Multiple comorbidities affect a large proportion of elderly population, and the presence of comorbidities influences clinical decision. However, the utility of comorbidities in fracture risk assessment has largely been ignored. In this study, we used a network analysis approach to define the relationship between comorbidities, and then to develop a predictive model that uses comorbidities network to determine the risk of fracture for an individual.
The study included 2591 women and 1524 men aged 50 years and older (average age: 68 yr) who were participants of the Dubbo Osteoporosis Epidemiology Study. At baseline, 96 comordibities were we ascertained by using a structured questionnaire. Bone mineral density at the femoral neck was measured at baseline. The incidence of fragility fractures was ascertained during the follow-up period (1990-2015). We used a network-based analysis to examine the associations between multiple comorbidities. From the network of diseases, we derived a disease risk score (DRS) so that it has population mean of 0, and then used the score to predict the risk of future fracture.
The comorbidities network had 9024 linkages, and there were more links in osteoporosis compared with non-osteoporosis after adjusting for age and gender. Using relative risk (RR) as a metric of co-occurrence, individuals with a fracture were more likely to have osteopenia (RR=175), hypertension (RR=102), osteoporosis (RR=94), and osteoarthritis (RR=74). Each unit increase in DRS was associated with a >2-fold increase in the odds of fracture in women (odds ratio [OR] 2.31; 95%CI, 2.07-2.60) and in men (OR 2.66; 95%CI 2.23-3.20) after adjusting for age, prior fracture, and femoral neck BMD. The area under the ROC curve with DRS was 0.78, an increase from 0.67 from the model with age, femoral neck BMD and prior fracture. Thus, multimorbidity could be a powerful prognostic aid for fracture risk assessment in clinical setting.