Using data collected in 2007 for Addis Ababa Mortality Surveillance, the paper examines the effect of cause of death/type of illness on choice of health care in adults 12 years and above. The multinomial logit model using bootstrapped standard errors is used to investigate the relationship between dominant type of treatment and the covariates: cause of death, gender, age, education, occupation, ethnicity and religion. Availability of water, television and telephone in the household are used as a proxy for economic status. After controlling for duration of illness (exposure), type of illness, gender and marital status are significant. Those who die of HIV/TB and cancer behave similar in way they seek help, and have high likelihood of using traditional healers as the first point for help compared to those who died as a result of other illnesses. Thus the study concludes that cause of death; gender and marital status affect choice of health service.