External validation of a COPD prediction model using population-based primary care data: a nested case- control study

Bright I. Nwaru, Colin R. Simpson, Aziz Sheikh, Daniel Kotz*

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

Emerging models for predicting risk of chronic obstructive pulmonary disease (COPD) require external validation in order to assess their clinical value. We validated a previous model for predicting new onset COPD in a different database. We randomly drew 38,597 case-control pairs (total N = 77,194) of individuals aged >= 35 years and matched for sex, age, and general practice from the United Kingdom Clinical Practice Research Datalink database. We assessed accuracy of the model to discriminate between COPD cases and non-cases by calculating area under the receiver operator characteristic (ROCAUC) for the prediction scores. Analogous to the development model, ever smoking (OR 6.70; 95% CI 6.41-6.99), prior asthma (OR 6.43; 95% CI 5.85-7.07), and higher socioeconomic deprivation (OR 2.90; 95% CI 2.72-3.09 for highest vs. lowest quintile) increased the risk of COPD. The validated prediction scores ranged from 0-5.71 (ROCAUC 0.66; 95% CI 0.65-0.66) for males and 0-5.95 (ROCAUC 0.71; 95% CI 0.70-0.71) for females. We have confirmed that smoking, prior asthma, and socioeconomic deprivation are key risk factors for new onset COPD. Our model seems externally valid at identifying patients at risk of developing COPD. An impact assessment now needs to be undertaken to assess whether this prediction model can be applied in clinical care settings.

Original languageEnglish
Article number44702
Number of pages7
JournalScientific Reports
Volume7
DOIs
Publication statusPublished - 17 Mar 2017

Keywords

  • OBSTRUCTIVE PULMONARY-DISEASE
  • PROGNOSTIC MODEL
  • RISK
  • ALGORITHM
  • MORTALITY

Fingerprint

Dive into the research topics of 'External validation of a COPD prediction model using population-based primary care data: a nested case- control study'. Together they form a unique fingerprint.

Cite this