Artificial neural network-boosted Cardiac Arrest Survival Post-Resuscitation In-hospital (CASPRI) score accurately predicts outcome in cardiac arrest patients treated with targeted temperature management

Post-Cardiac arrest outcome prediction using machine learning: A systematic review and meta-analysis

BACKGROUND

Early and reliable prognostication in post-cardiac arrest patients remains challenging, with various factors linked to return of spontaneous circulation (ROSC), survival, and neurological results. Machine learning and deep learning models show promise in improving these predictions. This systematic review and meta-analysis evaluates how effective these approaches are in predicting clinical outcomes at different time points using structured data.

METHODS

This study followed PRISMA guidelines, involving a comprehensive search across PubMed, Scopus, and Web of Science databases until March 2024. Studies aimed at predicting ROSC, survival (or mortality), and neurological outcomes after cardiac arrest through the application of machine learning or deep learning techniques with structured data were included. Data extraction followed the guidelines of the CHARMS checklist, and the bias risk was evaluated using PROBAST tool. Models reporting the AUC metric with 95 % confidence intervals were incorporated into the quantitative synthesis and meta-analysis.

RESULTS

After extracting 2,753 initial records, 41 studies met the inclusion criteria, yielding 97 machine learning and 16 deep learning models. The pooled AUC for predicting favorable neurological outcomes (CPC 1 or 2) at hospital discharge was 0.871 (95 % CI: 0.813 - 0.928) for machine learning models and 0.877 (95 % CI: 0.831-0.924) across deep learning algorithms. For survival prediction, this value was found to be 0.837 (95 % CI: 0.757-0.916). Considerable heterogeneity and high risk of bias were observed, mainly attributable to inadequate management of missing data and the absence of calibration plots. Most studies focused on pre-hospital factors, with age, sex, and initial arrest rhythm being the most frequent features.

CONCLUSION

Predictive models utilizing AI-based approaches, including machine and deep learning models exhibit enhanced effectiveness compared to previous regression algorithms, but significant heterogeneity and high risk of bias limit their dependability. Evaluating state-of-the-art deep learning models tailored for tabular data and their clinical generalizability can enhance outcome prediction after cardiac arrest.

Role of artificial intelligence in predicting neurological outcomes in postcardiac resuscitation

Being an extremely high mortality rate condition, cardiac arrest cases have rightfully been evaluated via various studies and scoring factors for effective resuscitative practices and neurological outcomes postresuscitation. This narrative review aims to explore the role of artificial intelligence (AI) in predicting neurological outcomes postcardiac resuscitation. The methodology involved a detailed review of all relevant recent studies of AI, different machine learning algorithms, prediction tools, and assessing their benefit in predicting neurological outcomes in postcardiac resuscitation cases as compared to more traditional prognostic scoring systems and tools. Previously, outcome determining clinical, blood, and radiological factors were prone to other influencing factors like limited accuracy and time constraints. Studies conducted also emphasized that to predict poor neurological outcomes, a more multimodal approach helped adjust for confounding factors, interpret diverse datasets, and provide a reliable prognosis, which only demonstrates the need for AI to help overcome challenges faced. Advanced machine learning algorithms like artificial neural networks (ANN) using supervised learning by AI have improved the accuracy of prognostic models outperforming conventional models. Several real-world cases of effective AI-powered algorithm models have been cited here. Studies comparing machine learning tools like XGBoost, AI Watson, hyperspectral imaging, ChatGPT-4, and AI-based gradient boosting have noted their beneficial uses. AI could help reduce workload, healthcare costs, and help personalize care, process vast genetic and lifestyle data and help reduce side effects from treatments. Limitations of AI have been covered extensively in this article, including data quality, bias, privacy issues, and transparency. Our objectives should be to use more diverse data sources, use interpretable data output giving process explanation, validation method, and implement policies to safeguard patient data. Despite the limitations, the advancements already made by AI and its potential in predicting neurological outcomes in postcardiac resuscitation cases has been quite promising and boosts a continually improving system, albeit requiring close human supervision with training and improving models, with plans to educate clinicians, the public and sharing collected data.

Validation of CASPRI, GO-FAR, PIHCA scores in predicting favorable neurological outcomes after in-hospital cardiac arrest; A five-year three center retrospective study in IRAN

BACKGROUND

Predicting neurological outcomes following in-hospital cardiac arrest is crucial for guiding subsequent clinical treatments. This study seeks to validate the effectiveness of the CASPRI, GO-FAR, and PIHCA tools in predicting favorable neurological outcomes after in-hospital cardiac arrest.

METHOD

This retrospective study utilized a Utstein-style structured form to review the medical records of patients who experienced in-hospital cardiac arrest between March 2018 and March 2023. Predictors were examined using multivariable logistic regression, and the validity of the tools was assessed using ROC curves. Statistical analysis was conducted using SPSS version 25 software.

RESULTS

Out of the 1100 patients included in the study, 42 individuals (3.8%) achieved a favorable neurological outcome. multivariable regression analysis revealed that age, respiratory failure, resuscitation shift, duration of renal failure, and CPC score 24 h before cardiac arrest were significantly associated with favorable neurological outcomes. The predictive abilities of the CASPRI, GO-FAR, and PIHCA scores were calculated as 0.99 (95% CI, 0.98-1.00), 0.98 (95% CI, 0.97-0.99), and 0.96 (95% CI, 0.94-0.99) respectively. A statistically significant difference was observed in the predictive abilities of the CASPRI and PIHCA scores (P = 0.001), while the difference between CASPRI and GO-FAR did not reach significance (P = 0.057). Additionally, there was no significant difference between the predictive abilities of GO-FAR and PIHCA scores (P = 0.159).

CONCLUSION

The study concludes that CASPRI and GO-FAR scores show strong potential as objective measures for predicting favorable neurological outcomes post-cardiac arrest. Integrating these scores into clinical decision-making may enhance treatment and care strategies, in the Iranian healthcare context.

A comparison of Simplified Acute Physiology Score II and Sepsis-related Organ Failure Assessment Score for prediction of mortality after Intensive Care Unit cardiac arrest

BACKGROUND

This study investigates the predictive value and suitable cutoff values of the Sepsis-related Organ Failure Assessment Score (SOFA) and Simplified Acute Physiology Score II (SAPS-II) to predict mortality during or after Intensive Care Unit Cardiac Arrest (ICU-CA).

METHODS

In this secondary analysis the ICU database of a German university hospital with five ICU was screened for all ICU-CA between 2016-2019. SOFA and SAPS-II were used for prediction of mortality during ICU-CA, hospital-stay and one-year-mortality. Receiver operating characteristic curves (ROC), area under the ROC (AUROC) and its confidence intervals were calculated. If the AUROC was significant and considered "acceptable," cutoff values were determined for SOFA and SAPS-II by Youden Index. Odds ratios and sensitivity, specificity, positive and negative predictive values were calculated for the cutoff values.

RESULTS

A total of 114 (78 male; mean age: 72.8±12.5 years) ICU-CA were observed out of 14,264 ICU-admissions (incidence: 0.8%; 95% CI: 0.7-1.0%). 29.8% (N.=34; 95% CI: 21.6-39.1%) died during ICU-CA. SOFA and SAPS-II were not predictive for mortality during ICU-CA (P>0.05). Hospital-mortality was 78.1% (N.=89; 95% CI: 69.3-85.3%). SAPS-II (recorded within 24 hours before and after ICU-CA) indicated a better discrimination between survival and death during hospital stay than SOFA (AUROC: 0.81 [95% CI: 0.70-0.92] vs. 0.70 [95% CI: 0.58-0.83]). A SAPS-II-cutoff-value of 43.5 seems to be suitable for prognosis of hospital mortality after ICU-CA (specificity: 87.5%, sensitivity: 65.6%; SAPS-II>43.5: 87.5% died in hospital; SAPS-II<43.5: 65.6% survived; odds ratio:13.4 [95% CI: 3.25-54.9]). Also for 1-year-mortality (89.5%; 95% CI: 82.3-94.4) SAPS-II showed a better discrimination between survival and death than SOFA: AUROC: 0.78 (95% CI: 0.65-0.91) vs. 0.69 (95% CI: 0.52-0.87) with a cutoff value of the SAPS-II of 40.5 (specificity: 91.7%, sensitivity: 64.3%; SAPS-II>40.5: 96.4% died; SAPS-II<40.5: 42.3% survived; odd ratio: 19.8 [95% CI: 2.3-168.7]).

CONCLUSIONS

Compared to SOFA, SAPS-II seems to be more suitable for prediction of hospital and 1-year-mortality after ICU-CA.

Plasma extracellular vesicle synaptic proteins as biomarkers of clinical progression in patients with Parkinson's disease

Synaptic dysfunction plays a key role in Parkinson's disease (PD), and plasma extracellular vesicle (EV) synaptic proteins are emerging as biomarkers for neurodegenerative diseases. Assessment of plasma EV synaptic proteins for their efficacy as biomarkers in PD and their relationship with disease progression was conducted. In total, 144 participants were enrolled, including 101 people with PD (PwP) and 43 healthy controls (HCs). The changes in plasma EV synaptic protein levels between baseline and 1-year follow-up did not differ significantly in both PwP and HCs. In PwP, the changes in plasma EV synaptic protein levels were significantly associated with the changes in Unified Parkinson's Disease Rating Scale (UPDRS)-II and III scores. Moreover, PwP with elevated levels (first quartile) of any one plasma EV synaptic proteins (synaptosome-associated protein 25, growth-associated protein 43 or synaptotagmin-1) had significantly greater disease progression in UPDRS-II score and the postural instability and gait disturbance subscore in UPDRS-III than did the other PwP after adjustment for age, sex, and disease duration. The promising potential of plasma EV synaptic proteins as clinical biomarkers of disease progression in PD was suggested. However, a longer follow-up period is warranted to confirm their role as prognostic biomarkers.

Predicting in-hospital cardiac arrest outcomes: CASPRI and GO-FAR scores

It is important to predict the neurological prognoses of in-hospital cardiac arrest (IHCA) patients immediately after recovery of spontaneous circulation (ROSC) to make further critical management. The aim of this study was to confirm the usefulness of the Cardiac Arrest Survival Post-Resuscitation In-hospital (CASPRI) and Good Outcome Following Attempted Resuscitation (GO-FAR) scores for predicting the IHCA immediately after the ROSC. This is a retrospective analysis of patient data from a tertiary general hospital located in South Korea. A total of 488 adult patients who had IHCA and achieved sustained ROSC from September 2016 to August 2021 were analyzed to compare effectiveness of the CASPRI and GO-FAR scores related to neurologic prognosis. The primary outcome was Cerebral Performance Category (CPC) score at discharge, defined as a CPC score of 1 or 2. The secondary outcomes were survival-to-discharge and normal neurological status or minimal neurological damage at discharge. Of the 488 included patients, 85 (20.8%) were discharged with good prognoses (CPC score of 1 or 2). The area under the receiver operating characteristic curve of CASPRI score for the prediction of a good neurological outcome was 0.75 (95% CI 0.69-0.81), whereas that of GO-FAR score was 0.67 (95% CI 0.60-0.73). The results of this study show that these scoring systems can be used for timely and satisfactory prediction of the neurological prognoses of IHCA patients after ROSC.

Therapeutic hypothermia in patients after cardiac arrest: A systematic review and meta-analysis of randomized controlled trials

OBJECTIVE

Targeted temperature management (TTM) with therapeutic hypothermia (TH) has been used to improve neurological outcomes in patients after cardiac arrest; however, several trials have reported conflicting results regarding its effectiveness. This systematic review and meta-analysis assessed whether TH was associated with better survival and neurological outcomes after cardiac arrest.

METHOD

We searched online databases for relevant studies published before May 2023. Randomized controlled trials (RCTs) comparing TH and normothermia in post-cardiac-arrest patients were selected. Neurological outcomes and all-cause mortality were assessed as the primary and secondary outcomes, respectively. A subgroup analysis according to initial electrocardiography (ECG) rhythm was performed.

RESULT

Nine RCTs (4058 patients) were included. The neurological prognosis was significantly better in patients with an initial shockable rhythm after cardiac arrest (RR = 0.87, 95% confidence interval [CI] = 0.76-0.99, P = 0.04), especially in those with earlier TH initiation (<120 min) and prolonged TH duration (≥24 h). However, the mortality rate after TH was not lower than that after normothermia (RR = 0.91, 95% CI = 0.79-1.05). In patients with an initial nonshockable rhythm, TH did not provide significantly more neurological or survival benefits (RR = 0.98, 95% CI = 0.93-1.03 and RR = 1.00, 95% CI = 0.95-1.05, respectively).

CONCLUSION

Current evidence with a moderate level of certainty suggests that TH has potential neurological benefits for patients with an initial shockable rhythm after cardiac arrest, especially in those with faster TH initiation and longer TH maintenance.

Risk factor identification and prediction models for prolonged length of stay in hospital after acute ischemic stroke using artificial neural networks

Background

Accurate estimation of prolonged length of hospital stay after acute ischemic stroke provides crucial information on medical expenditure and subsequent disposition. This study used artificial neural networks to identify risk factors and build prediction models for a prolonged length of stay based on parameters at the time of hospitalization.

Methods

We retrieved the medical records of patients who received acute ischemic stroke diagnoses and were treated at a stroke center between January 2016 and June 2020, and a retrospective analysis of these data was performed. Prolonged length of stay was defined as a hospital stay longer than the median number of days. We applied artificial neural networks to derive prediction models using parameters associated with the length of stay that was collected at admission, and a sensitivity analysis was performed to assess the effect of each predictor. We applied 5-fold cross-validation and used the validation set to evaluate the classification performance of the artificial neural network models.

Results

Overall, 2,240 patients were enrolled in this study. The median length of hospital stay was 9 days. A total of 1,101 patients (49.2%) had a prolonged hospital stay. A prolonged length of stay is associated with worse neurological outcomes at discharge. Univariate analysis identified 14 baseline parameters associated with prolonged length of stay, and with these parameters as input, the artificial neural network model achieved training and validation areas under the curve of 0.808 and 0.788, respectively. The mean accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of prediction models were 74.5, 74.9, 74.2, 75.2, and 73.9%, respectively. The key factors associated with prolonged length of stay were National Institutes of Health Stroke Scale scores at admission, atrial fibrillation, receiving thrombolytic therapy, history of hypertension, diabetes, and previous stroke.

Conclusion

The artificial neural network model achieved adequate discriminative power for predicting prolonged length of stay after acute ischemic stroke and identified crucial factors associated with a prolonged hospital stay. The proposed model can assist in clinically assessing the risk of prolonged hospitalization, informing decision-making, and developing individualized medical care plans for patients with acute ischemic stroke.

Identifying Risk Factors for Prolonged Length of Stay in Hospital and Developing Prediction Models for Patients with Cardiac Arrest Receiving Targeted Temperature Management

Background

Prolonged length of stay (LOS) following targeted temperature management (TTM) administered after cardiac arrest may affect healthcare plans and expenditures. This study identified risk factors for prolonged LOS in patients with cardiac arrest receiving TTM and explored the association between LOS and neurological outcomes after TTM.

Methods

The retrospective cohort consisted of 571 non-traumatic cardiac arrest patients aged 18 years or older, treated with cardiopulmonary resuscitation (CPR), had a Glasgow Coma Scale score < 8, or were unable to comply with commands after the restoration of spontaneous circulation (ROSC), and received TTM less than 12 hours after ROSC. Prolonged LOS was defined as LOS beyond the 75th quartile of the entire cohort. We analyzed and compared relevant variables and neurological outcomes between the patients with and without prolonged LOS and established prediction models for estimating the risk of prolonged LOS.

Results

The patients with in-hospital cardiac arrest had a longer LOS than those with out-of-hospital cardiac arrest (p = 0.0001). Duration of CPR (p = 0.02), underlying heart failure (p = 0.001), chronic obstructive pulmonary disease (p = 0.008), chronic kidney disease (p = 0.026), and post-TTM seizures (p = 0.003) were risk factors for prolonged LOS. LOS was associated with survival to hospital discharge, and patients with the lowest and highest Cerebral Performance Category scores at discharge had a shorter LOS. A logistic regression model based on parameters at discharge achieved an area under the curve of 0.840 to 0.896 for prolonged LOS prediction, indicating the favorable performance of this model in predicting LOS in patients receiving TTM.

Conclusions

Our study identified clinically relevant risk factors for prolonged LOS following TTM and developed a prediction model that exhibited adequate predictive performance. The findings of this study broaden our understanding regarding factors associated with hospital stay and can be beneficial while making clinical decisions for patients with cardiac arrest who receive TTM.