Predicting long-term sickness absence among retail workers after four days of sick-listing

Abstract

OBJECTIVE: This study tested and validated an existing tool for its ability to predict the risk of long-term (ie, ≥6 weeks) sickness absence (LTSA) after four days of sick-listing. METHODS: A 9-item tool is completed online on the fourth day of sick-listing. The tool was tested in a sample (N=13 597) of food retail workers who reported sick between March and May 2017. It was validated in a new sample (N=104 698) of workers (83% retail) who reported sick between January 2020 and April 2021. LTSA risk predictions were calibrated with the Hosmer-Lemeshow (H-L) test; non-significant H-L P-values indicated adequate calibration. Discrimination between workers with and without LTSA was investigated with the area (AUC) under the receiver operating characteristic (ROC) curve. RESULTS: The data of 2203 (16%) workers in the test sample and 14 226 (13%) workers in the validation sample was available for analysis. In the test sample, the tool together with age and sex predicted LTSA (H-L test P=0.59) and discriminated between workers with and without LTSA [AUC 0.85, 95% confidence interval (CI) 0.83–0.87]. In the validation sample, LTSA risk predictions were adequate (H-L test P=0.13) and discrimination was excellent (AUC 0.91, 95% CI 0.90–0.92). The ROC curve had an optimal cut-off at a predicted 36% LTSA risk, with sensitivity 0.85 and specificity 0.83. CONCLUSION: The existing 9-item tool can be used to invite sick-listed retail workers with a ≥36% LTSA risk for expedited consultations. Further studies are needed to determine LTSA cut-off risks for other economic sectors.

Keywords

• prediction model
• roc analysis
• sick-listing
• external validation
• sick leave
• risk assessment
• sickness absence
• occupational health service
• long-term sickness absence