Survey Procedure
Participants were recruited to participate in a 2011 national study of the professional development of physicians-in-training. In September 2009, samples were drawn from the American Medical Association Physician Masterfile, which has a near-complete listing of students pursuing M.D. degrees at schools within the U.S. and its territories. To construct our target sample, we selected 960 third-year students from 24 allopathic medical schools in the U.S. using a two-stage sample design. In stage one, we selected schools with probabilities proportional to total enrollment so that the larger schools would have a greater chance of being included in the study. Data for allopathic medical school sampling was obtained from published reports. In stage two, we used simple random sampling to select a fixed number of students (40) from each selected school.
After a relevant literature review in medical ethics/professionalism, survey questions underwent expert review by colleagues, as well as pre-testing by a group of third-year medical students at one University in the U.S. Midwest. Quantitative data collection was conducted in two phases: administration of self-report Questionnaire 1 occurred between January and April 2011 and assessed demographic variables including gender, race/ethnicity, immigration history, specialty intention, social mission score of the medical school, and other information about medical school experiences (Time 1; students were third-year); a re-contact phase administering Questionnaire 2 occurred six to nine months later to those students who responded to Questionnaire 1 (Time 2; third-year students had become fourth-year). Participants were paid $5 for completion of the first questionnaire, $10 for completion of the second questionnaire. Participants in this study were generally third-year medical students (94.8 %). Using a combination of postal mailings and email links to the online versions of the questionnaires, we obtained a response rate of 63% for Time 1 (605/960). At Time 2 our sample size decreased to 499 participants.
Content Analysis
In the Time 2 Questionnaire, we asked the following open-ended survey item which had been adapted from a previous study(3): “Lastly, please describe a clinical experience you observed that, in your opinion, raised an ethical or professional issue. Then describe how you thought the situation should have been approached.” Out of a total of 499 respondents, only 144 respondents (28.9%) offered a legible response to this survey item.
We performed content analysis of each of the responses (n=144), which ranged from single-sentence answers to long paragraph responses. Three undergraduate student investigators (JK, MJ, CH) were provided an initial list of codes from the original Kaldjian taxonomy(3) which contained codes for 7 major themes (Decisions regarding treatment, Communication, Professional duties, Justice, Student-specific issues, Quality of care, and Miscellaneous) and sub-codes for 32 sub-themes (listed in Table 1). Using this list, these three student investigators independently coded each survey response and reached consensus on the ethical and professional issues deemed present. Coded text was entered into NVivo 12.1 (QSR International, 2018). One internal medicine fellow investigator (MH) and two faculty investigators (JY and LK), all of whom have been formally trained in clinical ethics or philosophy, then reviewed all the coding results to confirm agreement. Then through an iterative consensus-building process with the student investigators, further adjustments to major codes and sub-codes were finalized. In the minority of cases when coding of responses did not reach unanimous consensus within the team, the final decision was made by the faculty investigator with the most familiarity of the original taxonomy (LK). In this way, we developed a final list of major codes and sub-codes that provided a summary analysis of students’ responses.
We then re-examined students’ responses using a virtues taxonomy utilizing a list of virtues most commonly referenced in a study of medical oaths used in North American medical schools(14). Student investigators (JK, MJ, CH) were provided a list of these 16 virtues (listed in Table 2) and their corresponding Oxford English Dictionary definitions (Appendix). Then using similar procedures as noted above, we coded all the responses using the virtues categories through an iterative, consensus-building process. As noted above, when coding of responses did not reach unanimous consensus within the team, the final decision was made by the faculty investigator with the most familiarity of the original list of virtues (LK).
We approached data analysis with a post-positivist lens, by acknowledging the potential effects of our personal biases. Two authors (JK, CH) had previous training in qualitative data analysis, and together with the faculty investigator (JY) trained the rest of the student team to use these methods rigorously. Another investigator (LK), an experienced clinician-ethicist who was also well-versed in qualitative data analysis, helped guide theoretical discussions along with the other authors (JY and MH). Together, our combined expertise allowed for a careful examination of the data while our staged, iterative approach helped protect against potential biases.
This study was approved by the Institutional Review Board in January 2011. The illustrative examples in the supplementary table (see appendix, available online only) have been modified in non-essential respects to remove identifiable information.