Using a machine-learning model for pulmonary arterial hypertension can help pinpoint key features linked to future clinical worsening
Presenter: Hilary M. DuBrock, MD, associate professor, Division of Pulmonary and Critical Care, Mayo Clinic Rochester, MN.
Use of machine-learning models to identify clinical features in patients with pulmonary arterial hypertension associated with a future clinical worsening event. Pulmonary Vascular Disease Posters 2. Presented Oct 10, 2023.
https://journal.chestnet.org/article/S0012-3692(23)04859-6/fulltext
Machine-learning models based on electronic health record data have been used to identify features associated with future clinical worsening events in a cohort of real-world patients with pulmonary pulmonary arterial hypertension (PAH), according to research reported at the CHEST 2023 meeting.
In the best performing model, key features such as male sex, dyspnea symptoms, and associated connective tissue disease were linked with future clinical worsening, said presenter Hilary M. DuBrock, MD, an associate professor in the Division of Pulmonary and Critical Care at Mayo Clinic in Rochester, MN. Taken together, the findings indicate that machine-learning models can help identify clinical worsening among patients with PAH, which potentially could lead to earlier intervention and improved patient outcomes.
“Machine learning, when used in the appropriate clinical context, can be a helpful adjunctive tool in the longitudinal management of PAH patients,” Dr. DuBrock said.
Because PAH is incurable, management is often focused on delaying disease progression. Toward that end, better risk prediction for clinical worsening could optimize management and may facilitate early intervention, Dr. DuBrock noted.
In other recent studies, machine-learning algorithms have been used to predict clinical worsening among patients with heart failure and COVID-19. So in this study, Dr. DuBrock and colleagues sought to apply machine-leaning models in real-word PAH data to identify clinical features that are commonly observed in patients with PAH prior to a clinical worsening event.
The retrospective analysis was based on Mayo Clinic electronic health records of patients diagnosed with PAH between 2015 and 2019. All patients included had a PAH diagnosis confirmed by hemodynamic testing, plus a specific pulmonary disease-related International Classification of Disease code or a record of receiving PAH medication following a PAH diagnosis.
The data set included 2,442 patients diagnosed with PAH, of whom 1,023 experienced a clinical worsening event. Approximately 87% of patients where white, non-Hispanic, 51% were male, and the median age at diagnosis was 67.5 years.
Machine-learning models were trained on the data to identify associations between clinical features and a future clinical worsening event. Associations were analyzed at baseline (ie, within the first month of diagnosis) and during the progression window, meaning a specific time interval immediately before a clinical worsening event.
At baseline, the clinical features most clearly linked to future clinical worsening events were male sex (P < .05), PAH associated with connective tissue disease (P < .05), and a high erythrocyte distribution width value (P < .001), according to her report.
Within the 1-month progression window, features most strongly associated with the future clinical worsening event included dyspnea (P < .001), chronic respiratory disease (P < .01), chest pain, a primary care physician visit, and a cardiologist visit (all P < .05).
This study had some limitations, according to the researchers. For example, the real-world cohort may have included some patients with the more general condition of pulmonary hypertension rather than PAH, specifically. And the prediction algorithm could have been more refined if some key variables that are useful in identifying PAH worsening were documented, such as the 6-minute walk test.
Nevertheless, the researchers said their present study has important implications research and clinical practice, including how computational approaches can be used to analyze real-world and claims-based data. Furthermore, improved monitoring of patients at risk for clinical worsening could result in earlier clinical intervention, which may optimize patient outcomes and provide better support for shared decision-making practices.
Disclosures:
Hilary M. DuBrock, MD, reported research grants from Bayer Pharmaceuticals, consultancy fees from Janssen, Pharmaceutical Companies of Johnson & Johnson, and has served on advisory boards for Janssen Pharmaceutical Companies of Johnson & Johnson and United Therapeutics.
References:
Humbert M, Kovacs G, Hoeper MM, et al. 2022 ESC/ERS guidelines for the diagnosis and treatment of pulmonary hypertension [published correction appears in Eur Heart J 2023; 44(15):1312]. Eur Heart J 2022; 43(38):3618-3731. doi:10.1093/eurheartj/ehac237
Burger CD, Ghandour M, Padmanabhan Menon D, Helmi H, Benza RL. Early intervention in the management of pulmonary arterial hypertension: clinical and economic outcomes. Clinicoecon Outcomes Res 2017; 9:731-739. doi:10.2147/CEOR.S119117
Ru B, Tan X, Liu Y, et al. Comparison of machine learning algorithms for predicting hospital readmissions and worsening heart failure events in patients with heart failure with reduced ejection fraction: Modeling Study. JMIR Form Res 2023; 7:e41775. doi:10.2196/41775
Shamout FE, Shen Y, Wu N, et al. An artificial intelligence system for predicting the deterioration of COVID-19 patients in the emergency department. NPJ Digit Med 2021; 4(1):80. doi:10.1038/s41746-021-00453-0