Housing and Inequalities in US Life Expectancy, Child and Youth Mortality, and All-Cause and Cause-Specific Mortality, 1979-2020: Results from the National Longitudinal Mortality Study and the National Vital Statistics System

2.1. The National Longitudinal Mortality Study (NLMS)

To examine the association of housing tenure with all-cause and cause-specific mortality, we used the 1979-2011 NLMS data. The NLMS is a longitudinal dataset for examining socioeconomic, occupational, and demographic factors associated with all-cause and cause-specific mortality in the US.21,22 The NLMS was conducted by the National Heart, Lung, and Blood Institute in collaboration with the US Census Bureau, the National Cancer Institute, the National Institute on Aging, and the National Center for Health Statistics.21,22 The NLMS consisted of 39 Current Population Survey (CPS) and census cohorts between 1973 and 2011 whose survival (mortality) experiences were studied from 1979 to 2011.21 The CPS is a sample household and telephone interview survey of the civilian non-institutionalized population in the US and is conducted by the US Census Bureau to produce monthly national statistics on unemployment and the labor force. Data from death certificates on the fact of death and the cause of death were combined with the socioeconomic and demographic characteristics of the NLMS cohorts by means of the National Death Index.21,22 Detailed descriptions of the NLMS have been provided elsewhere.21,22

We used the public-use NLMS file to derive cohort-based mortality risks during 1979-2011, with a maximum mortality follow-up of 11 years or 4,018 days.21 Child (<18 years) mortality, youth (15-24) mortality, and all-cause and cause-specific mortality risks and life expectancy at birth were the outcome variables in the NLMS. Housing tenure (home ownership vs. renting) was the main independent variable. Differentials in mortality risks by housing tenure were adjusted by multivariate Cox proportional hazards regression for a number of sociodemographic covariates, such as age (continuous), sex (male, female), race/ethnicity (non-Hispanic White, non-Hispanic Black, Asian/Pacific Islander, Hispanic, other) marital status (currently married, widowed, divorced/separated, never married), nativity/immigrant status (US-born, foreign-born), rural-urban residence, educational attainment (<12, 12, 13-15, 16 years), occupation (professional and managerial; sales, clerical and technical support; service; craft and repair; laborer; military; unemployed and not in labor force), and income/poverty level (<100% of poverty level, 100%-149%,150-199%, 200-299%, 300-399%, 400-599%, and at or above 600% of poverty level).16,17,22

The 1979-2011 sample included 1,313,627 individuals aged ≥20 years at the baseline and 156,617 deaths during the 11-year mortality follow-up. In estimating the mortality risk, all those surviving beyond the 11-year follow-up (measured in days) during the follow-up period were treated as right-censored observations. Relative risks of mortality were estimated by hazard ratios derived from the Cox regression models.17 In the NLMS, age-adjusted mortality rates and life expectancy estimates by housing tenure were computed using the person-years approach.16 Follow-up of mortality and survivorship in the NLMS are not available beyond 2011.

2.2. Linkage of the National Vital Statistics System with the American Community Survey

To supplement the individual-level analysis, we conducted area-level analysis of 2011-2020 pooled mortality data by two housing variables, household crowding and housing stability. Specifically, we linked age, sex, race/ethnicity, and cause-specific mortality data from the 2011-2020 NVSS to the selected housing data from the 2015-2019 ACS at the county level using the common geocodes in the two databases.23–25 Household crowding was measured by the percentage of households with more than 1 person per room and grouped into 3 categories: counties with crowding <1.62% (low crowding), 1.62% to 4.84%, and ≥4.85% (high crowding). Housing stability was measured by the percentage of people living in the same house during the past year and was grouped into 3 categories: counties with housing stability <80% (low stability), 80% to 89.99%, and 90% to 100% (high stability).

We used county-level 2013 rural-urban continuum codes linked to the national mortality data to analyze metropolitan-nonmetropolitan patterns in mortality by household crowding and housing stability.3,26 Mortality rates were age-adjusted by the direct method using the age composition of the 2000 US population as the standard.

3.1. Housing Tenure, Life Expectancy, Child and Youth Mortality, and Cause-Specific Mortality: The National Longitudinal Mortality Study

Figure 1 shows life expectancy at birth by housing tenure, sex, race/ethnicity, nativity/immigrant status, and rural-urban residence, based on cohort life tables. During 1979-2011, homeowners had a life expectancy of 74.2 years, about 3.5 years longer than the life expectancy of 70.8 years for renters. The difference in life expectancy between homeowners and renters was larger for males than females (4.0 vs. 3.6 years). The difference in life expectancy between homeowners and renters was largest for American Indians/Alaska Natives (AIANs) (3.9 years), non-Hispanic Whites (3.8 years), and non-Hispanic Blacks (3.6 years) and smallest for Asian/Pacific Islanders (APIs) (2.5 years) and Hispanics (0.3 years). US-born homeowners had 4.4 years longer life expectancy than US-born renters. The difference in life expectancy between homeowners and renters was only 0.7 years among the foreign-born. The difference in life expectancy between homeowners and renters was 3.6 years among urban residents and 3.7 years among rural residents.

Close

Figure 1

Life Expectancy at Birth (Years) by Housing Tenure, Sex, Race/Ethnicity, Nativity/Immigrant Status, and Rural-Urban Residence, United States, 1979-2011, The National Longitudinal Mortality Study

Source: Data derived from the 1979-2011 National Longitudinal Mortality Study (NLMS). The life expectancy estimates roughly correspond to the midpoint of the study time period, i.e., 1995

Export to PPT

Homeownership is associated with reduced mortality risks regardless of education or income level (Figure 2a,b). During 1979-2011, education and income levels were strongly and consistently associated with reduced all-cause mortality rates for both homeowners and renters. Within each level of education and income, homeowners had significantly lower mortality rates than renters (Figure 2a,b).

Close

Figure 2

(a, b) Age-Sex-Adjusted All-Cause Mortality Rates by Housing Tenure, Educational Attainment, and Family Income/Poverty Level, US Population Aged 20+ Years, 1979-2011

Source: Data derived from the 1979-2011 National Longitudinal Mortality Study (NLMS). Differences in all-cause mortality rates by housing tenure and education and by housing tenure and family income were statistically significant at p<.01

Export to PPT

Table 1 presents the results of the age-adjusted and covariate-adjusted Cox models, showing all-cause and cause-specific mortality risks associated with housing tenure. Overall, renters had 28% higher all-cause mortality than homeowners, with the risk of mortality being 31% higher for male renters and 25% higher for female renters. After controlling for socioeconomic and demographic factors, the relative risk of mortality was reduced but was still 13% higher for renters than homeowners. The reduced risk of mortality associated with homeownership was largest for AIANs and non-Hispanic Whites. AIAN renters had a 40% higher age-adjusted risk and a 30% higher covariate-adjusted risk of mortality than AIAN homeowners.

Children aged <18 years living in rented homes had 18% higher mortality than children in owner-occupied homes, with differences in racial/ethnic composition, household income, and rural-urban residence accounting for the excess mortality. Similar patterns existed in youth mortality; male and female renters aged 15-24 had a 25% higher mortality risk than homeowners. Controlling for covariates accounted for differences in youth mortality (Table 1).

Homeownership was associated with a reduced risk of cardiovascular disease (CVD) mortality, with the impact of homeownership on reduced stroke mortality being significantly greater for men than for women (27% vs. 11%). Homeownership was associated with lower cancer mortality, and the association was particularly strong for cervical cancer mortality. Female renters had an 83% higher age-adjusted risk and 31% higher covariate-adjusted risk of cervical cancer mortality than female homeowners. Renters had higher mortality from stomach, liver, and esophageal cancer than homeowners, reflecting their higher risks of Helicobacter pylori infection, alcohol consumption, hepatitis infection, and smoking.27

Compared with homeowners, renters had 36-64% higher observed risks of mortality from diabetes, pneumonia and influenza, chronic obstructive pulmonary disease (COPD), cirrhosis, and kidney disease. Renters were at higher risk of infectious disease mortality; in age-sex-adjusted models, renters had 2.9 times higher risk of HIV/AIDS mortality, 3.0 times higher risk of tuberculosis mortality, and 1.6 times higher risk of sepsis mortality than homeowners. In age-sex-adjusted models, renters had 41% higher unintentional injury mortality, 27% higher suicide mortality, and 97% higher risk of homicide victimization, compared with homeowners (Table 1).

3.2. County-Level Associations between Household Crowding and Cause-Specific Mortality Rates

Table 2 shows age-adjusted all-cause and cause-specific mortality rates at the county level by household crowding (percentage of households with more than one person per room) during 2011 to 2020. Higher levels of household crowding in racial/ethnic minority communities were associated with higher all-cause mortality rates. For example, APIs and Hispanics in communities with higher levels of household crowding had 15 to 17% higher mortality rates than their counterparts in communities with lower levels of crowding. Similar relationships between household crowding and county-level mortality existed, particularly for AI/AN and non-metropolitan children and youth. AI/AN children and youth living in counties with high levels of household crowding had, respectively, 41% and 50% higher mortality rates than their counterparts living in counties with low levels of crowding. Children and youth living in non-metropolitan counties with high levels of household crowding had, respectively, 17% and 37% higher mortality rates than their counterparts living in counties with low levels of crowding.

The higher the level of household crowding, the higher the rate of mortality from hypertension. For example, non-Hispanic Whites living in areas with high levels of household crowding were 40% more likely to die from hypertension than Whites in areas with low levels of crowding. Consistent gradients in mortality by household crowding were found for several prominent causes of death, including cirrhosis, liver and stomach cancer, influenza and pneumonia, tuberculosis (TB), HIV/AIDS, and Alzheimer’s disease, with the magnitude of the association varying by race/ethnicity. Compared to those living in areas with low household crowding, those living in areas with high levels of crowding had a 15% higher risk of cirrhosis mortality, 32% higher risk of liver cancer mortality, 65% higher stomach cancer mortality, and 7% higher influenza and pneumonia mortality. Those living in areas with high levels of household crowding were, respectively, 2.4 and 2.9 times more likely to die from HIV/AIDS and TB than those living in areas with low crowding.

3.3. County-Level Associations between Housing Stability and Cause-Specific Mortality

Table 3 shows age-adjusted cause-specific mortality rates from 2011 to 2020 at the county level by housing stability levels (defined as the percentage of those living in the same house during the past year) stratified by race/ethnicity and metropolitan area. Low housing stability was associated with increased mortality from all causes combined and child mortality in rural, non-metropolitan areas, and from stroke, COPD, liver cirrhosis, suicide, and Alzheimer’s disease. Those living in areas of low housing stability had a 15% higher risk of stroke mortality than those living in areas with high housing stability. The stroke mortality risk associated with low housing stability increased by 34% for Blacks and 56% for APIs. The risk of cirrhosis mortality and suicide in the Hispanic population increased by 48% and 64% in areas with low housing stability, respectively.

4.1. Limitations

Our study lacked data on psychosocial factors (social integration), physical and mental health status at baseline, health behaviors (diet, smoking, drinking, obesity, and physical inactivity), neighborhood factors including housing conditions, and healthcare access that could partly explain the association between housing tenure and cause-specific mortality. Secondly, we considered only a dichotomous measure of housing tenure (homeowners vs. renters) due to data availability. A multifaceted nature of housing tenure that differentiates renters who receive rental or housing assistance from a government program from those who do not receive any rental assistance would have provided additional insight into why renters experience a higher risk of mortality than homeowners.31,36,37 Compared with homeowners and non-rental assistance renters, public housing residents and rental assistance renters have higher levels of social isolation, unemployment, lower education and incomes, and higher rates of obesity, smoking, secondhand smoke exposure at home, asthma, poorer health status, depressive symptoms, and hypertension.31,32,36,38 Thirdly, NLMS did not have mortality/survivorship follow-up data after 2011. Having more recent follow-up data would have allowed us to examine whether disparities in life expectancy and mortality between homeowners and renters have widened over time. Finally, additional housing variables such as household crowding, housing quality, housing cost burden, and housing stability, which are lacking in NLMS, are needed at the individual or household level that can allow for a more thorough investigation of housing and inequalities in life expectancy, mortality, and other health outcomes.