Mortality Rates and Life Expectancy of Veterans from 1980 to 2017, and ...

[Pages:30]Mortality Rates and Life Expectancy of Veterans from 1980 to 2017, and by Education, Income, and

Period of Service

Department of Veterans Affairs National Center for Veterans Analysis and Statistics

September 2021

EXECUTIVE SUMMARY

This report generates estimates of mortality rates and life expectancy for Veterans at ages 20 to 85, in the 8 to 10 years periods of 1980-89, 1990-99, 2000-09, and 2010-17, and compares them to the U. S. population. The expected life-years for Veterans at age 25 in the years 2010-17, an averaged estimate of the 8 years, is about 0.9 life-years shorter than those of U. S. population in 20145, the mid-year of the 8 years, for both males and females. The estimated education differentials in life expectancy for Veterans in 2010-17 are 0.2 to 0.8 life-years smaller than those of U. S. population in 1979-893. Much smaller income differentials in life expectancy, 0.2 to 3.6 life-years smaller for Veterans in 2010-17 are estimated. For example, life expectancy of Veterans at age 25 with household incomes less than $20,000 are estimated to be 4.3 and 2.8 less lifeyears than those of $50,000+ income, compared to 7.9 and 3.3 life-years in the U. S. population, thus 3.6 and 0.5 life-years smaller income differentials for male and female Veterans, respectively, shown in Tables 17 and 18. At age 55, World War 2 Veterans were estimated to live 21.4 and 26.1 life-years, 1.4 and 0.3 more life-years for Korean Conflict Veterans, and 3 and 1.5 more life-years for Vietnam Era Veterans for male and female, respectively. The report discusses missing older Veterans and other limitations in data which may have caused computing shorter life expectancy estimates.

Introduction

Veterans are a preselected group of healthy people from the general population. A soldier has to be fit both physically and mentally to endure the hardship in an adverse environment such as a battlefield. The healthy soldier/worker effect gradually subdued over time. This is similar to the selection effect of underwriting requirement provision in the insurance industries. It has been well-documented in epidemiology studies.1,2 The phenomenon of healthy soldier effect could be confirmed with the established database for U.S. Veterans Eligibility Trends and Statistics, USVETS, a data source for Veterans undertaken by the Data Governance and Analytics business line in the Department of Veterans Affairs. The mortality rates of recently released Veterans are expected to be lower than those of the U. S. population and the mortality rates will become very similar as Veterans get older.

Life expectancy, a measure summarizing the mortality experience over a lifetime, is an estimate of expected life-years at a specific age. Life expectancy is computed directly from the mortality rate at each age. Similar to the mortality rates, the life expectancy of Veterans should be close to the life expectancy of the U. S. population. The gaps in life expectancy by socioeconomic factors have been observed in the U. S. population3. This

2

paper attempts to measure the socioeconomic differentials in life expectancy such as education and income in Veterans and compare them with the U.S. population.

The study design bases on available data. For example, the socioeconomic differential estimates started in 2010. A direct comparison of mortality of Veterans to U. S. population from the literature may be difficult. This study draws conclusions based on the known facts from the current study and other studies in the literature.

Study Data

The Data Governance and Analytics (DGA) business line developed the Veterans Eligibility Trends and Statistics (USVETS) dataset in fiscal year 2018. USVETS is a data environment consisting mainly of data sources from the Veterans Benefit Administration, the Veterans Health Administration, the Department of Defense's (DoD) Defense Manpower Data Center, and other data sources including commercial data sources. This dataset contains information about individual Veterans including demographics, details of military service, VA benefit usage, and more. The dataset contains one record per Veteran. It includes all living and deceased Veterans. The dates of death of deceased Veterans include National Death Index (NDI) data obtained from VA Center of Excellence for Suicide Prevention, Joint VA and DoD Suicide Data Repository (SDR) - NDI, ; December 20, 2016.DGA uses its database to conduct statistical analysis, predictive modeling, and reporting. USVETS includes the software, hardware, and the associated processes that allow staff to create various USVETS work products. Like any other dataset, USVETS has its limitations and assumptions that users need to consider.

1. USVETS tends to be more accurate for underlying Veteran population below the age of 67 than for older Veterans due to missing DoD records for older Veterans. This limitation will gradually decrease over time.

2. USVETS is mixed data. It uses internal VA administrative and external commercial vendor data. For the internal VA administrative data, DGA does not have control over input data production or quality of these sources. Although DGA applies rigorous data quality techniques to ensure the high quality of these data, there may be flaws in data inputs that are not identified.

3. For the commercial vendor data, the contractor collects data from a variety of sources such as public records, surveys, and on-line and off-line registrations. The accuracy and completeness of the data are determined by these sources. The data may not be complete and in some cases the data may not be current due to timing of updates from these sources.

3

4. Like all other databases, USVETS relies on business rules that allow the data manager to select one data element over another data element for achieving validity, reliability and consistency. Thus, counts derived from USVETS may differ from other internal VA administrative databases.

5. Some of USVETS administrative variables may have special limitations based on how the data was collected and imputed. For certain variables USVETS may use commercial vendor data to supplement or replace VA administrative data.

6. USVETS is subject to continuous scrutiny, revision and refresh in the pursuit of producing the best Veteran data for Veteran analysis, planning and policy. Accordingly, USVETS versions may vary.

The data in the study is from the 2018 USVETS database, combining Veterans data from Department of Defense, Veterans Benefit Administration, Veterans Health Administration, and others, with a total of 39,372,027 administrative records (of living and deceased Veterans). We applied rigorous exclusion rules that reduced the number of records for this analysis to 33,189,277. The top three reasons for excluding a record were: (1) the Veteran died prior to 1980 because the study focuses on the years 1980 to 2017; (2) the Veteran's record did not have a separation date; and (3) the Veteran's identity information did not pass SSA validation test.

Table 1. shows the number of records by gender, living status and median birth year in 1980-2017. Those records without a separation date are older with a median birth year of 1931 versus those with a separation date and median birth year of 1946.

Table 1. Gender, Living Status and Median Birth Year, USVETS, 1980-2017

Total

Male Female Unknown

Living

Median Dead birth year

In Analysis 33,189,277 30,903,803 2,270,797 14,677 17,491,983 15,697,294

1946

(percent)

100

93

7

0

53

47

No Separation Date

(percent)

2,091,540 1,858,184 232,120

1,236 718,932

100

89

11

0

34

Source: U.S. Veterans Eligibility Trends and Statistics, 2018 Prepared by the National Center for Veterans Analysis and Statistics

1,372,608 66

1931

4

A consumer-data collecting company has provided socioeconomic data for 29,608,786 records in this analysis. By excluding Veterans who died before year 2010, we have a total of 21,181,619 records to compute the socioeconomic differentials in life expectancy in the years of 2010-2017. The education variable has 4 levels: `Completed high school,' `Completed college,' `Completed graduate school,' and `Attended vocational/technical.' Roughly 23 percent are unknown. These education levels are grouped into 3 categories: `None or Unknown', `High School', and `College or More;' `College or More' includes `Completed college', `Completed graduate school', and `Attended vocational/technical'.

The estimated household income of commercial vendor data has13 brackets: ? `Less than $15,000', ? `$15,000-$19,999', ? `$20,000-$29,999', ? '$30,000-$39,999', ? '$40,000-$49,999', ? '$50,000-$59,999', ? '$60,000-$69,999', ? '$70,000-$79,999', ? '$80,000-$89,999', ? '$90,000-$99,999', ? `$100,000-$124,999', ? `$125,000-$149,999', and ? `Greater than $149,999.'

Roughly 9 percent are unknown and grouped into 3 categories: `Less than $20,000 or Unknown', `$20,000-$49,999', and `$50,000 or More.' The unknown has been shown to be similar to the low income brackets and therefore is combined with the `Less than $20,000' category.

Table 2. shows the number of records by gender, living status and median birth year, by education and income categories, for the years 2010-2017. The group of no high school or unknown education and the groups of lower or unknown income have higher percentage of deaths. The relative high percentage of deaths for the group of `College or more' education can be explained by older ages of this group of the median birth year of 1952, compared to all groups of median of 1956.

5

Table 2. Gender, Living Status and Median Birth Year by Education and Income, USVETS, 2010-2017

All (percent) None or Unknown

(percent)

Total 21,181,619

100

4,911,463 100

Male 19,239,814

91

4,540,485 92

Female 1,941,805

9

370,978 8

Unknown 0 0

0 0

Living 17,298,944

82

3,898,874 79

High School 8,516,475 7,689,283 827,192

0 7,175,919

(percent) College or

More

(percent) < $20K or Unknown

(percent)

100

7,753,681 100

3,940,753 100

90

7,010,046 90

3,531,942 90

10

743,635 10

408,811 10

0

84

0 6,224,151

0

80

0 3,089,607

0

78

$20K-$49K 6,466,345 5,880,168 586,177

(percent)

100

91

9

0 5,021,148

0

78

$50K or More 10,774,521 9,827,704 946,817

(percent)

100

91

9

0 9,188,189

0

85

Source: U.S. Veterans Eligibility Trends and Statistics, 2018 Prepared by the National Center for Veterans Analysis and Statistics

Dead 3,882,675

18

1,012,589 21

1,340,556 16

1,529,530 20

851,146 22

1,445,197 22

1,586,332 15

Median birth year

1956

1957 1959

1952

1959 1954 1956

Cohorts of Period of Service

Veterans of various period of service would have different mortality experience. The study population are grouped into 5 cohorts of period of service: World War II, Korean Conflict, Vietnam Era, Gulf Wars, and Peace Time. Tables 3 and 4 showed the numbers of total and living records, percentage of living, VetPop estimate, median age, and percentile range of birth year by period of service, in the years 1980-2017, for male and female, respectively, where VetPop is for the 2018 Veterans Population Projection The numbers of living WWII, KC, and VNE Veterans at the end of study period are smaller than the VetPop estimates due to missing older Veteran records in USVETS. Regardless the living counts on 12/31/2017 would be slightly smaller than the VetPop estimates at early date of 9/30/3017. The age calculated at the beginning of the study,

6

1/1/1980, should be appropriate for all Veterans except the GW Veterans of median birth year of 1977, and the percentile range of birth year may show how ages are different among all cohorts of period of service.

Table 3. Living Status, Median Age, and Percentile Range of Birth Year by Period of Service, USVETS, 1980-2017, Male Veterans

Total

Living

% of living

VetPop estimate

Median age as of 1980

All

30,903,803 16,226,400 53 19,088,341

34

WWII

7,667,198 274,367

4 558,076

59

KC

3,198,255 660,631 21 1,355,521

48

VNE

7,810,591 5,139,917 66 6,408,586

32

GW

6,662,750 6,427,740 96 6,417,214

Peace Time 5,565,009 3,723,745 67 4,348,944

22

Source: U.S. Veterans Eligibility Trends and Statistics, 2018 Prepared by the National Center for Veterans Analysis and Statistics

Median birth year 1945 1920 1931 1947 1974 1957

10th percentile birth year

1917 1911 1923 1936 1957 1928

90th percentile birth year

1975 1926 1934 1954 1989 1965

Table 4. Living Status, Median Age, and Percentile Range of Birth Year by Period of Service, USVETS, 1980-2017, Female Veterans

Total

Living

Median % of VetPop age as of living estimate 1980

All

2,270,797 1,923,462 85 1,973,651

WWII

186,983

9,830

5 29,463

59

KC

50,521 11,961 24 43,006

49

VNE

207,294 163,114 79 221,155

29

GW

1,293,721 1,271,783 98 1,238,116

Peace Time 532,278 466,774 88 441,911

19

Source: U.S. Veterans Eligibility Trends and Statistics, 2018 Prepared by the National Center for Veterans Analysis and Statistics

Median 10th birth percentile year birth year

1966

1929

1920

1910

1930

1919

1950

1941

1977

1961

1960

1948

90th percentile birth year

1988 1924

1934 1955 1991 1967

Person-Years Method

Person-years are computed for each attained age and calendar year. For example, a Veteran born on March 19, 1949, would get 0.2 person-year at age 30 and year 1980,

7

0.8 person-year at age 31 and year 1980, 0.2 person-year at age 31 and year 1981, and so on if no death occurred in the years of study. The same person would only get 0.5 person-year at age 31 and year 1980 if he or she died on July 1, 1980, and he or she would get no further person-year contribution. Person-years are then combined at each age for all Veterans in the study from January 1, 1980, through December 31, 2017. The mortality rate at each age is computed by dividing the number of deaths at the age by the number of persons surviving to the age which is the sum of person-years and half of number of deaths at the age. It assumes uniform distribution of deaths during the year last for the age. Then, these rates are applied in a life table to compute the life expectancy at each age. The standard error of life expectancy shown in parentheses in the tables is computed based on Chiang's formula4.

Results

Mortality rate and life expectancy at each age, from ages 20 to 85, and each 8-10 years period, 1980-1989, 1990-1999, 2000-2009, and 2010-2017, were computed for male and female Veterans, respectively. Tables 5 and 6, and Figures 1 and 2, compare the Veteran mortality rate and life expectancy, before and after adjusted for older ages, in years 2010-2017 to those of U. S. population in 20145 . Only the mortality rate and life expectancy at ages 20, 25, 30... 85 are shown for demonstration purposes. Compared to the U. S. population, the mortality rates of Veterans are higher at older ages, and the life expectancy of Veterans are estimated to be 0.9-1.2 life-years shorter before adjusted for older ages using SSA mortality rates6 and 0.6-1.0 life-years shorter after the adjustment for males, and to be 1.7-1.9 life-years shorter before and 0.9-1.0 lifeyears shorter after the adjustment for female at age 25, 45, and 65. The tables and figures afterwards only use the estimate of mortality rate and life expectancy after the adjustment.

8

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download