Cardiopulmonary risk model for patients with chronic obstructive pulmonary disease

Study type

• Study type:
  Non-interventional study
• Scope of the study:
  Assessment of risk minimisation measure implementation or effectiveness

  Disease epidemiology

Non-interventional study

• Non-interventional study design:
  Cohort

Study drug and medical condition

• Medical condition to be studied:
  Chronic obstructive pulmonary disease

  Myocardial infarction

  Cardiac failure

Population studied

• Age groups:
  Adults (18 to < 46 years)

  Adults (46 to < 65 years)

  Adults (65 to < 75 years)

  Adults (75 to < 85 years)

  Adults (85 years and over)
• Estimated number of subjects:
  131581

Study design details

• Main study objective:
  To construct a risk model that predicts major cardiopulmonary events (MACRE) in patients with chronic obstructive pulmonary disease (COPD)
• Outcomes:
  The outcome will be MACRE, which is a composite of any of the following: • First heart failure hospitalization after the index date, • First diagnosis or hospitalization for myocardial infarction, or occurrence of coronary revascularization after the index date, • First severe COPD exacerbation, defined by COPD-related hospitalizations, after the index date, and • Non-cancer-related mortality.
• Data analysis plan:
  A predictive model will be constructed for the outcome (MACRE) using Fine-Gray competing risk regression, with cancer-related mortality as the only competing event. The hold-out method will be used, with a randomly selected 80% of the cohort being the development dataset and the remainder 20% being the validation dataset. Elastic net will be used for variable selection and regularization. Discrimination of the resultant model will be evaluated within the derivation cohort using Harrell’s C-index. Calibration will be assessed quantitatively using the Greenwood-Nam-D'Agostino test and/or the Brier score, as well qualitatively as using an overlaid plot of the predicted and observed (Kaplan-Meier) freedom from the composite outcome for each quintile as grouped by the predicted risk. Prognostic value of the predicted risks generated by the resultant model will be additionally evaluated by means of survival analysis.