Admission Laboratory Values Accurately Predict In-hospital Mortality: a Retrospective Cohort Study

Pathophysiological Features in Electronic Medical Records Sustain Model Performance under Temporal Dataset Shift

Access to real-world data streams like electronic medical records (EMRs) has accelerated the development of supervised machine learning (ML) models for clinical applications. However, few studies investigate the differential impact of particular features in the EMR on model performance under temporal dataset shift. To explain how features in the EMR impact models over time, this study aggregates features into feature groups by their source (e.g. medication orders, diagnosis codes and lab results) and feature categories based on their reflection of patient pathophysiology or healthcare processes. We adapt Shapley values to explain feature groups' and feature categories' marginal contribution to initial and sustained model performance. We investigate three standard clinical prediction tasks and find that while feature contributions to initial performance differ across tasks, pathophysiological features help mitigate temporal discrimination deterioration. These results provide interpretable insights on how specific feature groups contribute to model performance and robustness to temporal dataset shift.

Accurate correction model of blood potassium concentration in hemolytic specimens

BACKGROUND AND AIMS

The results of blood potassium can be seriously affected by specimen hemolysis which may interfere with clinicians' interpretation of test results. Redrawing blood and retesting may delay treatment time and it is not feasible for critically ill patients with difficulty in specimen collection. Therefore, it is significant to establish a mathematical model that can quickly correct the blood potassium concentration of hemolytic specimens.

MATERIALS AND METHODS

The residual blood samples from 107 patients at Peking University Third Hospital were collected to establish potassium correction model. Samples with different hemolysis indexes were obtained by ultrasonic crushing method. Blood potassium correction models of hemolysis specimens were established by linear regression and curve fitting using SPSS and MATLAB, respectively. In addition, blood samples from another 85 patients were used to verify the accuracy of the models and determine the optimal model.

RESULTS

Variation of potassium (ΔK) was 0.003HI-0.03 (R2 = 0.9749) in linear regression model which had high correlation in ΔK and HI, and the correction formula was Kcorrection = Khemolysis-0.003 × HI + 0.03. Average rate of potassium change (αaverage) was 0.003 ± 0.0002 mmol/L in curve fitting model, and correction formula was Kcorrection = Khemolysis-0.003 × HI, and both men and women can use the same correction model. The accuracy of linear regression model was 96.5 %, and there was statistical difference between the verification results and the measured values (p < 0.05), while the accuracy of curve fitting model was 100 %, and there was no statistical difference between the verification results and the measured values (p = 0.552). The model was validated in an independent set of samples and all were within the TEa of 6 % and the accuracy of 100 %.

CONCLUSIONS

Both linear regression and curve fitting models of potassium correction had high accuracy, and can effectively correct the potassium concentration of hemolytic specimens, while the curve fitting model have superior accuracy.

Clinical Data based XGBoost Algorithm for infection risk prediction of patients with decompensated cirrhosis: a 10-year (2012-2021) Multicenter Retrospective Case-control study

OBJECTIVES

To appraise effective predictors for infection in patients with decompensated cirrhosis (DC) by using XGBoost algorithm in a retrospective case-control study.

METHODS

Clinical data were retrospectively collected from 6,648 patients with DC admitted to five tertiary hospitals. Indicators with significant differences were determined by univariate analysis and least absolute contraction and selection operator (LASSO) regression. Further multi-tree extreme gradient boosting (XGBoost) machine learning-based model was used to rank importance of features selected from LASSO and subsequently constructed infection risk prediction model with simple-tree XGBoost model. Finally, the simple-tree XGBoost model is compared with the traditional logical regression (LR) model. Performances of models were evaluated by area under the receiver operating characteristic curve (AUROC), sensitivity, and specificity.

RESULTS

Six features, including total bilirubin, blood sodium, albumin, prothrombin activity, white blood cell count, and neutrophils to lymphocytes ratio were selected as predictors for infection in patients with DC. Simple-tree XGBoost model conducted by these features can predict infection risk accurately with an AUROC of 0.971, sensitivity of 0.915, and specificity of 0.900 in training set. The performance of simple-tree XGBoost model is better than that of traditional LR model in training set, internal verification set, and external feature set (P < 0.001).

CONCLUSIONS

The simple-tree XGBoost predictive model developed based on a minimal amount of clinical data available to DC patients with restricted medical resources could help primary healthcare practitioners promptly identify potential infection.

Using machine-learning methods to predict in-hospital mortality through the Elixhauser index: A Medicare data analysis

Accurate in-hospital mortality prediction can reflect the prognosis of patients, help guide allocation of clinical resources, and help clinicians make the right care decisions. There are limitations to using traditional logistic regression models when assessing the model performance of comorbidity measures to predict in-hospital mortality. Meanwhile, the use of novel machine-learning methods is growing rapidly. In 2021, the Agency for Healthcare Research and Quality published new guidelines for using the Present-on-Admission (POA) indicator from the International Classification of Diseases, Tenth Revision, for coding comorbidities to predict in-hospital mortality from the Elixhauser's comorbidity measurement method. We compared the model performance of logistic regression, elastic net model, and artificial neural network (ANN) to predict in-hospital mortality from Elixhauser's measures under the updated POA guidelines. In this retrospective analysis, 1,810,106 adult Medicare inpatient admissions from six US states admitted after September 23, 2017, and discharged before April 11, 2019 were extracted from the Centers for Medicare and Medicaid Services data warehouse. The POA indicator was used to distinguish pre-existing comorbidities from complications that occurred during hospitalization. All models performed well (C-statistics >0.77). Elastic net method generated a parsimonious model, in which there were five fewer comorbidities selected to predict in-hospital mortality with similar predictive power compared to the logistic regression model. ANN had the highest C-statistics compared to the other two models (0.800 vs. 0.791 and 0.791). Elastic net model and AAN can be applied successfully to predict in-hospital mortality.

Evaluation of stacked ensemble model performance to predict clinical outcomes: A COVID-19 study

BACKGROUND

The application of machine learning (ML) to analyze clinical data with the goal to predict patient outcomes has garnered increasing attention. Ensemble learning has been used in conjunction with ML to improve predictive performance. Although stacked generalization (stacking), a type of heterogeneous ensemble of ML models, has emerged in clinical data analysis, it remains unclear how to define the best model combinations for strong predictive performance. This study develops a methodology to evaluate the performance of "base" learner models and their optimized combination using "meta" learner models in stacked ensembles to accurately assess performance in the context of clinical outcomes.

METHODS

De-identified COVID-19 data was obtained from the University of Louisville Hospital, where a retrospective chart review was performed from March 2020 to November 2021. Three differently-sized subsets using features from the overall dataset were chosen to train and evaluate ensemble classification performance. The number of base learners chosen from several algorithm families coupled with a complementary meta learner was varied from a minimum of 2 to a maximum of 8. Predictive performance of these combinations was evaluated in terms of mortality and severe cardiac event outcomes using area-under-the-receiver-operating-characteristic (AUROC), F1, balanced accuracy, and kappa.

RESULTS

The results highlight the potential to accurately predict clinical outcomes, such as severe cardiac events with COVID-19, from routinely acquired in-hospital patient data. Meta learners Generalized Linear Model (GLM), Multi-Layer Perceptron (MLP), and Partial Least Squares (PLS) had the highest AUROC for both outcomes, while K-Nearest Neighbors (KNN) had the lowest. Performance trended lower in the training set as the number of features increased, and exhibited less variance in both training and validation across all feature subsets as the number of base learners increased.

CONCLUSION

This study offers a methodology to robustly evaluate ensemble ML performance when analyzing clinical data.

Multimorbidity and catastrophic health expenditure: Evidence from the China Health and Retirement Longitudinal Study

Background

Population aging accompanied by multimorbidity imposes a great burden on households and the healthcare system. This study aimed to determine the incidence and determinants of catastrophic health expenditure (CHE) in the households of old people with multimorbidity in China.

Methods

Data were obtained from the China Health and Retirement Longitudinal Study (CHARLS) conducted in 2018, with 3,511 old people (≥60 years) with multimorbidity responding to the survey on behalf of their households. CHE was identified using two thresholds: ≥10% of out-of-pocket (OOP) health spending in total household expenditure (THE) and ≥40% of OOP health spending in household capacity to pay (CTP) measured by non-food household expenditure. Logistic regression models were established to identify the individual and household characteristics associated with CHE incidence.

Results

The median values of THE, OOP health spending, and CTP reached 19,900, 1,500, and 10,520 Yuan, respectively. The CHE incidence reached 31.5% using the ≥40% CTP threshold and 45.6% using the ≥10% THE threshold. It increased by the number of chronic conditions reported by the respondents (aOR = 1.293-1.855, p < 0.05) and decreased with increasing household economic status (aOR = 1.622-4.595 relative the highest quartile, p < 0.001). Hospital admissions over the past year (aOR = 6.707, 95% CI: 5.186 to 8.674) and outpatient visits over the past month (aOR = 4.891, 95% CI: 3.822 to 6.259) of the respondents were the strongest predictors of CHE incidence. The respondents who were male (aOR = 1.266, 95% CI: 1.054 to 1.521), married (OR = 1.502, 95% CI: 1.211 to 1.862), older than 70 years (aOR = 1.288-1.458 relative to 60-69 years, p < 0.05), completed primary (aOR = 1.328 relative to illiterate, 95% CI: 1.079 to 1.635) or secondary school education (aOR = 1.305 relative to illiterate, 95% CI: 1.002 to 1.701), lived in a small (≤2 members) household (aOR = 2.207, 95% CI: 1.825 to 2.669), and resided in the northeast region (aOR = 1.935 relative to eastern, 95% CI: 1.396 to 2.682) were more likely to incur CHE.

Conclusion

Multimorbidity is a significant risk of CHE. Household CHE incidence increases with the number of reported chronic conditions. Socioeconomic and regional disparities in CHE incidence persist in China.

Parameters of the complete blood count predict in hospital mortality

INTRODUCTION

Mortality rates are used to evaluate the quality of hospital care after adjusting for disease severity and, commonly also, for age, comorbidity, and laboratory data with only few parameters of the complete blood count (CBC).

OBJECTIVE

To identify the parameters of the CBC that predict independently in-hospital mortality of acutely admitted patients.

POPULATION

All patients were admitted to internal medicine, cardiology, and intensive care departments at the Laniado Hospital in Israel in 2018 and 2019.

VARIABLES

Independent variables were patients' age, sex, and parameters of the CBC. The outcome variable was in-hospital mortality.

ANALYSIS

Logistic regression. In 2018, we identified the variables that were associated with in-hospital mortality and validated this association in the 2019 cohort.

RESULTS

In the validation cohort, a model consisting of nine parameters that are commonly available in modern analyzers had a c-statistics (area under the receiver operator curve) of 0.86 and a 10%-90% risk gradient of 0%-21.4%. After including the proportions of large unstained cells, hypochromic, and macrocytic red cells, the c-statistic increased to 0.89, and the risk gradient to 0.1%-29.5%.

CONCLUSION

The commonly available parameters of the CBC predict in-hospital mortality. Addition of the proportions of hypochromic red cells, macrocytic red cells, and large unstained cells may improve the predictive value of the CBC.

The Value Proposition for Pathologists: A Population Health Approach

The transition to a value-based payment system offers pathologists the opportunity to play an increased role in population health by improving outcomes and safety as well as reducing costs. Although laboratory testing itself accounts for a small portion of health-care spending, laboratory data have significant downstream effects in patient management as well as diagnosis. Pathologists currently are heavily engaged in precision medicine, use of laboratory and pathology test results (including autopsy data) to reduce diagnostic errors, and play leading roles in diagnostic management teams. Additionally, pathologists can use aggregate laboratory data to monitor the health of populations and improve health-care outcomes for both individual patients and populations. For the profession to thrive, pathologists will need to focus on extending their roles outside the laboratory beyond the traditional role in the analytic phase of testing. This should include leadership in ensuring correct ordering and interpretation of laboratory testing and leadership in population health programs. Pathologists in training will need to learn key concepts in informatics and data analytics, health-care economics, public health, implementation science, and health systems science. While these changes may reduce reimbursement for the traditional activities of pathologists, new opportunities arise for value creation and new compensation models. This report reviews these opportunities for pathologist leadership in utilization management, precision medicine, reducing diagnostic errors, and improving health-care outcomes.