Services Research Outcomes Study (SROS)

1. MORTALITY DURING THE POST-TREATMENT PERIOD
   • of Deaths Among Clients in the SROS Sample and in the U.S. Population
     • Table 3-13. Comparison of expected and observed deaths in the SROS client sample
   • of Known Deceased and Living in SROS Sample
     • Table 3-14. Comparison of treatment features of clients who were deceased and living
     • at followup

MORTALITY DURING THE POST-TREATMENT PERIOD

In addition to the 1,799 SROS sample respondents interviewed five years after treatment, 277 clients—about nine percent of the sample—were found to have died during the post-treatment period. In order to gain some value from this information, this section examines:

·510·Death rates in the client population compared with the corresponding (age-sex-race-adjusted) total resident U.S. population, including comparisons of overall death rates and rates for each of four demographic groups (i.e., white males, white females, black males, and black females); and·510

·510·Characteristics of clients who were known to be deceased and those who were known to be alive at the time of the SROS interviewing field period in 1995–1996.·510

Comparison of Deaths Among Clients in the SROS Sample and in the U.S. Population

Since death rates in the U.S. population are known to differ appreciably among groups sorted by age, sex, and race, it was necessary to control for these variables in order to compare usefully the mortality of treatment clients to that of the total resident population. One way to accomplish this was to use death rates that were specific for each age, sex, and race, which were readily available in annual Vital Statistics reports. The 1990 rates were used (in contrast to death rates averaged across all four years or from the midpoint year, 1993) because age-specific death rates have not changed significantly in the past few years, [ Garner P. Hudson, "Advance Report of Final Mortality Statistics, 1993."] and the decennial census in 1990 provided population denominators considered more accurate than the estimates of the population in noncensus years. The conclusions of the following analyses would not be affected by the small variations that would result from using adjusters from other years.

Estimation of the expected number of deaths in the sample involved multiplying the five-year age-sex-race-specific death rates of the United States by the size of the initial SROS sample in each age-sex-race category at the beginning of the sample period, using the following steps:

·510·The 1990 age-sex-race-specific death rates per 100,000 were multiplied by five to reflect the observed interval of five post-treatment years. ·510

·510·This proportion was multiplied by the sample size in each group, and the result was the expected number of deaths in each age-sex-race group. The resulting numbers of deaths for each age group and category were then summed to produce the number of each demographic subgroup that would be expected to die during the five-year period.·510

·510·This process was carried out separately for blacks and whites, and for males and females; the four totals sum to the total number of sample clients who would be expected to die during the period. ·510

The age distribution of the SROS sample included very few cases younger than 15 years or older than 60 years, and the number of Native Alaskans, American Indians, and Asian or Pacific Islanders in the sample was small. Hispanics are included within their self-identified racial subgroup. This analysis therefore refers to whites and blacks 15 to 60 years of age at the time of discharge from the 1989–1990 index treatment episode, and it includes Hispanics in both the expected and observed death totals. [ This procedure excludes from the analysis 53 individuals who died. They came from categories having such a small number of cases that results are not meaningful. The categories were: under age 15 or over age 60 at discharge, and certain minorities.]

Table 3-13 presents the comparison between expected and observed deaths in most of the client population. The number of deaths in the client population was 7.3 times higher than would be expected if the total population were matched to it in terms of age, sex, and race. Instead of the 31 expected deaths, the client group had 224 observed deaths.

Table 3-13. Comparison of expected and observed deaths in the SROS client sample
Population group*	Sample n	Annual Death Rate/100,000 Population**	Expected Deaths (5 years)	Observed Deaths (5 years)	Ratio of Observed to Expected Deaths
White males	1,112	377.6	13.3	108	8.1
White females	411	190.4	1.9	35	18.4
Black males	460	837.4	13.3	67	5.0
Black females	191	392.8	2.0	14	7.0
Total			30.5	224	7.3

*Population groups include Hispanics within self-designated white or black subgroups..

**Source of annual death rates: National Center for Health Statistics, Vital Statistics of the United States, 1991.

For subgroups of the SROS population:

·510·White males discharged from treatment in the SROS sample had 108 observed deaths, rather than the expected 13 deaths, giving this population about eight times as many deaths as expected.·510

·510·White females discharged from treatment in the SROS sample had 35 deaths, rather than the expected two deaths, giving this population nearly 18 times as many deaths as expected.·510

·510·Black males discharged from treatment in the SROS sample had 67 deaths, rather than the expected 13 deaths, giving this population about five times as many deaths as expected.·510

·510·Black females discharged from treatment in the SROS sample had 14 deaths, rather than the expected two deaths, giving this population seven times as many deaths as expected.·510

The high observed death rate in the client sample does not seem surprising in view of other information about the health status of the treatment group. For example, 33.5 percent of the weighted client population rated their health status as fair or poor in the five years before treatment; by way of comparison, Current Estimates from the National Health Interview Survey of 1990 reported that 9.5 percent of the total population reported itself to be in fair or poor health. Since the interviewed group did not include the part of the sample who had died, whose average health status before treatment could reasonably beassumed to be worse than the surviving group’s health status, it is not surprising to observe a death rate exceeding the ratio of 3.5 to 1 (for poor and fair health; i.e., 33.5/9.5) by a factor of two.

This analysis further indicates that observed deaths among white female clients exceeded their expected deaths by a much larger margin than the other client sex-race groups. This is partly determined by the low expected death rate among white females in the client population, 190 deaths per 100,000 lives, in comparison with white males at 378, black females at 393, and especially compared with the high expected mortality among black males at 837 per 100,000.

Comparison of Known Deceased and Living in SROS Sample

The 277 deceased clients in the SROS sample were weighted up to an estimated 93,000 individuals who died in the five years after discharge from the SROS episode. A comparison of continuous variables for the known dead and living is presented in Table 3-14.

Table 3-14. Comparison of treatment features of clients who were deceased and living at followup
Treatment Feature	Deceased (sample n)	Living (sample n)	Total (sample n)
Average months of treatment episode	2.9 (260)	3.8 (1,778)	3.7 (2,038)
Average age at discharge from treatment episode	41.3* (262)	30.3* (1,747)	31.6 (2,009)
Number of services during treatment	3.5 (248)	3.2 (1,698)	3.3 (1,946)
Number of lifetime treatment episodes	3.0 (188)	3.0 (937)	3.0 (1,125)
* Difference statistically significant at the 0.05 level.

A comparison of the living and deceased clients yielded the following results:

·510·Those who died in the five years after treatment were about ten years older at discharge from treatment than those who survived.·510

·510·The index length-of-stay of living versus deceased clients appeared to be shorter, but the difference was not statistically significant.·510

·510·The average number of specific treatment services rendered to living versus deceased clients in the SROS episode was the same.·510

·510·The average number of treatment episodes received by living versus deceased clients before the index treatment episode was the same.·510

A logistic regression was run to see whether any of the known demographic or treatment characteristics could be used to predict mortality in the five years after clients left the SROS index episode. Unfortunately, many of the predictor variables used in the regression analyses reported earlier in this chapter were either not available for the deceased or resulted in such a large number of missing cases as to make the analysis impractical (and in some instances misleading). The results shown were based on a model chosen to maximize consistency with the other analyses in this report without sacrificing accuracy due to lost sample size through missing data. The multiple r² was low at 0.11 or 11 percent of variance accounted for. Female respondents were one-half (0.50) as likely as males to have died—although, as shown above, this places them at much greater death rates than would be expected— while older respondents were much more likely to die than younger clients. Finally, although other service measures such as length of stay and number of services did not discriminate between living and deceased clients, those clients who completed their treatment program were 30 percent less likely (odds ratio of 0.70) to die than those who did not complete treatment.

This page was last updated on June 03, 2008.

SAMHSA, an agency in the Department of Health and Human Services, is the Federal Government's lead agency for improving the quality and availability of substance abuse prevention, addiction treatment, and mental health services in the United States.