Predicting high health care use in patients with spinal disorder in secondary care: model development and validation

Abstract Introduction: Spinal disorders are among the biggest contributors to health care utilization (HCU). Objective: To develop and externally validate a prediction model for high all-cause HCU (75th percentile during 1 year after the index date) among patients with spinal disorders visiting multidisciplinary secondary care clinics. Methods: We developed and internally validated the model using the Norwegian Neck and Back Registry, including patients registered between January 1, 2016, and December 31, 2020, linked with national health registries (N = 9092). For external validation, we used data from the Danish SpineData Registry, linked with national registries, for the same period (N = 34,853). We assessed Nagelkerke R 2 , discrimination (area under receiver operating characteristics curve [AUC]), and calibration (calibration-in-the-large [CITL], slope, and calibration plot). Results: The final model included sex, nationality, education, physical activity, smoking, prior HCU, work status, disability, health-related quality of life, medicine use, diagnosis, kinesiophobia, and comorbidity. It demonstrated acceptable discrimination (AUC 0.78, 95% confidence interval [CI], 0.77–0.78), an R 2 of 0.26, and good calibration after internal validation. Upon external validation, the model demonstrated excellent discrimination (AUC 0.81, 95% CI 0.80–0.81) and an R 2 of 0.31. The calibration slope was 1.08 (95% CI 1.06–1.11) and CITL was 0.16 (95% CI 0.12–0.19). Predicted probabilities closely matched observed probabilities across all deciles in internal validation, with slight underestimation of high HCU in the top 3 deciles during external validation. Conclusion: Overall, the model shows promise in predicting high HCU in patients with spinal disorders referred to secondary care but requires further testing and validation in implementation settings before recommendation.