Early Detection of Burnout - Healthcare Workers

Overview

Burnout is a public health issue. Healthcare workers are particularly at risk of burnout with occupational stress identified as the major risk factor. The "Health Work Environment" service is composed of physicians, nurses and psychologist with the aim of providing efficient and adapted care for healthcare workers at CHU of Clermont-Ferrand. In addition, they must ensure a role of primary, secondary and tertiary prevention. With regard to burnout, the majority of the work carried out concerns tertiary prevention, that is to say the care of a person in a situation of burnout. It would be necessary to carry out secondary prevention in order that people at risk of burnout can be detected earlier. However, there is currently no individual questionnaire to detect early burnout

Early detection of burnout was designed to provide a better understanding of warning sign of burnout in order to detect earlier workers at risk. In the present protocol, parameters are measured on several occasions (0, 1 week, at 6 month, at 12 months and then at each occupational visit within 5 years). Statistical analysis will be performed using Stata software (version 13; Stata-Corp, College Station, Tex., USA). All statistical tests will be two-sided and p\<0.05 will be considered significant. After testing for normal distribution (Shapiro-Wilk test), data will be treated either by parametric or non-parametric analyses according to statistical assumptions. Inter-groups comparisons will systematically be performed 1) without adjustment and 2) adjusting on factors liable to be biased between groups. Analysis will be performed using anova or Kruskal-Wallis (KW) tests. When appropriate (p\<0.05), a post-hoc test for multiple comparisons (Tukey-Kramer after anova and Dunn post KW) will be used. Comparisions of categorical variables will be performed using Chi-squared or Fischer test. Marascuillo's procedure will be performed for multiple comparisons. Relations between quantitative outcomes will be analyzed using correlation coefficients (Pearson or Spearman). Fisher's Z transformation and William's T2 statistic will be performed to compare correlations between variables and within a single group of subjects. Longitudinal data will be treated using mixt-model analyses in order to treat fixed effects group, time and group x time interaction taking into account between and within participant variability.