Heart failure and current smoking were the strongest risk factors for predicting ten-year mortality in COPD patients, a risk prediction model based on primary care data has found.

Data on 37,485 patients with COPD in the Clinical Practice Research Database were analysed. Patients were followed up for ten years. The main outcome was all-cause mortality.

Four types of predictors were examined, demographic and behavioural characteristics (e.g. exercise, smoking status, limitation of physical activity at diagnosis), comorbidities, and COPD severity. Because of the lack of spirometry data COPD severity was based on breathlessness, a history of COPD exacerbations and hospitalisation due to COPD.

The investigators developed and validated a machine learning-based prediction model of ten-year COPD mortality to help risk stratify patients.

Just over half the study population were male (53.1%), 86.2% were 56 years or older. Overall, 14.7% were never smokers, 63.3% former smokers and 22% current smokers. Mean age at COPD diagnosis was 68.

Approximately, a quarter (23.6%) of patients died within ten years of COPD diagnosis, 57.8% were men.

The mMRC dyspnoea score in the 12 months leading up to diagnosis was recorded in only 61.7% of participants but when it was, there was a strong correlation with the proportion of patients who died: 6.3% when the mMRC dyspnoea score was 0 or 1, 8.5% when the score was 2, and 12.5% when the score was 3 or 4.

In the 70% of patients for whom comorbidities were recorded, asthma, hypertension, and depression were the three most common.

The two risk factors which had the greatest effect on mortality were:

• Heart failure HR 1.92 (95% CI: 1.87-1.96)

• Current smoker HR 1.68 (95% CI: 1.66-1.71)

The other prognostic factors with large effect sizes that were also significant (>20% increased risk of mortality) were:

• Ischaemic heart disease HR 1.49 (95% CI: 1.45-1.54)

• Former smoker HR 1.49 (95% CI: 1.46-1.51)

• Male gender HR 1.26 (95% CI: 1.25-1.27)

• mMRC dyspnoea score ≥ 3 HR 1.33 (95% CI: 1.28-1.39)

• Breathlessness HR 1.25 (95% CI: 1.24-1.26)

• Physical activity limitation HR 1.37 (95% CI: 1.33-1.42)

 Hypertension, anxiety, a history of pulmonary TB, and depression had more modest effects on mortality.

Although a large and robust study the authors acknowledge that smoking data is often prone to error. They also acknowledge that as primary care data alone were used, hospitalisations may have been missed.

The study authors state: ‘While the model developed achieved strong internal validation performance, it is now important to externally validate this model before deploying it in clinical practice.’

The study emphasises that COPD patients with heart failure should be a priority for optimisation of control of both conditions. It offers further evidence for the importance of smoking cessation management in COPD patients and also identifies other adverse prognostic factors that may require action.

Dr Peter Saul