Prediction of Patient Outcomes through Social Determinants of Health: The Pulmonary Hypertension Association Registry (PHAR) Evaluation

Outcomes of patients with pulmonary arterial hypertension (PAH) may be associated with social determinants of health (SDOH) and other baseline patient characteristics. At present, there is no prognostic model to predict important patient outcomes in PAH based on SDOH. Utilizing information from the Pulmonary Hypertension Association Registry (PHAR), we derive a model (PHAR Evaluation or PHARE) to predict an important composite patient outcomes based on SDOH and other patient characteristics. Baseline data regarding SDOH from adult patients with PAH enrolled in the PHAR between 2015 and March 23, 2020, were included for analysis. We performed repeated measures logistic regression modeling with dichotomous outcome data (0 for no events, 1 for one or more events) to derive the PHARE. Here, 1275 consecutive adult patients enrolled in the PHAR from 47 participating centers were included. Variables included in our model are race, gender, ethnicity, household income, level of education, age, body mass index, drug use, alcohol use, marital status, and type of health insurance. Interaction effect between variables was analyzed and several interactions were also included in the PHARE. The PHARE shows a c‐statistic of 0.608 (p < 0.0001) with 95% confidence intervals (0.583, 0.632). Using SDOH and baseline characteristics from the PHAR, the PHARE correlates with our composite patient outcome. Further work evaluating the role of SDOH in prognostic modeling of PAH is indicated.

Elucidating tricuspid Doppler signal interpolation and its implication for assessing pulmonary hypertension

Doppler echocardiography plays a central role in the assessment of pulmonary hypertension (PAH). We aim to improve quality assessment of systolic pulmonary arterial pressure (SPAP) by applying a cubic polynomial interpolation to digitized tricuspid regurgitation (TR) waveforms. Patients with PAH and advanced lung disease were divided into three cohorts: a derivation cohort (n = 44), a validation cohort (n = 71), an outlier cohort (n = 26), and a non-PAH cohort (n = 44). We digitized TR waveforms and analyzed normalized duration, skewness, kurtosis, and first and second derivatives of pressure. Cubic polynomial interpolation was applied to three physiology-driven phases: the isovolumic phase, ejection phase, and "shoulder" point phase. Coefficients of determination and a Bland-Altman analysis was used to assess bias between methods. The cubic polynomial interpolation of the TR waveform correlated strongly with expert read right ventricular systolic pressure (RVSP) with R 2 > 0.910 in the validation cohort. The biases when compared to invasive SPAP measured within 24 h were 6.03 [4.33; 7.73], -2.94 [1.47; 4.41], and -3.11 [-4.52; -1.71] mmHg, for isovolumic, ejection, and shoulder point interpolations, respectively. In the outlier cohort with more than 30% difference between echocardiographic estimates and invasive SPAP, cubic polynomial interpolation significantly reduced underestimation of RVSP. Cubic polynomial interpolation of the TR waveform based on isovolumic or early ejection phase may improve RVSP estimates.

Multimodal fusion with deep neural networks for leveraging CT imaging and electronic health record: a case-study in pulmonary embolism detection

Recent advancements in deep learning have led to a resurgence of medical imaging and Electronic Medical Record (EMR) models for a variety of applications, including clinical decision support, automated workflow triage, clinical prediction and more. However, very few models have been developed to integrate both clinical and imaging data, despite that in routine practice clinicians rely on EMR to provide context in medical imaging interpretation. In this study, we developed and compared different multimodal fusion model architectures that are capable of utilizing both pixel data from volumetric Computed Tomography Pulmonary Angiography scans and clinical patient data from the EMR to automatically classify Pulmonary Embolism (PE) cases. The best performing multimodality model is a late fusion model that achieves an AUROC of 0.947 [95% CI: 0.946–0.948] on the entire held-out test set, outperforming imaging-only and EMR-only single modality models.

Hemodynamic trajectories and outcomes in patients with pulmonary arterial hypertension

The relative pulmonary to systemic pressure ratio (mean pulmonary arterial pressure/mean arterial pressure) has been proven to be valuable in cardiac surgery. Little is known on the prognostic value of baseline and trajectory of mean pulmonary arterial pressure/mean arterial pressure in pulmonary arterial hypertension. Patients with confirmed idiopathic, familial, drug and toxins, or connective tissue disease-related pulmonary arterial hypertension and at least one complete right heart catheterization were included and prospectively followed-up for 5.9 ± 4.03 years. Correlates of the primary end point (i.e. death or lung transplant need) during follow-up were determined using Cox regression modeling. Results showed that among the 308 patients included, 187 had at least one follow-up catheterization (median time between catheterizations: 2.16 (1.16–3.19) years). In the total cohort (mean age 47.3 ± 14.9 years, 82.8% of female and 58.1% in New York Heart Association class 3 or 4), mean pulmonary arterial pressure/mean arterial pressure (1.38 (1.07–1.77)) was associated with outcome (p = 0.01). Mean pulmonary arterial pressure/mean arterial pressure was incremental to a basic model (including right atrial pressure, systolic blood pressure, New York Heart Association class 3 or 4, and connective tissue disease) for outcome prediction, while mean pulmonary arterial pressure was not. In the 187 patients with a follow-up catheterization, both delta mean pulmonary arterial pressure and delta mean pulmonary arterial pressure/mean arterial pressure were associated with outcome (1.32 (1.11–1.58) and 1.31 (1.1–1.57) respectively, p < 0.01). Mean pulmonary arterial pressure and mean pulmonary arterial pressure/mean arterial pressure were both incremental to the basic model, while worsening in mean pulmonary arterial pressure or mean pulmonary arterial pressure/mean arterial pressure did not reach significance. In conclusion, mean pulmonary arterial pressure/mean arterial pressure at baseline prognosticates long-term outcome with a significant, albeit modest, incremental value to basic variables.

Quantifying the Influence of Wedge Pressure, Age, and Heart Rate on the Systolic Thresholds for Detection of Pulmonary Hypertension

Background The strong linear relation between mean (MPAP) and systolic (SPAP) pulmonary arterial pressure (eg, SPAP=1.62×MPAP) has been mainly reported in precapillary pulmonary hypertension. This study sought to quantify the influence of pulmonary arterial wedge pressure (PAWP), heart rate, and age on the MPAP-SPAP relation. Methods and Results An allometric equation relating invasive MPAP and SPAP was developed in 1135 patients with pulmonary arterial hypertension, advanced lung disease, chronic thromboembolic pulmonary hypertension, or left heart failure. The equation was validated in 60 885 patients from the United Network for Organ Sharing (UNOS) database referred for heart and/or lung transplant. The MPAP/SPAP longitudinal stability was assessed in pulmonary arterial hypertension with repeated right heart catheterization. The equation obtained was SPAP=1.39×MPAP×PAWP^-0.07×(60/heart rate)^0.12×age^0.08 (P<0.001). It was validated in the UNOS cohort (R²=0.93, P<0.001), regardless of the type of organ(s) patients were listed for (mean bias [-1.96 SD; 1.96 SD] was 0.94 [-8.00; 9.88] for heart, 1.34 [-7.81; 10.49] for lung and 0.25 [-16.74; 17.24] mm Hg for heart-lung recipients). Thresholds of SPAP for MPAP=25 and 20 mm Hg were lower in patients with higher PAWP (37.2 and 29.8 mm Hg) than in those with pulmonary arterial hypertension (40.1 and 32.0 mm Hg). In 186 patients with pulmonary arterial hypertension, the predicted MPAP/SPAP was stable over time (0.63±0.03 at baseline and follow-up catheterization, P=0.43). Conclusions This study quantifies the impact of PAWP, and to a lesser extent heart rate and age, on the MPAP-SPAP relation, supporting lower SPAP thresholds for pulmonary hypertension diagnosis in patients with higher PAWP for echocardiography-based epidemiological studies.

PENet-a scalable deep-learning model for automated diagnosis of pulmonary embolism using volumetric CT imaging

Pulmonary embolism (PE) is a life-threatening clinical problem and computed tomography pulmonary angiography (CTPA) is the gold standard for diagnosis. Prompt diagnosis and immediate treatment are critical to avoid high morbidity and mortality rates, yet PE remains among the diagnoses most frequently missed or delayed. In this study, we developed a deep learning model-PENet, to automatically detect PE on volumetric CTPA scans as an end-to-end solution for this purpose. The PENet is a 77-layer 3D convolutional neural network (CNN) pretrained on the Kinetics-600 dataset and fine-tuned on a retrospective CTPA dataset collected from a single academic institution. The PENet model performance was evaluated in detecting PE on data from two different institutions: one as a hold-out dataset from the same institution as the training data and a second collected from an external institution to evaluate model generalizability to an unrelated population dataset. PENet achieved an AUROC of 0.84 [0.82-0.87] on detecting PE on the hold out internal test set and 0.85 [0.81-0.88] on external dataset. PENet also outperformed current state-of-the-art 3D CNN models. The results represent successful application of an end-to-end 3D CNN model for the complex task of PE diagnosis without requiring computationally intensive and time consuming preprocessing and demonstrates sustained performance on data from an external institution. Our model could be applied as a triage tool to automatically identify clinically important PEs allowing for prioritization for diagnostic radiology interpretation and improved care pathways via more efficient diagnosis.

Phenotypically Silent Bone Morphogenetic Protein Receptor 2 Mutations Predispose Rats to Inflammation-Induced Pulmonary Arterial Hypertension by Enhancing the Risk for Neointimal Transformation

BACKGROUND

Bmpr2 (bone morphogenetic protein receptor 2) mutations are critical risk factors for hereditary pulmonary arterial hypertension (PAH) with approximately 20% of carriers developing disease. There is an unmet medical need to understand how environmental factors, such as inflammation, render Bmpr2 mutants susceptible to PAH. Overexpressing 5-LO (5-lipoxygenase) provokes lung inflammation and transient PAH in Bmpr2^+/- mice. Accordingly, 5-LO and its metabolite, leukotriene B₄, are candidates for the second hit. The purpose of this study was to determine how 5-LO-mediated pulmonary inflammation synergized with phenotypically silent Bmpr2 defects to elicit significant pulmonary vascular disease in rats.

METHODS

Monoallelic Bmpr2 mutant rats were generated and found phenotypically normal for up to 1 year of observation. To evaluate whether a second hit would elicit disease, animals were exposed to 5-LO-expressing adenovirus, monocrotaline, SU5416, SU5416 with chronic hypoxia, or chronic hypoxia alone. Bmpr2-mutant hereditary PAH patient samples were assessed for neointimal 5-LO expression. Pulmonary artery endothelial cells with impaired BMPR2 signaling were exposed to increased 5-LO-mediated inflammation and were assessed for phenotypic and transcriptomic changes.

RESULTS

Lung inflammation, induced by intratracheal delivery of 5-LO-expressing adenovirus, elicited severe PAH with intimal remodeling in Bmpr2^+/- rats but not in their wild-type littermates. Neointimal lesions in the diseased Bmpr2^+/- rats gained endogenous 5-LO expression associated with elevated leukotriene B₄ biosynthesis. Bmpr2-mutant hereditary PAH patients similarly expressed 5-LO in the neointimal cells. In vitro, BMPR2 deficiency, compounded by 5-LO-mediated inflammation, generated apoptosis-resistant and proliferative pulmonary artery endothelial cells with mesenchymal characteristics. These transformed cells expressed nuclear envelope-localized 5-LO consistent with induced leukotriene B₄ production, as well as a transcriptomic signature similar to clinical disease, including upregulated nuclear factor Kappa B subunit (NF-κB), interleukin-6, and transforming growth factor beta (TGF-β) signaling pathways. The reversal of PAH and vasculopathy in Bmpr2 mutants by TGF-β antagonism suggests that TGF-β is critical for neointimal transformation.

CONCLUSIONS

In a new 2-hit model of disease, lung inflammation induced severe PAH pathology in Bmpr2^+/- rats. Endothelial transformation required the activation of canonical and noncanonical TGF-β signaling pathways and was characterized by 5-LO nuclear envelope translocation with enhanced leukotriene B₄ production. This study offers an explanation of how an environmental injury unleashes the destructive potential of an otherwise silent genetic mutation.

Development and Performance of the Pulmonary Embolism Result Forecast Model (PERFORM) for Computed Tomography Clinical Decision Support

IMPORTANCE

Pulmonary embolism (PE) is a life-threatening clinical problem, and computed tomographic imaging is the standard for diagnosis. Clinical decision support rules based on PE risk-scoring models have been developed to compute pretest probability but are underused and tend to underperform in practice, leading to persistent overuse of CT imaging for PE.

OBJECTIVE

To develop a machine learning model to generate a patient-specific risk score for PE by analyzing longitudinal clinical data as clinical decision support for patients referred for CT imaging for PE.

DESIGN, SETTING, AND PARTICIPANTS

In this diagnostic study, the proposed workflow for the machine learning model, the Pulmonary Embolism Result Forecast Model (PERFORM), transforms raw electronic medical record (EMR) data into temporal feature vectors and develops a decision analytical model targeted toward adult patients referred for CT imaging for PE. The model was tested on holdout patient EMR data from 2 large, academic medical practices. A total of 3397 annotated CT imaging examinations for PE from 3214 unique patients seen at Stanford University hospitals and clinics were used for training and validation. The models were externally validated on 240 unique patients seen at Duke University Medical Center. The comparison with clinical scoring systems was done on randomly selected 100 outpatient samples from Stanford University hospitals and clinics and 101 outpatient samples from Duke University Medical Center.

MAIN OUTCOMES AND MEASURES

Prediction performance of diagnosing acute PE was evaluated using ElasticNet, artificial neural networks, and other machine learning approaches on holdout data sets from both institutions, and performance of models was measured by area under the receiver operating characteristic curve (AUROC).

RESULTS

Of the 3214 patients included in the study, 1704 (53.0%) were women from Stanford University hospitals and clinics; mean (SD) age was 60.53 (19.43) years. The 240 patients from Duke University Medical Center used for validation included 132 women (55.0%); mean (SD) age was 70.2 (14.2) years. In the samples for clinical scoring system comparisons, the 100 outpatients from Stanford University hospitals and clinics included 67 women (67.0%); mean (SD) age was 57.74 (19.87) years, and the 101 patients from Duke University Medical Center included 59 women (58.4%); mean (SD) age was 73.06 (15.3) years. The best-performing model achieved an AUROC performance of predicting a positive PE study of 0.90 (95% CI, 0.87-0.91) on intrainstitutional holdout data with an AUROC of 0.71 (95% CI, 0.69-0.72) on an external data set from Duke University Medical Center; superior AUROC performance and cross-institutional generalization of the model of 0.81 (95% CI, 0.77-0.87) and 0.81 (95% CI, 0.73-0.82), respectively, were noted on holdout outpatient populations from both intrainstitutional and extrainstitutional data.

CONCLUSIONS AND RELEVANCE

The machine learning model, PERFORM, may consider multitudes of applicable patient-specific risk factors and dependencies to arrive at a PE risk prediction that generalizes to new population distributions. This approach might be used as an automated clinical decision-support tool for patients referred for CT PE imaging to improve CT use.

Plasma levels of S100A4 in portopulmonary hypertension

We previously showed that a single nucleotide polymorphism in S100A4 was associated with portopulmonary hypertension (PPHTN) in patients with advanced liver disease. We aimed to determine the association between plasma levels of S100A4 and PPHTN. We performed a case-control study of patients with advanced liver disease. Cases with PPHTN had mean pulmonary artery pressure >25 mmHg, pulmonary vascular resistance >240 dynes s cm(-5) and pulmonary capillary wedge pressure </=15 mmHg. Controls with liver disease had right ventricular systolic pressure <40 mmHg and normal right atrial and ventricular morphology by echocardiography. Plasma samples were assayed for S100A4. The study sample included 14 cases with PPHTN and 32 controls with liver disease. There was no difference in mean age between cases and controls (p = 0.52). Seventy-nine percent of cases were female compared with 44% of controls (p = 0.03). There was no difference in S100A4 levels between cases and controls (p = 0.58). Both groups had significantly higher S100A4 levels than healthy volunteers (p <0.05). There was no significant difference in plasma levels of S100A4 between PPHTN patients and controls with liver disease, although liver disease itself was associated with increased S100A4 levels.