[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
Purchase Options:
[Skip to Content Landing]
Figure 1.  Estimated Change in Years of Life Lost by Cause of Death Between 2010 and 2017
Estimated Change in Years of Life Lost by Cause of Death Between 2010 and 2017

The graphs show the years of life lost between ages 25 and 84 years among non-Hispanic white and black US adults. The error bars indicate 95% CIs.

Figure 2.  Estimated Change in Years of Life Lost by Education and Cause of Death Between 2010 and 2017 Among Non-Hispanic White US Residents
Estimated Change in Years of Life Lost by Education and Cause of Death Between 2010 and 2017 Among Non-Hispanic White US Residents

The graphs show the years of life lost between ages 25 and 84 years. The error bars indicate 95% CIs. Education categories are as follows: low = high school graduate or lower, middle = some college education but no 4-year college degree, and high = 4-year college graduate or higher.

Figure 3.  Estimated Change in Years of Life Lost by Education and Cause of Death Between 2010 and 2017 Among Non-Hispanic Black US Residents
Estimated Change in Years of Life Lost by Education and Cause of Death Between 2010 and 2017 Among Non-Hispanic Black US Residents

The graphs show the years of life lost between ages 25 and 84 years. The error bars indicate 95% CIs. Education categories are as follows: low = high school graduate or lower, middle = some college education but no 4-year college degree, and high = 4-year college graduate or higher.

Table.  Expected Age at Death at Age 25 Years by Race, Sex, and Education, 2010-2017a
Expected Age at Death at Age 25 Years by Race, Sex, and Education, 2010-2017a
1.
Centers for Disease Control and Prevention. NCHS: death rates and life expectancy at birth. https://data.cdc.gov/NCHS/NCHS-Death-rates-and-life-expectancy-at-birth/w9j2-ggv5. Updated June 4, 2018. Accessed June 13, 2019.
2.
Murphy  SL, Xu  J, Kochanek  KD, Arias  E. Mortality in the United States, 2017. NCHS Data Brief No. 328. National Center for Health Statistics. https://www.cdc.gov/nchs/data/databriefs/db328-h.pdf. Published November 2018. Accessed January 6, 2019.
3.
Case  A, Deaton  A.  Rising morbidity and mortality in midlife among white non-Hispanic Americans in the 21st century.  Proc Natl Acad Sci U S A. 2015;112(49):15078-15083. doi:10.1073/pnas.1518393112PubMedGoogle ScholarCrossref
4.
Xu  J, Murphy  SL, Kochanek  KD, Bastian  B, Arias  E.  Deaths: final data for 2016.  Natl Vital Stat Rep. 2018;67(5):1-76.PubMedGoogle Scholar
5.
Case  A, Deaton  A.  Mortality and morbidity in the 21st century.  Brookings Pap Econ Act. 2017;2017:397-476. doi:10.1353/eca.2017.0005PubMedGoogle ScholarCrossref
6.
Meara  ER, Richards  S, Cutler  DM.  The gap gets bigger: changes in mortality and life expectancy, by education, 1981-2000.  Health Aff (Millwood). 2008;27(2):350-360. doi:10.1377/hlthaff.27.2.350PubMedGoogle ScholarCrossref
7.
Olshansky  SJ, Antonucci  T, Berkman  L,  et al.  Differences in life expectancy due to race and educational differences are widening, and many may not catch up.  Health Aff (Millwood). 2012;31(8):1803-1813. doi:10.1377/hlthaff.2011.0746PubMedGoogle ScholarCrossref
8.
Montez  JK, Hummer  RA, Hayward  MD, Woo  H, Rogers  RG.  Trends in the educational gradient of US adult mortality from 1986 to 2006 by race, gender, and age group.  Res Aging. 2011;33(2):145-171. doi:10.1177/0164027510392388PubMedGoogle ScholarCrossref
9.
Chetty  R, Stepner  M, Abraham  S,  et al.  The association between income and life expectancy in the United States, 2001-2014.  JAMA. 2016;315(16):1750-1766. doi:10.1001/jama.2016.4226PubMedGoogle ScholarCrossref
10.
Sasson  I.  Trends in life expectancy and lifespan variation by educational attainment: United States, 1990-2010.  Demography. 2016;53(2):269-293. doi:10.1007/s13524-015-0453-7PubMedGoogle ScholarCrossref
11.
Sasson  I.  Diverging trends in cause-specific mortality and life years lost by educational attainment: evidence from United States Vital Statistics data, 1990-2010.  PLoS One. 2016;11(10):e0163412. doi:10.1371/journal.pone.0163412PubMedGoogle ScholarCrossref
12.
Ho  JY.  The contribution of drug overdose to educational gradients in life expectancy in the United States, 1992-2011.  Demography. 2017;54(3):1175-1202. doi:10.1007/s13524-017-0565-3PubMedGoogle ScholarCrossref
13.
Ho  JY, Fenelon  A.  The contribution of smoking to educational gradients in US life expectancy.  J Health Soc Behav. 2015;56(3):307-322. doi:10.1177/0022146515592731PubMedGoogle ScholarCrossref
14.
Ruggles  S, Flood  S, Goeken  R,  et al.  IPUMS USA: Version 8.0. Minneapolis, MN: IPUMS; 2018. doi:10.18128/D010.V8.0
15.
Centers for Disease Control and Prevention. Vital Statistics online data portal. https://www.cdc.gov/nchs/data_access/vitalstatsonline.htm. Accessed January 14, 2019.
16.
US Census Bureau. American Community Survey: sample size and data quality. https://www.census.gov/acs/www/methodology/sample-size-and-data-quality/. Accessed July 13, 2019.
17.
US Census Bureau. American Community Survey accuracy of the data (2017). https://www2.census.gov/programs-surveys/acs/tech_docs/accuracy/ACS_Accuracy_of_Data_2017.pdf. Accessed July 13, 2019.
18.
Centers for Disease Control and Prevention. US Census populations with bridged race categories. https://www.cdc.gov/nchs/nvss/bridged_race.htm. Accessed April 19, 2019.
19.
Arias  E, Schauman  WS, Eschbach  K, Sorlie  PD, Backlund  E.  The validity of race and Hispanic origin reporting on death certificates in the United States.  Vital Health Stat 2. 2008;(148):1-23.PubMedGoogle Scholar
20.
Xu  J, Murphy  SL, Kochanek  KD, Arias  E.  Mortality in the United States, 2015. Hyattsville, MD: National Center for Health Statistics; 2016.
21.
Rostron  BL, Boies  JL, Arias  E.  Education reporting and classification on death certificates in the United States.  Vital Health Stat 2. 2010;(151):1-21.PubMedGoogle Scholar
22.
Alpérovitch  A, Bertrand  M, Jougla  E,  et al.  Do we really know the cause of death of the very old? comparison between official mortality statistics and cohort study classification.  Eur J Epidemiol. 2009;24(11):669-675. doi:10.1007/s10654-009-9383-2PubMedGoogle ScholarCrossref
23.
Tinetti  ME, McAvay  GJ, Murphy  TE, Gross  CP, Lin  H, Allore  HG.  Contribution of individual diseases to death in older adults with multiple diseases.  J Am Geriatr Soc. 2012;60(8):1448-1456. doi:10.1111/j.1532-5415.2012.04077.xPubMedGoogle ScholarCrossref
24.
Centers for Disease Control and Prevention. US standard certificate of death. https://www.cdc.gov/nchs/data/dvs/death11-03final-acc.pdf. Accessed April 19, 2019.
25.
Centers for Disease Control and Prevention (CDC).  Smoking-attributable mortality, years of potential life lost, and productivity losses: United States, 2000-2004.  MMWR Morb Mortal Wkly Rep. 2008;57(45):1226-1228.PubMedGoogle Scholar
26.
Andersen  PK, Canudas-Romo  V, Keiding  N.  Cause-specific measures of life years lost.  Demogr Res. 2013;29:1127-1152. doi:10.4054/DemRes.2013.29.41Google ScholarCrossref
27.
Diez Roux  AV.  Despair as a cause of death: more complex than it first appears.  Am J Public Health. 2017;107(10):1566-1567. doi:10.2105/AJPH.2017.304041PubMedGoogle ScholarCrossref
28.
Rigg  KK, Monnat  SM, Chavez  MN.  Opioid-related mortality in rural America: geographic heterogeneity and intervention strategies.  Int J Drug Policy. 2018;57:119-129. doi:10.1016/j.drugpo.2018.04.011PubMedGoogle ScholarCrossref
29.
Stein  EM, Gennuso  KP, Ugboaja  DC, Remington  PL.  The epidemic of despair among white Americans: trends in the leading causes of premature death, 1999-2015.  Am J Public Health. 2017;107(10):1541-1547. doi:10.2105/AJPH.2017.303941PubMedGoogle ScholarCrossref
30.
Ruhm  CJ. Deaths of Despair or Drug Problems? NBER Working Paper No. 24188. https://www.nber.org/papers/w24188. Accessed January 13, 2019.
31.
Masters  RK, Tilstra  AM, Simon  DH.  Mortality from suicide, chronic liver disease, and drug poisonings among middle-Aged US white men and women, 1980-2013.  Biodemography Soc Biol. 2017;63(1):31-37. doi:10.1080/19485565.2016.1248892PubMedGoogle ScholarCrossref
32.
Sasson  I.  Reply to trends in education-specific life expectancy, data quality, and shifting education distributions: a note on recent research.  Demography. 2017;54(3):1215-1219. doi:10.1007/s13524-017-0583-1PubMedGoogle ScholarCrossref
33.
Brown  DC, Lariscy  JT, Kalousová  L.  Comparability of mortality estimates from Social Surveys and Vital Statistics data in the United States.  Popul Res Policy Rev. 2018;(December). doi:10.1007/s11113-018-9505-1PubMedGoogle Scholar
34.
Dowd  JB, Hamoudi  A.  Is life expectancy really falling for groups of low socio-economic status? lagged selection bias and artefactual trends in mortality.  Int J Epidemiol. 2014;43(4):983-988. doi:10.1093/ije/dyu120PubMedGoogle ScholarCrossref
35.
Bound  J, Geronimus  AT, Rodriguez  JM, Waidmann  TA.  Measuring recent apparent declines in longevity: the role of increasing educational attainment.  Health Aff (Millwood). 2015;34(12):2167-2173. doi:10.1377/hlthaff.2015.0481PubMedGoogle ScholarCrossref
Original Investigation
August 27, 2019

Association Between Educational Attainment and Causes of Death Among White and Black US Adults, 2010-2017

Author Affiliations
  • 1Department of Sociology and Anthropology and the Herczeg Institute on Aging, Tel Aviv University, Tel Aviv, Israel
  • 2Department of Sociology and Population Research Center, University of Texas at Austin
JAMA. 2019;322(8):756-763. doi:10.1001/jama.2019.11330
Key Points

Question  Has there been an association between educational attainment and causes of death among US adults?

Findings  In this serial cross-sectional study that included 4 690 729 deaths among US adults, estimated life expectancy at age 25 years declined overall between 2010 and 2017; however, it declined among persons without a 4-year college degree and increased among college-educated persons. Much of the increasing educational differences in years of life lost may be related to deaths attributed to drug use.

Meaning  Increasing differences in life expectancy based on educational attainment have occurred between 2010 and 2017.

Abstract

Importance  There are substantial and increasing educational differences in US adult life expectancy. To reduce social inequalities in mortality, it is important to understand how specific causes of death have contributed to increasing educational differences in adult life expectancy in recent years.

Objective  To estimate the relationship of specific causes of death with increasing educational differences in adult life expectancy from 2010 to 2017.

Design, Setting, and Participants  Serial cross-sectional study of 4 690 729 deaths recorded in the US National Vital Statistics System in 2010 and 2017.

Exposures  Sex, race/ethnicity, and educational attainment.

Main Outcomes and Measures  Life expectancy at age 25 years and years of life lost between ages 25 and 84 years by cause of death.

Results  The analysis included a total of 2 211 633 deaths in 2010 and 2 479 096 deaths in 2017. Between 2010 and 2017, life expectancy at age 25 significantly declined among white and black non-Hispanic US residents from an expected age at death of 79.34 to 79.15 years (difference, −0.18 [95% CI, −0.23 to −0.14]). Greater decreases were observed among persons with a high school degree or less (white men: −1.05 years [95% CI, −1.15 to −0.94], white women: −1.14 years [95% CI, −1.24 to −1.04], and black men: −0.30 years [95% CI, −0.56 to −0.04]). White adults with some college education but no 4-year college degree experienced similar declines in life expectancy (men: −0.89 years [95% CI, −1.07 to −0.73], women: −0.59 years [95% CI, −0.77 to −0.42]). In contrast, life expectancy at age 25 significantly increased among the college-educated (white men: 0.58 years [95% CI, 0.42 to 0.73], white women: 0.78 years [95% CI, 0.57 to 1.00], and black women: 1.70 years [95% CI, 0.91 to 2.53]). The difference between high- and low-education groups increased from 2010 to 2017, largely because life-years lost to drug use increased among those with a high school degree or less (white men: 0.93 years [95% CI, 0.90 to 0.96], white women: 0.50 years [95% CI, 0.47 to 0.52], black men: 0.75 years [95% CI, 0.71 to 0.79], and black women: 0.28 years [95% CI, 0.25 to 0.31]).

Conclusions and Relevance  In this serial cross-sectional study, estimated life expectancy at age 25 years declined overall between 2010 and 2017; however, it declined among persons without a 4-year college degree and increased among college-educated persons. Much of the increasing educational differences in years of life lost may be related to deaths attributed to drug use.

Introduction

Life expectancy at birth in the United States decreased by 0.3 years from 2014 to 2017—the largest decline since 1993.1,2 Although originally attributed to rising midlife mortality among white non-Hispanic US residents, decreases in life expectancy have since been observed in other race/ethnic groups.3,4 The rise in mortality has been attributed to deaths due to unhealthy alcohol use, drug use, and suicide, which were not evenly shared across socioeconomic strata.5

Quiz Ref IDSince the late 1980s, differences in mortality rates and adult life expectancy have increased substantially between education and income groups.6-9 By 2010, the gap in life expectancy at age 25 years between high school noncompleters and college-educated US residents had ranged from 4.6 to 11.9 years across major race-sex groups.10 A total of 60% to 80% of the gap was attributed to cardiovascular diseases, smoking-related diseases, and external causes of death.11 Moreover, life expectancy among low-educated white adults declined from 1990 to 2010 predominantly due to external causes, which included (but did not disaggregate) drug use, alcohol use, and suicide. A subsequent study found that by 2011, drug use deaths accounted for 15% to 20% of the educational gradient in adult life expectancy among white non-Hispanic individuals.12 By contrast, educational differences in life expectancy attributed to smoking declined for men (but not women) from 1986 to 2006.13 All of those studies reported persistent and increasing socioeconomic differences in life expectancy, but also pointed to rapid changes in the causes of death underlying those differences.

Using Vital Statistics data, this study estimated the extent to which major causes of death were related to increasing educational differences in adult life expectancy from 2010 to 2017.

Methods
Data Sources

This study analyzed deidentified, government-issued public use data, which according to the Office for Human Research Protections regulations does not require institutional review board approval. All data were obtained from 2 sources: the number of deaths from the US National Vital Statistics System (NVSS) and midyear population estimates from the IPUMS American Community Survey (ACS).14,15 The ACS data, which replaced the long form of the decennial census, are provided by the US Census Bureau after imputing missing information (race: 1.5% in 2010 and 1.6% in 2017; educational attainment: 5.6% in 2010 and 8.5% in 2017).16 Although the true nonsampling error is unknown, the ACS has high total population coverage rates (96.6% in 2010; 91.6% in 2017) and low unit nonresponse rates (2.5% in 2010; 6.3% in 2017).16,17

Exposure

The main variables of interest were race/ethnicity, sex, and educational attainment. These measures were self-reported in the ACS and completed or verified on the death certificate (NVSS) by a funeral director. In both cases, race is recorded in fixed categories with an open-ended “other” category. However, because the ACS allows multiple-race categorization, some categories differ across the 2 data sources. The National Center for Health Statistics provides bridged-race population corrections to adjust for differences in race categorization across the 2 data sources, which were also applied in this study.18 The analysis was limited to white and black non-Hispanic individuals aged 25 years and older because other race/ethnic groups were either too small or less reliably identified in the NVSS.10,19

Quiz Ref IDEducational attainment was recoded into 3 categories based on completed years of schooling or highest degree attained: (1) low, defined as high school graduate or lower; (2) middle, defined as some college education but no 4-year college degree; and (3) high, defined as 4-year college graduate or higher. These education categories were chosen for 2 reasons. First, whereas in the past declines in life expectancy were observed only among those who did not graduate from high school, life expectancy has recently declined nationwide and among wider populations, including high school graduates.7,10,20 Second, using broad categories minimizes the potential misclassification bias that results from minor discrepancies in education reporting across the 2 data sources.21

Outcomes

Quiz Ref IDThe study examined 2 outcome variables: (1) life expectancy at age 25 years and (2) the mean number of years of life lost (YLL) between ages 25 and 84 years by cause of death. The study focused on YLL between ages 25 and 84 years because educational attainment is generally completed by age 25, at least at the college level, and because recorded causes of death among the oldest old are less reliable due to comorbidities.22,23

The underlying cause of death (“disease or injury that initiated the events resulting in death”) was determined by the medical professional who certified the death and recorded it on the death certificate.24 Causes of death followed the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10), which were recoded in this study into 9 categories: circulatory diseases, smoking-related diseases, cancer (excluding smoking-related), diabetes, drug use (ICD-10 codes: F11-F16, F19 Mental and behavioral disorders due to psychoactive substance use; X40-X44 Accidental poisoning by and exposure to drugs; Y10-Y14 Poisoning by and exposure to drugs, undetermined intent), alcohol use (ICD-10 codes: F10, K70, X45, Y15), suicide (ICD-10 codes: X60-X84, Y87.0), firearm deaths (ICD-10 codes: W32-W34, X93-X95, Y22-Y24, Y35.0; excluding intentional self-harm), and all remaining (“other”) causes of death. A complete list of ICD codes is available in eTable 1 in the Supplement.

Smoking-related causes were defined as those in which the smoking-attributable fraction of deaths exceeds 65% in both sexes combined: cancers of the lip, pharynx, oral cavity, esophagus, larynx, trachea, lung, and bronchus, as well as bronchitis, emphysema, and chronic airway obstruction.25 This category was designed to gauge trends in smoking behavior across race, sex, and education groups, rather than accurately capture all deaths due to smoking.

Statistical Analyses

Missing information on educational attainment in the NVSS data was imputed with Stata version 15’s multiple imputation routine (StataCorp), using multinomial regression on age, race, and sex. All subsequent statistical analyses were conducted in R version 3.5.2 (R Project for Statistical Computing). Multiple-decrement period life tables were constructed, in 5-year age groups, by race, sex, and educational attainment for 2010 and 2017. Life expectancy at age 25 years and YLLs between ages 25 and 84 years were estimated from the life tables. YLLs were decomposed by cause of death using a novel decomposition method that has several advantages (eMethods in the Supplement).26 First, it allows a decomposition of the absolute number of YLLs in a single point in time, as well as the change between 2 points in time, by cause of death. Second, it does not assume that causes of death are independent from one another. Third, the number of YLLs to each cause are additive to the total number of years lost to all-cause mortality.

Results
Study Population Characteristics

The study population in 2017 was 48.2% male and 51.8% female; 84.8% of the population was white and 15.2% was black. In the same year, 42.6% of the population consisted of the low-education group (compared with 46.9% in 2010), 23.8% of the middle-education group (compared with 23.5% in 2010), and 33.6% of the high-education group (compared with 29.5% in 2010). The analysis included a total of 2 211 633 deaths in 2010 (50.8% female and 12.2% black US adults) and 2 479 096 deaths in 2017 (49.4% female and 13.0% black US adults). Educational attainment was missing from the NVSS in 2.2% and 1.6% of death certificates in 2010 and 2017, respectively, and was imputed. The numbers of exposures in each of the race, sex, and education groups, as well as number of deaths by cause, are shown in eTable 2 in the Supplement.

Changes in Adult Life Expectancy by Education

The estimated life expectancy at age 25 years significantly declined in the study population between 2010 and 2017, from an expected age at death of 79.34 to 79.15 years (difference, −0.18 [95% CI, −0.23 to −0.14]). However, while life expectancy decreased significantly for white adults—from 77.55 to 77.31 years among men (difference, −0.24 [95% CI, −0.31 to −0.16]) and from 81.93 to 81.82 years among women (difference, −0.11 [95% CI, −0.18 to −0.04])—it had not changed significantly for black men and increased significantly among black women, from 79.23 to 79.86 years (difference, 0.63 [95% CI, 0.39 to 0.86]).

From 2010 to 2017, estimated educational differences in adult life expectancy increased for white and black men and women. In almost all race-sex groups, this was associated with both decreasing life expectancy among persons with less than a 4-year college education and increasing life expectancy among the college educated (Table).

Among white men, the expected age at death—conditional on survival to age 25 years—significantly declined in the low-education group from 74.52 to 73.47 years (difference, −1.05 [95% CI, −1.15 to −0.94]) and in the middle-education group from 77.81 to 76.92 years (difference, −0.89 [95% CI, −1.07 to −0.73]). However, the high-education group had gained 0.58 years (95% CI, 0.42 to 0.73), with the expected age at death increasing from 82.51 to 83.09 years. A similar pattern was observed among white women: the expected age at death decreased from 80.12 to 78.99 years (difference, −1.14 [95% CI, −1.24 to −1.04]) in the low-education group, decreased from 81.82 to 81.23 years (difference, −0.59 [95% CI, −0.77 to −0.42]) in the middle-education group, and increased from 85.47 to 86.25 years (difference, 0.78 [95% CI, 0.57 to 1.00]) in the high-education group.

The expected age at death (at age 25 years) significantly decreased for black men in the low-education group from 71.56 to 71.26 years (difference, −0.30 [95% CI, −0.56 to −0.04]), but had not changed significantly in the other education groups. Among black women, the expected age at death significantly increased in the low-education group by 0.36 years (95% CI, 0.06 to 0.67) from 77.94 to 78.30; however, the high-education group experienced a greater increase of 1.70 years (95% CI, 0.91 to 2.53) from 82.20 to 83.89, the largest increase among all race-sex groups.

Changes in YLLs by Cause of Death

Changes in the mean number of YLLs between ages 25 and 84 years, from 2010 to 2017, were estimated and decomposed into 9 cause-of-death groupings: circulatory diseases, cancer, smoking-related diseases, diabetes, drug use, alcohol use, suicide, firearm, and all other causes. The results are shown by race and sex in Figure 1.

The breakdown of estimated YLLs by cause of death shows that circulatory diseases, cancer, and smoking-related disease were the leading disorders associated with life years lost across all race-sex groups. Modest decreases in YLL to circulatory diseases were observed in all race-sex groups: white men, −0.08 (95% CI, −0.10 to −0.06); white women, −0.03 (95% CI, −0.05 to −0.02); black men, −0.08 (95% CI, −0.16 to −0.02); and black women, −0.15 (95% CI, −0.19 to −0.09). Greater decreases were observed in YLLs to smoking-related diseases, particularly among white men (−0.25 [95% CI, −0.26 to −0.23]) and black men (−0.37 [95% CI, −0.41 to −0.34]).

By contrast, YLLs to drug use, alcohol use, and suicide increased significantly in all 4 race-sex groups. Increases in YLLs to drug use were most pronounced—white men, 0.52 (95% CI, 0.50 to 0.53); white women, 0.21 (95% CI, 0.19 to 0.21); black men, 0.53 (95% CI, 0.50 to 0.55); and black women, 0.16 (95% CI, 0.14 to 0.17)—whereas increases in YLLs to alcohol use were smaller. For example, the largest increases in YLLs to alcohol use and suicide were observed among white men, estimated at 0.08 (95% CI, 0.06 to 0.08) and 0.07 (95% CI, 0.06 to 0.09), respectively.

Changes in YLLs by Cause of Death and Education

Changes in YLLs by cause of death, in each education group, are shown in Figure 2 (white men and women) and Figure 3 (black men and women). Among white adults of both sexes, YLLs to cancer and smoking-related diseases significantly decreased in all education groups, whereas YLLs to drug use increased across both sexes and all education groups. However, increases were larger among the low- and middle-education groups. Among white men, YLLs to drug use had increased by 0.93 (95% CI, 0.90 to 0.96) in the low-education group, by 0.53 (95% CI, 0.50 to 0.55) in the middle-education group, and by 0.11 (95% CI, 0.10 to 0.12) in the high-education group. Among white women, the same increases were estimated at 0.50 (95% CI, 0.47 to 0.52) in the low-education group, 0.24 (95% CI, 0.22 to 0.26) in the middle-education group, and 0.03 (95% CI, 0.02 to 0.04) in the high-education group. Smaller increases in YLLs to alcohol use and suicide were observed across education groups (with the exception of white and black women in the high-education group).

Similar patterns were observed among black adults, with significant decreases in YLLs to smoking-related diseases and, in some education groups, also in circulatory diseases and cancer. However, these decreases in YLLs were offset by increases due to drug use in both men (low education: 0.75 [95% CI, 0.71 to 0.79]; middle education: 0.36 [95% CI, 0.32 to 0.41]; high education: 0.10 [95% CI, 0.07 to 0.12]) and women (low education: 0.28 [95% CI, 0.25 to 0.31]; middle education: 0.12 [95% CI, 0.09 to 0.15]). Also noteworthy were YLLs to firearm deaths among low-educated black men, which significantly increased by 0.35 (95% CI, 0.30 to 0.41) between 2010 and 2017. Estimates for all causes of death, by race, sex, and education, are available in eTables 3 through 6 in the Supplement.

Educational differences in YLLs increased from 2010 to 2017 in all race-sex groups, with drug use deaths associated with the largest contribution to the increase (eTable 7 in the Supplement). Among white men, the difference in YLLs to all causes of death between the low- and high-education groups had increased by 1.40 years (95% CI, 1.31 to 1.50) from 2010 to 2017, of which 0.82 years (95% CI, 0.79 to 0.86) were attributed to drug use. Among white women, the same difference increased by 1.32 years (95% CI, 1.23 to 1.40), of which 0.47 years (95% CI, 0.44 to 0.50) were due to drug use. Among black men, the difference increased by 1.18 years (95% CI, 0.82 to 1.55), of which 0.65 years (95% CI, 0.61 to 0.70) were due to drug use. Among black women, the difference increased by 0.65 years (95% CI, 0.38 to 0.91), with 0.27 years (95% CI, 0.23 to 0.31) attributed to drug use.

Discussion

Quiz Ref IDThis study documented an overall decline in life expectancy at age 25 years between 2010 and 2017. Although adult life expectancy increased among college-educated persons, it was counterbalanced by a decline in life expectancy among persons without a 4-year college degree. An analysis of the causes of death for 2010-2017 suggests that much of the increasing educational differences in YLLs may be related to deaths attributed to drug use, particularly among white men and women. These findings are consistent with other research that has shown that drug use significantly contributed to increasing educational differences in adult life expectancy in the 1990s up to 2011.12 In addition, firearm deaths contributed to increased educational differences in life expectancy among black men.

The magnitude of the recent changes in educational differences in mortality shown here appears large given the brevity of the period. Inequality in the mean number of life years lost between ages 25 and 84 years increased substantially for white and black US residents over the 7-year study period. Among white men and women, there was increasing inequality largely related to rising mortality among all education groups without a 4-year college education, but especially among persons with a high school degree or less, while the college-educated group continued to improve. The finding that improving life expectancy was largely restricted to only well-educated US residents in the most recent period supports an increasing role of inequality in the national trend in life expectancy. Nevertheless, drug use mortality increased over the 7-year period even among the college educated.

Changes in adult life expectancy from 2010 to 2017 should be interpreted in the context of overall trends in longevity since 1990. From 1990 to 2010, life expectancy continued to increase among white adults despite increasing educational differences.10 Since 2010, however, adult life expectancy declined among white men and women, largely due to the decline in adult life expectancy among all non–college-educated groups. Among black adults, life expectancy improved despite increasing educational differences associated with the greater improvement in life expectancy among college-educated men and women. Nevertheless, increases in life expectancy among black adults were substantially reduced in the 2010-2017 period. Particularly among black men, the stagnation in adult life expectancy during those 7 years ended a 2-decade long improvement in life expectancy for this group.

Quiz Ref IDThe nature of the recent declines in adult life expectancy among white US residents and the increase in educational differences across all race-sex groups remains unclear. Prior research attributed the rise in mortality to “deaths of despair,” which consist of drug use, alcohol use, and suicide.5 These deaths, it was argued, collectively reflect self-destructive behaviors linked to the growing social isolation, economic distress, and downward social mobility that have characterized large segments of the US working class.27-29 However, deteriorating economic conditions at the county level appear to explain less than one-tenth of the rise in drug- and opioid-related deaths.30 Furthermore, consistent with prior research, this study finds that increases in life years lost to drug use were substantially larger than increases associated with alcohol use and suicide.31

Limitations

This study has several limitations. First, it relied on unlinked ACS and NVSS data to estimate educational differences in adult mortality. Unlinked data are less accurate than self-reported survey data and are prone to misclassification bias.21 However, general population surveys are insufficient in studying trends in cause-specific mortality and have been shown to overstate life expectancy, irrespective of education, because their populations are healthier than the general US population.32,33 By contrast, the NVSS records all deaths in the United States and allows the study of cause-specific mortality, which is important in understanding the contributing factors for national trends.

Second, changes in the proportions of the population in various education strata can introduce selection effects and exaggerate differences in mortality.34 If the low-education group diminished in size over time, it may consist of individuals who are more disadvantaged—for example, with respect to early-life social conditions—and their health could decline irrespective of educational attainment later in life. However, because this study focused on a short period of time in which the change in educational composition was small, selection effects would be minimal. Previous studies have shown that declines in life expectancy among low-education groups persisted even after adjusting for compositional changes.35

Third, the ACS data are subject to both sampling and nonsampling errors, as are all probability samples. Although the accuracy of the data is unknown, the ACS has high population coverage rates and low nonresponse rates, which likely minimize the potential error.17

Conclusions

In this serial cross-sectional study, estimated life expectancy at age 25 years declined overall between 2010 and 2017; however, it declined among persons without a 4-year college degree and increased among college-educated persons. Much of the increasing educational differences in years of life lost may be related to deaths attributed to drug use.

Back to top
Article Information

Corresponding Author: Isaac Sasson, PhD, Department of Sociology and Anthropology and the Herczeg Institute on Aging, Tel Aviv University, Tel Aviv 6997801, Israel (isasson@tauex.tau.ac.il).

Accepted for Publication: July 18, 2019.

Author Contributions: Dr Sasson had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Both authors.

Acquisition, analysis, or interpretation of data: Both authors.

Drafting of the manuscript: Sasson.

Critical revision of the manuscript for important intellectual content: Hayward.

Statistical analysis: Sasson.

Obtained funding: Sasson.

Conflict of Interest Disclosures: None reported.

Funding/Support: This research was supported by a grant from the German-Israeli Foundation for Scientific Research and Development (GIF) awarded to Dr Sasson and an infrastructure grant from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD; P2C HD042849) awarded to the Population Research Center at The University of Texas at Austin.

Role of the Funder/Sponsor: The GIF and the NICHD had no role in design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

References
1.
Centers for Disease Control and Prevention. NCHS: death rates and life expectancy at birth. https://data.cdc.gov/NCHS/NCHS-Death-rates-and-life-expectancy-at-birth/w9j2-ggv5. Updated June 4, 2018. Accessed June 13, 2019.
2.
Murphy  SL, Xu  J, Kochanek  KD, Arias  E. Mortality in the United States, 2017. NCHS Data Brief No. 328. National Center for Health Statistics. https://www.cdc.gov/nchs/data/databriefs/db328-h.pdf. Published November 2018. Accessed January 6, 2019.
3.
Case  A, Deaton  A.  Rising morbidity and mortality in midlife among white non-Hispanic Americans in the 21st century.  Proc Natl Acad Sci U S A. 2015;112(49):15078-15083. doi:10.1073/pnas.1518393112PubMedGoogle ScholarCrossref
4.
Xu  J, Murphy  SL, Kochanek  KD, Bastian  B, Arias  E.  Deaths: final data for 2016.  Natl Vital Stat Rep. 2018;67(5):1-76.PubMedGoogle Scholar
5.
Case  A, Deaton  A.  Mortality and morbidity in the 21st century.  Brookings Pap Econ Act. 2017;2017:397-476. doi:10.1353/eca.2017.0005PubMedGoogle ScholarCrossref
6.
Meara  ER, Richards  S, Cutler  DM.  The gap gets bigger: changes in mortality and life expectancy, by education, 1981-2000.  Health Aff (Millwood). 2008;27(2):350-360. doi:10.1377/hlthaff.27.2.350PubMedGoogle ScholarCrossref
7.
Olshansky  SJ, Antonucci  T, Berkman  L,  et al.  Differences in life expectancy due to race and educational differences are widening, and many may not catch up.  Health Aff (Millwood). 2012;31(8):1803-1813. doi:10.1377/hlthaff.2011.0746PubMedGoogle ScholarCrossref
8.
Montez  JK, Hummer  RA, Hayward  MD, Woo  H, Rogers  RG.  Trends in the educational gradient of US adult mortality from 1986 to 2006 by race, gender, and age group.  Res Aging. 2011;33(2):145-171. doi:10.1177/0164027510392388PubMedGoogle ScholarCrossref
9.
Chetty  R, Stepner  M, Abraham  S,  et al.  The association between income and life expectancy in the United States, 2001-2014.  JAMA. 2016;315(16):1750-1766. doi:10.1001/jama.2016.4226PubMedGoogle ScholarCrossref
10.
Sasson  I.  Trends in life expectancy and lifespan variation by educational attainment: United States, 1990-2010.  Demography. 2016;53(2):269-293. doi:10.1007/s13524-015-0453-7PubMedGoogle ScholarCrossref
11.
Sasson  I.  Diverging trends in cause-specific mortality and life years lost by educational attainment: evidence from United States Vital Statistics data, 1990-2010.  PLoS One. 2016;11(10):e0163412. doi:10.1371/journal.pone.0163412PubMedGoogle ScholarCrossref
12.
Ho  JY.  The contribution of drug overdose to educational gradients in life expectancy in the United States, 1992-2011.  Demography. 2017;54(3):1175-1202. doi:10.1007/s13524-017-0565-3PubMedGoogle ScholarCrossref
13.
Ho  JY, Fenelon  A.  The contribution of smoking to educational gradients in US life expectancy.  J Health Soc Behav. 2015;56(3):307-322. doi:10.1177/0022146515592731PubMedGoogle ScholarCrossref
14.
Ruggles  S, Flood  S, Goeken  R,  et al.  IPUMS USA: Version 8.0. Minneapolis, MN: IPUMS; 2018. doi:10.18128/D010.V8.0
15.
Centers for Disease Control and Prevention. Vital Statistics online data portal. https://www.cdc.gov/nchs/data_access/vitalstatsonline.htm. Accessed January 14, 2019.
16.
US Census Bureau. American Community Survey: sample size and data quality. https://www.census.gov/acs/www/methodology/sample-size-and-data-quality/. Accessed July 13, 2019.
17.
US Census Bureau. American Community Survey accuracy of the data (2017). https://www2.census.gov/programs-surveys/acs/tech_docs/accuracy/ACS_Accuracy_of_Data_2017.pdf. Accessed July 13, 2019.
18.
Centers for Disease Control and Prevention. US Census populations with bridged race categories. https://www.cdc.gov/nchs/nvss/bridged_race.htm. Accessed April 19, 2019.
19.
Arias  E, Schauman  WS, Eschbach  K, Sorlie  PD, Backlund  E.  The validity of race and Hispanic origin reporting on death certificates in the United States.  Vital Health Stat 2. 2008;(148):1-23.PubMedGoogle Scholar
20.
Xu  J, Murphy  SL, Kochanek  KD, Arias  E.  Mortality in the United States, 2015. Hyattsville, MD: National Center for Health Statistics; 2016.
21.
Rostron  BL, Boies  JL, Arias  E.  Education reporting and classification on death certificates in the United States.  Vital Health Stat 2. 2010;(151):1-21.PubMedGoogle Scholar
22.
Alpérovitch  A, Bertrand  M, Jougla  E,  et al.  Do we really know the cause of death of the very old? comparison between official mortality statistics and cohort study classification.  Eur J Epidemiol. 2009;24(11):669-675. doi:10.1007/s10654-009-9383-2PubMedGoogle ScholarCrossref
23.
Tinetti  ME, McAvay  GJ, Murphy  TE, Gross  CP, Lin  H, Allore  HG.  Contribution of individual diseases to death in older adults with multiple diseases.  J Am Geriatr Soc. 2012;60(8):1448-1456. doi:10.1111/j.1532-5415.2012.04077.xPubMedGoogle ScholarCrossref
24.
Centers for Disease Control and Prevention. US standard certificate of death. https://www.cdc.gov/nchs/data/dvs/death11-03final-acc.pdf. Accessed April 19, 2019.
25.
Centers for Disease Control and Prevention (CDC).  Smoking-attributable mortality, years of potential life lost, and productivity losses: United States, 2000-2004.  MMWR Morb Mortal Wkly Rep. 2008;57(45):1226-1228.PubMedGoogle Scholar
26.
Andersen  PK, Canudas-Romo  V, Keiding  N.  Cause-specific measures of life years lost.  Demogr Res. 2013;29:1127-1152. doi:10.4054/DemRes.2013.29.41Google ScholarCrossref
27.
Diez Roux  AV.  Despair as a cause of death: more complex than it first appears.  Am J Public Health. 2017;107(10):1566-1567. doi:10.2105/AJPH.2017.304041PubMedGoogle ScholarCrossref
28.
Rigg  KK, Monnat  SM, Chavez  MN.  Opioid-related mortality in rural America: geographic heterogeneity and intervention strategies.  Int J Drug Policy. 2018;57:119-129. doi:10.1016/j.drugpo.2018.04.011PubMedGoogle ScholarCrossref
29.
Stein  EM, Gennuso  KP, Ugboaja  DC, Remington  PL.  The epidemic of despair among white Americans: trends in the leading causes of premature death, 1999-2015.  Am J Public Health. 2017;107(10):1541-1547. doi:10.2105/AJPH.2017.303941PubMedGoogle ScholarCrossref
30.
Ruhm  CJ. Deaths of Despair or Drug Problems? NBER Working Paper No. 24188. https://www.nber.org/papers/w24188. Accessed January 13, 2019.
31.
Masters  RK, Tilstra  AM, Simon  DH.  Mortality from suicide, chronic liver disease, and drug poisonings among middle-Aged US white men and women, 1980-2013.  Biodemography Soc Biol. 2017;63(1):31-37. doi:10.1080/19485565.2016.1248892PubMedGoogle ScholarCrossref
32.
Sasson  I.  Reply to trends in education-specific life expectancy, data quality, and shifting education distributions: a note on recent research.  Demography. 2017;54(3):1215-1219. doi:10.1007/s13524-017-0583-1PubMedGoogle ScholarCrossref
33.
Brown  DC, Lariscy  JT, Kalousová  L.  Comparability of mortality estimates from Social Surveys and Vital Statistics data in the United States.  Popul Res Policy Rev. 2018;(December). doi:10.1007/s11113-018-9505-1PubMedGoogle Scholar
34.
Dowd  JB, Hamoudi  A.  Is life expectancy really falling for groups of low socio-economic status? lagged selection bias and artefactual trends in mortality.  Int J Epidemiol. 2014;43(4):983-988. doi:10.1093/ije/dyu120PubMedGoogle ScholarCrossref
35.
Bound  J, Geronimus  AT, Rodriguez  JM, Waidmann  TA.  Measuring recent apparent declines in longevity: the role of increasing educational attainment.  Health Aff (Millwood). 2015;34(12):2167-2173. doi:10.1377/hlthaff.2015.0481PubMedGoogle ScholarCrossref
×