Determinant Factors and Estimation for Early Retirement in Korean Workers

Abstract

Introduction: The increasing proportion of the ageing working population is recognized as a significant public health challenge. Although many studies have reported on early retirement of vulnerable populations, only a few of these studies have examined the different early retirement predictors based on the type of retirement among self-employed and regular paid workers. To date, there have been no studies focused on developing an early retirement prediction model. Accordingly, the primary purposes of this study were to analyze the eterminants of early retirement according to the retirement type among self-employed and regular paid workers, and to establish a prediction model of early retirement owing to the poor personal health status of the retirees.

Methods: This study analyzed the longitudinal data of 2,708 workers who had previously participated in the Korean Longitudinal Study of Ageing (KLoSA), from 2006 to 2014. The prevalence of early retirement was calculated according to the type of retirement. Multivariate Cox regression analyses were conducted to identify the predictors for early retirement according to the type of retirement. Moreover, the current study constructed a prediction model for early retirement due to poor personal health status using a Cox regression model.

Results: Over the 8-year follow-up, 314 workers retired early, including 31, 88, 124, and 71 because of sufficient economic status; leisure or volunteer activities; personal health problems; and family members’ health problems, housekeeping, or childcare, respectively. In all strata, multivariable analysis revealed that older and female workers were significantly more vulnerable towards early retirement. Sufficient economic status was an important determinant of voluntary early retirement (hazard ratio [HR], 1.22; 95% confidence interval [CI], .08–1.39). Significant predictors for early retirement among workers who retired due to poor personal health included hypertension (HR, 1.52; 95% CI, 1.01–2.28), abnormal body mass index (BMI) (HR, 1.60; 95% CI, 1.10–2.35), decreased grasping power index, and erceived health status. A prediction model designed to estimate the risk of early retirement because of poor personal health status—using age, sex, hypertension, diabetes, abnormal BMI, smoking, alcohol consumption, perceived health status, and working life expectancy—showed fair performance (areas under the curve = 0.784 [± 0.023] in the training dataset, 0.781 [± 0.047] in the test dataset, and 0.751 [± 0.029] in the external validation dataset from the English Longitudinal Study of Ageing).

Conclusion: The current study including both self-employed and regular paid workers revealed the specific determinants of shortened working life according to the type of early retirement. From these results, it seems that, to understand the process of shortened working life, different determinants of early retirement according to the main reasons of retiring should be considered. The performance of our prediction model for early retirement because of poor personal health remained stable even after external validation, suggesting the ossibility of prediction of unwanted early retirement. Taken together, our findings suggest a role of our prediction model as a preventive strategy of unwanted retirement from the workplace.