Risk Prediction of Emergency Department Visits in Patients With Lung Cancer Using Machine Learning: Retrospective Observational Study

Ensemble machine learning models for lung cancer incidence risk prediction in the elderly: a retrospective longitudinal study

BACKGROUND

Identifying high risk factors and predicting lung cancer incidence risk are essential to prevention and intervention of lung cancer for the elderly. We aim to develop lung cancer incidence risk prediction model in the elderly to facilitate early intervention and prevention of lung cancer.

METHODS

We stratified the population into six subgroups according to age and gender. For each subgroup, random forest, extreme gradient boosting, deep neural networks, support vector machine, multiple logistic regression and deep Q network (DQN) models were developed and validated. Models were trained and tested using samples from 2000 to 2015 and independent external validated through those from 2016 to 2019. The suitable model for lung cancer risk prediction and high risk factors identification was chosen based on internal validation and independent external validation.

RESULTS

The DQN model achieved the optimal prediction performance in stratified subgroups, with AUROC ranging from 0.937 to 0.953, recall ranging from 0.932 to 0.943, F2-score ranging from 0.929 to 0.946, precision ranging from 0.926 to 0.952, F1-score ranging from 0.933 to 0.963 and RMSE ranging from 0.21 to 0.27. SHAP values were supplied for model interpretability. High risk factors of lung cancer incidence were identified in the elderly. Men ≥ 65 carrying C > A/G > T mutation had the highest lung cancer incidence decrease of 39.5% after five years quitting in stratified elderly groups, which were 1.83 times more than women ≥ 65 not carrying C > A/G > T mutation.

CONCLUSIONS

The DQN model may be suitable for identifying high risk factors and predicting lung cancer risk with high performance. The proposed intervention and diagnosis pathways could be used for early screening and intervention before the occurrence of lung cancer, which could help oncologists develop targeted intervention strategies for the stratified elderly to reduce lung cancer incidence and improve therapeutic effect. Proposed method could also be used in predicting the risk of other chronic diseases to help conduct intervention and reduce incidence.

Adoption of the OMOP CDM for Cancer Research using Real-world Data: Current Status and Opportunities

Background

The Observational Medical Outcomes Partnership (OMOP) common data model (CDM) that is developed and maintained by the Observational Health Data Sciences and Informatics (OHDSI) community supports large scale cancer research by enabling distributed network analysis. As the number of studies using the OMOP CDM for cancer research increases, there is a growing need for an overview of the scope of cancer research that relies on the OMOP CDM ecosystem.

Objectives

In this study, we present a comprehensive review of the adoption of the OMOP CDM for cancer research and offer some insights on opportunities in leveraging the OMOP CDM ecosystem for advancing cancer research.

Materials and Methods

Published literature databases were searched to retrieve OMOP CDM and cancer-related English language articles published between January 2010 and December 2023. A charting form was developed for two main themes, i.e., clinically focused data analysis studies and infrastructure development studies in the cancer domain.

Results

In total, 50 unique articles were included, with 30 for the data analysis theme and 23 for the infrastructure theme, with 3 articles belonging to both themes. The topics covered by the existing body of research was depicted.

Conclusion

Through depicting the status quo of research efforts to improve or leverage the potential of the OMOP CDM ecosystem for advancing cancer research, we identify challenges and opportunities surrounding data analysis and infrastructure including data quality, advanced analytics methodology adoption, in-depth phenotypic data inclusion through NLP, and multisite evaluation.

Healthcare risk stratification model for emergency departments based on drugs, income and comorbidities: the DICER-score

BACKGROUND

During the last decade, the progressive increase in age and associated chronic comorbidities and polypharmacy. However, assessments of the risk of emergency department (ED) revisiting published to date often neglect patients' pharmacotherapy plans, thus overseeing the Drug-related problems (DRP) risks associated with the therapy burden. The aim of this study is to develop a predictive model for ED revisit, hospital admission, and mortality based on patient's characteristics and pharmacotherapy.

METHODS

Retrospective cohort study including adult patients visited in the ED (triage 1, 2, or 3) of multiple hospitals in Catalonia (Spain) during 2019. The primary endpoint was a composite of ED visits, hospital admission, or mortality 30 days after ED discharge. The study population was randomly split into a model development (60%) and validation (40%) datasets. The model included age, sex, income level, comorbidity burden, measured with the Adjusted Morbidity Groups (GMA), and number of medications. Forty-four medication groups, associated with medication-related health problems, were assessed using ATC codes. To assess the performance of the different variables, logistic regression was used to build multivariate models for ED revisits. The models were created using a "stepwise-forward" approach based on the Bayesian Information Criterion (BIC). Area under the curve of the receiving operating characteristics (AUCROC) curve for the primary endpoint was calculated.

RESULTS

851.649 patients were included; 134.560 (15.8%) revisited the ED within 30 days from discharge, 15.2% were hospitalized and 9.1% died within 30 days from discharge. Four factors (sex, age, GMA, and income level) and 30 ATC groups were identified as risk factors and combined into a final score. The model showed an AUCROC values of 0.720 (95%CI:0.718-0.721) in the development cohort and 0.719 (95%CI.0.717-0.721) in the validation cohort. Three risk categories were generated, with the following scores and estimated risks: low risk: 18.3%; intermediate risk: 40.0%; and high risk: 62.6%.

CONCLUSION

The DICER score allows identifying patients at high risk for ED revisit within 30 days based on sociodemographic, clinical, and pharmacotherapeutic characteristics, being a valuable tool to prioritize interventions on discharge.