State sequence analysis for a deeper understanding of treatment adherence patterns in fragility fracture patients

We compared two methods to describe adherence to osteoporotic drugs after fragility fractures. While both similarly predicted refracture risk, a novel approach using sequence analysis provided a clearer picture of patient behavior, helping clinicians better understand and address adherence issues in osteoporosis care.

Fragility fractures are common among individuals with osteoporosis, causing significant morbidity, mortality, and healthcare costs. Although effective pharmacological treatments exist, underdiagnosis and poor treatment adherence remain pervasive in clinical practice.

In order to select an effective methodology to describe and visually represent treatment adherence and correctly stratify patients with fragility fractures according to their adherence patterns, a conventional method using average PDC was compared to an alternative method, given by the state sequence analysis (SSA) and clustering procedure. Data on patients aged 50 or older who experienced fragility fractures between 2012 and 2017 were retrieved from healthcare utilization databases in Lombardy, Italy. Fine and Gray's model was employed to analyze the association between adherence (calculated by conventional and alternative methods) and refracture risk. Finally, the discriminatory power to predict outcomes was calculated for each approach.

Out of the 8976 patients included in this observational study, four different adherence groups were considered using the conventional method (very poor, poor, intermediate, and optimal), while three clusters (non-adherence, short-term adherence, and long-term adherence) were obtained from the SSA. Compared with non-adherent patients, those with long-term adherence were found to have a significantly reduced risk of combined death and refracture with both methods. Regarding discriminatory performance, the two approaches showed similar AUC, 0.646 and 0.644 for conventional and alternative methods, respectively.

Based on the SSA and cluster analysis, the alternative method does not significantly modify the prediction of the refracture risk but enhances the description and visualization of the adherence patterns.