Standardized mortality from cervical cancer: a measure of performance?

A retrospective cohort analysis, using data extracted from clinical notes, examined the validity of standardized mortality ratios (SMRs) for cervical cancer as an indicator of the quality of health services by exploring the mortality at five years of patients presenting with cervical cancer, identifying factors associated with survival, and studying the relationship of those factors to SMRs for the disease. The subjects were 1038 women aged 15-64 registered at the Yorkshire Cancer Registry as having invasive cervical cancer between 1979 and 1983. It was found that independent risk factors for mortality within five years of presentation were stage at diagnosis and smoking habit. Method of treatment had an effect that failed to reach statistical significance. Independent risk factors for late stage at presentation were absence of a history of cervical smear and increasing age. Considerable variation remained unexplained in both models. District SMRs were not related to five-year survival, stage at presentation or screening history. SMRs were related to proportions of smokers. It is concluded that SMRs are subject to considerable influence from a range of unidentified factors, as well as identified risk factors. They are too broad an indicator to be a valid measure of health service performance. The health service's ability to reduce case fatality does not outweigh the effects of incidence and severity. Better indicators would be survival to five years of those presenting at stage II or less and distribution of stage at presentation.