S Yang1, W D Leslie2, S N Morin3, L M Lix4. 1. Department of Epidemiology and Biostatistics, School of Public Health, Jilin University, 232-1163 Xinmin Street, Changchun, 130021, Jilin, China. shumanyang@jlu.edu.cn. 2. Department of Internal Medicine, University of Manitoba, C5121-409 Tache Ave, Winnipeg, Manitoba, R2H 2A6, Canada. bleslie@sbgh.mb.ca. 3. Department of Medicine, McGill University, Montreal, Quebec, Canada. 4. Department of Community Health Sciences, University of Manitoba, Winnipeg, Manitoba, Canada.
Abstract
Fracture risk scores generated from population-based administrative healthcare data showed comparable or better discrimination than the Fracture Risk Assessment Tool (FRAX) scores computed without bone mineral density for predicting incident major osteoporotic fracture. Administrative data may be useful to identify individuals at high fracture risk at the population level. PURPOSE: To evaluate the discrimination of fracture risk scores defined using inputs available from administrative data for predicting incident major osteoporotic fracture (MOF) and hip fracture (HF) alone. METHODS: Using the Manitoba Bone Mineral Density (BMD) Database (1997-2013), we identified 61,041 individuals aged 50 years or older with healthcare coverage following their first BMD test. We calculated two-modified FRAX)scores based on administrative data: FRAX-A and FRAX-A+. The FRAX-A modification used all FRAX inputs, except for BMD, body mass index, and parental HF, while the FRAX-A+ modification using all FRAX-A inputs plus a comorbidity score, number of hospitalizations in the 3 years prior to the BMD test, depression diagnosis, and dementia diagnosis. FRAX scores computed with BMD (i.e., FRAX [BMD]) and without BMD (i.e., FRAX [no-BMD]) were the comparators. RESULTS: During a mean of 7 years of follow-up, we identified 5306 (8.7%) incident MOF and 1532 (2.5%) incident HF. The c-statistic for MOF associated with FRAX-A was lower than FRAX (BMD) (0.655 vs 0.675; P < 0.05) and comparable to FRAX (no-BMD) (0.654; P = 0.07). The c-statistic for MOF using FRAX-A+ (0.663) was lower than FRAX (BMD) but higher than FRAX (no-BMD) (both P < 0.05). For predicting incident HF, c-statistics associated with FRAX-A (0.762) and FRAX-A+ (0.767) were lower than FRAX (BMD) (0.789) and FRAX (no-BMD) (0.773; both P < 0.05). CONCLUSIONS: FRAX-A and FRAX-A+ showed comparable or better discrimination than FRAX without BMD for predicting incident MOF, but slightly lower discrimination for HF alone.
Fracture risk scores generated from population-based administrative healthcare data showed comparable or better discrimination than the Fracture Risk Assessment Tool (FRAX) scores computed without bone mineral density for predicting incident major osteoporotic fracture. Administrative data may be useful to identify individuals at high fracture risk at the population level. PURPOSE: To evaluate the discrimination of fracture risk scores defined using inputs available from administrative data for predicting incident major osteoporotic fracture (MOF) and hip fracture (HF) alone. METHODS: Using the Manitoba Bone Mineral Density (BMD) Database (1997-2013), we identified 61,041 individuals aged 50 years or older with healthcare coverage following their first BMD test. We calculated two-modified FRAX)scores based on administrative data: FRAX-A and FRAX-A+. The FRAX-A modification used all FRAX inputs, except for BMD, body mass index, and parental HF, while the FRAX-A+ modification using all FRAX-A inputs plus a comorbidity score, number of hospitalizations in the 3 years prior to the BMD test, depression diagnosis, and dementia diagnosis. FRAX scores computed with BMD (i.e., FRAX [BMD]) and without BMD (i.e., FRAX [no-BMD]) were the comparators. RESULTS: During a mean of 7 years of follow-up, we identified 5306 (8.7%) incident MOF and 1532 (2.5%) incident HF. The c-statistic for MOF associated with FRAX-A was lower than FRAX (BMD) (0.655 vs 0.675; P < 0.05) and comparable to FRAX (no-BMD) (0.654; P = 0.07). The c-statistic for MOF using FRAX-A+ (0.663) was lower than FRAX (BMD) but higher than FRAX (no-BMD) (both P < 0.05). For predicting incident HF, c-statistics associated with FRAX-A (0.762) and FRAX-A+ (0.767) were lower than FRAX (BMD) (0.789) and FRAX (no-BMD) (0.773; both P < 0.05). CONCLUSIONS: FRAX-A and FRAX-A+ showed comparable or better discrimination than FRAX without BMD for predicting incident MOF, but slightly lower discrimination for HF alone.
Authors: Amit Gupta; Christina Maslen; Madhavi Vindlacheruvu; Richard L Abel; Pinaki Bhattacharya; Paul A Bromiley; Emma M Clark; Juliet E Compston; Nicola Crabtree; Jennifer S Gregory; Eleni P Kariki; Nicholas C Harvey; Eugene McCloskey; Kate A Ward; Kenneth E S Poole Journal: Ther Adv Musculoskelet Dis Date: 2022-03-28 Impact factor: 5.346