The survey was developed by the authors and was informed by existing United States Centers for Disease Control and Prevention (CDC) guidelines as well as recommendations on infection control precautions for healthcare workers, and suspected or confirmed COVID-19 positive patients. The survey, along with an informed consent agreement, were input into Qualtrics.xm software (Qualtrics, Provo UT) with the system logic set to prohibit progression of the study without consent. This survey was distributed by email individually to all 263 allopathic and osteopathic General Psychiatry Residency Program Coordinators and Directors individually, as well as to all programs via the American Association of Directors of Psychiatric Residency Training (AADPRT) Program Director listserv and Program Coordinator listservs, to distribute to resident trainees if deemed appropriate. Fellowship Directors were not directly contacted. The survey was open from April 17, 2020 to April 29, 2020 and anonymized data was retrieved on closure of the survey.
This study was found exempt by the IRB (IRB_00131976), with no formal ethics approval required due to the anonymous nature of the survey and its adherence to national guidelines, however, informed consent was still required prior to completion of the survey. After consenting, respondents were then directed to a sixteen-question survey (see supplementary materials). The survey included socio-demographic questions relevant to training (region of program as defined by the AADPRT), year in training, number of trainees per class, inpatient sites requiring coverage, and the number of COVID-19 cases in the state in which they are practicing at the time the survey was completed. Study-specific questions regarding perceived departmental changes such as consult policy changes, call changes (including several approaches), policy implementations for suspected and confirmed COVID-19 patients, and policies involving PPE comprised the next section of the survey. Several trainee concerns were also examined via Likert scales, which, for the purposes of statistical comparison, were treated as an ordinal approximation of a continuous variable as all outcomes had at least five categories or were summed across multiple questions13–16. These scales measured trainee concern for contraction and spread of COVID-19 to patients and family members, perceived PPE shortage, trainee burnout, perceived overall risk to trainees compared to attending clinicians, trainee comfort in communicating concerns to leadership, and overall satisfaction of department response. Trainee respondents were also encouraged to communicate any further comments at the end of the survey.
Statistical analyses evaluated trainee concerns, including perceived program responsiveness, implemented changes, and comfort level in communicating with the leadership. Briefly, primary outcomes were converted to numeric scores across one or more Likert scales, including satisfaction of department response by trainees (5 levels, scored 0–4, with higher values indicating higher satisfaction), perceived risk of trainee burnout (two 4-level scales merged, scored 0–6, with higher scores indicating greater perceived risk), and perceived trainee infection risk (five 4 to 5-level scales merged, scored 0–16, with higher scores indicating higher perceived risk). We examined the relationship between these scored outcomes and trainee region, number of residents per class, number of hospitals covered by trainees, perceived changes to consult requirements for trainees (yes/no), perceived changes to rotations or call expectations for trainees (yes/no), perceived overall change to trainee workload (converted to a numerical sum of changes with + 1 for each perceived increase in workload and − 1 for each decrease in workload), perceived trainee changes in responsibility toward COVID negative and COVID positive patients (converted to a numerical sum of responsibilities for each patient group, + 1 for each additional responsibility), perceived changes in PPE policy by trainees (yes/no), and trainee comfort level communicating concerns to leadership (5-level scale, scored 0–4 with higher values indicating increased comfort).
All statistical analyses were performed using the R statistical framework17 utilizing simple linear regression. Outputs were collected of Beta (β), standard error (SE), and p-value (P). All p-values were subjected to conservative Bonferroni correction for a total of 29 comparisons, with a significance threshold of 0.0017 (0.05/29 test) required for study-wide significance.
Any survey responses that were not answered by a given respondent or demonstrated clear numerical inaccuracy (such as PGY class sizes larger than any U.S. psychiatry programs) were declared as missing within the analyses. As numerical results regarding the current number of COVID-19 cases in each state were highly variable, even within the same AADPRT region, this question was excluded from the analyses. Results from the statistical evaluation of trainee responses were plotted as boxplots within R using the ggplot2 library18. Finally, free text comments provided as part of the survey were independently rated as subjectively positive, negative, mixed, or neutral between two raters (JO, TGJ). Only statements that were agreed upon as positive, negative, or mixed were included as examples in the discussion.