Inequalities in Life Expectancy and All-Cause Mortality in the United States by Levels of Happiness and Life Satisfaction: A Longitudinal Study

2.1. Data

The data source for this study was the 2001 National Health Interview Survey (NHIS) prospectively linked to the National Death Index (NDI).13 As a nationally representative, cross-sectional household interview survey, NHIS provides demographic, socio-economic, and health characteristics for the civilian, non-institutionalized population of the US. The National Center for Health Statistics (NCHS) developed public-use versions of NHIS linked with death certificate records from the NDI. We used the public-use linked mortality file, including the mortality follow-up data from the date of survey participation in 2001 through December 31, 2014.14

2.2. Sample

The study sample was restricted to adults aged 18 years and older in the 2001 NHIS sample adult file. The sample ineligible for mortality follow-up was eliminated from the analysis. The final pooled eligible sample size without missing values was 30,377. Missing data in covariates, ranging from 0 to 448, were excluded from the analysis but used for weights, using the Stata svy, subpop option. For missing values for poverty status (21.67%), alcohol consumption (1.63%), and social support (1.76%), we created missing covariate categories to prevent omission of many observations from the analysis.

2.3. Measurement

We examined two outcomes, life expectancy and all-cause mortality. Follow-up time for individuals who died was defined by the number of months from the month/year of interview to the month/year of death. Since NHIS-NDI provides only the quarter of death, we assumed that death occurred in the middle month of the quarter, February, May, August, or November.15

2.4. Happiness and Life Satisfaction

The independent variables were happiness and life satisfaction. The level of happiness was measured by responses to the question, “During the past 30 days, how often did you feel happy?” Happiness was categorized by none or a little of the time, some of the time, most of the time, and all of the time. Life satisfaction was measured by responses to the question, “In general, how satisfied are you with your life?” Life satisfaction was categorized by very dissatisfied or dissatisfied, satisfied, and very satisfied.

2.5. Covariates

Based on the previous literature, we selected the following covariates in adjusted models: age, gender, race/ethnicity, nativity/immigrant status, marital status, social support, education, poverty status, homeownership, region, activity limitation, body mass index (BMI), smoking status, and alcohol consumption.3

2.6.Analytic Approach

We computed age-adjusted mortality rates per 100,000 person-years of exposure by happiness and life satisfaction level and other covariates. In computing person-years for age-specific and age-adjusted mortality rates, deaths were assumed to occur uniformly throughout the follow-up period rather than being clustered in the first part of the period. We used the direct standardization method for age adjustment using the age distribution of the 2000 US Standard Population. Rate ratios and 95% confidence intervals (CIs) were used to compare mortality rates for categorical variables. We used the standard life table methodology to calculate life expectancy at age 18 and subsequent ages along with standard errors.16 Specifically, life expectancy was estimated by converting age-specific mortality rates (for 14 age groups: 18-24, 25-29, 30-34, 35-39, 40-44, 45-49, 50-54, 55-59, 60-64, 65-69, 70-74, 75-79, 80-84, and ≥85 years) into life-table probabilities of dying. We tested for differences in life expectancy estimates by happiness/life satisfaction levels across gender and age groups.

Cox proportional hazards regression models were used to derive relative risks of mortality, controlling for individual characteristics. The model assumes that hazard rates are a log-linear function of parameter for the effects of covariates.17,18 Individuals surviving beyond the follow-up period were treated as right-censored observations.

Complex survey design procedures were used to account for clustering, multiple stages of selection, and disproportionate sampling. To correct the bias from the ineligible adults for linkage to the NDI due to insufficient identifying data, we used eligibility-adjusted weights developed by NCHS, instead of the standard sample weight.19 All analyses were conducted by Stata 15.20

3.1. Descriptive Statistics

Table 1 provides descriptive statistics of individual characteristics for adults aged ≥18 years. Approximately 4.4% of the sample responded that they felt happy none or a little of the time during the past 30 days. About 5.9% of the sample responded that they were very dissatisfied or dissatisfied with their life in general. The mean age was 44.9 years. The study sample consisted of more females (52.1%), non-Hispanic Whites (73.9%), US-born adults (86.4%), high school graduates (31.6%), currently married (58.5%), adults with incomes at or above 400% of the poverty threshold (31.9%), homeowners (70.5%), Southern residents (37.2%), adults without activity limitation (85.2%), adults with BMI ≤ 25 (40.4%), non-smokers (54.6%), and current drinkers (62.8%).

3.2. Estimated Life Expectancies at Selected Ages at Baseline

Table 2 presents estimated life expectancies for both male and female adults. For adults at age 18, the estimated life expectancy was 54.10 years. The estimated life expectancies for adults with lower levels of happiness or life satisfaction were shorter than those for adults with higher levels of happiness or life satisfaction, regardless of age or gender. At age 18, adults reporting happiness none or a little of the time had 7.45 years shorter life expectancy, compared with adults reporting happiness all of the time. Adults experiencing happiness some of the time had 2.48 years shorter life expectancy at age 18, compared with adults reporting happiness all of the time. Adults who were happy none or a little of the time had 6.44 and 3.71 years shorter life expectancy at ages 45 and 65, respectively, compared with adults who were happy all of the time. Both males and females who were happy none or a little of the time had substantially lower life expectancy than those who were happy all of the time.

Regarding life satisfaction, we found that at age 18, adults who were very dissatisfied or dissatisfied with their life in general had 8.9 years shorter life expectancy than those who were very satisfied. Adults who were satisfied with their life had 2.82 years shorter life expectancy than those who were very satisfied with their life. Adults at ages 45 and 65 who were very dissatisfied or dissatisfied had shorter life expectancy by 7.19 and 3.59 years, respectively, compared with adults who were very satisfied with their life. Both males and females who were very dissatisfied or dissatisfied had lower life expectancies than those who were very satisfied with their life.

Patterns in life expectancy associated with happiness and life satisfaction differed between men and women (Table 2). The impact of happiness on life expectancy was greater for women than men, particularly at older ages. At age 45, women being happy none or a little of the time had 7 year shorter life expectancy than women who were happy all of the time, while men being happy none or a little of the time had 6 years shorter life expectances than men who were happy all of the time. However, life satisfaction had a stronger impact on life expectancy of men than that of women. Compared with their counterparts who reported high life satisfaction, men with low life satisfaction had 10 years shorter life expectancy, whereas women with low life satisfaction had 8 years shorter life expectancy.

Figure 1 provides age-specific life expectancies by levels of happiness and life satisfaction. Consistent with the findings above, these curves indicate higher life expectancies for adults with lower levels of happiness and life satisfaction at each age. For example, at age 65, the difference between the highest and lowest levels of happiness was 3.7 years (16.4 vs. 12.7 years); the difference between the highest and lowest levels of life satisfaction was 3.6 years (16.2 vs. 12.6 years).

Close

Figure 1

(a, b). Age-specific life expectancies by levels of happiness and life satisfaction, United States, 2001-2014

Export to PPT

3.3. Cox Proportional Hazards Models

Table 3 provides age-adjusted mortality rates per 100,000 person-years by happiness and life satisfaction categories, and age-adjusted and covariate-adjusted hazard ratios (HRs). Adults with lower levels of happiness and life satisfaction had higher age-adjusted mortality rates. The age-adjusted mortality rate for adults with “none or a little of the time” happiness was 2,367.98 per 100,000 person-years, 53% higher than the mortality rate of 1,550.22 for adults with “all of the time” happiness. In Cox regression models, controlling for age, the mortality risk was 82% higher (HR=1.82; 95% CI: 1.59,2.08) in adults with “none or a little of the time” happiness, compared with adult with “all of the time” happiness. After controlling for socioeconomic and demographic covariates, the mortality risk was 71% higher (HR=1.71; 95% CI=1.50,1.96; data not shown) in adults with “none or a little of the time” happiness, compared to adults with “all of the time” happiness. After controlling for all covariates, the mortality risk was 36% higher (HR=1.36; 95% CI=1.17,1.57) in adults with “none or a little of the time” happiness, compared to those with “all of the time” happiness.

The age-adjusted mortality rate for adults who were very dissatisfied or dissatisfied with their life was 2,336.71 per 100,000 person-years, 62% higher than the rate of 1,446.59 for adults who were very satisfied (Table 3). In Cox models, controlling for age, the mortality risk was 107% higher (HR=2.07; 95% CI=1.80,2.38) in adults who were very dissatisfied or dissatisfied with their life, compared with adults who were very satisfied. After controlling for socioeconomic and demographic covariates, the mortality risk was 71% higher (HR=1.71; 95% CI=1.50,1.95; data not shown) in adults who were very dissatisfied or dissatisfied, compared with adults who were very satisfied. After controlling for all covariates, the mortality risk was 39% higher (HR=1.39; 95% CI=1.20,1.60) in adults who were very dissatisfied or dissatisfied with their life, compared with adults who were very satisfied.

4.1. Limitations

There are several limitations in this study. First, our study only contains NHIS sample eligible for linkage to the NDI. Excluding sample ineligible for linkage may lead to biased mortality estimates. To address this bias, we used the adjusted original sampling weight to account for the NHIS-NDI mismatches.19 Second, there could be omitted-variable bias. While our Cox regression models controlled for self-reported health status including BMI and activity limitation, there could be potential confounders such as neighborhood environments or periods of economic downturn. Third, our measures of happiness and life satisfaction might not capture all aspects of subjective well-being. In further studies, the realization of personal potential and the fulfillment of life goals2 or combined scale of multiple items22, or specified measures of happiness in such domains as marriage/family, job, and financial situation23 could be used as an alternative measures. Fourth, since the NHIS excludes the institutionalized population, who may have lower levels of happiness/life satisfaction and higher mortality, mortality risks associated with happiness/life satisfaction in this study might be underestimated. Fifth, all the covariates in the NHIS-NDI database were time-fixed at the baseline as of the survey date. Several of the covariates such as socioeconomic status, health status, behavioral risk factors, and happiness/life satisfaction could have varied over the long mortality follow-up period of 14 years, which would have influenced their estimated impact on all-cause mortality. Further studies are needed to evaluate the temporal robustness of happiness/life satisfaction patterns in all-cause mortality using longitudinal datasets with time-varying covariates.