Advertisement

Identification of Patient Prescribing Predicting Cancer Diagnosis Using Boosted Decision Trees

  • Josephine FrenchEmail author
  • Cong Chen
  • Katherine Henson
  • Brian Shand
  • Patrick Ferris
  • Josh Pencheon
  • Sally Vernon
  • Meena Rafiq
  • David Howe
  • Georgios Lyratzopoulos
  • Jem Rashbass
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 11526)

Abstract

Machine learning has potential to identify patterns in pre-diagnostic prescribing that act as an early signal of cancer diagnosis. Danish studies using classical regression models have shown that prescribing of particular drugs increases in the months prior to lung and colorectal cancer diagnosis. The aim of this case-control study is to assess the potential for machine learning to extend these findings to identify combinations of prescriptions that might act as pre-cancer signals. We use a boosted trees approach to analyse prescriptions data from NHS Business Services Authority linked to English cancer registry data to classify individuals into two classes: cancer patients and controls. We then identify the drugs that contributed the most to the classification decisions in the models. To the best of our knowledge, this is the first study utilising machine learning to find pre-diagnostic primary-care-prescription-related indicators of cancer diagnosis in England. We assess two feature selection approaches using text categorisation methods alone and in combination with clinical domain knowledge. Matched samples of controls (ten controls for each patient) to control for age are used throughout. We train models for matched cohorts of 6,770 lung cancer patients and 5,869 colorectal cancer patients starting the cancer pathway for the first time between January and March 2016. The models outperform classical methods by AUC, AUC-PR, and F\(_{0.5}\) score, showing strong potential for using machine learning to extract signals from this dataset to aid earlier diagnosis. Our findings confirm the Danish studies.

Keywords

Cancer Boosted trees Feature selection Clinical input 

References

  1. 1.
    World Health Organisation Cancer Fact Sheet. https://www.who.int/news-room/fact-sheets/detail/cancer. Accessed 3 Jan 2019
  2. 2.
    Pottegård, A., Hallas, J.: New use of prescription drugs prior to a cancer diagnosis. Pharmacoepidemiol. Drug Saf. 26, 223–227 (2017)CrossRefGoogle Scholar
  3. 3.
    Guldbrandt, L.M., Møller, H., Jakobsen, E., Vedsted, P.: General practice consultations, diagnostic investigations, and prescriptions in the year preceding a lung cancer diagnosis. Cancer Med. 6, 79–88 (2017)CrossRefGoogle Scholar
  4. 4.
    Hansen, P.L., Hjertholm, P., Vedsted, P.: Increased diagnostic activity in general practice during the year preceding colorectal cancer diagnosis. Int. J. Cancer 137, 615–624 (2015)CrossRefGoogle Scholar
  5. 5.
    Henson, K., et al.: Cohort profile: prescriptions dispensed in the community linked to the national cancer registry in England. BMJ Open 8, e020980 (2018)CrossRefGoogle Scholar
  6. 6.
    Joint Formulary Committee: British National Formulary Vol 68 and 72. BMJ Group and Pharmaceutical Press, London (2014, 2016)Google Scholar
  7. 7.
    Largeron, C., Moulin, C., Géry, M.: Entropy based feature selection for text categorization. In: ACM Symposium on Applied Computing TaiChung, pp. 924–928 (2011)Google Scholar
  8. 8.
    Chen, T., Guestrin, C.: XGBoost: a scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, San Francisco, pp. 785–794 (2016)Google Scholar

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  • Josephine French
    • 1
    • 2
    Email author
  • Cong Chen
    • 1
    • 2
  • Katherine Henson
    • 2
  • Brian Shand
    • 1
    • 2
  • Patrick Ferris
    • 1
    • 2
  • Josh Pencheon
    • 2
  • Sally Vernon
    • 2
  • Meena Rafiq
    • 4
  • David Howe
    • 1
    • 2
  • Georgios Lyratzopoulos
    • 2
    • 3
  • Jem Rashbass
    • 1
    • 2
  1. 1.Health Data Insight CICCambridgeUK
  2. 2.National Cancer Registration and Analysis Service, Public Health EnglandLondonUK
  3. 3.ECHO (Epidemiology of Cancer Healthcare and Outcomes) Group, Department of Behavioural Science and HealthUniversity College LondonLondonUK
  4. 4.University College London Institute of Health InformaticsLondonUK

Personalised recommendations