Objective: We aimed to predict mortality rate from the findings of annual health checkups for men and women.
Method: The subjects were 31,053 men and 61,224 women who were living in Ibaraki prefecture (Japan), aged 40 to 79 years, without history of any stroke and coronary heart disease, and who participated in annual health checkups in 1993. They were followed until the end of 2001, with a systemic review of resident registration and death certificates. The Cox's proportional hazards model with step-down procedure was used to estimate predictive model.
Results: During the 8.0 years follow-up, there were 5260 deaths (710 from stroke, 389 from coronary heart disease and 2,322 from cancer). The predictive factors for all causes were advanced age, high systolic blood pressure, medication for hypertension, low serum HDL cholesterol, high or low serum creatinine, high AST or ALT, diabetes, low body mass index, current smoking, heavy drinking, and urinary protein among men. The predictive factors for cardiovascular disease were advanced age, high systolic blood pressure, medication for hypertension, low or high serum total cholesterol, low serum HDL cholesterol, high serum creatinine, diabetes, low body mass index, current smoker, and urinary protein, and those for cancer were advanced age, medication for hypertension, low serum HDL cholesterol, low serum creatinine, high AST or ALT, diabetes, low body mass index, current smoking, heavy drinking and urinary protein. Furthermore, those for stroke were advanced age, high systolic blood pressure, medication for hypertension, low serum HDL cholesterol, high serum creatinine, high AST or ALT, low body mass index, current smoking, while for coronary heart disease they were advanced age, high systolic blood pressure, high serum total cholesterol, low serum HDL cholesterol, diabetes, current smoking, and urinary protein among men. For women, similar predictive factors were obtained, although some of them did not reach statistical significance.
Conclusion: We could construct predictive models for 5-year mortality rate from results of annual health checkups. These findings should prove useful for computerized health education on the prevention of stroke, coronary heart disease, and cancer.