Risk prediction and modification in a clinical and database population of people with chronic obstructive pulmonary disease

COPD is a leading cause of hospital admission and healthcare utilisation in the UK. The DOSE score can be used to risk stratify COPD patients based on clinical components routinely measured in Primary Care. Resource is often focussed on patients with high symptom burden, presenting repeatedly to medical professionals. We hypothesised that a database approach could be used to identify lower risk COPD patients by DOSE score and a subgroup at higher risk of clinical deterioration could be further identified using their documented clinical characteristics. We hypothesised that early specialist clinical review of COPD patients would reduce their subsequent symptom burden and future health care utilisation. In a real-world COPD database cohort of 13,608 Primary Care records, we risk stratified patients over four years by DOSE score. We showed clinical characteristics could identify a subgroup at increased risk of poor health outcomes with recent pneumonia and a raised eosinophil count showing the strongest individual associations. Logistic regression modelling determined the combination of characteristics most strongly associated with poor health outcomes. In a feasibility study, 120 patients deemed low risk by DOSE score were identified by Primary Care electronic record search. 76 patients were randomised to an intervention of respiratory specialist review in their Primary Care Practice. The findings suggested that proactive speciality review in a modified form warrants further research with modification of endpoint measures and further development of the recruitment process to avoid recruitment bias. In conclusion, we established the DOSE score can be administrated within a database approach and clinical characteristics can be used to further risk stratify COPD patients within a database. Pro-active specialist review of patients low risk by DOSE score is feasible for further study with some modification.