The basic steps in computing an individual's life table are fairly simple:

1. If nothing is known about the individual except the age and sex, a standard government life table applies (National Center for Health Statistics, 1997). The life table will specify the mortality rate m(x) for each age x.
2. Certain risk factors have a known effect on mortality. For example, heavy alcohol consumption increases mortality by amounts that have been well documented in the medical literature. The increases are excess death rates, or EDRs. Thus if we know that the individual is a heavy drinker we can increase the age-specific rates m(x) appropriately.
3. This new schedule of mortality rates can be used to construct a new life table, from which the life expectancy and other quantities of interest can at once be obtained.
4. In some cases the medical literature specifies how the excess risk changes over the lifetime. For example, patients with acute renal failure or cancer have a very large EDR during the first few post-onset years, but these gradually diminish if the patient survives long enough. The point is that a whole schedule of age- specific EDRs is needed, not just the EDR at the current age.
5. It should be noted that the EDRs can be negative. This will occur for persons with healthy life styles -- for example, those who are slightly underweight or who exercise regularly.
6. In most cases there will be several factors to consider simultaneously. For example, a diabetic patient may also be a smoker. The literature specifies EDRs for each factor, but how should they be combined?

7. We have consulted hundreds of medical and actuarial articles to determine the EDRs for the various risk factors. This is an ongoing process, however, as new data and medical research findings are constantly being published. The sources currently used for the software's schedules of EDRs are given in the references.
8. Once the schedule of EDRs is determined, and the corresponding life table computed, all other quantities of interest are immediately available. For example, the expected present value of payments and services specified by a life care plan can be computed from the life table.

This trio of factors deserves separate consideration here. They need to be considered simultaneously because they are strongly interrelated. For example, college educated persons are much less likely to smoke than persons with high school education or less, and the college educated include much lower proportions of Hispanics and Blacks (Richards & Abele 1999, pp. 38 and 43). Thus if, for instance, the increased mortality associated with smoking and with low education were simply added there would be some double counting.

Each of the three factors is related to mortality. Smoking results in a substantial increase in death rates, as has been documented by a large research literature (see summaries in Singer 1976, pp.38-40, and Lew & Gajewski 1990, pp.3-7 to 3-11). Mortality rates also vary by race, the difference being sufficient that the National Center for Health Statistics (NCHS) publishes race-specific life tables (NCHS 1997, Anderson 1999). Finally, it has been shown that mortality rates differ according to education level, the most educated groups having the lowest mortality rates (Richards & Barry 1999). This is not surprising because education often serves as a proxy for socioeconomic status.

There do not seem to be published tables on the "three way interaction" -- i.e., broken down by smoking, education and race simultaneously (for each age and sex). Some assumption must therefore be made. The most plausible appears to be that race and smoking have independent effects once education is controlled for. Under this assumption the appropriate procedure is to make adjustments for race and education simultaneously, then for smoking and education simultaneously, and to combine the two adjustments. This procedure has been followed here.

1. Smoking and Education

2. Race and Education