Table 2.
Rank-ordered resident preferences for seeking support for work-related stress and/or burnout (if you were feeling stressed or burned out from your work, from whom would you seek support?): frequency counts, best-worst scoring, multinomial logistic analysis, and latent class segmentation rank ordering.
| Surveyed factors | Aggregate data | Latent class segmentation (rank order) | ||||||||
|
|
Number of times selected (maximum possible: 308) | Best-worst score | MNLa utility score (SE)b | Rank order | Segment 1: “Open to formal help,” (39.4%c) | Segment 2: “Not open to therapy,” (26.8%) | Segment 3: “Open to isolating” (25.9%) | Segment 4: “Formal help-seeking“ (7.9%) | ||
|
|
Best choice | Worst choice |
|
|
|
|
|
|
|
|
| Speaking with my other peers that are still in residency training | 219 | 2 | 0.705 | 0.077 (0.15)A | 1 | 2 | 1 | 2 | 1 | |
| Speaking with my family and friends outside of work | 215 | 5 | 0.682 | 0.000 (N/Ad)A | 2 | 1 | 2 | 1 | 3 | |
| A counselor or therapist one-on-one | 36 | 53 | −0.055 | −3.01 (0.211)B | 3 | 3 | 6 | 4 | 2 | |
| Residency-sponsored peer support group, (like Intern Forum, but not necessarily just for interns) | 30 | 49 | −0.062 | −3.088 (0.212)B | 4 | 4 | 5 | 5 | 6 | |
| Speaking with supportive attending physicians not directly involved in the administration | 5 | 71 | −0.215 | −3.404 (0.213)B,C | 5 | 5 | 4 | 6 | 5 | |
| Speaking with my administration (chief residents or program directors) | 23 | 130 | −0.347 | −3.919 (0.215)C | 6 | 6 | 3 | 7 | 4 | |
| No one, I don’t like seeking support from others | 11 | 229 | −0.708 | −5.043 (0.224)D | 7 | 7 | 7 | 3 | 7 | |
aMNL: multinomial logit.
bFour statistically significant groupings (A-D) were found for the 7 surveyed factors when comparing multinomial logit utility scores with Tukey honest significant difference posthoc testing. Mean utility scores followed by the same letter did not differ significantly (Tukey honest significant difference test, P>.05); exact P values for multiple comparisons are shown in Table S1 in Multimedia Appendix 1.
cThe latent number of groups displayed was based on the lowest Bayesian information criterion. Owing to the probabilistic nature of the latent class method, respondents do not wholly belong to one group or another, although most respondents (74/77, 96%) had >90% probability of membership to a single group.
dN/A: not applicable.