A Decision Support Tool for Predicting Patients at Risk of Readmission: A Comparison of Classification Trees, Logistic Regression, Generalized Additive Models, and Multivariate Adaptive Regression Splines

Using machine learning methods to predict all-cause somatic hospitalizations in adults: A systematic review

AIM

In this review, we investigated how Machine Learning (ML) was utilized to predict all-cause somatic hospital admissions and readmissions in adults.

METHODS

We searched eight databases (PubMed, Embase, Web of Science, CINAHL, ProQuest, OpenGrey, WorldCat, and MedNar) from their inception date to October 2023, and included records that predicted all-cause somatic hospital admissions and readmissions of adults using ML methodology. We used the CHARMS checklist for data extraction, PROBAST for bias and applicability assessment, and TRIPOD for reporting quality.

RESULTS

We screened 7,543 studies of which 163 full-text records were read and 116 met the review inclusion criteria. Among these, 45 predicted admission, 70 predicted readmission, and one study predicted both. There was a substantial variety in the types of datasets, algorithms, features, data preprocessing steps, evaluation, and validation methods. The most used types of features were demographics, diagnoses, vital signs, and laboratory tests. Area Under the ROC curve (AUC) was the most used evaluation metric. Models trained using boosting tree-based algorithms often performed better compared to others. ML algorithms commonly outperformed traditional regression techniques. Sixteen studies used Natural language processing (NLP) of clinical notes for prediction, all studies yielded good results. The overall adherence to reporting quality was poor in the review studies. Only five percent of models were implemented in clinical practice. The most frequently inadequately addressed methodological aspects were: providing model interpretations on the individual patient level, full code availability, performing external validation, calibrating models, and handling class imbalance.

CONCLUSION

This review has identified considerable concerns regarding methodological issues and reporting quality in studies investigating ML to predict hospitalizations. To ensure the acceptability of these models in clinical settings, it is crucial to improve the quality of future studies.

On the importance of interpretable machine learning predictions to inform clinical decision making in oncology

Machine learning-based tools are capable of guiding individualized clinical management and decision-making by providing predictions of a patient's future health state. Through their ability to model complex nonlinear relationships, ML algorithms can often outperform traditional statistical prediction approaches, but the use of nonlinear functions can mean that ML techniques may also be less interpretable than traditional statistical methodologies. While there are benefits of intrinsic interpretability, many model-agnostic approaches now exist and can provide insight into the way in which ML systems make decisions. In this paper, we describe how different algorithms can be interpreted and introduce some techniques for interpreting complex nonlinear algorithms.

Sustainable Intensification Practices Reduce Food Deficit for the Best- and Worst-Off Households in Ethiopia and Mozambique

An adequate food supply is widely recognized as a necessary condition for social development as well as a basic human right. Food deficits are especially common among semi-subsistence farming households in eastern and southern Africa and farm productivity is widely regarded as the locus for enhancing household food outcomes. However, knowledge gaps surrounding benefits associated with climate smart, productivity-enhancing technologies require attention. This study evaluates benefits associated with sustainable intensification farm management practices (crop residue retention, minimum tillage, manure application and use of herbicides, pesticides, fertilizer, and improved seeds) for household calorie and protein supplies and demonstrates their scope across households with high-, moderate- and low- likelihoods of calorie and protein deficits. Household-level calorie and protein deficits were estimated from survey data on food production, acquisition and consumption for households in Ethiopia and Mozambique. Multinomial logistic models were used to identify drivers of household food deficit status and logistic model trees established “rules of thumb” to classify households by food deficit status as low, moderate or high likelihood. In Ethiopia, especially wet seasons were associated with a high likelihood of a food deficit while especially dry seasons were associated with a high likelihood of food deficit in Mozambique. The practices associated with sustainable intensification and related technologies substantially enhanced food outcomes in groups with a high- and a low-likelihood of food deficit, and associated benefits were high for the best-off households. Benefits associated with sustainable intensification technologies were not observed for households with a moderate likelihood of a food deficit and some technologies even increased risk. The sustainable intensification practices assessed here were associated with improved food outcomes yet benefits were limited in scope for households of intermediate status. Thus, there is a need to expand the technical options available to reduce food deficit.

Prediction of Chronic Obstructive Pulmonary Disease Stages Using Machine Learning Algorithms

Identifying chronic obstructive pulmonary disease (COPD) severity stages is of great importance to control the related mortality rates and reduce the associated costs. This study aims to build prediction models for COPD stages and, to compare the relative performance of five machine learning algorithms to determine the optimal prediction algorithm. This research is based on data collected from a private hospital in Egypt for the two calendar years 2018 and 2019. Five machine learning algorithms were used for the comparison. The F1 score, specificity, sensitivity, accuracy, positive predictive value and negative predictive value were the performance measures used for algorithms comparison. Analysis included 211 patients’ records. Our results show that the best performing algorithm in most of the disease stages is the PNN with the optimal prediction accuracy and hence it can be considered as a powerful prediction tool used by decision makers in predicting severity stages of COPD.

Operational Modeling with Health Economics to Support Decision Making for COPD Patients

OBJECTIVE

To assess the impact of interventions for improving the management of chronic obstructive pulmonary disease (COPD), specifically increased use of pulmonary rehabilitation (PR) on patient outcomes and cost-benefit analysis.

DATA SOURCES

We used the national Hospital Episode Statistics (HES) datasets in England, local data and experts from the hospital setting, National Prices and National Tariffs, reports and the literature around the effectiveness of PR programs.

STUDY DESIGN

The COPD pathway was modeled using discrete event simulation (DES) to capture the patient pathway to an adequate level of detail as well as randomness in the real world. DES was further enhanced by the integration of a health economic model to calculate the net benefit and cost of treating COPD patients based on key sets of interventions.

DATA COLLECTION/EXTRACTION METHODS

A total of 150 input parameters and 75 distributions were established to power the model using the HES dataset, outpatient activity data from the hospital and community services, and the literature.

PRINCIPAL FINDINGS

The simulation model showed that increasing referral to PR (by 10%, 20%, or 30%) would be cost-effective (with a benefit-cost ratio of 5.81, 5.95, and 5.91, respectively) by having a positive impact on patient outcomes and operational metrics. Number of deaths, admissions, and bed days decreased (ie, by 3.56 patients, 4.90 admissions, and 137.31 bed days for a 30% increase in PR referrals) as well as quality of life increased (ie, by 5.53 QALY among 1540 patients for the 30% increase).

CONCLUSIONS

No operational model, either statistical or simulation, has previously been developed to capture the COPD patient pathway within a hospital setting. To date, no model has investigated the impact of PR on COPD services, such as operations, key performance, patient outcomes, and cost-benefit analysis. The study will support policies around extending availability of PR as a major intervention.

A predictive analytics framework for identifying patients at risk of developing multiple medical complications caused by chronic diseases

Chronic diseases often cause several medical complications. This paper aims to predict multiple complications among patients with a chronic disease. The literature uses single-task learning algorithms to predict complications independently and assumes no correlation among complications of chronic diseases. We propose two methods (independent prediction of complications with single-task learning and concurrent prediction of complications with multi-task learning) and show that medical complications of chronic diseases can be correlated. We use a case study and compare the performance of these two methods by predicting complications of hypertrophic cardiomyopathy on 106 predictors in 1078 electronic medical records from April 2009-April 2017, inclusive. The methods are implemented using logistic regression, artificial neural networks, decision trees, and support vector machines. The results show multi-task learning with logistic regression improves the performance of predictions in terms of both discrimination and calibration.

Using machine-learning methods to support health-care professionals in making admission decisions

BACKGROUND

Large tertiary hospitals usually face long waiting lines; patients who want to receive hospitalization need to be screened in advance. The patient admission screening process involves a health-care professional ranking patients by analyzing registration information.

OBJECTIVE

The purpose of this study was to develop a machine-learning approach to screening, using historical data and the experience of health-care professionals to develop a set of screening rules to help health-care professionals prioritize patient needs automatically.

METHODS

We used five machine-learning methods to sequence and predict elective patients: logistic regression (LR), random forest (RF), gradient-boosting decision tree (GBDT), extreme gradient boosting (XGBoost), and an ensemble model of the four models.

RESULTS

The results indicate that all of the five models showed a good prioritization performance with high predictive values. In particular, XGBoost had the best predictive performance compared with others in terms of the area under the receiver operating characteristic curve (AUC), with the AUC values of LR, RF, GBDT, XGBoost, and the ensemble model being 0.881, 0.816, 0.820, 0.901, and 0.897, respectively.

CONCLUSION

The results reported here indicate that machine-learning techniques can be valuable for automating the screening process. Our model can assist health-care professionals in automatically evaluating less complex cases by identifying important factors affecting patient admission.

A novel risk classification system for 30-day mortality in children undergoing surgery

A simple, objective and accurate way of grouping children undergoing surgery into clinically relevant risk groups is needed. The purpose of this study, is to develop and validate a preoperative risk classification system for postsurgical 30-day mortality for children undergoing a wide variety of operations. The National Surgical Quality Improvement Project-Pediatric participant use file data for calendar years 2012-2014 was analyzed to determine preoperative variables most associated with death within 30 days of operation (D30). Risk groups were created using classification tree analysis based on these preoperative variables. The resulting risk groups were validated using 2015 data, and applied to neonates and higher risk CPT codes to determine validity in high-risk subpopulations. A five-level risk classification was found to be most accurate. The preoperative need for ventilation, oxygen support, inotropic support, sepsis, the need for emergent surgery and a do not resuscitate order defined non-overlapping groups with observed rates of D30 that vary from 0.075% (Very Low Risk) to 38.6% (Very High Risk). When CPT codes where death was never observed are eliminated or when the system is applied to neonates, the groupings remained predictive of death in an ordinal manner.

Higher Quality and Lower Cost from Improving Hospital Discharge Decision Making

This paper reports research on improving decisions about hospital discharges - decisions that are now made by physicians based on mainly subjective evaluations of patients' discharge status. We report an experiment on uptake of our clinical decision support software (CDSS) which presents physicians with evidence-based discharge criteria that can be effectively utilized at the point of care where the discharge decision is made. One experimental treatment we report prompts physician attentiveness to the CDSS by replacing the default option of universal "opt in" to patient discharge with the alternative default option of "opt out" from the CDSS recommendations to discharge or not to discharge the patient on each day of hospital stay. We also report results from experimental treatments that implement the CDSS under varying conditions of time pressure on the subjects. The experiment was conducted using resident physicians and fourth-year medical students at a university medical school as subjects.