The impact on colorectal cancer survival of cases registered by 'death certificate only': implications for national survival rates.

This paper describes the effect of including death certificate only (DCO) registrations on 5 year relative survival rates for colorectal cancer in four district health authorities (DHAs) in south-east England. A retrospective case note study was set up to examine all cases of colorectal cancer listed in the Thames Cancer Registry (TCR) as having been diagnosed in 1983 and 1988 and resident in one of four districts, A, B, C and D. A total of 673 sets of cases notes were requested from all hospitals within the four districts, including 150 sets on DCO cases. Of 465 (69%) sets of case notes retrieved, 378 (72.3%) were non-DCO cases. Of these, 14 were excluded from survival analysis because of missing dates of diagnosis or death in the notes. Eighty-seven (58.0%) sets of case notes were retrieved on DCO registrations, of which seven were excluded because no date of diagnosis was available in the notes. Retrieval rates on case note registrations varied by DHA of residence: 73.3% in DHA A, 96.6% in DHA B, 34.5% in DHA C and 79.2% in DHA D. The corresponding figures for DCO registrations were 63.5%, 69.0%, 7.4% and 76.2%. Cumulative relative 5 year survival rates by DHA of residence were calculated first for cases registered from case notes and then for all cases including those registered solely from a death certificate. The total number of cases used in the survival analysis was 444 (18% DCOs). In all four DHAs, 5 year survival decreased with the inclusion of DCO registrations: by 9.1% in district A (from 52.8 to 43.7), by 4.5% in district B (from 59.6 to 55.1), by 4.8% in district C (from 80.0 to 75.2) and by 7.6% in district D (from 31.4 to 23.8). The overall reduction in survival was 8.6%. The exclusion of death certificate only registrations from survival data is an important source of bias. Using TCR data, we compared DCO proportions for colorectal cancer with other sites. DCO proportions were shown to vary by tumour site and survival time. The DCO registration is an important quality measure of ascertainment and follow-up. OPCS should publish DCO proportions by registry area and cancer site. Registries should implement DCO monitoring as part of quality improvement programmes.