Karanam, V

Abstract Rationale Per the 2021 Sepsis Campaign, it is recommended that hospitals use a performance improvement program for sepsis screening, utilizing machine learning to improve specificity. Around 20% of sepsis at the Mayo Clinic, Rochester site is hospital acquired. As inpatient sepsis is usually detected by extreme physiologic derangement or clinical concern, we seek to leverage risk factors beyond traditional SIRS and sepsis criteria to create a machine and deep learning based risk stratification model to improve the precision of sepsis monitoring. Methods We designed a retrospective observational case control study, including patients diagnosed with sepsis as an inpatient at the Mayo Clinic, Rochester hospital and compared them to patients without a sepsis diagnosis over the last 7 years. Sepsis diagnosis was labeled by a combination of sepsis ICD codes and initial lactate elevation. We excluded transfers from outside hospitals, direct admissions for sepsis and patients diagnosed with sepsis in the emergency department. Data was extracted using Mayo Data Explorer, and data preparation and validation were performed using JMP Statistical Software and Python libraries. Feature selection was critical as we examined all data from previous contacts with healthcare, environmental factors, and dynamic trends of inpatient objective data; amongst 5,000 options of lab tests and vital measurements, we selected near 100 to reduce model complexity. We used XG Boost to train our model with a baseline cohort of 100 sepsis patients and 1000 patients without sepsis. Results We created a model that categorically classified patients into three sepsis risk level profiles (low-risk, medium-risk, high-risk) with performance metrics of a specificity of 97%, sensitivity of 65% and area under the receiver operating characteristic curve of 95%. Amongst laboratory data, lactate and creatinine had the highest feature importance (figure below). Our model also identified variables with the least importance to further refine our architecture. Conclusions Our model is one of the first sepsis risk stratification models whose algorithm is optimized for non-ICU inpatient wards, while the majority of similar studies were conducted on ICU and emergency department patients due to limited data availability. The model’s high specificity will allow for earlier sepsis recognition. By including our full cohort of 7,500 patients with sepsis and 35,000 patients without sepsis, we hope to improve class imbalance and increase sensitivity. By integrating transformer architecture, we seek to eliminate variables of low significance to create a model with more practical use, while maintaining high specificity. This abstract is funded by: None

Abstract Rationale Portopulmonary hypertension (POPH) and hepatopulmonary syndrome (HPS) are two distinct pulmonary vascular complications of liver disease. POPH is characterized by pulmonary hypertension and elevated pulmonary vascular resistance while HPS is characterized by intrapulmonary vasodilation and hypoxemia. POPH and HPS reflect opposing pulmonary vascular responses with markedly different therapeutic and transplant implications. Accurate distinction on imaging may aid clinical decision-making and risk stratification. We aimed to quantitatively characterize and compare chest CT imaging features of POPH and HPS compared to controls with isolated portal hypertension (PoH). Methods Three cohorts of HPS, POPH, and PoH were included (n = 5 per cohort). POPH patients with intrapulmonary vasodilation on contrast echocardiography were excluded. Contrast-enhanced chest CTs with slice thickness ≤ 1.5mm obtained closest to the time of diagnosis were selected (mean 2.7 years following diagnosis). CT images were analyzed using a deep learning algorithm developed in-house that segmented and computed volumes of central, interlobar, peripheral, upper peripheral, and lower peripheral pulmonary arteries (PAs); total, upper, and lower pulmonary veins (PVs); and the four cardiac chambers. Resulting volumes were indexed by body-surface area. One-way analysis of variance (ANOVA) was performed with a post-hoc pairwise Mann-Whitney U test for comparison. Results Patients with HPS were 80% male, had a mean age of 48.6±17.4, and mean MELD-Na of 15.4±3.4. The POPH cohort was 20% male, had a mean age of 56.8±11.9, and mean MELD-Na of 13.4±4.0. The PoH cohort was 100% male, had a mean age of 62.8±4.8, and mean MELD-Na of 12.8±8.8. HPS patients had smaller central PA volumes (mean 26.4±1.8 mm2/m2) compared to both PoH (mean 38.6±9.8 mm2/m2, p = 0.03) and POPH patients (44.8±10.8 mm2/m2, p = 0.008). HPS patients, however, had larger lower peripheral PA (mean 30.7±1.9 mm2/m2) and PV (mean 28.3±6.7 mm2/m2) volumes compared to PoH (PAs mean 20.1±4.8 mm2/m2, p = 0.008; PVs mean 18.2±2.8 mm2/m2, p = 0.02) and larger lower peripheral PA volumes compared to POPH patients (PAs mean 22.1±5.1 mm2/m2, p = 0.03). There was no significant difference in volumes between POPH and PoH for all segments. Conclusion Compared to PoH and POPH, HPS is associated with lower central PA volume and higher peripheral PA volume in the lower lungs. Volumes were similar between POPH and PoH; however, this may reflect limited statistical power given the small sample size. Larger studies are needed to further delineate imaging distinctions between POPH, HPS, and PoH. This abstract is funded by: NHLBI