Lung cancer prediction using machine learning on data from a symptom e-questionnaire for never smokers, formers smokers and current smokers

Elinor Nemlander; Andreas Rosenblad; Eliya Abedi; Simon Ekman; Jan Hasselström; Lars E Eriksson; Axel C Carlsson

doi:10.1371/journal.pone.0276703

Lung cancer prediction using machine learning on data from a symptom e-questionnaire for never smokers, formers smokers and current smokers

PLoS One. 2022 Oct 21;17(10):e0276703. doi: 10.1371/journal.pone.0276703. eCollection 2022.

Authors

Elinor Nemlander^{1

2}, Andreas Rosenblad^{1

3

4}, Eliya Abedi^{1

2}, Simon Ekman⁵, Jan Hasselström^{1

2}, Lars E Eriksson^{6

7

8}, Axel C Carlsson^{1

2}

Affiliations

¹ Division of Family Medicine and Primary Care, Department of Neurobiology, Care Sciences and Society, Karolinska Institutet, Solna, Sweden.
² Academic Primary Health Care Centre, Region Stockholm, Stockholm, Sweden.
³ Regional Cancer Centre Stockholm-Gotland, Region Stockholm, Stockholm, Sweden.
⁴ Division of Clinical Diabetology and Metabolism, Department of Medical Sciences, Uppsala University, Uppsala, Sweden.
⁵ Thoracic Oncology Centre, Karolinska University Hospital, Dept of Oncology-Pathology, Karolinska Institutet, Stockholm, Sweden.
⁶ Division of Nursing, Department of Neurobiology, Care Sciences and Society, Karolinska Institutet, Huddinge, Sweden.
⁷ School of Health and Psychological Sciences, City, University of London, London, United Kingdom.
⁸ Medical Unit Infectious Diseases, Karolinska University Hospital, Huddinge, Sweden.

Abstract

Purpose: The aim of the present study was to investigate the predictive ability for lung cancer of symptoms reported in an adaptive e-questionnaire, separately for never smokers, former smokers, and current smokers.

Patients and methods: Consecutive patients referred for suspected lung cancer were recruited between September 2014 and November 2015 from the lung clinic at the Karolinska University Hospital, Stockholm, Sweden. A total of 504 patients were later diagnosed with lung cancer (n = 310) or no cancer (n = 194). All participants answered an adaptive e-questionnaire with a maximum of 342 items, covering background variables and symptoms/sensations suspected to be associated with lung cancer. Stochastic gradient boosting, stratified on smoking status, was used to train and test a model for predicting the presence of lung cancer.

Results: Among never smokers, 17 predictors contributed to predicting lung cancer with 82% of the patients being correctly classified, compared with 26 predictors with an accuracy of 77% among current smokers and 36 predictors with an accuracy of 63% among former smokers. Age, sex, and education level were the most important predictors in all models.

Conclusion: Methods or tools to assess the likelihood of lung cancer based on smoking status and to prioritize investigative and treatment measures among all patients seeking care with diffuse symptoms are much needed. Our study presents risk assessment models for patients with different smoking status that may be developed into clinical risk assessment tools that can help clinicians in assessing a patient's risk of having lung cancer.

Publication types

Research Support, Non-U.S. Gov't

MeSH terms

Humans
Lung Neoplasms* / diagnosis
Lung Neoplasms* / epidemiology
Lung Neoplasms* / etiology
Machine Learning
Smokers*
Smoking / adverse effects
Surveys and Questionnaires

Grants and funding

The present project was funded by Vetenskapsrådet (ref #2016–01712 and #2019–01222)Vårdalstiftelsen(ref #2014–0044), the Strategic Research Area Health Care Science (SFO-V, ref. #2–2764/2018 and 2020), Cancerföreningen i Stockholm (ref #191092), Sjöbergstiftelsen(ref #2022-01-11:7), Astra Zeneca (unrestricted grant), Zero vision cancer and Stiftelsen Einar Belvén. a) Please clarify the sources of funding (financial or material support) for your study. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. No salaries were paid by any of the funding organizations to any of the authors.