A Bayesian Network Approach to Lung Cancer Screening: Assessing the Impact of Data Quantity, Quality, and the Combination of Data from Danish Electronic Health Records
Authors:
- Florian van DaalenFDFlorian van Daalen
- Margrethe Høstgaard Bang HenriksenMHMargrethe Høstgaard Bang Henriksen
- Torben Frøstrup HansenTorben Frøstrup Hansen
- Lars Henrik JensenLJLars Henrik Jensen
- Claus Lohman BrasenCBClaus Lohman Brasen
- Ole HilbergOHOle Hilberg
- Martin Ask Klausholt AndersenMAMartin Ask Klausholt Andersen
- Elise HumerfeltEHElise Humerfelt
- Leonard WeeLWLeonard Wee
- Inigo BermejoIBInigo Bermejo
Open Access
Abstract
Background/Objectives: Lung cancer (LC) is the leading cause of cancer mortality, making early diagnosis essential. While LC screening trials are underway globally, optimal prediction models and inclusion criteria are still lacking. This study aimed to develop and evaluate Bayesian Network (BN) models for LC risk prediction using a decade of data from Denmark. The primary goal was to assess BN performance on datasets varying in size and completeness, simulate real-world screening scenarios, and identify the most valuable data sources for LC screening. Methods: The study included 38,944 patients evaluated for LC, with 11,284 (29%) diagnosed. Data on comorbidities, medications, and general practice were available for the entire cohort, while laboratory results, smoking habits, and other variables were only available for subsets. The cohort was divided into four subsets based on data availability, and BNs were trained and validated across these subsets using cross-validation and external validation. To determine the optimal combination of variables, all possible data combinations were evaluated on the samples that contained all the variables (n = 5587). Results: A model trained on the small, complete dataset (AUC 0.78) performed similarly on a larger dataset with 21% missing data (AUC 0.78). Performance dropped when 39% of data were missing (AUC 0.67), resulting in informative variables missing completely in the dataset. Laboratory results and smoking data were the most informative, significantly outperforming models based only on age and smoking status (AUC 0.70). Conclusions: BN models demonstrated moderate to strong predictive performance, even with incomplete data, highlighting the potential value of incorporating laboratory results in LC screening programs.Keywords
Funding Information
- The Dagmar Marshall Foundation (R198-A14299)
- The Region of Southern Denmark (R198-A14299)
- The University of Southern Denmark (R198-A14299)
- Lilly and Herbert Hansen Foundation (R198-A14299)
- Hede Nielsen Family Foundation (R198-A14299)
- The Beckett Foundation (R198-A14299)
- The Danish National Research Center for Lung Cancer, Danish Cancer Society (R198-A14299)
This publication has 37 references indexed in Scilit:
- P1.14-04 European Lung Cancer Screening Implementation: 4-IN-THE-LUNG-RUN trialJournal of Thoracic Oncology, 2023
- Lung cancer mortality reduction by LDCT screening: UKLS randomised trial results and international meta-analysisThe Lancet Regional Health - Europe, 2021
- Comparative performance of lung cancer risk models to define lung screening eligibility in the United KingdomBritish Journal of Cancer, 2021
- Bayesian networks in healthcare: Distribution by medical conditionArtificial Intelligence in Medicine, 2020
- Do Current Lung Cancer Screening Guidelines Apply for Populations With High Prevalence of Granulomatous Disease? Results From the First Brazilian Lung Cancer Screening Trial (BRELT1)The Annals of Thoracic Surgery, 2016
- Exploiting missing clinical data in Bayesian network modeling for predicting medical problemsJournal of Biomedical Informatics, 2008
- Baseline results of the Depiscan study: A French randomized pilot trial of lung cancer screening comparing low dose CT scan (LDCT) and chest X-ray (CXR)Lung Cancer, 2007
- Lung Cancer Screening Using Multi-Slice Thin-Section Computed Tomography and Autofluorescence BronchoscopyJournal of Thoracic Oncology, 2006
- Bayesian Network ClassifiersMachine Learning, 1997
- A Bayesian method for the induction of probabilistic networks from dataMachine Learning, 1992