Mortality in co-morbidity (II)--excess death rates derived from a follow-up study on 10,025 subjects divided into 4 groups with or without depression and diabetes mellitus.

BACKGROUND: Most mortality follow-up (FU) studies focus on excess mortality in a single risk factor or impairment. However, many persons in the general population with 1 important medical risk factor are likely to have co-morbidity in the form of other risk factors, some minor, but others that may be of major significance. Logically, with 2 major significant risk factors present, the combined excess mortality may be smaller or greater than the sum of the individual mortality rates, or may be nearly the same as the sum. When 2 major co-morbid risk factors are present, it is important to know which of these 3 possibilities is present. In the first article of this series, we analyzed the results of 57 individual impairments in the Multiple Medical Impairment Study2 (MMIS). We found that, when elevated blood pressure (EBP) was the comorbid impairment, the excess mortality outcomes were divided almost equally among these 3 possibilities. ARTICLE BY EGEDE ET AL: This informative 2005 article has utilized data from an 11-year FU of subjects in the 1971-1975 National Health and Nutrition Examination Survey, from which mean annual mortality rates per 1000 have been presented in 4 groups of subjects: Group 1 with no depression (D) or diabetes mellitus (DM), Group 2 with D only, Group 3 with DM only and Group 4 with both D and DM present. Total subjects numbered 10,025, of whom 70.2% were in Group 1, 22.7% in Group 2, 4.5% in Group 3, and only 2.6% in Group 4. In addition to mean age, proportions were given in each group for sex and race, 3 additional demographic and 8 additional medical risk factors. Two different models were used to calculate hazard ratios, by the Cox proportional hazards method, for total mortality rates and rate for death rates due to coronary heart disease (CHD). The unadjusted mortality rates (q) were given for each group as the ratio of deaths (d) to 1000 person-years of exposure (E). In obtaining hazard ratios the authors used the mortality rate of Group 1 as the reference or expected rate (q') for adjusting the rates in the other groups to derive the hazard ratios in the 2 adjustment models employed. METHODOLOGY OF CURRENT ARTICLE: For Groups 2-4, with 1 or both of the impairments present, we have estimated an adjusted mortality rate, q(a), by multiplying the reference q' (19.1 per 1000 per year) by the appropriate decimal hazard ratio given in Table 2 of the article. For each of the impaired groups, 2-4, the corresponding adjusted EDR has been derived as EDR = q(a) - q' = q(a) - 19.1. We use EDR values as a difference between mortality rates instead of a ratio of rates because EDR values when age/sex/race-adjusted are directly additive and do not require weighting. RESULTS: In Model 1, with adjustment for age, sex, race and 4 other demographic factors, annual EDR values were 6.5, 16.8 and 48.5 per 1000, respectively, in Groups 2, 3 and 4. In Model 2, with all factors in Model 1 and additional medical risk factors, such as heart disease, hypertension and cancer, EDRs were reduced to 3.8, 16.8 and 28.6, respectively in the D, DM and D+DM groups. CONCLUSION: When group mortality differences were adjusted (for other demographic and medical factors as well the basic factors of age, sex and race), EDR in Group 4 subjects, with both D and DM present exceeded the sum of EDRs in Group 2 (D alone) and Group 3 (DM alone) by 83% in Model 1 and by 39% in Model 2. We conclude that the authors of this study have provided convincing evidence that excess mortality measured as EDR is greater in subjects with both depression and diabetes mellitus present than the sum of the EDRs in the groups when each impairment is present alone. This particular combination of impairments has a strong synergistic effect on excess mortality.