Importance: In aetiological research, investigators using death certificate data have traditionally extracted underlying cause of mortality alone. With multi-morbidity being increasingly common, more than one condition is often compatible with the manner of death. Using contributory cause plus underlying would also have some analytical advantages but their combined utility is largely untested.

Objective: To compare the relative utility of cause of death data extracted from the underlying field versus any location on the death certificate (underlying and contributing combined), we undertook analyses of two large cohort studies where we compared the relationships of three known risk factors (cigarette smoking, low educational attainment, and hypertension) to health outcomes using each approach to mortality ascertainment.

Design, Setting, and Participants: Two prospective cohort study collaborations: UK Biobank (N=502,655), and the Health Survey for England (HSE; 15 studies) and the Scottish Health Surveys (SHS; 3 studies) collaboration (N=193,873).

Main Outcomes and Measures: Death from cardiovascular disease, cancer, dementia, and injury. To summarise our results, for each risk factor–mortality endpoint combination, we computed a ratio of hazard ratio (RHR) by dividing the effect estimate for underlying cause by that for any mention.

Results: In Biobank there were 14,421 deaths during a mean of 7 years of follow up; in HSE-SHS, a mean of 10 years of mortality surveillance gave rise to 21,314 deaths. Established associations between our risk factors and death outcomes were essentially the same irrespective of the placement of cause on the death certificate. Results from each study were also mutually supportive. For having ever smoked cigarettes (versus never), the RHRs were around unity (p-value for difference≥0.09), as they were for hypertension (versus none) (p-value>0.36). For no university education (versus greater), the RHRs ranged between 0.73 and 1.30 such that, again, there was no differences in the location of the death data (p-value>0.2).

Conclusions and Relevance: Risk factor–endpoint associations were not sensitive to the placement of data on the death certificate. This has implications for the examination of risk factors for the occurrence of causes of death where there may be too few events to compute reliable effect estimates based on the underlying field alone.