Methods
Description of Study Populations
This study was designed to compare the general happiness and career satisfaction of primary care fellowship alumni from a moderate-sized rural-serving medical institution with an associated teaching hospital (900 beds), to a general population of primary care physicians in North Carolina. Fellows were chosen from an alumni database from the Brody School of Medicine (BSOM) primary care fellowship programme that spanned from 1995–2011. Primary care physicians were chosen randomly using the North Carolina State Medical Board database of practitioners (2011 edition), and choosing physicians who self-designated no fellowship, specialty, or additional training beyond general family medicine, paediatrics, or internal medicine. These general practitioners were chosen from a similar age demographic as fellowship alumni, with limitation to those who indicated a primary care specialty and who had graduated since 1980 (less than 60 years of age).
Survey Instrument
A written survey instrument was sent to 120 BSOM fellowship graduates. The survey instrument consisted of 12 practise-type and demographic questions, 11 questions regarding their fellowship experience, and 3 previously validated psychological instruments. Scale 1 used was the Satisfaction with Life Scale developed by Deiner, which is a 5-item Likert scale survey that measures the respondents' satisfaction with life as a whole, and does not contain any domain-specific items. It is scored on a scale of 5–35, with higher scores equating to increased satisfaction. Scale 2 used was the Perceived Stress Scale-4 item version developed by Cohen, and is a survey designed to measure overall stress levels of adults. It results in a stress scale of 0–16, with higher scores equating to higher levels of perceived stress. Scale 3 used was the Subjective Happiness Scale developed by Lyubomirsky, and is a 5-item survey designed to measure global subjective happiness of the respondent. It results in a combined and averaged score from 1 to 7, with higher scores equating to higher levels of subjective happiness. Each of the three chosen survey instruments have been validated extensively and used widely for their respective measures. It should be noted however, that the three scales have not been used extensively in measuring medical professionals and are not meant to assess any psychological conditions. The instrument was tested for content validity, readability and clarity with a sample of faculty (MD, PharmD, and PhD), and professional staff (n=7) at the Brody School of Medicine and East Carolina University.
A second survey, containing the same three measures, was completed in conjunction with the alumni survey. It was sent to a stratified random sample of active Family Medicine physicians in North Carolina (1150), with a return rate of 21.57% or 248 (203 useable) surveys. This sample produces a margin of error of 2.7%. Any survey with completed psychological scales was included in the final analysis. Surveys missing demographic information were used if the scales were completed, but were not included when comparing groups by demographic variables.
Analysis
The 120 surveys were sent on a single occasion to the last known addresses of alumni from the BSOM Faculty Development Fellowship. No follow-up reminders or second mailings were performed. Of the total surveys mailed, 53 surveys were returned (return rate of 44%), and 50 were complete and useable. Surveys were considered adequate if the physician had responded to the three psychological scales. Surveys were entered into a database and analysed using the Statistical Package for the Social Sciences V.19.0 (Armonk, New York, IBM). Descriptive statistics were used to present practice demographics, t tests (a 0.05) to determine group differences, linear regression used on scores and demographic variables, and ANOVA to determine coefficient of determination between groups.
Data were used to calculate overall means for the three psychological scores to determine general population responses. Within-group analysis was performed for each group, with gender, marriage and children status used as variables. Between and subgroup group analysis consisting of t test and ANOVA was used to determine if significant differences existed between the fellowship-trained and general physicians populations in general, as well as within subgroups. Linear regression was used to determine the strength of the various significant relationships among subgroups.