Decision tree model for predicting in-hospital cardiac arrest among patients admitted with acute coronary syndrome

Application of Machine Learning for Patients With Cardiac Arrest: Systematic Review and Meta-Analysis

BACKGROUND

Currently, there is a lack of effective early assessment tools for predicting the onset and development of cardiac arrest (CA). With the increasing attention of clinical researchers on machine learning (ML), some researchers have developed ML models for predicting the occurrence and prognosis of CA, with certain models appearing to outperform traditional scoring tools. However, these models still lack systematic evidence to substantiate their efficacy.

OBJECTIVE

This systematic review and meta-analysis was conducted to evaluate the prediction value of ML in CA for occurrence, good neurological prognosis, mortality, and the return of spontaneous circulation (ROSC), thereby providing evidence-based support for the development and refinement of applicable clinical tools.

METHODS

PubMed, Embase, the Cochrane Library, and Web of Science were systematically searched from their establishment until May 17, 2024. The risk of bias in all prediction models was assessed using the Prediction Model Risk of Bias Assessment Tool.

RESULTS

In total, 93 studies were selected, encompassing 5,729,721 in-hospital and out-of-hospital patients. The meta-analysis revealed that, for predicting CA, the pooled C-index, sensitivity, and specificity derived from the imbalanced validation dataset were 0.90 (95% CI 0.87-0.93), 0.83 (95% CI 0.79-0.87), and 0.93 (95% CI 0.88-0.96), respectively. On the basis of the balanced validation dataset, the pooled C-index, sensitivity, and specificity were 0.88 (95% CI 0.86-0.90), 0.72 (95% CI 0.49-0.95), and 0.79 (95% CI 0.68-0.91), respectively. For predicting the good cerebral performance category score 1 to 2, the pooled C-index, sensitivity, and specificity based on the validation dataset were 0.86 (95% CI 0.85-0.87), 0.72 (95% CI 0.61-0.81), and 0.79 (95% CI 0.66-0.88), respectively. For predicting CA mortality, the pooled C-index, sensitivity, and specificity based on the validation dataset were 0.85 (95% CI 0.82-0.87), 0.83 (95% CI 0.79-0.87), and 0.79 (95% CI 0.74-0.83), respectively. For predicting ROSC, the pooled C-index, sensitivity, and specificity based on the validation dataset were 0.77 (95% CI 0.74-0.80), 0.53 (95% CI 0.31-0.74), and 0.88 (95% CI 0.71-0.96), respectively. In predicting CA, the most significant modeling variables were respiratory rate, blood pressure, age, and temperature. In predicting a good cerebral performance category score 1 to 2, the most significant modeling variables in the in-hospital CA group were rhythm (shockable or nonshockable), age, medication use, and gender; the most significant modeling variables in the out-of-hospital CA group were age, rhythm (shockable or nonshockable), medication use, and ROSC.

CONCLUSIONS

ML represents a currently promising approach for predicting the occurrence and outcomes of CA. Therefore, in future research on CA, we may attempt to systematically update traditional scoring tools based on the superior performance of ML in specific outcomes, achieving artificial intelligence-driven enhancements.

TRIAL REGISTRATION

PROSPERO International Prospective Register of Systematic Reviews CRD42024518949; https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=518949.

Risk prediction of arrhythmia after percutaneous coronary intervention in patients with acute coronary syndrome: A systematic review and meta-analysis

The purpose of study was to evaluate the predictive performance of models for the development of arrhythmias in patients with acute coronary syndrome after percutaneous coronary intervention. Two researchers screened the literature according to CHARMS, and assessed the risk of bias and applicability based on PROBAST. A total of 44 studies were included in the review, comprising 62 models, of which 30 models identified as having a low risk of bias, and only 7 studies combined other machine learning algorithms. A meta-analysis of some of the studies combined gave an AUC of 0.813 (95 % CI 0.791 to 0.835), and a meta-analysis of the models with low bias among them gave an AUC of 0.803 (95 % CI 0.768 to 0.837). The performance of the integrated models was satisfactory overall, but the modelling approach was homogeneous. The external validation of the existing models should be incorporated to enhance their extrapolation.

Radiomics model based on computed tomography images for prediction of radiation-induced optic neuropathy following radiotherapy of brain and head and neck tumors

Purpose

We aimed to build a machine learning-based model to predict radiation-induced optic neuropathy in patients who had treated head and neck cancers with radiotherapy.

Materials and methods

To measure radiation-induced optic neuropathy, the visual evoked potential values were obtained in both case and control groups and compared. Radiomics features were extracted from the area segmented which included the right and left optic nerves and chiasm. We integrated CT image features with dosimetric and clinical data subsequently, ranked 5 supervised ML models Bernoulli Naive Bayes, Decision Tree, Gradient Boosting Decision Trees, K-Nearest Neighbor, and Random Forest on 4 input datasets to predict radiation-induced visual complications classifiers by implementing 5-fold cross-validation. The F1 score, accuracy, sensitivity, specificity, and area under the ROC curve were compared to access prediction capability.

Results

radiation-induced optic neuropathy affected 31 % of the patients. 856 radiomic characteristics were extracted and selected from each segmented area. Decision Tree and Random Forest with the highest AUC (97 % and 95 % respectively) among the five classifiers were the most powerful algorithms to predict radiation-induced optic neuropathy. Chiasm with higher sensitivity and precision was able to predict radiation-induced optic neuropathy better than right or left optic nerve or combination of all radiomic features.

Conclusion

We found that combination of radiomic, dosimetric, and clinical factors can predict radiation-induced optic neuropathy after radiation treatment with high accuracy. To acquire more reliable results, it is recommended to conduct visual evoked potential tests before and after radiation therapy, with multiple follow-up courses and more patients.

Integrating EPSOSA-BP neural network algorithm for enhanced accuracy and robustness in optimizing coronary artery disease prediction

Coronary artery disease represents a formidable health threat to middle-aged and elderly populations worldwide. This research introduces an advanced BP neural network algorithm, EPSOSA-BP, which integrates particle swarm optimization, simulated annealing, and a particle elimination mechanism to elevate the precision of heart disease prediction models. To address prior limitations in feature selection, the study employs single-hot encoding and Principal Component Analysis, thereby enhancing the model's feature learning capability. The proposed method achieved remarkable accuracy rates of 93.22% and 95.20% on the UCI and Kaggle datasets, respectively, underscoring its exceptional performance even with small sample sizes. Ablation experiments further validated the efficacy of the data preprocessing and feature selection techniques employed. Notably, the EPSOSA algorithm surpassed classical optimization algorithms in terms of convergence speed, while also demonstrating improved sensitivity and specificity. This model holds significant potential for facilitating early identification of high-risk patients, which could ultimately save lives and optimize the utilization of medical resources. Despite implementation challenges, including technical integration and data standardization, the algorithm shows promise for use in emergency settings and community health services for regular cardiac risk monitoring.

Application of multi-feature-based machine learning models to predict neurological outcomes of cardiac arrest

Cardiac arrest (CA) is a major disease burden worldwide and has a poor prognosis. Early prediction of CA outcomes helps optimize the therapeutic regimen and improve patients' neurological function. As the current guidelines recommend, many factors can be used to evaluate the neurological outcomes of CA patients. Machine learning (ML) has strong analytical abilities and fast computing speed; thus, it plays an irreplaceable role in prediction model development. An increasing number of researchers are using ML algorithms to incorporate demographics, arrest characteristics, clinical variables, biomarkers, physical examination findings, electroencephalograms, imaging, and other factors with predictive value to construct multi-feature prediction models for neurological outcomes of CA survivors. In this review, we explore the current application of ML models using multiple features to predict the neurological outcomes of CA patients. Although the outcome prediction model is still in development, it has strong potential to become a powerful tool in clinical practice.

Blood laboratory analyses preceding in-hospital cardiac arrest: A matched case-control study

BACKGROUND

Whether blood laboratory analyses differ in patients who later suffer in-hospital cardiac arrest (IHCA) compared to other hospitalised patients remains unknown. The aim of this study was to describe pre-arrest sampling frequencies, results, and trends in blood laboratory analyses in patients with IHCA compared to controls.

METHODS

This study was a matched case-control study using national registries in Denmark. Cases were defined as patients with IHCA from 2017 to 2021. Controls were defined as hospitalised patients and were matched on age, sex, and date and length of admission. Data on a total of 51 different blood laboratory analyses were obtained. The laboratory analyses of primary interest were lactate, sodium, potassium, and haemoglobin. The index time for cases was defined as the time of cardiac arrest, and a corresponding index time was defined for controls based on the time to cardiac arrest for their corresponding case. Blood sampling frequencies were reported for blood laboratory analyses obtained either within the last 24 h before the index time or between the time of hospital admission and the index time. Blood sampling results were reported for blood laboratory analyses obtained within the last 24 h before the index time.

RESULTS

A total of 9268 cases and 92,395 controls were included in this study. Cases underwent more frequent sampling of all blood laboratory analyses compared to controls. This higher sampling frequency was more pronounced for lactate compared to sodium, potassium, or haemoglobin. The last measured lactate was higher in cases (median [IQR]: 2.3 [1.3, 4.9]) compared to controls (median [IQR]: 1.3 [0.9, 2.0]). Differences in sodium, potassium, and haemoglobin were negligible. The proportion of abnormally elevated levels of lactate and potassium increased as time to cardiac arrest decreased; no such effect was seen in controls. No temporal trend was evident for sodium or haemoglobin.

CONCLUSIONS

Patients with IHCA undergo more frequent blood sampling prior to IHCA and have higher levels of lactate compared to matched controls.

[Classification tree obtained by artificial intelligence for the prediction of heart failure after acute coronary syndromes]

BACKGROUND

Coronary heart disease is the leading cause of heart failure (HF), and tools are needed to identify patients with a higher probability of developing HF after an acute coronary syndrome (ACS). Artificial intelligence (AI) has proven to be useful in identifying variables related to the development of cardiovascular complications.

METHODS

We included all consecutive patients discharged after ACS in two Spanish centers between 2006 and 2017. Clinical data were collected and patients were followed up for a median of 53months. Decision tree models were created by the model-based recursive partitioning algorithm.

RESULTS

The cohort consisted of 7,097 patients with a median follow-up of 53months (interquartile range: 18-77). The readmission rate for HF was 13.6% (964 patients). Eight relevant variables were identified to predict HF hospitalization time: HF at index hospitalization, diabetes, atrial fibrillation, glomerular filtration rate, age, Charlson index, hemoglobin, and left ventricular ejection fraction. The decision tree model provided 15 clinical risk patterns with significantly different HF readmission rates.

CONCLUSIONS

The decision tree model, obtained by AI, identified 8 leading variables capable of predicting HF and generated 15 differentiated clinical patterns with respect to the probability of being hospitalized for HF. An electronic application was created and made available for free.

Lifestyle Influence on Mild Cognitive Impairment Progression: A Decision Tree Prediction Model Study

Purpose

This study assessed the influences of different lifestyle on mild cognitive impairment (MCI) progression and established a decision tree prediction model to analyse their predictive significance on MCI progression incidence.

Patients and Methods

From October 2015 to February 2020,330 patients with MCI were recruited, and demographic and lifestyle information collected. They were followed up for 19.04 ± 10.227 months. Cognitive function was assessed using the Mini-Mental State Examination Scale every 6 months, and they were divided into MCI stable group and MCI progression group.

Results

The Kaplan Meier survival analysis showed an overall cohort survival rate of 33.2%; the annual conversion rate of MCI progression was 20%. Physical exercise, social engagement, high-fat diet, age, napping, and tea drinking were decision tree prediction model nodes. Hobbies were the most important factor for predicting MCI progression. The MCI progression probability rates were: with hobbies 26.829% (44 cases), without hobbies 57.831% (96 cases); for those withot hobbies, with physical exercise 43.077% (28 cases) without physical exercise 72.340% (68 cases); for those without hobbies with physical exercise and social engagement 20.000% (4 cases), without social engagement 53.333% (24 cases); for those without hobbies, physical exercises and social engagement and with nap habits 48.485% (16 cases), without nap habits 66.667% (8 cases). The decision tree prediction model AUC for predicting the MCI progression receiver operating characteristic curve was 0.737 (95% confidence interval: 0.685-0.785) (75.71% sensitivity, 71.75% specificity, P < 0.001.

Conclusion

Hobbies, physical exercise, social engagement, napping, and drinking tea can help prevent MCI progression, while a high-fat diet may exacerbate MCI progression. In this study the rule with the lowest MCI progress probability for those who had hobbies, high-fat diet, and social engagement. And the decision tree model had good prediction efficiency.

Physical activity, sleep disorders, and type of work in the prevention of cognitive function decline in patients with hypertension

BACKGROUND

Hypertensive patients are likelier to have cognitive function decline (CFD). This study aimed to explore physical activity level, sleep disorders, and type of work that influenced intervention effects on cognitive function decline in hypertensive patients and to establish a decision tree model to analyze their predictive significance on the incidence of CFD in hypertensive patients.

METHODS

This cross-sectional study recruited patients with essential hypertension from several hospitals in Shandong Province from May 2022 to December 2022. Subject exclusion criteria included individuals diagnosed with congestive heart failure, valvular heart disease, cardiac surgery, hepatic and renal dysfunction, and malignancy. Recruitment is through multiple channels such as hospital medical and surgical outpatient clinics, wards, and health examination centers. Cognitive function was assessed using the Mini-Mental State Examination (MMSE), and sleep quality was assessed using the Pittsburgh Sleep Quality Index (PSQI). Moreover, we obtained information on the patients' type of work through a questionnaire and their level of physical activity through the International Physical Activity Questionnaire (IPAQ).

RESULTS

The logistic regression analysis results indicate that sleep disorder is a significant risk factor for CFD in hypertension patients(OR:1.85, 95%CI:[1.16,2.94]), mental workers(OR:0.12, 95%CI: [0.04,0.37]) and those who perform both manual and mental workers(OR: 0.5, 95%CI: [0.29,0.86]) exhibit protective effects against CFD. Compared to low-intensity, moderate physical activity(OR: 0.53, 95%CI: [0.32,0.87]) and high-intensity physical activity(OR: 0.26, 95%CI: [0.12,0.58]) protects against CFD in hypertension patients. The importance of predictors in the decision tree model was ranked as follows: physical activity level (54%), type of work (27%), and sleep disorders (19%). The area under the ROC curves the decision tree model predicted was 0.72 [95% CI: 0.68 to 0.76].

CONCLUSION

Moderate and high-intensity physical activity may reduce the risk of developing CFD in hypertensive patients. Sleep disorders is a risk factor for CFD in hypertensive patients. Hypertensive patients who engage in mental work and high-intensity physical activity effectively mitigate the onset of CFD in hypertensive patients.

Pre-arrest comorbidity burden and the future risk of out-of-hospital cardiac arrest in Korean adults

OBJECTIVE

To investigate the impact of pre-arrest comorbidities on future out-of-hospital cardiac arrest (OHCA) development using a nationwide dataset.

METHODS

This population-based, matched case-control study used the national health insurance claims data relevant to OHCA in South Korea from January 2009 to December 2018. Case patients were randomly matched to controls by age, sex and date of cardiac arrest. Controls were defined as patients who did not experience OHCA based on claim codes in national health screening data. The comorbidity burden was assessed using the Charlson Comorbidity Index (CCI).

RESULTS

A total of 191 370 OHCA patients were matched to 347 568 controls. The mean CCI in the case group was 3.76, which was significantly higher than that in the control group (1.75, p<0.001). Overall, OHCA was 1.35 (95% CI 1.34 to 1.35) times more likely to occur with every 1 point increase in the CCI. All other comorbidities constituting the CCI were associated with the OHCA risk (p<0.001). Patients with CCI ≥3 presented an OR of 3.71 (95% CI 3.67 to 3.76) for the risk of OHCA occurrence. This association was more pronounced in patients aged <70 years than in those aged ≥70 years (OR (95% CI) 16.07 (15.48 to 16.68) vs 6.50 (6.33 to 6.68)).

CONCLUSION

A high burden of pre-arrest comorbidity was associated with a higher risk of OHCA development, which was more pronounced in patients with less advanced age.

A decision tree model for predicting intravenous immunoglobulin resistance and coronary artery involvement in Kawasaki disease

Objectives

This study aims to develop a new algorithm for predicting intravenous immunoglobulin (IVIG) resistance and coronary artery involvement in Kawasaki disease (KD) through decision tree models.

Methods

Medical records of children hospitalized for KD were analysed retrospectively. We compared the clinical characteristics, and the laboratory data in the groups with IVIG resistance and coronary artery dilatations (CADs) in KD patients. The decision tree models were developed to predict IVIG resistance and CADs.

Results

A total 896 patients (511 males and 385 females; 1 month-12 years) were eligible. IVIG resistance was identified in 111 (12.3%) patients, and CADs were found in 156 (17.4%). Total bilirubin and nitrogen terminal- pro-brain natriuretic peptide (NT-proBNP) were significantly higher in IVIG resistant group than in IVIG responsive group (0.62 ± 0.8 mg/dL vs 1.38 ± 1.4 mg/dL and 1231 ± 2136 pg/mL vs 2425 ± 4459 mL, respectively, P < 0.01). Also, CADs were more developed in the resistant group (39/111; 14.9% vs. 117/785; 35.1%, P < 0.01). The decision tree for predicting IVIG resistance was classified based on total bilirubin (0.7 mg/mL, 1.46 mg/dL) and NT-proBNP (1561 pg/mL), consisting of two layers and four nodes, with 86.2% training accuracy and 90.5% evaluation accuracy. The Receiver Operating Characteristic (ROC) evaluated the predictive ability of the decision tree, and the area under the curve (AUC) (0.834; 95% confidence interval, 0.675–0.973; P < 0.05) showed relatively higher accuracy. The group with CADs had significantly higher total bilirubin and NT-proBNP levels than the control group (0.64 ± 0.82 mg/dL vs 1.04 ± 1.14 mg/dL and 1192 ± 2049 pg/mL vs 2268 ± 4136 pg/mL, respectively, P < 0.01). The decision trees for predicting CADs were classified into two nodes based on NT-proBNP (789 pg/mL) alone, with 83.5% training accuracy and 90.3% evaluation accuracy.

Conclusion

A new algorithm decision tree model presents for predicting IVIG resistance and CADs in KD, confirming the usefulness of NT-proBNP as a predictor of KD.

The Ostomy Skin Tool 2.0: a new instrument for assessing peristomal skin changes

BACKGROUND

Peristomal skin complications (PSCs) are frequently reported postoperative complications. PSCs can present visibly or as symptoms such as pain, itching or burning sensations.

AIM

To develop a new tool that can capture a range of sensation symptoms together with visible complications and an objective assessment of discolouration in the peristomal area.

METHOD

Consensus from qualitative interviews with health professionals and people with an ostomy, and input from expert panels, formed the basis of a patient-reported outcome (PRO) questionnaire. A decision tree model was used to define a combined score including PRO and objectively assessed discolouration area.

FINDINGS

Six elements were included in the PRO questionnaire and four health states representing different severity levels of the peristomal skin were defined.

CONCLUSION

The Ostomy Skin Tool 2.0 is a sensitive tool that can be used to follow changes in the peristomal skin on a regular basis and thereby help prevent severe PSCs.

Fuzzy Decision Tree Based Method in Decision-Making of COVID-19 Patients’ Treatment

A new method in decision-making of timing of tracheostomy in COVID-19 patients is developed and discussed in this paper. Tracheostomy is performed in critically ill coronavirus disease (COVID-19) patients. The timing of tracheostomy is important for anticipated prolonged ventilatory wean when levels of respiratory support were favorable. The analysis of this timing has been implemented based on classification method. One of principal conditions for the developed classifiers in decision-making of timing of tracheostomy in COVID-19 patients was a good interpretation of result. Therefore, the proposed classifiers have been developed as decision tree based because these classifiers have very good interpretability of result. The possible uncertainty of initial data has been considered by the application of fuzzy classifiers. Two fuzzy classifiers as Fuzzy Decision Tree (FDT) and Fuzzy Random Forest (FRF) have been developed for the decision-making in tracheostomy timing. The evaluation of proposed classifiers and their comparison with other show the efficiency of the proposed classifiers. FDT has best characteristics in comparison with other classifiers.

Artificial Intelligence: A Shifting Paradigm in Cardio-Cerebrovascular Medicine

The future of healthcare is an organic blend of technology, innovation, and human connection. As artificial intelligence (AI) is gradually becoming a go-to technology in healthcare to improve efficiency and outcomes, we must understand our limitations. We should realize that our goal is not only to provide faster and more efficient care, but also to deliver an integrated solution to ensure that the care is fair and not biased to a group of sub-population. In this context, the field of cardio-cerebrovascular diseases, which encompasses a wide range of conditions-from heart failure to stroke-has made some advances to provide assistive tools to care providers. This article aimed to provide an overall thematic review of recent development focusing on various AI applications in cardio-cerebrovascular diseases to identify gaps and potential areas of improvement. If well designed, technological engines have the potential to improve healthcare access and equitability while reducing overall costs, diagnostic errors, and disparity in a system that affects patients and providers and strives for efficiency.

Data mining in clinical big data: the frequently used databases, steps, and methodological models

Many high quality studies have emerged from public databases, such as Surveillance, Epidemiology, and End Results (SEER), National Health and Nutrition Examination Survey (NHANES), The Cancer Genome Atlas (TCGA), and Medical Information Mart for Intensive Care (MIMIC); however, these data are often characterized by a high degree of dimensional heterogeneity, timeliness, scarcity, irregularity, and other characteristics, resulting in the value of these data not being fully utilized. Data-mining technology has been a frontier field in medical research, as it demonstrates excellent performance in evaluating patient risks and assisting clinical decision-making in building disease-prediction models. Therefore, data mining has unique advantages in clinical big-data research, especially in large-scale medical public databases. This article introduced the main medical public database and described the steps, tasks, and models of data mining in simple language. Additionally, we described data-mining methods along with their practical applications. The goal of this work was to aid clinical researchers in gaining a clear and intuitive understanding of the application of data-mining technology on clinical big-data in order to promote the production of research results that are beneficial to doctors and patients.

Application of a decision tree model in the early identification of severe patients with severe fever with thrombocytopenia syndrome

Background

Severe fever with thrombocytopenia syndrome (SFTS) is a serious infectious disease with a fatality of up to 30%. To identify the severity of SFTS precisely and quickly is important in clinical practice.

Methods

From June to July 2020, 71 patients admitted to the Infectious Department of Joint Logistics Support Force No. 990 Hospital were enrolled in this study. The most frequently observed symptoms and laboratory parameters on admission were collected by investigating patients’ electronic records. Decision trees were built to identify the severity of SFTS. Accuracy and Youden’s index were calculated to evaluate the identification capacity of the models.

Results

Clinical characteristics, including body temperature (p = 0.011), the size of the lymphadenectasis (p = 0.021), and cough (p = 0.017), and neurologic symptoms, including lassitude (p<0.001), limb tremor (p<0.001), hypersomnia (p = 0.009), coma (p = 0.018) and dysphoria (p = 0.008), were significantly different between the mild and severe groups. As for laboratory parameters, PLT (p = 0.006), AST (p<0.001), LDH (p<0.001), and CK (p = 0.003) were significantly different between the mild and severe groups of SFTS patients. A decision tree based on laboratory parameters and one based on demographic and clinical characteristics were built. Comparing with the decision tree based on demographic and clinical characteristics, the decision tree based on laboratory parameters had a stronger prediction capacity because of its higher accuracy and Youden’s index.

Conclusion

Decision trees can be applied to predict the severity of SFTS.

A predictive framework in healthcare: Case study on cardiac arrest prediction

Data-driven healthcare uses predictive analytics to enhance decision-making and personalized healthcare. Developing prognostic models is one of the applications of predictive analytics in medical environments. Various studies have used machine learning techniques for this purpose. However, there is no specific standard for choosing prediction models for different medical purposes. In this paper, the ISAF framework was proposed for choosing appropriate prediction models with regard to the properties of the classification methods. As one of the case study applications, a prognostic model for predicting cardiac arrests in sepsis patients was developed step by step through the ISAF framework. Finally, a new modified stacking model produced the best results. We predict 85 % of heart arrest cases one hour before the incidence (sensitivity> = 0.85) and 73 % of arrest cases 25 h before the occurrence (sensitivity> = 0.73). The results indicated that the proposed prognostic model has significantly improved the prediction results compared to the two standard systems of APACHE II and MEWS. Furthermore, compared to previous research, the proposed model has extended the prediction interval and improved the performance criteria.

Analysis of hospitalization costs related to fall injuries in elderly patients

BACKGROUND

With the aging world population, the incidence of falls has intensified and fall-related hospitalization costs are increasing. Falls are one type of event studied in the health economics of patient safety, and many developed countries have conducted such research on fall-related hospitalization costs. However, China, a developing country, still lacks large-scale studies in this area.

AIM

To investigate the factors related to the hospitalization costs of fall-related injuries in elderly inpatients and establish factor-based, cost-related groupings.

METHODS

A retrospective study was conducted. Patient information and cost data for elderly inpatients (age ≥ 60 years, n = 3362) who were hospitalized between 2016 and 2019 due to falls was collected from the medical record systems of two grade-A tertiary hospitals in China. Quantile regression (QR) analysis was used to identify the factors related to fall-related hospitalization costs. A decision tree model based on the chi-squared automatic interaction detector algorithm for hospitalization cost grouping was built by setting the factors in the regression results as separation nodes.

RESULTS

The total hospitalization cost of fall-related injuries in the included elderly patients was 180479203.03 RMB, and the reimbursement rate of medical benefit funds was 51.0% (92039709.52 RMB/180479203.03 RMB). The medical material costs were the highest component of the total hospitalization cost, followed (in order) by drug costs, test costs, treatment costs, integrated medical service costs and blood transfusion costs The QR results showed that patient age, gender, length of hospital stay, payment method, wound position, wound type, operation times and operation type significantly influenced the inpatient cost (P < 0.05). The cost grouping model was established based on the QR results, and age, length of stay, operation type, wound position and wound type were the most important influencing factors in the model. Furthermore, the cost of each combination varied significantly.

CONCLUSION

Our grouping model of hospitalization costs clearly reflected the key factors affecting hospitalization costs and can be used to strengthen the reasonable control of these costs.

Influence of Lifestyles on Mild Cognitive Impairment: A Decision Tree Model Study

Objective

To explore the effects of different lifestyle choices on mild cognitive impairment (MCI) and to establish a decision tree model to analyse their predictive significance on the incidence of MCI.

Methods

Study participants were recruited from geriatric and physical examination centres from October 2015 to October 2019: 330 MCI patients and 295 normal cognitive (NC) patients. Cognitive function was evaluated by the Mini-Mental State Examination Scale (MMSE) and Clinical Dementia Scale (CDR), while the Barthel Index (BI) was used to evaluate life ability. Statistical analysis included the χ² test, logistic regression, and decision tree. The ROC curve was drawn to evaluate the predictive ability of the decision tree model.

Results

Logistic regression analysis showed that low education, living alone, smoking, and a high-fat diet were risk factors for MCI, while young age, tea drinking, afternoon naps, social engagement, and hobbies were protective factors for MCI. Social engagement, a high-fat diet, hobbies, living condition, tea drinking, and smoking entered all nodes of the decision tree model, with social engagement as the root node variable. The importance of predictive variables in the decision tree model showed social engagement, a high-fat diet, tea drinking, hobbies, living condition, and smoking as 33.57%, 27.74%, 22.14%, 11.94%, 4.61%, and 0%, respectively. The area under the ROC curve predicted by the decision tree model was 0.827 (95% CI: 0.795~0.856).

Conclusion

The decision tree model has good predictive ability. MCI was closely related to lifestyle; social engagement was the most important factor in predicting the occurrence of MCI.

Research on diagnosis-related group grouping of inpatient medical expenditure in colorectal cancer patients based on a decision tree model

BACKGROUND

In 2018, the diagnosis-related groups prospective payment system (DRGs-PPS) was introduced in a trial operation in Beijing according to the requirements of medical and health reform. The implementation of the system requires that more than 300 disease types pay through the DRGs-PPS for medical insurance. Colorectal cancer (CRC), as a common malignant tumor with high prevalence in recent years, was among the 300 disease types.

AIM

To investigate the composition and factors related to inpatient medical expenditure in CRC patients based on disease DRGs, and to provide a basis for the rational economic control of hospitalization expenses for the diagnosis and treatment of CRC.

METHODS

The basic material and cost data for 1026 CRC inpatients in a Grade-A tertiary hospital in Beijing during 2014-2018 were collected using the medical record system. A variance analysis of the composition of medical expenditure was carried out, and a multivariate linear regression model was used to select influencing factors with the greatest statistical significance. A decision tree model based on the exhaustive χ² automatic interaction detector (E-CHAID) algorithm for DRG grouping was built by setting chosen factors as separation nodes, and the payment standard of each diagnostic group and upper limit cost were calculated. The correctness and rationality of the data were re-evaluated and verified by clinical practice.

RESULTS

The average hospital stay of the 1026 CRC patients investigated was 18.5 d, and the average hospitalization cost was 57872.4 RMB yuan. Factors including age, gender, length of hospital stay, diagnosis and treatment, as well as clinical operations had significant influence on inpatient expenditure (P < 0.05). By adopting age, diagnosis, treatment, and surgery as the grouping nodes, a decision tree model based on the E-CHAID algorithm was established, and the CRC patients were divided into 12 DRG cost groups. Among these 12 groups, the number of patients aged ≤ 67 years, and underwent surgery and chemotherapy or radiotherapy was largest; while patients aged > 67 years, and underwent surgery and chemotherapy or radiotherapy had the highest medical cost. In addition, the standard cost and upper limit cost in the 12 groups were calculated and re-evaluated.

CONCLUSION

It is important to strengthen the control over the use of drugs and management of the hospitalization process, surgery, diagnosis and treatment to reduce the economic burden on patients. Tailored adjustments to medical payment standards should be made according to the characteristics and treatment of disease types to improve the comprehensiveness and practicability of the DRGs-PPS.

Decision tree model for predicting in-hospital cardiac arrest among patients admitted with acute coronary syndrome

BACKGROUND

In-hospital cardiac arrest (IHCA) may be preventable, with patients often showing signs of physiological deterioration before an event. Our objective was to develop and validate a simple clinical prediction model to identify the IHCA risk among cardiac arrest (CA) patients hospitalized with acute coronary syndrome (ACS).

HYPOTHESIS

A predicting model could help to identify the risk of IHCA among patients admitted with ACS.

METHODS

We conducted a case-control study and analyzed 21 337 adult ACS patients, of whom 164 had experienced CA. Vital signs, demographic, and laboratory data were extracted from the electronic health record. Decision tree analysis was applied with 10-fold cross-validation to predict the risk of IHCA.

RESULTS

The decision tree analysis detected seven explanatory variables, and the variables' importance is as follows: VitalPAC Early Warning Score (ViEWS), fatal arrhythmia, Killip class, cardiac troponin I, blood urea nitrogen, age, and diabetes. The development decision tree model demonstrated a sensitivity of 0.762, a specificity of 0.882, and an area under the receiver operating characteristic curve (AUC) of 0.844 (95% CI, 0.805 to 0.849). A 10-fold cross-validated risk estimate was 0.198, while the optimism-corrected AUC was 0.823 (95% CI, 0.786 to 0.860).

CONCLUSIONS

We have developed and internally validated a good discrimination decision tree model to predict the risk of IHCA. This simple prediction model may provide healthcare workers with a practical bedside tool and could positively impact decision-making with regard to deteriorating patients with ACS.