Observational Cost-Effectiveness Analysis Using Routine Data: Admission and Discharge Care Bundles for Patients with Chronic Obstructive Pulmonary Disease.

BACKGROUND

Chronic obstructive pulmonary disease (COPD) is a prevalent respiratory disease, and accounts for a substantial proportion of unplanned hospital admissions. Care bundles for COPD are a set of standardised, evidence-based interventions that may improve outcomes in hospitalised COPD patients. We estimated the cost effectiveness of care bundles for acute exacerbations of COPD using routinely collected observational data.

METHODS

Data were collected from implementation (n = 7) and comparator (n = 7) acute hospitals located in England and Wales. We conducted a difference-in-difference cost-effectiveness analysis using a secondary care (i.e. hospital) perspective to examine the effect on National Health Service (NHS) costs and 90-day mortality of implementing care bundles compared with usual care for patients admitted to hospital with an acute exacerbation of COPD. Adjusted models included as covariates patient age, sex, deprivation, ethnicity and seasonal effects and mixed effects for site.

RESULTS

Outcomes and baseline characteristics of up to 12,532 patients were analysed using both complete case and multiply imputed models. Implementation of bundles varied. COPD care bundles were associated with slightly lower secondary care costs, but there was no evidence that they improved outcomes once adjustments were made for site and baseline covariates. Care bundles were unlikely to be cost effective for the NHS with an estimated net monetary benefit per 90-day death avoided from an adjusted multiply imputed model of -£1231 (95% confidence interval - £2428 to - £35) at a high cost-effectiveness threshold of £50,000 per 90-day death avoided.

CONCLUSION AND RECOMMENDATIONS

Care bundles for COPD did not appear to be cost effective, although this finding may have been influenced by unmeasured variations in bundle implementation and other potential confounding factors.