Abstract
Improved prediction modeling in osteoarthritis (OA) may encourage risk reduction through calculation of individual and population lifetime risks. There are currently no prediction models for hand OA. Thus, we aimed to 1) develop a prediction model for hand OA in men and 2) to contrast its discriminative performance to a prediction model for lung cancer and chronic obstructive pulmonary disease (COPD). We included 40,118 men aged 18 years undergoing mandatory conscription in Sweden 1969-70. Incident hand OA and lung cancer/COPD were obtained from diagnostic codes in the Swedish National Patient Register 1987-2010, i.e., until subjects were 59 years of age. We studied the strongest candidate predictors from five domains; socioeconomic, local biomechanical, systemic, lifestyle-related and general health factors, using logistic regression with backward elimination of candidate predictors with P>0.2 to determine final models. To avoid overfitting we used bootstrapping. The strongest predictors for hand OA were body mass index (BMI), elbow flexor strength, systolic blood pressure, lower education and sleep problems. We observed excellent agreement between observed and predicted values, yet the discrimination was moderate (Area Under the Curve [AUC]=0.62, 95% CI=0.58-0.64). The discrimination in the prediction model for lung cancer/COPD was good (AUC=0.74, 95% CI=0.72-0.76). This prediction model for hand OA was capable of discriminating between persons with and without hand OA to a similar extent that has been previously reported for knee OA. Still, prediction of OA is more challenging than for chronic pulmonary disease.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have