Survival analysis using the stress check data to predict employee turnover

ストレス　チェック　ニ　ヨル　タイショク　ヨソク　モデル

タカハラ, リュウジ

Takahara, Ryuji

オオサカ ダイガク ダイガクイン ニンゲン カガク ケンキュウカ タイジン シャカイ シンリガク ケンキュウシツ

ストレスチェック

職業性ストレス簡易調査票

集団分析

マルチレベル生存モデル

turnover

stress check program

brief job stress questionnaire

group analysis

multilevel survival analysis

It is recommended to use the job stress assessment diagram as group analysis when using the brief job stress questionnaire, the de facto standard of the stress check program in Japan. However, in order to clarify how to improve the work environment, it is thought to be effective to conduct multivariate analysis using not only subjective answers from the questionnaire but also objective indicators of the workplace. In this paper, models predict to turnover were constructed to test the effectiveness of the above-mentioned multivariate analysis. Multilevel survival analyses using stress check, attendance and demographics data in a nationwide multi branch service company were conducted. Since some employees did not respond to the stress check questionnaire, we examined two models, one using individual-level stress check results and one without. In the former model, the variables of having a meaningful job, group-level satisfaction of job and life, overtime, tenure, employment status and group size were shown to have significant effects on turnover. In the latter model, group-level vigor, group-level somatic symptoms, overtime, age, tenure, employment status, and group-level previous year's turnover rate of part-time employees were shown to have significant effects on turnover. The models showed sufficient discrimination and indicated the possibility of the prediction of objective events such as turnover by using stress check data.