Abstract
BackgroundIn health services research, composite scores to measure changes in health-seeking behaviour and uptake of services do not exist. We describe the rationale and analytical considerations for a composite primary outcome for primary care research. We simulate its use in a large hypothetical population and use it to calculate sample sizes. We apply it within the context of a proposed cluster randomised controlled trial (RCT) of a Community Health Worker (CHW) intervention.MethodsWe define the outcome as the proportion of the services (immunizations, screening tests, stop-smoking clinics) received by household members, of those that they were eligible to receive. First, we simulated a population household structure (by age and sex), based on household composition data from the 2011 England and Wales census. The ratio of eligible to received services was calculated for each simulated household based on published eligibility criteria and service uptake rates, and was used to calculate sample size scenarios for a cluster RCT of a CHW intervention. We assume varying intervention percentage effects and varying levels of clustering.ResultsAssuming no disease risk factor clustering at the household level, 11.7% of households in the hypothetical population of 20,000 households were eligible for no services, 26.4% for 1, 20.7% for 2, 15.3% for 3 and 25.8% for 4 or more. To demonstrate a small CHW intervention percentage effect (10% improvement in uptake of services out of those who would not otherwise have taken them up, and additionally assuming intra-class correlation of 0.01 between households served by different CHWs), around 4,000 households would be needed in each of the intervention and control arms. This equates to 40 CHWs (each servicing 100 households) needed in the intervention arm. If the CHWs were more effective (20%), then only 170 households would be needed in each of the intervention and control arms.ConclusionsThis is a useful first step towards a process-centred composite score of practical value in complex community-based interventions. Firstly, it is likely to result in increased statistical power compared with multiple outcomes. Second, it avoids over-emphasis of any single outcome from a complex intervention.
Highlights
In health services research, composite scores to measure changes in health-seeking behaviour and uptake of services do not exist
We have been exploring the possibility of embedding such a service into local GP practices in deprived communities in North Wales [21]. It is in this context that we describe the rationale and analytical considerations for this process composite outcome score
The main exception is middle-aged men, who are only eligible for an intervention if they smoke, have diabetes or are otherwise considered highrisk and in need of the seasonal flu vaccination
Summary
Composite scores to measure changes in health-seeking behaviour and uptake of services do not exist. We simulate its use in a large hypothetical population and use it to calculate sample sizes. We apply it within the context of a proposed cluster randomised controlled trial (RCT) of a Community Health Worker (CHW) intervention. There are many challenges in designing studies of complex interventions [1], one of which is to decide upon appropriate outcome measures. When undertaking a randomised trial, and other evaluation studies, it is desirable to find a single primary outcome measure so that statistically robust conclusions about the success of the intervention can be made. An alternative may be to collect many different measures to assess the effectiveness of the intervention, yet this may require statistical adjustments such as the Bonferroni correction [4]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
