenience sample. Interview and focus group scripts were developed from a critical literature review and a retrospective review of administrative data from a community teaching hospital to maintain consistency of questions. Three qualitative researchers (G.Y.D., K.C., and G.D.B.) conducted the focus group discussions and interviews and analyzed the data. Interviewees and focus group participants were selected from the study population. There were 51 focus group participants and 3 interview participants. Interviews and focus group discussions were documented through audio recordings and descriptive analytic field notes, and they ranged from 30 to 60 minutes.
Nine focus group sessions and 3 interviews were conducted. Each focus group was composed of 4 to 6 participants from the study population who volunteered to participate. Participants from the regional association of osteopathic physicians were excluded from the focus groups. Interviews, focus group discussions, and descriptive analytic field notes were transcribed and coded, and underlying themes that highlighted residency choices or preference factors were derived. Coding was done manually as well as using NVivo 10 software (QSR International). Comparisons were made between the manually extracted codes, categories, and themes and those generated by NVivo version 10 to extract the 10 final themes used in the survey. The 3 qualitative researchers independently examined the transcripts and developed codes, categories, and themes with eventual comparison to the NVivo results. This process was iterative, involving constant comparisons and consensus building until such time that no new themes emerged. At this point of saturation, further categories and themes became repetitive and redundant. The 10 factors of residency choice derived were societal and health policy changes, performance during clerkship, the matching process, financial considerations, mentor or preceptor influences, rotations, personality traits or self-concept, location preference or geography, work/life balance, and characteristics of the program.
Quantitative data were obtained from the survey that was developed based on data from the review of the literature, interviews, and focus group discussions. Experts in GME with psychometric training reviewed the survey items for face and content validity. The survey included 75 items, and responses were rated on a 5-point Likert scale ranging from 1 (“strongly disagree”) to 5 (“strongly agree”). Demographic and other attributive variables, such as age, current training status, and whether participants succeeded in securing their first-choice residency program were also collected. Participants were also asked to rank the 10 previously defined factors that contribute to residency selection on a scale of 1 (“most important”) to 10 (“least important”). Although these scales were ordinal, they were used as mean summated scales in the final data analysis. When used in that context, they are treated as continuous variables. The survey was administered electronically via SurveyMonkey to the same population from which the participants in the qualitative phase were drawn. The link to the survey was sent to contacts within each of the participating institutions to distribute to potential participants in fall 2013 through spring 2014. Over this period, 3 follow-up reminders were sent to the participants.
Survey results were analyzed using descriptive statistics involving frequencies and percentages for categorical variables, as well as means and SDs for continuous variables. Analysis of variance was used for inferential statistical analysis to gauge group differences with regard to the mean scores of the residency choice factors delineated. Partially completed surveys were included in the analysis. Item analysis was conducted to determine the internal consistency reliability of the survey items (Cronbach α=0.886). Statistical significance was set at P≤.05.