Stratifying patients can lead to smarter care

Segmenting high-risk patient populations according to sociodemographic and medical features may help improve care and lead to more efficient resource allocation, according to a recent Singapore study.

“Segmentation using medical and socioeconomic factors may be replicated by other health systems, forming the foundation for population-level health resource planning and tailored transitional care interventions,” researchers said.

Accessing the hospital-to-home transitional care programme (TCP) by the Singapore Ministry of Health, researchers identified 752 patients (52.4 percent female), who were then divided into distinct subgroups using latent class analysis.

Three segments of the study sample emerged: class 1 (frail, cognitively impaired and physically dependent), class 2 (prefrail but mostly physical independent) and class 3 (physically independent). Medical and socioeconomic characteristics differed significantly among the three subgroups. Class 1, for instance, fared significantly worse than the other two classes in terms of employment, dependency in activities of daily life, cognition, frailty and comorbidity. [BMC Health Serv Res 2019;19:931]

This pattern extended to the outcomes of interest. The 30-day hospital readmission rate in class 1 was 17.5 percent, which further climbed to 34.0 percent by 90 days. This was significantly greater than in classes 2 (18.3 percent and 26.6 percent) and 3 (10.7 percent vs 18.7 percent; p=0.027 and p=0.004, respectively).

Similarly, the mortality rates at 30 (6.8 percent vs 3.7 percent and 0.9 percent; p=0.005) and 90 (21.4 percent vs 8.4 percent and 2.1 percent; p<0.001) days were highest in class 1.

Multivariable logistic regression analysis confirmed these findings. Compared with the physically independent group, patients in class 1 were at the greatest risk of hospital readmission at 90 days (adjusted odds ratio [OR], 2.04, 95 percent confidence interval [CI], 1.21–3.46; p=0.008).

The same was true for mortality at 30 days (adjusted OR, 6.92, 95 percent CI, 1.76–27.21; p=0.006) and at 90 days (adjusted OR, 11.51, 95 percent CI, 4.57–29.02; p<0.001).

Notably, the risk of 30-day rehospitalization was only significant in class 2 patients (adjusted OR, 1.67, 95 percent CI, 1.03–2.71; p=0.037) and not in class 1 (adjusted OR, 1.61, 95 percent CI, 0.84–3.08; p=0.153). The other outcomes were also significantly more likely to occur in class 2 than in class 3, though to a lesser degree.

The present study showed that further dividing high-risk populations may lead to distinct subgroups, each with its own features and health outcomes, the researchers said. This demonstrates “the utility of population segmentation in prognosticating patients and highlighting cohorts of patients that require differing tiers of care.”

“Such an understanding would be fundamental for health policymakers and clinicians to make informed decisions on targeted health interventions for each class, allowing for optimal resource allocation and better health outcomes in a resource-strapped environment,” they added.