Factors Associated with Secondary Pulmonary Hypertension Among Hospitalized Females: An Artificial Neural Network Analysis of a National US Cohort

Background: Non-group 1 pulmonary hypertension, also known as secondary pulmonary hypertension (SPH), is predominantly observed among females. However, there is a significant lack of data concerning factors associated with hospitalization among patients diagnosed with SPH. This study aims to provide clinicians with vital insights for the identification of high-risk groups and for the more effective management of contributory risk factors within the female population affected by SPH. Methods: Using the 2019 National Inpatient Sample, we identified female admissions with SPH (n = 648,190), accounting for 3.8% of the total 17,236,228 female admissions. An Artificial Neural Network (ANN) analysis was conducted to evaluate predictive factors. We randomly allocated 3,319,543 patients into training and testing datasets at a ratio of 70:30, comprising 2,323,696 (70%) for training and 995,847 (30%) for testing, to calibrate and validate the performance of the ANN algorithm. Model performance was assessed by comparing misclassification rates between training and testing sets and by the area under the receiver operating characteristic curve (AUC); only internal validation was performed. Results: Females hospitalized with SPH were generally of older age, with a median of 75 years compared to 58 years, and more frequently identified as White (67.7% versus 65.5%) or Black (20.5% versus 15.5%) relative to those without SPH. They also demonstrated a higher prevalence of most atherosclerotic cardiovascular disease (ASCVD) risk factors or their equivalents, including complicated hypertension (50.6% versus 17.8%), diabetes with chronic complications (30.6% versus 13.7%), and hyperlipidemia (50.8% versus 29.2%), as well as other comorbidities such as COPD (43.4% versus 20.2%) and CKD (43.3% versus 14.0%), and exhibited increased all-cause mortality (4.5% versus 1.8%) (p < 0.001). Our ANN model achieved an AUC of 0.823, indicating good predictive capability. The rates of incorrect predictions were comparable in both the testing and training cohorts, at 3.8% each. The factors most strongly associated with a coded SPH diagnosis included age at admission, complicated hypertension, chronic kidney disease, chronic obstructive pulmonary disease, uncomplicated hypertension, prior VTE, race, arthropathies, and AIDS. Conclusions: Our ANN model identified demographic and comorbidity factors associated with a coded SPH diagnosis among hospitalized females, with good discrimination (AUC = 0.823). Because the model classifies the presence of an existing diagnosis rather than predicting future hospitalization, and was validated only internally, external and prospective validation is required before clinical application. Once validated, these factors could support individualized, sex-specific risk stratification for high-risk female populations, consistent with the goals of personalized medicine.