Hospitalization Risk among Early Adopters of Respiratory Syncytial Virus Vaccines: A Machine Learning Approach

Background: The US FDA recently approved two vaccines for the prevention of severe respiratory syncytial virus (RSV) disease in adults ≥60 years of age. Limited data on patients receiving the vaccine exist.

Objectives: To explore the baseline medical conditions associated with adult RSV hospitalization and predict the risk of hospitalization among early adopters of RSV vaccines.

Methods: Patients ≥60 years of age presenting with a primary diagnosis for RSV in the inpatient hospital setting (index) from 1/1/2016 through 9/30/2023 were selected from a large U.S. dataset that includes administrative and claims data for over 170 million patients across commercial payors, Medicare Advantage and Medicaid. RSV cases were propensity score matched 1:2 to controls with no history of RSV hospitalization (random index date assigned) based on age, sex, race, Census region, Charlson Comorbidity Index, and calendar quarter of index to control for seasonality. All diagnoses from a period of continuous enrollment starting one year prior to index were included in an XGBoost classification model for predicting RSV hospitalization. The trained model was then used to estimate risk of future RSV hospitalization among a separate cohort of patients ≥60 years of age with evidence of RSV vaccine administration from 10/01/2023-11/30/2023 (N=12,773).

Results: 6,316 patients with an RSV hospitalization were matched to 12,270 controls. The XGBoost model was able to predict RSV hospitalization with an area under the ROC curve of 83.26% (95% CI: 81.93%-84.59%). Top predictors included chronic lung disease (chronic obstructive pulmonary disease [COPD], asthma), evidence of prior acute respiratory infections/failure, obesity, hyperlipidemia, atherosclerosis and immunodeficiency disorders. Using a prediction threshold equivalent to the prevalence of RSV hospitalization in the model training sample (>34%), the model predicted future RSV hospitalization among 56.48% of patients receiving an RSV vaccine (N=7,214/12,773). Among vaccinated patients, the most common baseline diagnosis was hyperlipidemia (38.78%) followed by vascular disease (32.04%) and type 2 diabetes mellitus (21.01%). Approximately 1/6 of vaccinated patients had a diagnosis for asthma or COPD, while only 1.85% had documented immunodeficiencies.

Conclusions: A machine learning model identified known clinical predictors of severe RSV and estimated a high likelihood of future RSV-related hospitalization in approximately half of immunized older adults. The opportunity for increased vaccine adoption may exist for those with chronic lung disease and weakened immune systems.