[ABE-L] Seminário Conjunto PIPGEs - ICMC/UFSCar - 15/12/2014 - 14h00 - Emmanuel Lesaffre

[Cibele Russo]

Divulgando:


*Seminário Conjunto PIPGEs - ICMC/UFSCar - 15/12/2014 - 14h00*

*LOCAL:* Auditório Luiz Antonio Favaro (sala 4-111) – ICMC - USP, São
Carlos, SP.

*TÍTULO:* Exploring burnout in a large European nurse survey using a
multilevel covariance regression model

*Palestrante:* Emmanuel Lesaffre (Department of Biostatistics, Erasmus MC,
Rotterdam, the Netherlands, L-Biostat, KULeuven, Leuven, Belgium)

*Resumo:*  We propose a novel modeling approach that can model both the
mean structure and the covariance structure with a mixed effects model in a
multivariate context. We called this the multilevel covariance regression
(MCR) model. When the dimension of the response is high, a joint model of a
multilevel factor analytic (MFA) model and an MCR model (MHOF model) is
then proposed.
We applied the MCR model to data from the RN4CAST (Sermeus et al. 2011) FP7
project which involves 33,731 registered nurses in 2,169 nursing units in
486 hospitals in 12 European countries. The MHOF model was applied to the
Belgium part of the project. As response we have taken in the first
analysis the historically derived three burnout dimensions (Maslach and
Jackson, 1981), while the MHOF model is based on the raw data, i.e. the
responses to the 22-item questionnaire. The three burnout dimensions are
emotional exhaustion (EE), depersonalization (DP) and personal
accomplishment (PA). Applying the MHOF model to burnout could address the
following questions simultaneously: 1) is the burnout structure the same as
the commonly used structure by Maslach and Jackson? 2) how much variation
of burnout could be explained by the level-specific fixed and random
effects? 3) do the variances and correlations among burnout stay constant
across level-specific characteristics and units at each level?
We opted for the Bayesian approach as our estimating method for the MCR and
MHOF models. The JAGS (just another Gibbs sampler) MCMC (Markov chain Monte
Carlo) program was used through the R package rjags. Most parameters were
assigned a non-informative prior except for the fixed and random effects in
the factor loadings in the MCR part. These parameters were assigned a
mixture prior respectively to overcome the "flipping states" issue in
Bayesian context. Model comparison was done using the pseudo Bayes factor
(PSBF).
-------------- Próxima Parte ----------
Um anexo em HTML foi limpo...
URL: <https://lists.ime.usp.br/archives/abe/attachments/20141203/452ff76b/attachment.html>