Main objectiveSickness absence duration for shoulder lesion patients is difficult to prognosticate, and scientific evidence for the sick-listing practice is lacking. Our objective was to develop a clinically implementable prediction model for the duration of a sickness absence spell due to shoulder lesions. MethodsAll new sickness absence spells due to shoulder lesions (ICD-10-code: M75) issued in the period January 2010-June 2012 that were longer than 14 days were identified through the nationwide sickness absence insurance register. Information on predictors was linked from four other nationwide registers. Piecewise-constant hazards regression models were fitted to predict duration of sickness absence. The model was developed and validated using split sample validation. Variable selection was based on log-likelihood loss ranking when excluding a variable from the model. The model was evaluated using calibration plots and the c-statistic. Results20 049 sickness absence spells were identified, of which 34% lasted >90 days. Predictors included in the model were age, sex, geographical region, occupational status, educational level, birth country, specialized healthcare at start of the spell, number of sickness absence days in the last 12 months, and specialized healthcare the last 12 months, before start date of the index sickness absence spell. The model was satisfactorily specified and calibrated. Overall c-statistic was 0.54 (95% CI 0.53-0.55). C-statistic for predicting durations >90, >180, and >365 days was 0.61, 0.66, and 0.74, respectively. SignificanceThe model can be used to predict the duration of sickness absence due to shoulder lesions. Covariates had limited predictive power but could discriminate the very long sickness absence spells from the rest.
A prognostic model for predicting the duration of 20,049 sickness absence spells due to shoulder lesions in a population-based cohort in Sweden
Frumento, Paolo;
2023-01-01
Abstract
Main objectiveSickness absence duration for shoulder lesion patients is difficult to prognosticate, and scientific evidence for the sick-listing practice is lacking. Our objective was to develop a clinically implementable prediction model for the duration of a sickness absence spell due to shoulder lesions. MethodsAll new sickness absence spells due to shoulder lesions (ICD-10-code: M75) issued in the period January 2010-June 2012 that were longer than 14 days were identified through the nationwide sickness absence insurance register. Information on predictors was linked from four other nationwide registers. Piecewise-constant hazards regression models were fitted to predict duration of sickness absence. The model was developed and validated using split sample validation. Variable selection was based on log-likelihood loss ranking when excluding a variable from the model. The model was evaluated using calibration plots and the c-statistic. Results20 049 sickness absence spells were identified, of which 34% lasted >90 days. Predictors included in the model were age, sex, geographical region, occupational status, educational level, birth country, specialized healthcare at start of the spell, number of sickness absence days in the last 12 months, and specialized healthcare the last 12 months, before start date of the index sickness absence spell. The model was satisfactorily specified and calibrated. Overall c-statistic was 0.54 (95% CI 0.53-0.55). C-statistic for predicting durations >90, >180, and >365 days was 0.61, 0.66, and 0.74, respectively. SignificanceThe model can be used to predict the duration of sickness absence due to shoulder lesions. Covariates had limited predictive power but could discriminate the very long sickness absence spells from the rest.File | Dimensione | Formato | |
---|---|---|---|
Frumento_1218384.pdf
accesso aperto
Tipologia:
Versione finale editoriale
Licenza:
Creative commons
Dimensione
855.23 kB
Formato
Adobe PDF
|
855.23 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.