Pre-stroke glycemic control is associated with early neurologic deterioration in acute atrial fibrillation-related ischemic stroke

Interpretable Machine Learning Model Predicting Early Neurological Deterioration in Ischemic Stroke Patients Treated with Mechanical Thrombectomy: A Retrospective Study

Early neurologic deterioration (END) is a common and feared complication for acute ischemic stroke (AIS) patients treated with mechanical thrombectomy (MT). This study aimed to develop an interpretable machine learning (ML) model for individualized prediction to predict END in AIS patients treated with MT. The retrospective cohort of AIS patients who underwent MT was from two hospitals. ML methods applied include logistic regression (LR), random forest (RF), support vector machine (SVM), and extreme gradient boosting (XGBoost). The area under the receiver operating characteristic curve (AUC) was the main evaluation metric used. We also used Shapley Additive Explanation (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) to interpret the result of the prediction model. A total of 985 patients were enrolled in this study, and the development of END was noted in 157 patients (15.9%). Among the used models, XGBoost had the highest prediction power (AUC = 0.826, 95% CI 0.781–0.871). The Delong test and calibration curve indicated that XGBoost significantly surpassed those of the other models in prediction. In addition, the AUC in the validating set was 0.846, which showed a good performance of the XGBoost. The SHAP method revealed that blood glucose was the most important predictor variable. The constructed interpretable ML model can be used to predict the risk probability of END after MT in AIS patients. It may help clinical decision making in the perioperative period of AIS patients treated with MT.

Interpretable machine learning for early neurological deterioration prediction in atrial fibrillation-related stroke

We aimed to develop a novel prediction model for early neurological deterioration (END) based on an interpretable machine learning (ML) algorithm for atrial fibrillation (AF)-related stroke and to evaluate the prediction accuracy and feature importance of ML models. Data from multicenter prospective stroke registries in South Korea were collected. After stepwise data preprocessing, we utilized logistic regression, support vector machine, extreme gradient boosting, light gradient boosting machine (LightGBM), and multilayer perceptron models. We used the Shapley additive explanation (SHAP) method to evaluate feature importance. Of the 3,213 stroke patients, the 2,363 who had arrived at the hospital within 24 h of symptom onset and had available information regarding END were included. Of these, 318 (13.5%) had END. The LightGBM model showed the highest area under the receiver operating characteristic curve (0.772; 95% confidence interval, 0.715–0.829). The feature importance analysis revealed that fasting glucose level and the National Institute of Health Stroke Scale score were the most influential factors. Among ML algorithms, the LightGBM model was particularly useful for predicting END, as it revealed new and diverse predictors. Additionally, the effects of the features on the predictive power of the model were individualized using the SHAP method.

High HbA1c is associated with decreased 6-month survival and poor outcomes after out-of-hospital cardiac arrest: a retrospective cohort study

Background

To evaluate the associations between glycated hemoglobin (HbA1c) at admission and 6-month mortality and outcomes after out-of-hospital cardiac arrest (OHCA) treated by hypothermic targeted temperature management (TTM).

Methods

This single-center retrospective cohort study included adult OHCA survivors who underwent hypothermic TTM from December 2011 to December 2019. High HbA1c at admission was defined as a level higher than 6%. Poor neurological outcomes were defined as cerebral performance category scores of 3–5. The primary outcome was 6-month mortality. The secondary outcome was the 6-month neurological outcome. Descriptive statistics, log-rank tests, and multivariable regression modeling were used for data analysis.

Results

Of the 302 patients included in the final analysis, 102 patients (33.8%) had HbA1c levels higher than 6%. The high HbA1c group had significantly worse 6-month survival (12.7% vs. 37.5%, p < 0.001) and 6-month outcomes (89.2% vs. 73.0%, p = 0.001) than the non-high HbA1c group. Kaplan-Meier analysis and the log-rank test showed that the survival time was significantly shorter in the patients with HbA1c > 6% than in those with HbA1c ≤6%. In the multivariable logistic regression analysis, HbA1c > 6% was independently associated with 6-month mortality (OR 5.85, 95% CI 2.26–15.12, p < 0.001) and poor outcomes (OR 4.18, 95% CI 1.41–12.40, p < 0.001).

Conclusions

This study showed that HbA1c higher than 6% at admission was associated with increased 6-month mortality and poor outcomes in OHCA survivors treated with hypothermic TTM. Poor long-term glycemic management may have prognostic significance after cardiac arrest.

Acute Neurological Deterioration in Large Vessel Occlusions and Mild Symptoms Managed Medically

Background and Purpose—

It is unclear which factors predict acute neurological deterioration in patients with large vessel occlusion and mild symptoms. We aim to evaluate the frequency, timing, and potential predictors of acute neurological deterioration ≥4 National Institutes of Health Stroke Scale (NIHSS) points in medically managed patients with large vessel occlusion and mild presentation.

Methods—

Single-center retrospective study of patients with consecutive minor stroke (defined as NIHSS score of ≤5 on presentation) and large vessel occlusion from January 2014 to December 2017. Primary outcome was acute neurological deterioration ≥4 NIHSS points during the hospitalization. Secondary outcomes included ΔNIHSS (defined as discharge minus admission NIHSS score).

Results—

Among 1133 patients with acute minor strokes, 122 (10.6%) had visible occlusions on computed tomography angiography/magnetic resonance angiography. Twenty-four (19.7%) patients had ≥4 points deterioration on NIHSS at a median of 3.6 (1–16) hours from arrival. No clinical or radiological predictors of acute neurological deterioration ≥4 NIHSS points were observed on multivariable analysis. Rescue endovascular thrombectomy was performed more often in the ones with acute neurological deterioration ≥4 NIHSS points compared with patients with no deterioration (54% versus 0%; P <0.001). Acute neurological deterioration ≥4 NIHSS points was associated with ΔNIHSS ≥4 points (33% versus 4.9%; P <0.01) and a trend toward lower independence rates at discharge (50% versus 70%; P =0.06) compared with the group with no deterioration. In patients with any degree of neurological worsening, patients who underwent rescue thrombectomy were more likely to be independent at discharge (73% versus 38%; P =0.02) and to have a favorable ΔNIHSS (−2 [−3 to 0] versus 0 [−1 to 6]; P =0.05) compared with the ones not offered rescue thrombectomy.

Conclusions—

Acute neurological deterioration ≥4 NIHSS points was observed in a fifth of patients with large vessel occlusion and mild symptoms, occurred very early in the hospital course, impacted functional outcomes, and could not be predicted by any of the studied clinical and radiological variables. Rescue thrombectomy was associated with improved clinical outcomes at discharge in patients with neurological deterioration.

Clinical Conundrums in the Management of Diabetic Ketoacidosis in the Elderly

One of the dreaded life-threatening complications of diabetes mellitus (DM) is diabetic ketoacidosis (DKA). American Diabetic Association (ADA) came out with 2018 guidelines on the management of DM and its complications, but these are woefully silent on the clinical conundrums that accompany DKA in elderly patients. In elderly patients, DKA is often complicated by sepsis, atrial fibrillation, polypharmacy, nonketotic hyperosmolar states, atypical clinical presentations, acute kidney injury (AKI), dementia, and noncompliance with medications. Here in we highlight these conundrums that need to be addressed to improve morbidity and mortality in elderly patients.

The association between long-term glycaemic control, glycaemic gap and neurological outcome of in-hospital cardiac arrest in diabetics: A retrospective cohort study

AIM

Resuscitation guidelines do not recommend a target blood glucose (BG) level specifically tailored for diabetics experiencing an in-hospital cardiac arrest (IHCA). The glycosylated haemoglobin (HbA1c) level may be associated with neurological prognosis and used to identify the optimal BG level for diabetic IHCA patients.

METHODS

This study was a retrospective study in a single medical centre. Patients with an IHCA between 2006 and 2015 were screened. The estimated average glucose (eAG) level was converted from the HbA1c level measured within three months prior to the IHCA. The minimum glycaemic gap was calculated from the post-resuscitation minimum BG level minus the eAG level.

RESULTS

A total of 141 patients were included in this study. The mean HbA1c was 7.2% (corresponding eAG: 160.2 mg/dL [8.9 mmol/L]). Multivariable logistic regression analysis indicated an eAG level of less than 196 mg/dL (10.9 mmol/L; corresponding HbA1c: 8.5%) was positively associated with a favourable neurological outcome at hospital discharge (odds ratio [OR]: 5.12, 95% confidence interval [CI]: 1.11-23.70; p-value = 0.04). An absolute minimum glycaemic gap of less than 70 mg/dL (3.9 mmol/L) was also positively associated with a favourable neurological outcome (OR: 5.41, 95% CI: 1.41-20.78; p-value = 0.01).

CONCLUSION

For diabetic patients, poor long-term glycaemic control correlated with worse neurological recovery following an IHCA. The HbA1c-derived average BG level could be used as a reference point for glycaemic management during the early stage of post-cardiac arrest syndrome. The glycaemic gap could be used to identify the optimal glycaemic range around the reference point.

Mean Platelet Volume Is Associated with Early Neurological Deterioration in Patients with Branch Atheromatous Disease: Involvement of Platelet Activation

BACKGROUND

The most attentive clinical problem in patients with branch atheromatous disease (BAD) is early neurological deterioration (END). Although the platelet activation (PA) is involved in pathogenesis, the relationship between PA and END has remained unclear. We investigated clinical data including mean platelet volume (MPV, fL) as a marker for PA to identify clinically useful biomarkers for END.

METHODS

A total of 64 patients with BAD were investigated retrospectively, and divided into 2 groups based on whether neurologic symptoms deteriorated or not: BAD with and without END (END and non-END). The END was defined as patients with point increase of 1 or greater in the National Institutes of Health Stroke Scale (NIHSS); non-END was defined as those without such increase. Clinical features such as NIHSS, modified Rankin scale (mRS), laboratory data including MPV, lesion size (LS, mm) on admission, and treatments were compared between the 2 groups.

RESULTS

Of 64 patients, 17 cases had an END. The median values of NIHSS, mRS, MPV, and LS on admission were significantly greater in END than in non-END (P < .05, respectively). There was no correlation of MPV with NIHSS, mRS and LS, respectively. The median values of MPV were significantly higher in END than in non-END and control (P < .05, respectively). A receiver operating characteristic curve indicated a value of 10.1 as cutoff level for MPV to discriminate between END and non-END.

CONCLUSIONS

High MPV values on admission may be an independent biomarker for END. Physicians should pay more careful attention to END in BAD showing MPV values higher than 10.1 on admission.