Paradox Between Immigrant Advantages in Morbidity and Mortality: Dynamic Patterns and Tentative Explanations

Abstract

Recent research indicates that immigrants are more likely to experience chronic conditions and disabilities than natives at older ages, yet they continue to exhibit lower overall mortality, thus suggesting a morbidity–mortality paradox. We utilize the IPUMS National Health Interview Survey 2002–2018 with linked mortality data through 2019 (n = 405,270) to comprehensively investigate how this paradox unfolds with age for various groups of immigrants. The analysis shows that immigrants’ advantages in chronic conditions and disabilities narrow or even disappear at old ages, whereas their mortality advantages continuously increase with age. These patterns exist for immigrants of different ethnoracial, sex, and educational groups. The decomposition analysis reveals that the narrowing disability gap is due to immigrants’ increasing prevalence of mental illness and diabetes, shrinking advantages in lung diseases and musculoskeletal conditions, and increasing vulnerability to the disabling effects of major chronic conditions. However, immigrants are less likely to die from chronic diseases and disabilities, and this advantage strengthens with age, widening the nativity gap in mortality risk with age. We suggest that health-based selection might simultaneously postpone the onset of chronic diseases and disabilities to later ages for immigrants and better enable them to weather the mortality consequences of the diseases and disabilities.

Introduction

Immigrants, on the whole, have better health and lower mortality rates than their native-born counterparts (Elo et al. 2004; Guillot et al. 2018; Hummer and Gutin 2018; Lariscy et al. 2015; Riosmena et al. 2013). Scholars have attributed these differences to healthy individuals’ greater tendency to migrate, immigrants’ cultural practices that promote healthy behavior and social support, and unhealthy immigrants’ higher rate of return to their home countries (Akresh and Frank 2008; Arenas et al. 2015; Kimbro 2009). Nevertheless, immigrants’ morbidity and disability advantages tend to diminish with age (Boen and Hummer 2019; Hayward et al. 2014; Melvin et al. 2014; Sheftel and Heiland 2018). At the same time, immigrants persistently experience lower mortality than natives (Arias et al. 2010; Mehta et al. 2016; Zheng and Yu 2022a). Taken together, these findings suggest a morbidity–mortality paradox for older immigrants relative to their native-born counterparts: although older immigrants become more likely to experience chronic conditions and disabilities than same-age natives, they continue to exhibit lower mortality rates.

Despite evidence that a morbidity–mortality paradox exists, how this paradox unfolds over ages remains an open question. To our knowledge, no study has directly contrasted changes in the nativity gaps in morbidity and mortality across ages and informed whether these gaps move in parallel, albeit to different extents, or diverge in specific ways. The argument for a morbidity–mortality paradox does not require a divergence in mortality hazards between immigrants and natives with age. However, it would make the paradox even more perplexing and worthy of attention if the changes in the nativity gap in mortality follow a pattern with age opposite that of changes in the gap in morbidity. In this study, we leverage a uniquely large and diverse dataset spanning a wide age range to investigate how the nativity gaps in chronic diseases, disabilities, and mortality evolve with age. Improving on prior studies that primarily focused on Hispanic immigrants in the United States (Boen and Hummer 2019; Hayward et al. 2014; Sheftel and Heiland 2018), our analysis encompasses a broader range of the foreign-born population. With data on immigrants of diverse ethnoracial backgrounds, we can show whether the morbidity–mortality paradox is generalizable and how the age dynamics of morbidity and mortality risks vary across immigrant groups.

Our study further contributes to the literature by exploring the mechanisms behind the morbidity–mortality paradox. We conduct a decomposition analysis to assess how the prevalence and disabling effects of various chronic diseases contribute to the evolving disability gap between immigrants and natives and how the prevalence and mortality effects of chronic diseases and disabilities contribute to the evolving mortality gap. We follow the progression of morbidity framework, which posits that the progression typically commences with physiological dysregulation, followed by the diagnosis of diseases, functional decline, disability, and eventually death (Crimmins et al. 2010; Verbrugge and Jette 1994). Guided by this framework and prior research, we consider chronic conditions as the primary cause of disability (Freedman et al. 2007) and both chronic diseases and disabilities as major determinants of death (Bauer et al. 2014; Kuper et al. 2024). If immigrants, with age, become more likely to develop certain chronic diseases with considerable disabling consequences but relatively low mortality effects or such chronic diseases yield greater disability consequences for immigrants than the native-born even at the same level of prevalence, we can potentially explain a shrinking immigrant advantage in disability without an equivalent trend in mortality. Alternatively, if the rising prevalence of chronic diseases among older immigrants is not limited to diseases with low mortality effects and the diseases do not disable immigrants more, then the shift in the disability gap might merely reflect the delayed onset of chronic diseases among immigrants. Thus, our decomposition analysis not only covers the principal pathway of the disablement process and mortality but also delves into the mechanisms and clarifies the intricate relationship between chronic conditions, disability, and mortality among immigrants and natives.

Background

Age Dynamics of Immigrants’ Relative Morbidity and Mortality

Several studies found that older Hispanic immigrants have a greater tendency to experience disabilities and other health problems than U.S.-born individuals (Boen and Hummer 2019; Hayward et al. 2014; Sheftel and Heiland 2018). Although younger Hispanic immigrants tend to have lower disability rates than non-Hispanic native-born Whites, their rates increase at a much faster pace with age. By their mid-50s or 60s, Hispanic immigrants have a higher prevalence of disabilities or difficulties managing daily activities than native-born Whites (Hayward et al. 2014; Sheftel and Heiland 2018). Even when education, major life stressors, and health behavior are controlled for, foreign-born Hispanics aged 50 or older still face more functional limitations than native-born Whites and, to a lesser extent, native-born Hispanics (Boen and Hummer 2019). In contrast, immigrants’ mortality advantage does not diminish in old age. Older foreign-born individuals persistently have higher life expectancies than U.S.-born individuals (Arias et al. 2010; Dupre et al. 2012; Hayward et al. 2014; Mehta et al. 2016). Some studies also documented a persistent or even widening mortality advantage over Whites with age among Hispanics (Hummer et al. 2004; Markides and Eschbach 2005) but not among Asian Americans (Elo et al. 2004), although not all Hispanics or Asian Americans in such studies are immigrants.

The aforementioned research points to the possibility that immigrants’ morbidity and mortality advantages over native-born individuals change in different directions with age, at least among Hispanics. Nonetheless, a detailed and systematic investigation of the age dynamics of nativity gaps in morbidity and mortality remains absent. In particular, we lack evidence on whether immigrants’ mortality advantage is declining, constant, or increasing with age. The documented lower mortality rate of older immigrants does not necessarily mean that their advantage over natives remains constant (e.g., Arias et al. 2010). This advantage can decrease with age, just like immigrants’ morbidity advantage (Sheftel and Heiland 2018), but the convergence could be so slow that immigrants continue to have an advantage into very old age. In a handful of studies that compared foreign- and native-born individuals by age, the analyses focused only on life expectancies or five-year survival rates (e.g., Dupre et al. 2012; Mehta et al. 2016), making it difficult to determine how the mortality gap changes with age. Moreover, prior mortality-related comparisons generally did not consider other confounders that could differentiate immigrants from the native-born, such as their socioeconomic conditions and health behavior. Thus, it is unclear whether any age-based variation simply reflects the different sociodemographic compositions of immigrants versus natives by age.

Equally important, no previous study has directly compared immigrants’ morbidity and mortality advantages with age. Because immigrants’ composition and selectivity change over time (Borjas 1995; Hamilton et al. 2015; Zheng and Yu 2022a), different directions of changes in their relative mortality and morbidity advantages shown in separate studies could simply reflect the studies’ different sample populations or selected age ranges. Therefore, using the same sample to examine changes in the nativity gaps in morbidity and mortality is imperative.

Finally, virtually all evidence about older immigrants’ relative rates of chronic diseases and disability is derived from the Hispanic population (Angel et al. 2014; Boen and Hummer 2019; Hayward et al. 2014; Sheftel and Heiland 2018). Immigrants of different ethnoracial identities are heterogeneous in their health-based selectivity, socioeconomic conditions, experiences of discrimination, and health behavior (Bosdriesz et al. 2013; Huang et al. 2011; Singh and Siahpush 2002; Villarreal and Tamborini 2018; Zheng and Yu 2022b). These factors can affect immigrants’ chances of suffering from chronic diseases and disability relative to U.S.-born individuals. Understanding the generalizability of the immigrant morbidity–mortality paradox requires a comprehensive analysis of different immigrant groups.

Explanations for the Immigrant Morbidity–Mortality Paradox and Their Implications

The literature on immigrant health has suggested several possible explanations for the immigrant morbidity–mortality paradox. These explanations have somewhat different implications for how chronic diseases, disabilities, and mortality among the foreign-born population will change with age relative to those among native-born individuals. The differing explanations also inform how shifts in the prevalences and consequences of chronic diseases with age could contribute to changes in the nativity gaps in morbidity and mortality with age.

First, the “wear and tear” perspective posits that immigrants, particularly Hispanic immigrants, tend to be overrepresented in physically demanding occupations, such as construction, service, and agriculture work (Kochhar 2005; Toussaint-Comeau 2006). Consequently, they face an increased risk of work-related injuries and nonfatal musculoskeletal conditions, increasing the likelihood of disability (Hayward et al. 2014). However, because musculoskeletal damage is typically nonfatal and because Hispanic immigrants often exhibit a lower risk of conditions such as coronary heart disease, stroke, and lung disease (Blue and Fenelon 2011; Crimmins et al. 2004; Elo et al. 2004; Hummer and Gutin 2018; Zhang et al. 2012), their overall mortality rate remains lower.

If the occupational hazard is the sole or predominant reason for the paradox, the prevalence of musculoskeletal conditions, in particular, will increase more with age among immigrants than natives. Because immigrants’ greater occupational wear and tear would make their musculoskeletal diseases more severe and more likely to result in disability, we can expect growing contributions of the prevalence and disabling impact of musculoskeletal conditions to immigrants’ diminishing disability advantage (or reversal and increasing disadvantage) over time. Further, although musculoskeletal diseases are generally not fatal (Case and Paxson 2005), prolonged wear and tear can lead to disability that might reduce, or even reverse, immigrants’ mortality advantage over natives in old age. Disability increases the risk of an unhealthy lifestyle and fatal diseases, such as cancer and cardiovascular diseases (CVDs), suicide, and unintentional injuries (Forman-Hoffman et al. 2015; Kuper et al. 2024). Disability also has an independent effect on mortality (Majer et al. 2011). Therefore, we should observe that the prevalence and mortality impact of musculoskeletal diseases and disability increasingly contribute to the nativity gap in mortality, ultimately narrowing this gap with age.

Second, immigrants’ better health endowments represent another potential mechanism. Because individuals must be relatively healthy to migrate in the first place, immigrants are selectively healthy (Bosdriesz et al. 2013; Guillot et al. 2018; Landale et al. 2006; Riosmena et al. 2017). Healthier individuals tend to experience the postponement of disease and disability onset to later ages. The earlier onset of chronic diseases and disabilities among the native-born population relative to the foreign-born might lead to an initial widening in middle age as the native-born begin to experience diseases. Subsequently, the nativity gap in disease and disability prevalence in old age might narrow when the foreign-born start to face the onset of health problems. A similar argument has been employed to explain the widening socioeconomic status differences in health from youth to middle age and the diminishing of such differences relatively late in life (House et al. 1994).

If the argument about health-based selection and immigrants’ better health endowments holds, then immigrants might fare better than natives even as they experience various major diseases and disabilities. Immigrants might better withstand the mortality consequences of many chronic conditions and disabilities—especially at older ages, when such health issues become more relevant. If so, the nativity gap in mortality will widen with age. Because the differential mortality consequences should be more relevant at older ages, the overall contribution of the mortality impacts of chronic diseases and disabilities to the nativity gap in mortality will also increase with age. At the same time, because the age-specific shifts in the nativity differences in chronic disease and disability prevalences merely reflect the delayed onset of diseases and disabilities for immigrants, they will likely contribute little to the change in the mortality gap between immigrants and natives with age.

Third, the mortality selection mechanism also operates under the assumption of the initial selection of healthy immigrants. According to the theory of population heterogeneity, populations consist of individuals or subpopulations varying in physiological vulnerability to mortality, commonly referred to as frailty (Vaupel et al. 1979; Vaupel and Yashin 1987). Mortality tends to remove frailer individuals from the population at earlier ages, leaving stronger individuals to survive to older ages (Vaupel and Missov 2014; Yashin et al. 2002; Zheng 2014, 2020; Zheng and Cheng 2018). Consequently, the average health and mortality of a given birth cohort would increasingly reflect those of robust individuals as the cohort ages. In the context of immigrant–native differences, it is plausible that unhealthy natives will experience earlier mortality, leaving a higher proportion of robust individuals at old ages in the native population than in the foreign-born population. The differential change in frailty composition could contribute to the appearance of relatively better health of the native population as it ages. At the same time, the foreign-born population's lower mortality rates at younger ages might enable a larger proportion of frail immigrants to survive to older ages, increasing their risk of developing diseases and experiencing disability relative to the native population. This process can result in morbidity and disability convergences and crossovers between foreign- and native-born populations. If mortality selection is the primary mechanism at play, we should expect a convergence in mortality as well, which differs from the expectation of the delayed onset mechanism. As discussed earlier, immigrants could have overall lower mortality at old ages than natives even if the nativity gap in mortality narrows with age. Because the native population will include an increasingly lower proportion of frail individuals with age, we should also observe that immigrants and natives converge (or cross over) in both the prevalence and impacts of all diseases—not just musculoskeletal diseases, as suggested by the wear and tear account.

The final mechanism centers on the selective pattern of return migration, commonly known as the salmon bias (Arenas et al. 2015; Guillot et al. 2023; Palloni and Ewbank 2004). The salmon bias argument contends that unhealthy immigrants are more likely to choose to return to their home countries, making older immigrants especially likely to be relatively healthy.¹ Because surveys likely capture immigrants before their returns but cannot track the deaths of really unhealthy immigrants after their returns, the estimate of the nativity gap in mortality might be exaggerated. The salmon bias has been used to explain a persistent or even increasing mortality gap between non-Hispanic Whites and Hispanics with age (Markides and Eschbach 2005), but it is unclear how it could account for the possible immigrant–native convergences or crossovers in morbidity and disability. Perhaps the immigrants remaining in the host country, although not especially unhealthy to begin with, adopt the less healthy lifestyle and behavior of the native-born as their stay extends (Abraido-Lanza et al. 2005; Antecol and Bedard 2006; Finch et al. 2001; Kimbro 2009), thereby narrowing the nativity gap in morbidity with age. A joint mechanism with the salmon bias and unhealthy assimilation arguments could predict a divergence in the nativity gaps in morbidity and mortality with age.

To summarize, the literature offers four possible explanations for the morbidity–mortality paradox, with somewhat different implications for how shifts in immigrants’ relative disease prevalence and consequences with age contribute to the changes in the nativity gaps in disability and mortality. At least one mechanism—the wear and tear argument—also emphasizes immigrants’ uneven exposure and vulnerability to different major diseases. By differentiating various diseases and decomposing the pervasiveness and impact of each disease across age groups, we can gain insights into why immigrants’ relative advantages in morbidity and mortality relative to natives change with age.

Methods

Data

We used the IPUMS National Health Interview Survey (NHIS) data from 2002 to 2018, with linked mortality records through the end of 2019 (Blewett et al. 2018).² The NHIS is an annual, cross-sectional, multistage probability sample survey conducted by the National Center for Health Statistics, focusing on the noninstitutionalized civilian U.S. population. The dataset offers a large and diverse sample, making it suitable for comparing immigrants of different ethnoracial identities. The survey data were linked to death records in the National Death Index through probabilistic record-matching methods, which used 13 criteria to ascertain the vital status of each respondent. This linkage allows us to know when the NHIS respondents died, assuming that they died before the end of 2019. We pooled the NHIS respondents from 2002 to 2018 and restricted the sample to individuals aged 25–84 at the time of the survey. We set age 25 as the lower bound to ensure the completion of education, given that we distinguish between those with versus those without college degrees in the analysis. Respondents aged 85 or older were excluded because the NHIS top-coded age at 85, making it impossible to differentiate individuals in this age category. We further limited the sample to sample adults, who provided answers to the disease and disability-related questions. Within this sample, 96.6% had eligible mortality records,³ resulting in a final sample of 428,094 individuals. We used the same sample to estimate the nativity gaps in mortality and morbidity to ensure that differences in immigrants’ relative morbidity and mortality advantages are not due to missing death records of return immigrants.

Unlike earlier studies (Boen and Hummer 2019; Hayward et al. 2014; Sheftel and Heiland 2018), our sample size is sufficiently large to divide both the foreign- and native-born into five ethnoracial groups—non-Hispanic Whites (Whites hereafter), non-Hispanic Blacks (Blacks hereafter), non-Hispanic Asians (Asians hereafter), Hispanics of Mexican origin, and other Hispanics. The Asians in our sample consist of Chinese, Filipinos, and Asian Indians. We excluded other Asian ethnic groups because such groups were very small or disproportionately foreign- or native-born in the NHIS data. After we eliminated respondents in other ethnoracial groups (4.6% of the sample with eligible mortality records; n = 19,674) and those with missing data on covariates (0.7%; n = 3,150), the analytic sample contains 337,070 native-born individuals: 259,129 are Whites, 53,767 are Blacks, 2,472 are Asians, 17,007 are Hispanics of Mexican origin, and 4,695 are other Hispanics. The sample includes 68,200 foreign-born individuals: 12,605 are Whites, 6,031 are Blacks, 10,498 are Asians, 23,550 are Hispanics of Mexican origin, and 15,516 are other Hispanics.

To show the nativity gaps in chronic diseases and disabilities across age groups, we divided the person-level data by their age at the survey time and conducted stratified analyses by age group. For the decomposition analysis of mortality hazards (described later), we transformed the dataset into a person-year format, resulting in 3,539,252 observations. This person-year format spanned from the year of the interview to the year of the respondent's death or 2019, whichever was earlier. The NHIS dataset provided information on the date of birth, interview, and death, enabling us to calculate the time-varying attained age for each respondent.

Measurements

The outcomes of interest in our study encompass a range of chronic diseases, disability status, and mortality. Chronic diseases include cancer, CVDs, lung diseases, mental illness, diabetes, liver disease, kidney disease, and musculoskeletal conditions. Respondents were asked whether a doctor or other health professional had ever diagnosed them with each condition. CVDs comprise hypertension, angina, coronary heart disease, heart attack, any other heart conditions or diseases, and stroke; lung diseases include asthma, emphysema, and bronchitis; and musculoskeletal conditions consist of arthritis/rheumatism, lower back pain, and neck pain. Mental illness was assessed using the Kessler 6 (K6) scale, a six-item instrument asking respondents how often they felt sad, nervous, restless, hopeless, worthless, or that everything was an effort during the preceding 30 days; responses range from none of the time (scored as 0) to all of the time (scored as 4). Responses to the six items were summed to yield a K6 score between 0 and 24. A score of 7 or higher on this scale indicates moderate or severe mental distress (Kessler et al. 2003; Pratt 2009; Prochaska et al. 2012). For each disease type, we created a binary variable, with a value of 1 indicating the presence of the chronic disease. Among the 402,345 individuals with valid information, 281,660 had at least one chronic disease. Within this subgroup, 244,402 individuals were natives (73.0% of the native sample), and 37,258 were immigrants (55.2% of the immigrant sample).

Disability was assessed through two measures: functional limitations and activity limitations. We considered experiencing any functional or activity limitation as an indicator of disability, although we examined functional and activity limitations separately in our analysis to be mindful about the different ways in which individuals may be hampered. Functional limitations refer to an individual's intrinsic limitations in body function, irrespective of the environment. The NHIS respondents were asked to report difficulties in performing a range of tasks without any special equipment: (1) pushing or pulling large objects; (2) going out for activities, such as shopping, movies, or sporting events; (3) participating in social activities; (4) engaging in leisure activities at home; (5) walking a quarter of a mile; (6) walking up 10 steps without resting; (7) standing on feet for roughly 2 hours; (8) sitting for 2 hours; (9) stooping, bending, or kneeling; (10) reaching up over the head; (11) using fingers to grasp or handle small objects; and (12) lifting or carrying something as heavy as 10 pounds. Individuals experiencing limitations in any of these tasks were classified as having functional limitations (1 = any functional limitation, 0 = not limited in any way). Among the 404,332 individuals with valid information, 155,750 faced at least one functional limitation, with 138,803 being natives (41.3% of natives) and 16,947 being immigrants (24.9% of immigrants).

Activity limitations capture an individual's need for assistance from others in performing (1) activities of daily living (ADLs), such as eating, bathing, dressing, and getting around inside the home; and (2) instrumental activities of daily living (IADLs), which include everyday household chores, conducting necessary business, shopping, or getting around for other purposes. The activity limitation measure also included work limitations and limitations in any other activities due to physical, mental, or emotional problems. Individuals reporting any of these limitations were categorized as having activity limitations (1 = any activity limitation, 0 = not limited in any way). Unlike the functional limitation measure, the activity limitation measure is influenced by individuals’ social contexts and living environments. Among the 404,991 individuals with valid activity limitation information, 75,206 had activity limitations; 68,489 were natives (20.3% of natives), and 6,717 were immigrants (9.9% of immigrants).

The third outcome variable is mortality status. By the end of 2019, a total of 45,082 NHIS respondents in our sample had died; 41,254 were natives (12.2% of the native-born), and 3,828 were immigrants (5.6% of the foreign-born). With precise information on respondents’ date of death, we can accurately determine the exact age at death. In the person-year data, we coded mortality status as 1 if a respondent died at the observed age and as 0 if they were alive.

The main predictor in our study is nativity status, measured by respondents’ birthplace. We further differentiated the foreign-born by race and ethnicity (Whites, Blacks, Asians, Hispanics of Mexican origin, and other Hispanics), given the potential variations in their selectivity and legal status composition. For a meaningful comparison, we also categorized the native-born population into the same five groups. Because health and mortality outcomes can vary by sex and education and because the wear and tear mechanism is particularly applicable to men and those without a bachelor's degree, we further measured respondents’ sex (male or female) and education (college degree or not) and examined the nativity gaps in morbidity and mortality by these characteristics. In addition, we divided immigrants by their years since immigration (YSI): less than 10 years or 10 and more years. Studies suggest that the health advantage of immigrants may diminish over time as they reside longer in the destination country, potentially adopting unhealthy behaviors and diets (Antecol and Bedard 2006). Such unhealthy assimilation could erode immigrants’ morbidity and mortality advantages.

We included several control variables—age at the survey, birth cohort, marital status, smoking status, body mass index (BMI), and health insurance coverage—in our models to account for other factors that may contribute to the nativity differences in morbidity, disability, and mortality. We divided respondents into seven cohorts based on birth years: 1918–1929, 1930–1939, 1940–1949, 1950–1959, 1960–1969, 1970–1979, and 1980–1993. The inclusion of birth cohort ensures that the age pattern observed is not due to different cohort compositions across age groups.⁴ We measured marital status with four categories: married, widowed, divorced or separated, and never married. Smoking status was also measured with four categories: never a smoker, former smoker, current smoker, and unknown smoking status. We created the BMI measure with six categories: underweight (<18.5 kg/m²), normal weight (18.5–24.9 kg/m²), overweight (25–29.9 kg/m²), Class I obese (30–34.9 kg/m²), Class II/III obese (>35 kg/m²), and unknown BMI status. Three categories were used to capture respondents’ health insurance status: covered by health insurance, not covered by health insurance, and unknown. For further information about our analytic sample and measures, Table A1 presents the descriptive statistics of key variables by nativity, race, and ethnicity (all tables and figures designated with an “A” appear in the online appendix).

Analytic Strategy

To observe the age patterns of morbidity and mortality gaps, we began by using person-level data to conduct a descriptive analysis comparing the percentages having any chronic conditions, reporting any functional limitations, experiencing any activity limitations, and having deceased by the last date of mortality follow-up by nativity status for six 10-year age groups (25–34, 35–44, 45–54, 55–64, 65–74, and 75–84). We then explored the mechanisms behind the immigrant morbidity–mortality paradox in several steps. First, we established the link between chronic conditions and disability by using linear probability models (LPMs) to predict the probability of disability given each chronic condition, as in Case and Paxson (2005). Second, we used the Blinder–Oaxaca regression decomposition method (Blinder 1973; Oaxaca 1973) to estimate the respective contributions of the prevalence (distribution) and disabling effect of each chronic disease to the nativity gap in disability (i.e., having any functional or activity limitations) by 10-year age group. Third, we estimated LPMs to establish how chronic conditions and disabilities are connected to mortality. Finally, we conducted a decomposition analysis estimating the respective contributions of the prevalence and mortality impacts of chronic diseases and disability (i.e., functional limitation, activity limitation) to the nativity gap in mortality hazards for the same 10-year age groups. Unlike the disability analysis, in which person-level data are employed, we used the person-year sample and survival analysis to estimate mortality hazards for those of a given age at the survey time. We used time-varying attained age as the time metric to estimate the baseline hazard function, accounting for left truncation and right-censoring (Singer and Willett 1993).

The Blinder–Oaxaca decomposition method is widely used to examine outcome differences between two groups and quantify the contributions of differences in the distributions of measured covariates and the differences in the regression coefficients of these covariates (e.g., O'Donnell et al. 2008; Stanley and Jarrell 1998). Although both disability and mortality are binary variables, we used LPMs for ease of interpretation (Case and Paxson 2005; Jann 2008; Long 1997; Wooldridge 2003). Because we used LPMs, the Blinder–Oaxaca decomposition results indicate the probability scale. For example, in the disability analysis, the decomposition results inform us of the contributions of predictors to the nativity gap in the probability of functional or activity limitations. In the mortality analysis, the decomposition results indicate the contributions of predictors to the difference in the hazard probability of death by nativity status.

Our models adjust for age, sex, race and ethnicity, education, marital status, smoking, BMI, and health insurance. Although these additional controls generally did not substantially alter the estimates of contributions of chronic diseases or disability, we included them to ensure that our estimates would not be influenced by nativity differences in socioeconomic status, health behaviors, and health insurance coverage.

Results

Age-Specific Nativity Gaps in Diseases, Limitations, and Mortality

Figure 1 displays the age pattern of the probability of having any major chronic diseases, disabilities, and mortality stratified by race and ethnicity, nativity, and YSI. The blue lines represent native-born individuals, while the green lines depict the foreign-born. Overall, immigrants are less likely to report chronic diseases than native-born individuals for most age groups, but the nativity gap narrows with age (panel a). In some cases, the line for the foreign-born crosses over that for the native-born, although the crossover point appears fairly late. For example, among Hispanics of Mexican origin, the foreign-born do not lose their advantage in disease prevalence until after age 80. Immigrants with different YSI exhibit similar patterns, although the patterns for those with YSI less than 10 years are less stable, particularly at older ages, mainly owing to the smaller sample size. We further examine the pattern by stratifying both immigrants and natives by sex and education (see Figure A1). Similarly, the nativity gap in diseases converges or crosses over with age regardless of sex or educational level. However, the nativity gap tends to be smaller among women and individuals with a bachelor's degree.

Panels b and c of Figure 1 display the percentage reporting any functional and activity limitations, respectively, within each age group. For both types of limitations, the nativity gap appears to widen from early to middle adulthood and then narrows (Whites and Blacks) or crosses over (Asians, Hispanics of Mexican origin, and other Hispanics) at older ages. For the ethnoracial groups that display a crossover pattern, the crossover occurs after age 70 or later. Thus, despite immigrants’ diminishing disability advantage over natives with age, they are advantaged for most of their lives. Breaking down by duration of residence produces similar patterns, although the patterns for immigrants with a YSI of less than 10 years are somewhat unstable because of the small sample size. The comparable patterns for the two YSI groups indicate that the narrowing of the disability gap between the foreign- and native-born does not primarily result from immigrants’ assimilation to unhealthy lifestyles as they stay longer in the host country. Figures A2 and A3 also exhibit similar patterns across educational and sex groups, suggesting that occupational wear and tear, which is more relevant to men with no college degree, might not be the only mechanism shaping the disability convergence with age.

Panel d of Figure 1 depicts the nativity differences in the percentage of respondents of each age group who died by the last date of mortality follow-up. Contrary to the patterns observed for chronic diseases and disabilities, the nativity gap in mortality widens across age groups. This widening pattern is consistent across almost all ethnoracial, YSI, sex, and educational groups (see Figure A4). Thus, regardless of their sociodemographic characteristics, immigrants experience a growing mortality advantage with age. Because the results so far indicate an immigrant–native convergence in diseases and disability and divergence in mortality with age in virtually all sociodemographic groups, we simply compare the foreign-born with the native-born, without any further distinction, in the remaining analyses.

Contributors to the Narrowing Nativity Gap in Disability

To understand the factors contributing to the narrowing nativity gap in disability with age, we first examine how the various chronic conditions are linked to functional limitations for immigrants and natives separately. We focus on functional limitations because the effects of chronic conditions on activity limitations can be moderated by individuals’ access to personal and environmental resources. Figure 2 illustrates the impact of each chronic condition on the probability of having any functional limitations by nativity, based on coefficients from separate LPMs predicting this probability for immigrants and natives. The solid 45-degree line represents equal magnitudes of the regression coefficients for both groups; that is, the increase in the probability of reporting a functional limitation in the presence of a given chronic condition is the same for immigrants and natives. Circles falling below the solid line mean that the chronic condition more likely disables immigrants than natives, whereas those above the line indicate that the condition has a greater disabling impact on the native-born than the foreign-born. The figure shows that musculoskeletal conditions and mental illnesses are the most debilitating, especially for the native-born population. Other diseases, such as cancer and liver and kidney diseases, are less disabling but tend to affect immigrants slightly more than natives.

Next, Table 1 presents the decomposition of the contributions of prevalence and the disabling effects of eight major chronic conditions to this gap by age group. We adjust for demographic and health-related factors but omit the results for the controls to conserve space (see Table A2 for full decomposition results). The multivariate results are similar to those shown in the descriptive figures. Among those aged 25–34, the probability of functional limitations is 10.63% for the foreign-born—8.10 percentage points lower than the corresponding rate among U.S.-born adults (18.73%). This advantage for foreign-born individuals widens until age 64 and gradually narrows in older age groups. Nevertheless, even in the oldest age group (75–84), the probability of functional limitations remains approximately 3.89 percentage points lower among immigrants than natives. Thus, immigrants, on average, do not suffer more from disability than the native-born, even in old age, although their advantage in functional limitations reduces by 4.21 percentage points with age (= −3.89 – −8.10).

The lower part of Table 1 demonstrates contributors to the nativity gap within each age group. Among those aged 25–34, immigrants’ advantage in functional limitations comes from a lower prevalence of CVD (contributing to 5.8% of the lower level of functional limitations), lung disease (7.6%), mental illness (12.2%), and musculoskeletal conditions (24.9%), as well as smaller disabling effects of mental illness (6.22%) and musculoskeletal conditions (8.4%). The widening gap up to age 64 results from immigrants’ continuous advantages in both the prevalence and disabling effects of these chronic conditions. From age 65 onward, however, the foreign-born become more likely to have mental illness and diabetes and suffer more from the disabling effects of these two diseases. Although they are still less likely to have cancer, CVD, lung disease, or musculoskeletal conditions at older ages, they are more likely to experience disability from these conditions.

The last column in Table 1, further visualized in Figure 3, indicates the contributors to the change in nativity gap between the youngest and oldest age groups. The overall decreasing immigrant advantage in functional limitations from ages 25–34 to 75–84 is attributable to a growing prevalence of mental illness and diabetes (contributing to 39.9% and 7.3% of the decreasing advantage, respectively). Even though immigrants have a lower prevalence of lung diseases and musculoskeletal conditions than natives across age groups, these advantages greatly decrease with age, contributing to approximately 8.4% and 20.5% of diminishing advantage in functional limitations, respectively. In addition, immigrants increasingly suffer from the disabling effects of mental illness, diabetes, cancer, CVD, lung disease, and musculoskeletal conditions, also contributing to their diminishing advantage in functional limitations.

We conduct the same decomposition analysis for activity limitations, with the full results presented in Table A3. Similar to functional limitations, immigrants’ initial advantage in activity limitations increases up to age 65 and decreases afterward, although even those in the oldest age group (75–84) remain slightly advantaged in this regard. The decomposition analysis indicates that with age, immigrants become more likely to have mental illness and diabetes and to suffer from these conditions’ disabling consequences. They also become more likely to develop activity limitations from other chronic conditions despite their continuously lower likelihood of experiencing the conditions. Overall, the decreasing advantage of immigrants in activity limitations from ages 25–34 to 75–84 is attributed to the rising prevalence of mental illness and diabetes (contributing to 55.5% and 10.8% of the decreasing advantage, respectively) and the growing disabling effects of mental illness, diabetes, cancer, CVD, and musculoskeletal conditions.

Contributors to the Widening Nativity Gap in Mortality

To establish the connections of chronic conditions and disabilities to mortality, we first illustrate the impacts of chronic conditions and disabilities on the hazard probability of mortality by nativity (Figure 4). Kidney disease is among the deadliest conditions, especially for the native-born population. Activity limitations represent severe forms of disability and are highly consequential for mortality. Cancer is particularly deadly for the foreign-born. Other diseases and limitations increase the risk of death more for natives than immigrants. Musculoskeletal conditions are not positively associated with mortality when we examine immigrants or natives as a whole, with no distinction by age.

Table 2 presents key results from the decomposition of the contributions of prevalence and mortality effects of eight major chronic diseases, functional limitations, and activity limitations to the nativity gap in mortality (full results are shown in Table A4). In the 25–34 age group, mortality risk is roughly 0.05 percentage points lower for immigrants than natives. This advantage for immigrants continuously increases up to ages 75–84, when their mortality risk is approximately 2.04 percentage points lower. Altogether, the mortality advantage of the foreign-born grows by 1.99 percentage points. This advantage mainly stems from a lower prevalence of cancer, CVD, and functional and activity limitations, as well as better mortality consequences of major diseases and disability. The exception is cancer; immigrants are more likely to die from it than natives.⁵

Figure 5 visualizes the information from the last column of Table 2. Changes in the prevalence of chronic conditions and disability with age do not contribute to immigrants’ growing mortality advantage over natives. The widening immigrant–native gap in mortality mainly originates from foreign-born individuals’ increasing survival advantages from major chronic diseases and disabilities: immigrants weather the mortality consequences of major diseases and disabilities better than natives. For example, the increasing survival advantage in CVD, lung diseases, mental illness, diabetes, and functional and activity limitations among the foreign-born contributed to roughly 2.7%, 4.2%, 9.5%, 7.1%, 21.1%, and 36.2% of immigrants’ widening mortality advantage, respectively, across ages (80.8% in total). Immigrants’ increasing mortality risk from cancer, however, offsets the widening mortality advantages by 11.3%.⁶

Alternative Explanations and Checks

Nativity Difference in Health Reporting

The analysis thus far indicates that the growing disabling effects of CVD, lung diseases, mental illness, and diabetes with age do not result in similar increases in mortality effects for immigrants relative to natives. One possible explanation is that immigrants might be more health-pessimistic than natives, making them more likely to report limitations as they experience chronic conditions with age, even if their conditions are milder. This likelihood of reporting mild conditions and limitations could explain their continued survival advantage over natives as they age. To test this possibility, we examine the impact of chronic conditions on the probability of reporting fair or poor health by nativity and age group (25–34 vs. 75–84), as shown in Figure 6. The measure of self-rated health is a binary indicator of whether self-rated health is fair or poor (hereafter referred to as “poor” health) versus good, very good, or excellent.

Among individuals aged 25–34, natives are more likely to report poor health in the presence of most chronic conditions (except cancer) than immigrants, although only the differences associated with lung and liver diseases are statistically significant. In this sense, immigrants are not more health-pessimistic. Among individuals aged 75–84, immigrants are more likely to report poor health in the presence of five of eight chronic conditions, with four of these associations being statistically significant. However, the larger coefficient associated with cancer among immigrants does not necessarily mean they are more health-pessimistic in the presence of this disease, given that cancer is increasingly more fatal for immigrants than natives at older ages (Table 2 and Figure 5). This result could indicate that cancer for older immigrants tends to be more severe or advanced, making it more disabling and deadlier. The larger coefficients associated with CVD, diabetes, and musculoskeletal conditions could imply that immigrants are more health-pessimistic in the presence of these diseases, but this interpretation needs to be reconciled with the nonsignificant differences associated with mental illness and lung diseases. If immigrants systematically factor milder forms of chronic conditions into their reports of limitations and poor health, this pattern would be consistent across all diseases, especially mental illness, which has the greatest impact on the probability of reporting poor health in old age. Overall, we do not find consistent evidence to support the conjecture that elderly immigrants are more pessimistic about their health. Instead, our results align more with the finding that immigrants are increasingly more likely disabled by certain chronic conditions at older ages, which enhances their probability of reporting poor health.

Nativity Gaps in Disability and Mortality Among Hispanics

Because previous studies of older immigrants’ declining disability advantage focus on just Hispanics, we perform a separate analysis of Hispanics to ensure that our results are not driven by our immigrant group's different composition (Tables A5–A7). Overall, Hispanics display age patterns in the nativity gaps in disability and mortality that are similar to those shown for the full sample. The contributors to the narrowing gaps in limitations, especially functional limitations, are also similar to those for immigrants as a whole (Table 1 and Table A3). Compared with their native-born coethnics, Hispanic immigrants face a higher risk of mental illness and shrinking advantages in lung and musculoskeletal diseases at older ages. The increasing vulnerability to the disabling effects of cancer, lung diseases, diabetes, and musculoskeletal conditions contributes to the narrowing and, ultimately, the reversal of their advantage in functional limitations. At the same time, Hispanic immigrants’ mortality advantage rises with age, primarily owing to their growing survival advantages from mental illness, diabetes, musculoskeletal conditions, and activity limitations.

Salmon Bias

To test the salmon bias as a possible explanation for the widening mortality gap between immigrants and natives, we conducted a sensitivity analysis that assumed all the immigrants without eligible death records (n = 2,796) in the NHIS are return migrants and included them in examining chronic diseases and disability.⁷Figure A5 shows the results for the nativity gaps in chronic diseases and limitations. The patterns are similar to those shown in Figure 1.

To include individuals with missing death records in our mortality analysis, we created three hypothetical scenarios in which a varying proportion of immigrants without eligible mortality records is assumed to be deceased by 2019, the end of the mortality follow-up. Among individuals with eligible mortality linkage, 12.2% of natives and 5.6% of immigrants died by 2019. We assumed that the immigrants without eligible mortality records were slightly more likely (15%) than natives to die, were four times more likely (50%) than natives to die, or all (100%) died by 2019. The corresponding results are presented in Figure A6. With these rather exaggerated assumptions about return migrants’ mortality, immigrants’ mortality advantage over natives still widens with age. To account for the differential mortality by age, we created a fourth scenario in which immigrants without eligible mortality linkage were twice as likely to die as the natives of their same age group (with the assumed death rate ranging from 2% at ages 25–29 to 100% at ages 80–84). This adjustment removes the unrealistically high mortality rate among young adults, and the nativity gap in mortality continues to widen with age. Thus, the salmon bias explains little of the immigrant morbidity–mortality paradox.

Discussion and Conclusions

Recent studies indicate that despite their lower mortality level, older Hispanic immigrants have higher disability and physiological dysregulation levels than natives (Boen and Hummer 2019; Hayward et al. 2014; Sheftel and Heiland 2018), implying a morbidity–mortality paradox when comparing immigrants with the native-born. Our study shows an even more perplexing pattern: immigrants’ advantages in chronic conditions and disability narrow or disappear at old ages, but their mortality advantage continuously grows. These opposite directions of change are observed for immigrants of different ethnoracial identities, sex groups, and educational levels.

Our decomposition analysis finds that the narrowing gap in disability is attributable to immigrants’ increasing prevalence of mental illness and diabetes, shrinking advantages in lung diseases and musculoskeletal conditions, and growing vulnerability to the disabling effects of many chronic conditions (cancer, CVD, lung diseases, mental illness, diabetes, and musculoskeletal conditions). Although immigrants’ advantages in the prevalence of chronic diseases and disabilities decrease with age, they are less likely to die from most of them (except cancer) than natives. Immigrants’ lower likelihood of dying from many chronic diseases becomes more pronounced with age, leading to a widening nativity gap in mortality risk.

Our evidence that immigrants are increasingly at risk of developing musculoskeletal conditions and being disabled from them with age is consistent with the wear and tear argument. Surprisingly, however, immigrants of all age groups are less likely to report musculoskeletal diseases than natives. This finding holds even if we focus only on Hispanic immigrants. We also find that the disability rates of immigrants and natives converge with age regardless of race and ethnicity, sex, and educational level, even though some groups of immigrants—for example, highly educated Whites and Asians—are relatively unlikely to work in occupations where wear and tear is common.

In addition, diseases other than musculoskeletal ones also contribute to the shrinking nativity gap in disability. The increasing prevalence of mental illness and diabetes and vulnerability to their disabling effects are significant contributors to the narrowing disability gap. Although the prevalence of cancer, CVD, and lung diseases is lower among immigrants, as the wear and tear argument claims, the contributions of these diseases to disability increase more for immigrants than for natives. These increases also explain immigrants’ diminishing disability advantage. Meanwhile, immigrants’ survival advantage amplifies at older ages, mainly because they can better weather typically fatal diseases and disability, despite their diminishing advantages in disease and disability rates. Altogether, our results suggest that the explanatory power of the wear and tear argument to the morbidity–mortality paradox in immigrant advantages is probably limited.

The morbidity–mortality paradox is also unlikely a joint result of immigrants’ unhealthy assimilation and the salmon bias. The unhealthy assimilation argument does not hold because (1) immigrants with long and short durations of residence alike experience convergences (or crossovers) in chronic diseases and disability with the native-born; and (2) the mortality gap widens with age, even though unhealthy assimilation should lead immigrants’ and natives’ mortality risks to converge, too. As for the salmon bias argument, our robustness check using fairly exaggerated assumptions about potential return migrants’ chances of death still indicates that the nativity gap in mortality would widen with age. This widening pattern is also contradictory to the prediction of the mortality selection mechanism. If mortality selection indeed removes frail individuals from the native-born population faster than from the foreign-born population, these two populations should become increasingly similar in both morbidity and mortality.

On the whole, our findings are probably most consistent with the explanation focusing on the health-based selection of immigrants and their delayed onset of diseases. Our finding that a morbidity–mortality paradox exists for all races and ethnicities, sex groups, and educational levels suggests that this paradox relates to common factors that influence all immigrants, such as health-based selection. Our evidence of the initially widening and then narrowing nativity gaps in morbidity and disability with age is congruent with the argument that immigrants’ better health endowments help postpone the onset of diseases to later ages. Moreover, our results from the decomposition analysis show that immigrants are better at weathering many fatal diseases, which also supports the notion that immigrants have better health endowments. Because differences in such endowments are likely more relevant to mortality at older ages, we also find that immigrants’ advantage in surviving fatal diseases becomes more pronounced with age. This change ultimately explains the widening mortality gap between foreign- and native-born individuals. Because the relatively low mortality among immigrants can further prolong the lives of frail immigrants with diseases and disability, the nativity difference in disability shrinks or reverses with age. This process, along with the postponement of disease onset to older ages, might partially explain why diseases become more disabling for immigrants as they age.

If the better health endowment is the primary mechanism shaping the paradox, we would expect this paradox to be more pronounced among more selected immigrants. To provide additional support, we follow Riosmena et al. (2017) in using smoking status and height as proxies for health endowment. We assume that never-smokers or immigrants who are taller than the average of their group are more selected than current/former smokers or shorter ones.⁸Figure A7 shows that despite their greater advantages in early adulthood, immigrants with no smoking history encounter a steeper increase in diseases and disabilities than immigrants who are current or former smokers. Consistent with the health endowment argument, these immigrants simultaneously experience a more pronounced increase in mortality advantage over natives. The results for height are more mixed, with taller immigrants experiencing a slightly steeper increase in diseases and disabilities with age but a smaller increase in mortality advantage than shorter immigrants (Figure A8). It is possible that height is not a reliable indicator of superior health endowment, given that some research has found that shorter people have a lower mortality rate (Khetan et al. 2022; Samaras and Elrick 2002; Sohn 2016).

Although our additional analysis provides some support for the health-based selection argument, it might not be the only reason for the changes in immigrants’ relative morbidity and mortality with age. For example, this perspective cannot explain why immigrants become increasingly more likely than natives to experience mental illness at older ages. Perhaps aside from occupational wear and tear, long-term exposure to discrimination and other unfavorable socioeconomic experiences also harm immigrants’ psychological well-being and old-age health.

One limitation of this study is our reliance on self-reported chronic conditions and disability assessments. Immigrants might generally have more restricted access to health care than U.S.-born individuals, but they could become more aware of their chronic illness after gaining access to Medicare at older ages, altering their reports. We are nevertheless skeptical that unequal access accounts for our results. First, we controlled for health insurance coverage, even though the measure does not fully capture changes in this coverage for natives and immigrants with age. Second, if Medicare were the primary driver, we would have observed a discrete step function in chronic condition diagnosis at age 65, the Medicare eligibility age. Contrary to this expectation, our analysis reveals a rather continuous age pattern. Third, if the convergence in the nativity gap in chronic condition prevalence were solely due to improved health care access for immigrants at older ages—leading to higher rates of physician diagnosis—we would not observe the faster increase in disabling effects of chronic conditions among immigrants as they age, nor would we expect a convergence in disability rates. When we replaced chronic condition reporting with self-rated health—a measure less affected by health care access and physician diagnosis—we still observed a convergence in the nativity health gap with age (see Figure A9).

Of course, it is possible that the poorer health care for immigrants leads to a real—not just reported—steeper increase in chronic diseases and disabilities relative to natives. This alternative interpretation, however, would need to be reconciled with the fact that immigrants’ mortality advantage does not diminish at older ages. Although immigrants’ poorer health care alone seems unlikely to explain the morbidity–mortality paradox, we do not dispute that immigrants face limited access to health care and preventive services. In fact, the differing health care experiences could explain the growing detrimental impacts of cancer and musculoskeletal conditions for immigrants. Compared with natives, immigrants might be diagnosed with cancer at more advanced stages, making it more disabling and deadlier. Similarly, a later diagnosis could explain why musculoskeletal conditions become more disabling and offset some of the mortality advantage for immigrants at older ages. Still, poorer health care cannot explain the increasing survival advantage for immigrants regarding other major diseases.

Another possibility is that foreign- and native-born individuals use different criteria to assess their health conditions and limitations. Our use of the measure of functional limitations, which are less influenced by health evaluation thresholds or by surrounding environments, decreases the impact of this possibility. Our direct test of the likelihood of differential health assessment, presented in Figure 6, also does not find sufficient and consistent support for the hypothesis that immigrants systematically have a more pessimistic assessment of their health, suggesting that using self-reported morbidity measures does not pose severe problems for our main arguments.

We also acknowledge that immigrants of different ethnoracial and sex groups might have diverse experiences in the United States, including varying patterns of earnings assimilation, labor market reception, discrimination, and access to healthy diets and medical care (Bosdriesz et al. 2013; Villarreal and Tamborini 2018; Zheng and Yu 2022b)—all of which could contribute to differences in morbidity and mortality patterns. Our decomposition analysis does not further distinguish among immigrants because the morbidity–mortality paradox is evident across most groups and because our study focuses on exploring mechanisms behind the paradox rather than explaining subgroup differences. However, we demonstrated that the morbidity and disability rates of some immigrant groups surpass those of natives at older ages, whereas those of others do not. Future research could further investigate how the varied U.S. experiences among immigrant groups elucidate the group differences observed in our study. Lastly, our analyses are based on pooled cross-sectional data and synthetic cohorts. We have controlled for birth cohorts to obtain net age patterns, but future efforts utilizing large longitudinal data could help confirm our findings.

Beyond the morbidity–mortality paradox, results from our study have two broader implications. First, because our study points to the health-based selection as the primary mechanism behind the immigrant morbidity–mortality paradox, it also suggests that the health and mortality advantages observed among them will likely diminish in the second generation. Unlike their parents, children of foreign-born parents likely grow up in environments that endorse different and potentially unfavorable health behaviors, facilitate weaker family and community support, and subject them to lifelong social discrimination (Eschbach et al. 2007; Rodríguez et al. 2008). All these environmental factors are likely to weaken the health endowment of the second generation, which could shape their lifetime health and mortality trajectories.

Second, our finding of first-generation immigrants’ high likelihood of developing mental illness and diabetes at older ages deserves special attention from researchers and policymakers. Although some research has recognized immigrants’ enhanced risk of diabetes (Crimmins et al. 2004), the literature tends to assume that immigrants have better psychological well-being than natives owing to their stronger family and social support (except for Boen and Hummer 2019). Our study shows that the rising prevalence of mental illness among immigrants accounts for 39.9% and 55.5% of their decreasing advantages in functional and activity limitations, respectively, with age. The critical role of mental illness in immigrants’ aging process implies that prioritizing treatments of immigrants’ mental health issues is a crucial step in improving their old-age well-being.

Acknowledgments

An early version of this article was presented at the Office of Population Research at Princeton University, the Research Hub of Population Studies at the University of Hong Kong, and the Department of Sociology at Case Western Reserve University. We thank Dale Dannefer, Eric Fong, Noreen Goldman, Tod Hamilton, Arun Hendi, Robert Hummer, Jessica Kelley, Sanyu Mojola, Rebecca Schut, Haoming Song, Ming Wen, and Yu Xie for useful comments and suggestions. This publication was supported by grants P2CHD058484 and P2CHD041022, funded by the National Institutes of Health. The article's contents are solely the authors’ responsibility and do not necessarily represent the official views of the National Institutes of Health or the Department of Health and Human Services.

Notes

References