Multimorbidity clusters potentially superior to individual diseases for stratifying fracture risk in older people: a nationwide cohort study

RATIONALE: Comorbidities are common in fracture patients, but the interaction between fracture and comorbidities remains unclear. This study aimed to define specific multimorbidity clusters in older adults and quantify the association between the multimorbidity clusters and fracture risk.

METHODS: This nationwide cohort study includes 1.7 million adults in Denmark aged ≥50 years who were followed from 2001 through 2014 for an incident low-trauma fracture. Chronic diseases and fractures were identified from the Danish National Hospital Discharge Register. Latent class analysis and Cox's regression were conducted to define the clusters and quantify fracture risk, respectively.

RESULTS: The study included 793 815 men (age: 64 ± 10) and 873 524 women (65.5 ± 11), with a third having ≥1 chronic disease. The pre-existent chronic diseases grouped individuals into low-multimorbidity (80.3% in men, 83.6% in women), cardiovascular (12.5%, 10.6%), malignant (4.1%, 3.8%), diabetic (2.4%, 2.0%) and hepatic clusters (0.7%, men only). These clusters distinguished individuals with advanced, complex, or late-stage disease from those having earlier-stage disease. During a median follow-up of 14 years (IQR: 6.5, 14), 95 372 men and 212 498 women sustained an incident fracture. The presence of multimorbidity was associated with a significantly greater risk of fracture, independent of age and sex. Importantly, the multimorbidity clusters had the highest discriminative performance in assessing fracture risk, whereas the strength of their association with fracture risk equalled or exceeded that of both the individual chronic diseases most prevalent in each cluster and of counts-based comorbidity indices.

CONCLUSIONS: Future fracture prevention strategies should take comorbidities into account. Multimorbidity clusters may provide greater insight into fracture risk than individual diseases or counts-based comorbidity indices.