Sample size calculations for intervention trials in primary care randomizing by primary care group: an empirical illustration from one proposed intervention trial

Because of the central role of the general practice in the delivery of Brit ish primary care, intervention trials in primary care often use the practic e as the unit of randomization. The creation of primary care groups (PCGs) in April 1999 changed the organization of primary care and the commissionin g of secondary care services. PCGs will directly affect the organization an d delivery of primary, secondary and social care services. The PCG therefor e becomes an appropriate target for organizational and educational interven tions. Trials testing these interventions should involve randomization by P CG. This paper discusses the sample size required for a trial in primary ca re assessing the effect of a falls prevention programme among older people. In this trial PCGs will be randomized, The sample size calculations involv e estimating intra-PCG correlation in primary outcome: fractured femur rate for those 65 years and over. No data on fractured femur rate were availabl e at PCG level. PCGs are, however, similar in size and often coterminous wi th local authorities. Therefore, intra-PCG correlation in fractured femur r ate was estimated from the intra-local authority correlation calculated fro m routine data. Three alternative trial designs are considered. In the firs t design, PCGs are selected for inclusion in the trial from the total popul ation of England (eight regions). In the second design, PCGs are selected f rom two regions only. The third design is similar to the second except that PCGs are stratified by region and baseline value of fracture rate. Intracl uster correlation is estimated for each of these designs using two methods: an approximation which assumes cluster sizes are equal and an alternative method which takes account of the fact that cluster sizes vary. Estimates o f sample size required vary between 26 and 7 PCGs in each intervention grou p, depending on the trial design and the method used to calculate sample si ze. Not unexpectedly, stratification by baseline value of the outcome varia ble decreases the sample size required. In our analyses, geographic restric tion of the population to be sampled reduces between-cluster variability in the primary outcome. This leads to an increase in precision. When allowanc e for variable cluster size is made, the increase in precision is not as gr eat as would be expected with equal cluster sizes. This paper highlights th e usefulness of routine data in work of this kind, and establishes one of t he essential prerequisites for our proposed trial and other trials using pr imary outcomes with similar between-PCG variation: a feasible sample size. Copyright (C) 2001 John Wiley & Sons, Ltd.