Does individual programme size affect screening performance? Results from the United Kingdom NHS breast screening programme.
OBJECTIVE: The size (number of women screened) of the 95 individual NHS breast screening programmes (NHSBSPs) varies by a factor of 10. This study investigates the impact of size on the performance of individual programmes. METHOD: Data were collated from the 95 United Kingdom screening programmes on the standard statistical returns for the past 5 years (1 April 1995-31 March 2000). Additional information was obtained from questionnaires. The number of women screened between 1 April 1999 and 31 March 2000 determined the size of a programme. The bottom 25% were defined as small, the middle 50% as medium, and the top 25% as large. On average large programmes screened about four times as many women as small programmes and medium programmes about twice as many. Performance was evaluated using cancer detection rates, referral rates for assessment, and positive predictive value (PPV) of assessment using PPV referral diagrams. RESULTS: The performance of smaller programmes was shown to be marginally poorer than medium and large sized programmes in that they detected fewer cancers and had a lower PPV. The smallest 25% of programmes had an invasive cancer detection rate 13% less than the medium and large programmes. However, if these programmes had an equivalent detection rate to the medium/large programmes the national detection rate would only increase by about 2%. This is because the 75% of programmes described as medium and large screen about 90% of all women. It is therefore important to place the clinical importance of these findings in context when considering any envisaged possible solutions. CONCLUSIONS: Although the performance of smaller programmes was shown to be poorer than that of the larger programmes, it is not clear from this study exactly why this is so. A likely contributory factor based on experience of evaluating the NHSBSP is that performance problems in larger programmes have been easier to detect by quality assurance staff. The size of the small programmes and the few screen detected cancers (and inherent statistical instability in detection rates) mean that problems are difficult to identify. As a consequence small programmes which are genuinely performing marginally below specific standards are likely to receive less attention than larger programmes, and even under close scrutiny the causes are less likely to be found.