Can right ventricular assessments improve triaging of low risk pulmonary embolism?

Machine Learning-Based Prediction of Pulmonary Embolism Prognosis Using Nutritional and Inflammatory Indices

PURPOSE

This study aimed to create and assess machine learning (ML) models that utilize nutritional and inflammatory indices, focusing on the advanced lung cancer inflammation index (ALI) and neutrophil-to-albumin ratio (NAR), to improve the prediction accuracy of PE prognosis.

PATIENTS AND METHODS

We conducted a retrospective analysis of 312 patients, comprising 254 survivors and 58 non-survivors. The Boruta algorithm was used to identify significant variables, and four ML models (XGBoost, Random Forest, Logistic Regression, and SVM) were employed to analyze the clinical data and assess the performance of the models.

RESULTS

The XGBoost model, with optimal hyperparameters, achieved the best performance, with an accuracy of 0.882, an F1-score of 0.889, a precision of 0.917, a sensitivity of 0.863, a specificity of 0.905, and an AUC of 0.873. Analysis of feature importance indicated that the most critical predictors across models were respiratory failure, log-transformed ALI, albumin level, age, diastolic blood pressure, and NAR.

CONCLUSION

The ML-based prediction models effectively predicted the prognosis of PE, with the XGBoost model exhibiting good performance. Respiratory failure, ALI, albumin level, age, diastolic blood pressure, and NAR were correlated with PE prognosis.

Disposition of emergency department patients with acute pulmonary embolism after ambulance arrival

Objective

Most outpatients with pulmonary embolism (PE) are diagnosed in the emergency department (ED). The relationship between means of arrival, site of diagnosis, and disposition in ED patients with PE is unknown. We compared discharge home between patients arriving by emergency medical services (EMS) and those arriving by other means. Within the EMS cohort, we compared those with a recent PE diagnosis in the outpatient clinic setting to those who were diagnosed with PE in the ED.

Methods

This study was a secondary analysis of a retrospective cohort that included all adult, non-pregnant ED patients treated for acute PE across 21 community EDs from January 2013 to April 2015. The primary outcome was discharge home within 24 h of ED registration; we also examined mortality. We described associations with patient arrival method and other patient characteristics.

Results

Among 2996 ED patient encounters with acute PE, 644 (21.5%) arrived by EMS. This group had a lower frequency of discharge (9.2% vs 26.4%) and higher 30-day all-cause mortality (8.7% vs 3.1%) than their counterparts (p < 0.001 for both). These associations remained after adjusting for confounding variables. Among the EMS cohort, 14 patients (2.2%) arrived with a PE diagnosis recently made in the outpatient setting.

Conclusion

Patients with PE who arrived at the ED by EMS were less likely to be discharged home within 24 h and more likely to die within 30 days than those who arrived by other means. Less than 3% of the EMS group had been diagnosed with PE before ED arrival.

Quality of life 1 month after acute pulmonary embolism in emergency department patients

OBJECTIVE

The Pulmonary Embolism Quality-of-Life (PEmb-QoL) questionnaire assesses quality of life (QoL) after pulmonary embolism (PE). We aimed to determine whether any clinical or pathophysiologic features of PE were associated with worse PEmb-QoL scores 1 month after PE.

METHODS

In this prospective multicenter registry, we conducted PEmb-QoL questionnaires. We determined differences in QoL domain scores for four primary variables: clinical deterioration (death, cardiac arrest, respiratory failure, hypotension requiring fluid bolus, catecholamine support, or new dysrhythmia), right ventricular dysfunction (RVD), PE risk stratification, and subsequent rehospitalization. For overall QoL score, we fit a multivariable regression model that included these four primary variables as independent variables.

RESULTS

Of 788 PE patients participating in QoL assessments, 156 (19.8%) had a clinical deterioration event, 236 (30.7%) had RVD of which 38 (16.1%) had escalated interventions. For those without and with clinical deterioration, social limitations had mean (±SD) scores of 2.07 (±1.27) and 2.36 (±1.47), respectively (p = 0.027). For intensity of complaints, mean (±SD) scores for patients without RVD (4.32 ± 2.69) were significantly higher than for those with RVD with or without reperfusion interventions (3.82 ± 1.81 and 3.83 ± 2.11, respectively; p = 0.043). There were no domain score differences between PE risk stratification groups. All domain scores were worse for patients with rehospitalization versus without. By multivariable analysis, worse total PEmb-QoL scores with effect sizes were subsequent rehospitalization 11.29 (6.68-15.89), chronic obstructive pulmonary disease (COPD) 8.17 (3.91-12.43), and longer index hospital length of stay 0.06 (0.03-0.08).

CONCLUSIONS

Acute clinical deterioration, RVD, and PE severity were not predictors of QoL at 1 month post-PE. Independent predictors of worsened QoL were rehospitalization, COPD, and index hospital length of stay.

Comparing predictive performance of pulmonary embolism risk stratification tools for acute clinical deterioration

Objectives

Existing pulmonary embolism (PE) risk scores were developed to predict death within weeks, but not more proximate adverse events. We determined the ability of 3 PE risk stratification tools (simplified pulmonary embolism severity index [sPESI], 2019 European Society of Cardiology guidelines [ESC], and PE short-term clinical outcomes risk estimation [PE-SCORE]) to predict 5-day clinical deterioration after emergency department (ED) diagnosis of PE.

Methods

We analyzed data from six EDs on ED patients with confirmed PE. Clinical deterioration was defined as death, respiratory failure, cardiac arrest, new dysrhythmia, sustained hypotension requiring vasopressors or volume resuscitation, or escalated intervention within 5 days of PE diagnosis. We determined sensitivity and specificity of sPESI, ESC, and PE-SCORE for predicting clinical deterioration.

Results

Of 1569 patients, 24.5% had clinical deterioration within 5 days. sPESI, ESC, and PE-SCORE classifications were low-risk in 558 (35.6%), 167 (10.6%), and 309 (19.6%), respectively. Sensitivities of sPESI, ESC, and PE-SCORE for clinical deterioration were 81.8 (78, 85.7), 98.7 (97.6, 99.8), and 96.1 (94.2, 98), respectively. Specificities of sPESI, ESC, and PE-SCORE for clinical deterioration were 41.2 (38.4, 44), 13.7 (11.7, 15.6), and 24.8 (22.4, 27.3). Areas under the curve were 61.5 (59.1, 63.9), 56.2 (55.1, 57.3), and 60.5 (58.9, 62.0). Negative predictive values were 87.5 (84.7, 90.2), 97 (94.4, 99.6), and 95.1 (92.7, 97.5).

Conclusions

ESC and PE-SCORE were better than sPESI for detecting clinical deterioration within 5 days after PE diagnosis.

Sustainability of a Clinical Decision Support Intervention for Outpatient Care for Emergency Department Patients With Acute Pulmonary Embolism

IMPORTANCE

Physicians commonly hospitalize patients presenting to the emergency department (ED) with acute pulmonary embolism (PE), despite eligibility for safe outpatient management. Risk stratification using electronic health record-embedded clinical decision support systems can aid physician site-of-care decision-making and increase safe outpatient management. The long-term sustainability of early improvements after the cessation of trial-based, champion-led promotion is uncertain.

OBJECTIVE

To evaluate the sustainability of recommended site-of-care decision-making support 4 years after initial physician champion-led interventions to increase outpatient management for patients with acute PE.

DESIGN, SETTING, AND PARTICIPANTS

This retrospective cohort study was conducted in 21 US community hospitals in an integrated health system. Participants included adult patients presenting to the ED with acute PE. Study sites had participated in an original decision-support intervention trial 4 years prior to the current study period: 10 sites were intervention sites, 11 sites were controls. In that trial, decision support with champion promotion resulted in significantly higher outpatient management at intervention sites compared with controls. After trial completion, all study sites were given continued access to a modified decision-support tool without further champion-led outreach. Data were analyzed from January 2019 to February 2020.

EXPOSURES

ED treatment with a modified clinical decision support tool.

MAIN OUTCOMES AND MEASURES

The main outcome was frequency of outpatient management, defined as discharge home directly from the ED, stratified by the PE Severity Index. The safety measure of outpatient care was 7-day PE-related hospitalization.

RESULTS

This study included 1039 patients, including 533 (51.3%) women, with a median (IQR) age of 65 (52-74) years. Nearly half (474 patients [45.6%]) were rated lower risk on the PE Severity Index. Overall, 278 patients (26.8%) were treated as outpatients, with only four 7-day PE-related hospitalizations (1.4%; 95% CI, 0.4%-3.6%). The practice gap in outpatient management created by the earlier trial persisted in the outpatient management for patients with lower risk: 109 of 236 patients (46.2%) at former intervention sites vs 81 of 238 patients (34.0%) at former control sites (difference, 12.2; [95% CI, 3.4-20.9] percentage points; P = .007), with wide interfacility variation (range, 7.1%-47.1%).

CONCLUSIONS AND RELEVANCE

In this cohort study, a champion-led, decision-support intervention to increase outpatient management for patients presenting to the ED with acute pulmonary embolism was associated with sustained higher rates of outpatient management 4 years later. The application of our findings to improving sustainability of practice change for other clinical conditions warrants further study.