Clinical Phenotyping of Out-of-Hospital Cardiac Arrest Patients With Shockable Rhythm - Machine Learning-Based Unsupervised Cluster Analysis

Association between sub-phenotypes identified using latent class analysis and neurological outcomes in patients with out-of-hospital cardiac arrest in Japan

BACKGROUND

In patients who experience out-of-hospital cardiac arrest (OHCA), it is important to assess the association of sub-phenotypes identified by latent class analysis (LCA) using pre-hospital prognostic factors and factors measurable immediately after hospital arrival with neurological outcomes at 30 days, which would aid in making treatment decisions.

METHODS

This study retrospectively analyzed data obtained from the Japanese OHCA registry between June 2014 and December 2019. The registry included a complete set of data on adult patients with OHCA, which was used in the LCA. The association between the sub-phenotypes and 30-day survival with favorable neurological outcomes was investigated. Furthermore, adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were estimated by multivariate logistic regression analysis using in-hospital data as covariates.

RESULTS

A total of, 22,261 adult patients who experienced OHCA were classified into three sub-phenotypes. The factor with the highest discriminative power upon patient's arrival was Glasgow Coma Scale followed by partial pressure of oxygen. Thirty-day survival with favorable neurological outcome as the primary outcome was evident in 66.0% participants in Group 1, 5.2% in Group 2, and 0.5% in Group 3. The 30-day survival rates were 80.6%, 11.8%, and 1.3% in groups 1, 2, and 3, respectively. Logistic regression analysis revealed that the ORs (95% CI) for 30-day survival with favorable neurological outcomes were 137.1 (99.4-192.2) for Group 1 and 4.59 (3.46-6.23) for Group 2 in comparison to Group 3. For 30-day survival, the ORs (95%CI) were 161.7 (124.2-212.1) for Group 1 and 5.78 (4.78-7.04) for Group 2, compared to Group 3.

CONCLUSIONS

This study identified three sub-phenotypes based on the prognostic factors available immediately after hospital arrival that could predict neurological outcomes and be useful in determining the treatment strategy of patients experiencing OHCA upon their arrival at the hospital.

Overview and future prospects of out-of-hospital cardiac arrest registries in Japan

Aim

Out-of-hospital cardiac arrest (OHCA) is a life-threatening emergency with high mortality. The "chain of survival" is critical to improving patient outcomes. To develop and enhance this chain of survival, measuring and monitoring the resuscitation processes and outcomes are essential for quality assurance. In Japan, several OHCA registries have successfully been implemented at both local and national levels. We aimed to review and summarise the conception, strengths, and challenges of OHCA registries in Japan.

Method and results

The following representing registries in Japan were reviewed: the All-Japan Utstein registry, the Utstein Osaka Project/the Osaka-CRITICAL study, the SOS-KANTO study, the JAAM-OHCA study, and the SAVE-J II study. The All-Japan Utstein registry, operated by the Fire and Disaster Management Agency of Japan and one of the largest nationwide population-based registries in the world, collects data concerning all patients with OHCA in Japan, excluding in-hospital data. Other research- and hospital-based registries collect detailed out-of-hospital and in-hospital data. The Osaka-CRITICAL study and the SOS-KANTO study are organized at regional levels, and hospitals in the Osaka prefecture and in the Kanto area participate in these registries. The JAAM-OHCA study is managed by the Japanese Association of Acute Medicine and includes 107 hospitals throughout Japan. The Save-J II study focuses on patients with OHCA treated with extracorporeal cardiopulmonary resuscitation.

Conclusion

Each OHCA registry has its own philosophy, strengths, perspectives, and challenges; however, all have been successful in contributing to the improvement of emergency medical service (EMS) systems through the quality improvement process. These registries are expected to be further utilized to enhance EMS systems and improve outcomes for patients with OHCA, while also contributing to the field of resuscitation science.

Derivation of Coagulation Phenotypes and the Association with Prognosis in Traumatic Brain Injury: A Cluster Analysis of Nationwide Multicenter Study

BACKGROUND

The pathogenesis and pathophysiology of traumatic coagulopathy during traumatic brain injury is not well understood, and the appropriate treatment strategy for this condition has not been established. This study aimed to evaluate the coagulation phenotypes and their effect on prognosis in patients with isolated traumatic brain injury.

METHODS

In this multicenter cohort study, we retrospectively analyzed data from the Japan Neurotrauma Data Bank. Adults with isolated traumatic brain injury (head abbreviated injury scale > 2; abbreviated injury scale of any other trauma < 3) who were registered in the Japan Neurotrauma Data Bank were included in this study. The primary outcome was the association of coagulation phenotypes with in-hospital mortality. Coagulation phenotypes were derived using k-means clustering with coagulation markers, including prothrombin time international normalized ratio (PT-INR), activated partial thromboplastin time (APTT), fibrinogen (FBG), and D-dimer (DD) on arrival at the hospital. Multivariable logistic regression analyses were conducted to calculate the adjusted odds ratios of coagulation phenotypes with their 95% confidence intervals (CIs) for in-hospital mortality.

RESULTS

In total, 556 patients were enrolled and five coagulation phenotypes were identified. The median (interquartile range) score for the Glasgow Coma Scale was 6 (4-9). Cluster A (n = 129) had the closest to normal coagulation values; cluster B (n = 323) had a mild high DD phenotype; cluster C (n = 30) had a prolonged PT-INR phenotype with a higher frequency of antithrombotic medication in elderly patients than in younger patients; cluster D (n = 45) had a low amount of FBG, high DD, and prolonged APTT phenotype with a high incidence of skull fracture; and cluster E (n = 29) had a low amount of FBG and extremely high DD phenotype with high energy trauma and a high incidence of skull fracture. In the multivariable logistic regression analysis, the association of clusters B, C, D, and E with in-hospital mortality yielded the corresponding adjusted odds ratios of 2.17 (95% CI 1.22-3.86), 2.61 (95% CI 1.01-6.72), 10.0 (95% CI 4.00-25.2), and 24.1 (95% CI 7.12-81.3), respectively, relative to cluster A.

CONCLUSIONS

This multicenter, observational study identified five different coagulation phenotypes of traumatic brain injury and showed associations of these phenotypes with in-hospital mortality.

Joint Modeling of Social Determinants and Clinical Factors to Define Subphenotypes in Out-of-Hospital Cardiac Arrest Survival: Cluster Analysis

Background

Machine learning clustering offers an unbiased approach to better understand the interactions of complex social and clinical variables via integrative subphenotypes, an approach not studied in out-of-hospital cardiac arrest (OHCA).

Objective

We conducted a cluster analysis for a cohort of OHCA survivors to examine the association of clinical and social factors for mortality at 1 year.

Methods

We used a retrospective observational OHCA cohort identified from Medicare claims data, including area-level social determinants of health (SDOH) features and hospital-level data sets. We applied k-means clustering algorithms to identify subphenotypes of beneficiaries who had survived an OHCA and examined associations of outcomes by subphenotype.

Results

We identified 27,028 unique beneficiaries who survived to discharge after OHCA. We derived 4 distinct subphenotypes. Subphenotype 1 included a distribution of more urban, female, and Black beneficiaries with the least robust area-level SDOH measures and the highest 1-year mortality (2375/4417, 53.8%). Subphenotype 2 was characterized by a greater distribution of male, White beneficiaries and had the strongest zip code-level SDOH measures, with 1-year mortality at 49.9% (4577/9165). Subphenotype 3 had the highest rates of cardiac catheterization at 34.7% (1342/3866) and the greatest distribution with a driving distance to the index OHCA hospital from their primary residence >16.1 km at 85.4% (8179/9580); more were also discharged to a skilled nursing facility after index hospitalization. Subphenotype 4 had moderate median household income at US $51,659.50 (IQR US $41,295 to $67,081) and moderate to high median unemployment at 5.5% (IQR 4.2%-7.1%), with the lowest 1-year mortality (1207/3866, 31.2%). Joint modeling of these features demonstrated an increased hazard of death for subphenotypes 1 to 3 but not for subphenotype 4 when compared to reference.

Conclusions

We identified 4 distinct subphenotypes with differences in outcomes by clinical and area-level SDOH features for OHCA. Further work is needed to determine if individual or other SDOH domains are specifically tied to long-term survival after OHCA.

Comparing outcomes of out-of-hospital cardiac arrest patients with initial shockable rhythm in Singapore and Osaka using population-based databases

BACKGROUND

Previous research indicated outcomes among refractory out-of-hospital cardiac arrest (OHCA) patients with initial shockable rhythm were different in Singapore and Osaka, Japan, possibly due to the differences in access to extracorporeal cardiopulmonary resuscitation. However, this previous study had a risk of selection bias. To address this concern, this study aimed to evaluate the outcomes between Singapore and Osaka for OHCA patients with initial shockable rhythm using only population-based databases.

METHODS

This was a secondary analysis of two OHCA population-based databases in Osaka and Singapore, including adult OHCA patients with initial shockable rhythm. A machine-learning-based prediction model was derived from the Osaka data (n = 3088) and applied to the PAROS-SG data (n = 2905). We calculated the observed-expected ratio (OE ratio) for good neurological outcomes observed in Singapore and the expected derived from the data in Osaka by dividing subgroups with or without prehospital ROSC.

RESULTS

The one-month good neurological outcomes in Osaka and Singapore among patients with prehospital ROSC were 70% (791/1,125) and 57% (440/773), and among patients without prehospital ROSC were 10% (196/1963) and 2.8% (60/2,132). After adjusting patient characteristics, the outcome in Singapore was slightly better than expected from Osaka in patients with ROSC (OE ratio, 1.067 [95%CI 1.012 to 1.125]), conversely, it was worse than expected in patients without prehospital ROSC (OE ratio, 0.238 [95%CI 0.173 to 0.294]).

CONCLUSION

This study showed the outcomes of OHCA patients without prehospital ROSC in Singapore were worse than expected derived from Osaka data even using population-based databases. (249/250 words).

Explainable artificial intelligence in emergency medicine: an overview

Artificial intelligence (AI) and machine learning (ML) have potential to revolutionize emergency medical care by enhancing triage systems, improving diagnostic accuracy, refining prognostication, and optimizing various aspects of clinical care. However, as clinicians often lack AI expertise, they might perceive AI as a "black box," leading to trust issues. To address this, "explainable AI," which teaches AI functionalities to end-users, is important. This review presents the definitions, importance, and role of explainable AI, as well as potential challenges in emergency medicine. First, we introduce the terms explainability, interpretability, and transparency of AI models. These terms sound similar but have different roles in discussion of AI. Second, we indicate that explainable AI is required in clinical settings for reasons of justification, control, improvement, and discovery and provide examples. Third, we describe three major categories of explainability: pre-modeling explainability, interpretable models, and post-modeling explainability and present examples (especially for post-modeling explainability), such as visualization, simplification, text justification, and feature relevance. Last, we show the challenges of implementing AI and ML models in clinical settings and highlight the importance of collaboration between clinicians, developers, and researchers. This paper summarizes the concept of "explainable AI" for emergency medicine clinicians. This review may help clinicians understand explainable AI in emergency contexts.

Outcome assessment for out-of-hospital cardiac arrest patients in Singapore and Japan with initial shockable rhythm

BACKGROUND

Singapore and Osaka in Japan have comparable population sizes and prehospital management; however, the frequency of ECPR differs greatly for out-of-hospital cardiac arrest (OHCA) patients with initial shockable rhythm. Given this disparity, we hypothesized that the outcomes among the OHCA patients with initial shockable rhythm in Singapore were different from those in Osaka. The aim of this study was to evaluate the outcomes of OHCA patients with initial shockable rhythm in Singapore compared to the expected outcomes derived from Osaka data using machine learning-based prediction models.

METHODS

This was a secondary analysis of two OHCA databases: the Singapore PAROS database (SG-PAROS) and the Osaka-CRITICAL database from Osaka, Japan. This study included adult (18-74 years) OHCA patients with initial shockable rhythm. A machine learning-based prediction model was derived and validated using data from the Osaka-CRITICAL database (derivation data 2012-2017, validation data 2018-2019), and applied to the SG-PAROS database (2010-2016 data), to predict the risk-adjusted probability of favorable neurological outcomes. The observed and expected outcomes were compared using the observed-expected ratio (OE ratio) with 95% confidence intervals (CI).

RESULTS

From the SG-PAROS database, 1,789 patients were included in the analysis. For OHCA patients who achieved return of spontaneous circulation (ROSC) on hospital arrival, the observed favorable neurological outcome was at the same level as expected (OE ratio: 0.905 [95%CI: 0.784-1.036]). On the other hand, for those who had continued cardiac arrest on hospital arrival, the outcomes were lower than expected (shockable rhythm on hospital arrival, OE ratio: 0.369 [95%CI: 0.258-0.499], and nonshockable rhythm, OE ratio: 0.137 [95%CI: 0.065-0.235]).

CONCLUSION

This observational study found that the outcomes for patients with initial shockable rhythm but who did not obtain ROSC on hospital arrival in Singapore were lower than expected from Osaka. We hypothesize this is mainly due to differences in the use of ECPR.

AI and machine learning in resuscitation: Ongoing research, new concepts, and key challenges

Aim

Artificial intelligence (AI) and machine learning (ML) are important areas of computer science that have recently attracted attention for their application to medicine. However, as techniques continue to advance and become more complex, it is increasingly challenging for clinicians to stay abreast of the latest research. This overview aims to translate research concepts and potential concerns to healthcare professionals interested in applying AI and ML to resuscitation research but who are not experts in the field.

Main text

We present various research including prediction models using structured and unstructured data, exploring treatment heterogeneity, reinforcement learning, language processing, and large-scale language models. These studies potentially offer valuable insights for optimizing treatment strategies and clinical workflows. However, implementing AI and ML in clinical settings presents its own set of challenges. The availability of high-quality and reliable data is crucial for developing accurate ML models. A rigorous validation process and the integration of ML into clinical practice is essential for practical implementation. We furthermore highlight the potential risks associated with self-fulfilling prophecies and feedback loops, emphasizing the importance of transparency, interpretability, and trustworthiness in AI and ML models. These issues need to be addressed in order to establish reliable and trustworthy AI and ML models.

Conclusion

In this article, we overview concepts and examples of AI and ML research in the resuscitation field. Moving forward, appropriate understanding of ML and collaboration with relevant experts will be essential for researchers and clinicians to overcome the challenges and harness the full potential of AI and ML in resuscitation.