In a poster presented at AMCP 2022, researchers presented their experience in developing a predictive model for identifying patients with pulmonary arterial hypertension (PAH) who are at increased risk for hospitalization, which is associated with adverse outcomes. The team, led by C. Zhang, concluded that their predictive model “showed acceptable performance to predict one-year hospitalization and could potentially be used to facilitate timely escalation of treatment, potentially improving clinical and economic outcomes.”
The investigators utilized retrospective claims data from the IBM MarketScan Commercial Claims and Encounters and Medicare Supplemental databases from 2008 to 2020. The study included patients who had begun treatment for PAH and were followed for one year. The initial risk factors assessed were age, sex, U.S. geographic region, health plan type, index year, PAH drug class, relevant individual comorbidities, symptoms, procedures, Quan-Charlson comorbidity index (CCI) score, claims-based frailty index (CFI), and PAH-related and non-related health care resource utilization and cost.
The full cohort consisted of 3,872 patients, among which 1,502 and 950 patients experienced all-cause and PAH-related hospitalization, respectively. The training cohort was randomly selected from the full study population, and the remaining patients were used as the validation cohort. The authors reported that their final models “had a c-statistic of 0.64 for all-cause and 0.61 for PAH-related hospitalization in the validation sample, respectively.” For all-cause hospitalization, the researchers selected CCI score ≥2, CFI score ≥1, hemoptysis, malaise/fatigue, history of PAH-related hospitalization/non-PAH-related emergency room visits, and higher total non-PAH-related outpatient cost. For PAH-related hospitalization, the authors selected: female, CCI score ≥4, CFI score ≥1, portal hypertension, dyspnea, and history of PAH-related hospitalization.