EUVIMEDEUVIMED entdecken
Multizentrische StudieBeobachtungsstudie

Filling the Gaps in Health Data: Using a Machine Learning Approach to Augment Partially Observed Variables Such as Smoking in Claims Data

Stefan Franzen iD, Evangelos Chandakas, Sam Hillman, Kirsty Rhodes, Clementine Nordon

Pharmacoepidemiology and Drug Safety · 2026

DOI: 10.1002/pds.70322 ↗

Kurz gefasst

Multizentrische Studie in der Zeitschrift Pharmacoepidemiology and Drug Safety, 2026. Verfasst von Stefan Franzen und 4 weiteren Autor:innen. Beteiligte Einrichtung: BPM Evidence Statistics, AstraZeneca, Gothenburg, Sweden. Hauptthemen laut MEDLINE: Insurance Claim Review, Machine Learning, Smoking. Die Arbeit nennt 33 Literaturangaben. Gefördert von: AstraZeneca.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Purpose

Missing information is common in real‐world claims data, particularly on behavioral confounders, for example, smoking. Often one category of the variable, “yes” is partially observed while the other “no” remains completely missing—a pattern we call missing with truncation. A common way to handle these missing values is to naïvely treat missing values as absence of the risk factor, which may lead to substantial misclassification. Standard multiple imputation is impossible as only one level of the variable is observed.

Methods

A case study was conducted using data from the NOVELTY study, including 12 224 people with physician diagnosed asthma and/or COPD (NCT02760329). From this cohort, 9733 patients with complete information were included. This dataset was split into two where the first part was used to train an imputation model and the second part was used to evaluate the imputations based on the model (1) when used to impute a truncated and amputated smoking variable against the naïvely classifying missing as “no” (2) when varying the percent smokers retained, q .

Results

The accuracy of approaches (1) and (2) was 0.79 and 0.43, respectively; for q = 90%, the accuracy of approaches (1) and (2) was 0.89 and 0.94, respectively. Transfer learning showed better accuracy than the naïve approach when the percentage of true smokers being recorded as smokers was < 80%.

Conclusions

The added value of transfer learning was greatest when low proportions of true ever‐smokers were recorded, with its advantage depending on both the true prevalence of true smokers and the predictive model's performance.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Insurance Claim ReviewMachine LearningSmokingAdultAgedFemaleHumansMaleMiddle AgedAsthmaCohort StudiesDatabases, FactualPulmonary Disease, Chronic Obstructive

Stichwörter der Autor:innen

claims dataimputationmachine learningmissing datasecondary datatransfer learning

Autor:innen und Institutionen

  • BPM Evidence Statistics, AstraZeneca, Gothenburg, Sweden.
  • Oncology Data Analytics, AstraZeneca, Cambridge, UK.
  • Clinical Data Science, AstraZeneca, Barcelona, Spain.
  • BPM Evidence Statistics, AstraZeneca, Cambridge, UK.
  • Oncology Outcomes Research, AstraZeneca, Cambridge, UK.

ORCID-Profile: Stefan Franzen

Förderung

  • AstraZeneca

Zitieren

APA

Franzen, S., Chandakas, E., Hillman, S., Rhodes, K., & Nordon, C. (2026). Filling the Gaps in Health Data: Using a Machine Learning Approach to Augment Partially Observed Variables Such as Smoking in Claims Data. Pharmacoepidemiology and Drug Safety. https://doi.org/10.1002/pds.70322

Vancouver

Franzen S, Chandakas E, Hillman S, Rhodes K, Nordon C. Filling the Gaps in Health Data: Using a Machine Learning Approach to Augment Partially Observed Variables Such as Smoking in Claims Data. Pharmacoepidemiology and Drug Safety. 2026. doi:10.1002/pds.70322

BibTeX
@article{Franzen2026,
  author = {Franzen, Stefan and Chandakas, Evangelos and Hillman, Sam and Rhodes, Kirsty and Nordon, Clementine},
  title = {Filling the Gaps in Health Data: Using a Machine Learning Approach to Augment Partially Observed Variables Such as Smoking in Claims Data},
  journal = {Pharmacoepidemiology and Drug Safety},
  year = {2026},
  doi = {10.1002/pds.70322}
}
Kostenlos starten →