Burnout in ICU caregivers: A multicenter study of factors associated to centers

Transcription

1 Burnout in ICU caregivers: A multicenter study of factors associated to centers Paolo Merlani, Mélanie Verdon, Adrian Businger, Guido Domenighetti, Hans Pargger, Bara Ricou and the STRESI+ group Online Data Supplement 1

2 Electronic file E1: Description of the survey The questionnaire filled by all the heads of the 74 ICU described the characteristics of the center, such as type, the speaking specificity, size, number of patients per year, length of stay, the list of nurses, physicians and nurse assistants and the nurse/patient ratio and 8 or 12 hours shift, duration of intern rotation and mortality rate. The questionnaires sent to all caregivers asked for demographic data with personal characteristics such as the marital situation, having children, living alone or with a partner, years of experience in ICU and the activity rate at job. The occurrence of stress was based on a national study of Swiss workers, led by the Swiss State Secretariat of Economic Affairs (1). We evaluated the burnout with the Maslach Burnout Inventory (MBI) in its Fontaine Frenchversion (2). The MBI is widely used and is validated for the staff members in human service institutions and health care occupations, including nursing, social work, psychology, etc. It is composed of 22 questions, measuring the 3 dimensions of burnout: emotional exhaustion, depersonalization and lack of personal accomplishment. The emotional exhaustion refers to feelings of being overextended and depleted of one s emotional and physical resources. The depersonalization refers to a negative, callous, or excessively detached response to various aspects of the job. The lack of personal accomplishment refers to feelings of incompetence and a lack of achievement and productivity at work (3). The questions assess the frequency and range from 0 = never to 6 = everyday. Each respondent obtains a score in each dimension of burnout from low, medium to high degree. A high score in the dimensions of emotional exhaustion and/or of depersonalization means a high degree of burnout. In the personal accomplishment dimension, a low score relates to high degree of burnout. We defined burnout as a high MBI score, according to previous publications in French population (4): high level of burnout is defined by an MBI score of 8 to 34, a moderate level is defined by a score from 21 to 9, and a low level of burnout is indicated by an MBI score of 45 to 22. 2

3 Compared to the questionnaire used previously (5) we did not investigate the impact of stress on the daily life and on health. We also omitted the locally developed part of the questionnaire investigating concerns at work. References: 1. Secrétariat d'etat à l'économie SECO SUVA Promotion Santé Suisse. Un programme pour diminuer et prévenir le stress au travail destiné aux collaboratrices et collaborateurs et aux cadres Last downloaded the at 2. Canouï P, Mauranges A. Le syndrome d'épuisement professionnel des soignants: de l'analyse du burn-out aux réponses. Masson, Paris; Maslach C, Schaufeli WB, Leiter MP. Job burnout. Annu Rev Psychol 2001;52: Embriaco N, Azoulay E, Barrau K, Kentish N, Pochard F, Loundou A, Papazian L. High level of burnout in intensivists: prevalence and associated factors. Am J Respir Crit Care Med 2007;175: Verdon M, Merlani P, Perneger T, Ricou B. Burnout in a surgical ICU team. Intensive Care Med 2008;34: Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol, 1996;49:

4 Electronic file E2: detailed power-analysis and statistics: Power size: Sample size estimation was based initially on a logistic regression model on the assumption that when a multivariate logistic model is developed, 5-10 events should be available for each predictor variable considered (6). Assuming that about 20 predictor variables would be considered for modeling, and that 30% of caregivers would either be in stress or burnout, we needed to include at least 1000 of them. Statistical analysis: Stata statistical software, release 11.0 (Stata Corporation, College Station, TX, USA) and Statview (SAS institute Inc, Cary, NC, USA) were used. In the univariate analysis, risk factors for both high degree of burnout and stress were evaluated with two-tailed Fisher's exact tests for dichotomous variables, larger Chi-squared Test for Independence, unpaired t-tests or Mann-Whitney-U tests for continuous variables, as appropriate. The complete list of analyzed variables can be found in the online data supplement table E2. Odds ratios (OR) with 95% confidence intervals were calculated considering all participants in order to estimate the effect size of risk factors associated with a high degree of burnout and stress. Some continuous variables were dichotomized (age) or assembled (nurses and physicians) according to the precise univariate analysis of the risk factors. The linearity of the effect on the outcome was tested when appropriate. Multivariate mixed effect logistic regressions with random effects on the intercept accounting for centers concerning a high degree of burnout or stress were then performed considering all individuals in order to obtain adjusted estimates of the ORs and to identify factors independently associated with the outcomes. All predictors at a 0.05 alpha level in the univariate analysis of burnout were entered into both (stress and burnout) mixed-effect logistic regression models of the multivariate analysis. In this way we forced some of the variables into the high degree of stress model (type of caregiver and proportion of women) and into the high degree of burnout multivariate analysis model (type of hospital, pediatric vs. adult ICU, number of beds, length 4

5 of stay, German-speaking hospital vs. others and living alone). Due to collinearity we discharged from the multivariate models some variables that were associated in the univariate analysis with one of the outcomes (for example years of ICU experience and years of work in the ICU, number of nurses/nurse-assistants/physicians and number of beds, leading position in the ICU and age). Others were discharged for the same reason after a bi- or trivariate model (nurse/patient ratio during the day or night with collinearity with type of hospital, of ICU and number of beds). We further performed separate models for burnout including the variable feeling stressed and a model analyzing only female nurses in order to show the effect of the proportion of female nurses in this population. We then assessed the sensitivity, specificity, positive and negative predictive values of stress to predict a high degree of burnout. We compared centers with a low (<50 percentile) overall burnout with centers with a high overall burnout, by an appropriate univariate analysis and a multivariate logistic regression including all variables associated with centers with a high burnout with a p<0.10. Variables are expressed as mean ± standard deviation (SD) if not specified otherwise. A p value less than 0.05 was considered statistically significant. 5

6 Table E1 List of all variables analyzed for association with burnout and stress, in all caregivers and in the 3 subgroups (nurse assistants, nurses and physicians) SD : Standard Deviation. tot : total population. N-A : Nurse-assistant. N : Nurses. Phys : Physicians : 6

7 Table E2. Overview of the percentage of burnout in Nurse-assistants, Nurses and Physicians in the 3 linguistic regions Burnout: N/N (%) Total Nurse-assistant Nurses Physicians p (Nuse assistant vs Italian vs German) Total 874/2996 (29.2) 48/117 (41.0) 683/2415 (28.3) 143/459 (31.2) French speaking 219/833(26.3) 17/52 (32.7) 166/649 (25.6) 36/128 (28.1) 0.48 Italian speaking 66/200 (33.0) 4/16 (25.0) 51/150 (34.0) 11/33 (33.3) 0.77 German speaking 589/1963 (30.0) 27/49 (55.1) 466/1616 (28.8) 96/298 (32.2) P (French vs Italian vs German) Tests: Multiple Chi-square test, 7

8 Table E3. Uni- and multivariate mixed-effects logistic regression for factors (including feeling stressed ) associated with burnout, with random effect by centers Univariate analysis Multivariate analysis N=2910 OR 95% CI OR 95% CI p Center-related factors University (non university) Pediatric ICUs (adult ICUs) Number of beds (per bed) German-speaking Switzerland (Italian-, French-) <0.001 Proportion of females in the nurse team (per %) Proportion of females in the medical team (per %) Patients-related factors Mortality rate (per %) Length of stay (per day) Caregivers-related factors Nurse-assistant (nurse, physician) Gender: male <0.001 No child Living alone Age 40 years Feeling stressed <0.001 OR: Odds Ratio, CI: Confidence Interval, ICU: Intensive Care Unit Random-effect parameter (center): Estimate 0.000, Standard error Likelihood ratio tested vs logistic regression: p=1.0 8

9 Table E4. Uni- and multivariate mixed-effects logistic regression for factors associated with burnout, with random effect by centers, including only female nurses. Univariate analysis Multivariate analysis n=1923 OR 95% CI OR 95% CI p Center-related factors University (non university) Pediatric ICUs (adult ICUs) Number of beds (per bed) German-speaking Switzerland (Italian-, French-) Proportion of females in the nurse team (per %) Proportion of females in the medical team (per %) Patients-related factors Mortality rate (per %) Length of stay (per day) Caregivers-related factors Nurse-assistant (nurse, physician) Gender: male No child Living alone Age 40 years OR: Odds Ratio, CI: Confidence Interval, ICU: Intensive Care Unit Random-effect parameter (center): Estimate 0.050, Standard error Likelihood ratio tested vs logistic regression: p=0.06 9

10 Table E5. Uni- and multivariate mixed-effects logistic regression for factors associated with stress, with random effect by centers Univariate analysis Multivariate analysis n=2966 OR 95% CI OR 95% CI p Center-related factors University (non university) Pediatric ICUs (adult ICUs) Number of beds (per bed) German-speaking Switzerland (Italian-, French-) Proportion of females in the nurse team (per %) Proportion of females in the medical team (per %) Patients-related factors Mortality rate (per %) Length of stay (per day) Caregivers-related factors Nurse-assistant (nurse, physician) Gender: male No child <0.001 Living alone Age 40 years OR: Odds Ratio, CI: Confidence Interval, ICU: Intensive Care Unit Random-effect parameter (center): Estimate 0.296, Standard error Likelihood ratio vs logistic regression: p<

11 Table E6. Overview of different multivariate mixed-effects logistic regressions with random effect by centers, for factors associated with burnout and stress, in different populations. Multivariate analysis Burn-Out Syndrome Multivariate analysis Stress Population: Stress included in the multivariate model: Overall (without stress) Overall (with stress) Female Nurses (without stress) Overall - N OR OR OR OR Center-related factors University (non university) Pediatric ICUs (adult ICUs) Number of beds (per bed) German-speaking Switzerland (Italian-, French-) Proportion of females in the nurse team (per %) Proportion of females in the medical team (per %) Patients-related factors Mortality rate (per %) Length of stay (per day) Caregivers-related factors Nurse-assistant (nurse, physician) Gender: male No child Living alone Age 40 years Feeling stressed Random-effect parameter (center) Estimate (Standard error) (0.013) (0.000) (0.040) (0.081) OR: Odds ratio, ICU: intensive care unit, Black bold: significant OR, Gray not bold: not significant OR 11

12 Table E7. Uni- and multivariate logistic regression for factors associated with centers with higher proportion of caregivers with high burnout (higher than median proportion of burnout). Univariate analysis Multivariate analysis n=74 OR 95% CI OR 95% CI p Pediatric ICUs (adult ICUs) German-speaking Switzerland (Italian-, French-) Proportion of females in the nurse team (per %) Proportion of females in the medical team (per %) Mortality rate (per %) Number of patients per year (per patient) OR: Odds Ratio, CI: Confidence Interval, ICU: Intensive Care Unit 12