Article Text

Unfavorable social determinants of health and risk of mortality in adults with diabetes: findings from the National Health Interview Survey
  1. Ryan Chang1,
  2. Jerrin Philip2,
  3. Umair Javed3,
  4. Anoop Titus4,
  5. Syed Karam Gardezi5,
  6. Harun Kundi2,
  7. Raman Yousefzai6,
  8. Adnan A Hyder7,
  9. Elias Mossialos8,
  10. Khurram Nasir2,9,10,
  11. Zulqarnain Javed2,4,9
  1. 1Baylor College of Medicine, Houston, Texas, USA
  2. 2Houston Methodist Academic Institute, Houston, Texas, USA
  3. 3Combined Military Hospital Lahore, Lahore, Pakistan
  4. 4Houston Methodist DeBakey Heart and Vascular Center, Houston, Texas, USA
  5. 5TJ Samson Community Hospital, Glasgow, Kentucky, USA
  6. 6AT Still University Kirksville College of Osteopathic Medicine, Kirksville, Missouri, USA
  7. 7George Washington University Milken Institute School of Public Health, Washington, DC, USA
  8. 8Health Policy, The London School of Economics and Political Science, London, UK
  9. 9Center for Cardiovascular Computational Health and Precision Medicine, Houston Methodist Hospital, Houston, Texas, USA
  10. 10Division of Cardiovascular Prevention and Wellness, Houston Methodist DeBakey Heart and Vascular Center, Houston, Texas, USA
  1. Correspondence to Ryan Chang; rchang2019{at}


Introduction Understanding the role of social determinants of health as predictors of mortality in adults with diabetes may help improve health outcomes in this high-risk population. Using population-based, nationally representative data, this study investigated the cumulative effect of unfavorable social determinants on all-cause mortality in adults with diabetes.

Research design and methods We used data from the 2013–2018 National Health Interview Survey, linked to the National Death Index through 2019, for mortality ascertainment. A total of 47 individual social determinants of health were used to categorize participants in quartiles denoting increasing levels of social disadvantage. Poisson regression was used to report age-adjusted mortality rates across increasing social burden. Multivariable Cox proportional hazards models were used to assess the association between cumulative social disadvantage and all-cause mortality in adults with diabetes, adjusting for traditional risk factors.

Results The final sample comprised 182 445 adults, of whom 20 079 had diabetes. In the diabetes population, mortality rate increased from 1052.7 per 100 000 person-years in the first quartile (Q1) to 2073.1 in the fourth quartile (Q4). In multivariable models, individuals in Q4 experienced up to twofold higher mortality risk relative to those in Q1. This effect was observed similarly across gender and racial/ethnic subgroups, although with a relatively stronger association for non-Hispanic white participants compared with non-Hispanic black and Hispanic subpopulations.

Conclusions Cumulative social disadvantage in individuals with diabetes is associated with over twofold higher risk of mortality, independent of established risk factors. Our findings call for action to screen for unfavorable social determinants and design novel interventions to mitigate the risk of mortality in this high-risk population.

  • Mortality
  • Diabetes Mellitus, Type 2
  • Public Health
  • Vulnerable Populations

Data availability statement

Data are available in a public, open access repository. All NHIS data are de-identified and publicly available at

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Adverse social determinants of health (SDOH) are associated with increased risk of diabetes and mortality; however, it remains unknown how cumulative social burden affects mortality risk in people with diabetes.


  • This study found that, in adults with diabetes, those experiencing the highest level of social disadvantage—assessed using 47 individual SDOH—had over twofold higher mortality risk, compared with those with the most favorable SDOH profiles.


  • Our findings highlight SDOH as a robust predictor of mortality in adults with diabetes, independent of traditional risk factors, and demonstrate the need to improve SDOH screening in order to identify high-risk patient populations and inform efforts to address SDOH and mitigate persistent disparities in mortality.


Over 1 in 10 adults in the USA suffer from diabetes, a leading cause of morbidity and mortality globally.1 2 Diabetes is an established risk factor for cardiovascular disease, chronic kidney disease, and vision disability, and often occurs alongside traditional risk factors including obesity and hypertension.3 Diabetes, with its numerous complications, significantly increases a person’s risk of morbidity and mortality and is among the top 10 leading causes of death in the USA.4

Increasing evidence points to social determinants of health (SDOH) as upstream drivers of persistent inequities in diabetes and associated cardiovascular events and all-cause mortality.5 6 Multiple unfavorable SDOH, such as low income, poor access to healthcare, and food insecurity, among others, have been associated with diabetes and contribute to worse health outcomes.7 8 However, existing approaches to assess mortality risk are predominantly based on “clinical” risk prediction algorithms which account for traditional measures of socioeconomic well-being, such as education or income, without a holistic assessment of cumulative social disadvantage across multiple SDOH domains.5 Current frameworks to capture SDOH burden are either based on solitary measures of individual SDOH or indices that capture a limited number of social determinants without due attention to the interconnected nature of SDOH across established domains.6 9 10 In addition, such indices have mostly been developed in subgroups without diabetes or in the general population,11 and predict non-mortality endpoints.12 To our knowledge, comprehensive SDOH indices have not been used to capture cumulative social disadvantage and assess mortality risk in adults with diabetes.

SDOH do not exist in silos; rather, they interact with each other and impact distal health outcomes. Despite the known association between SDOH and diabetes, SDOH and mortality, and diabetes and mortality, the extent to which cumulative social disadvantage may explain variation in mortality risk in adults with diabetes has not been previously investigated. Similarly, racial/ethnic and gender disparities in both diabetes and mortality are well documented.13 However, possible demographic disparities in the association between SDOH and mortality in adults with diabetes have not been studied to date.

In this population-based study, we sought to examine the impact of cumulative social disadvantage—measured using a comprehensive, validated SDOH framework—on all-cause mortality in a nationally representative sample of US adults with diabetes. We also assessed potential racial/ethnic and gender disparities in the SDOH–mortality association.


Data source and study design

This study used data from the 2013–2018 National Health Interview Survey (NHIS), linked to the National Death Index (NDI) for adults ≥18 years. The NHIS, collected by the National Center for Health Statistics in the Centers for Disease Control and Prevention, comprises a series of annual cross-sectional national surveys, which feature complex, multistage sampling to provide estimates on the non-institutionalized US population.14 15 The NHIS questionnaires collect information at the household, family, and personal levels, and are divided into four core components: household composition, family core, sample child core, and sample adult core.14 Data from NHIS 2013–2018 were used due to the richness of SDOH information collected during this time, resulting in a comprehensive list of 47 SDOH variables.16 17

We used NDI data for mortality follow-up through 2019, which represent the most current data on mortality that are linkable to the NHIS.18 With over 100 million death records, the NDI is the most complete source of mortality data in the USA, containing all death records for all 50 states and District of Columbia starting from 1979.19 The NDI uses the following identifiers for linkage to NHIS: social security number, first name, last name, father’s surname, date of birth, sex, race, and state/country of birth and residence. In the NDI, participants are followed quarterly per year for mortality ascertainment. Participant eligibility for linkage and inclusion in the linked data set was contingent on the availability and completeness of all aforementioned identifiers.

In this study, baseline was defined as the index quarter/year of survey participation for each NDI-linked NHIS participant, and October–December 2019 was used as the quarter/year of last follow-up. Additional methodological considerations for the NHIS-NDI data linkage have been reported in detail elsewhere.19–21


This study included all 2013–2018 NHIS participants ≥18 years of age with available death record information in the NDI (n=190 113). Participants with an invalid response to the diabetes screening question, unavailable information on death status, or insufficient identifying data (n=7668) were not eligible for linkage to NHIS and were therefore excluded from the study population.

Primary exposure

Social disadvantage, measured using quartiles of aggregate SDOH burden, was the primary exposure variable. We adapted the SDOH framework proposed by the Kaiser Family Foundation (KFF)22 and Healthy People 2030 (HP2030),10 organizing available SDOH information into six domains: (1) economic stability; (2) neighborhood, physical environment, and social cohesion; (3) community and social context; (4) food; (5) education; and (6) healthcare system. The KFF and HP2030 frameworks have been used and adapted widely in research and policy implementation settings to study disparities in diverse health outcomes. A list of 47 individual SDOH was constructed across the six domains (online supplemental eTable 1). Each SDOH was classified as either favorable or unfavorable and assigned a value of 0 or 1, respectively (eg, “0” for middle/high-income and “1” for low-income). An aggregate SDOH index was created by combining the 47 individual SDOH and dividing the range of values of the resulting index into quartiles, with the most favorable (ie, lowest) SDOH scores in the first quartile (Q1) and the most unfavorable (ie, highest) scores in the fourth quartile (Q4).

Supplemental material


Diabetes was assessed using the following question: “have you ever been told by a doctor or other health professional that you have diabetes (or sugar diabetes)?” Participants who responded “yes” were classified as having diabetes.


All-cause mortality, ascertained using death certificate records in the NDI, was the outcome event of interest. A participant was defined as dead if identified as “deceased” in the NDI during the study follow-up period. Cause of death was determined using the International Classification of Diseases, Ninth and Tenth Revisions.


Relevant covariates included sex (male and female); race/ethnicity (non-Hispanic white (NHW), non-Hispanic black (NHB), non-Hispanic Asian, Hispanic, and other); cardiovascular risk factor (CRF) profile, including hypertension, diabetes mellitus, smoking, and obesity (body mass index ≥30 kg/m2); and comorbidity profile, including atherosclerotic cardiovascular disease (ASCVD), chronic obstructive pulmonary disease, cancer, and chronic kidney disease. ASCVD was defined as a composite of coronary heart disease, angina, myocardial infarction, and stroke. CRF conditions and comorbidities were constructed as binary (yes/no) variables and aggregated to create separate indices, categorized as 0 (optimal), 1–2 (average), or ≥3 (poor) for CRF and 0, 1, or ≥2 comorbidities. All covariates were self-reported and obtained from the NHIS Sample Adult Core Questionnaire.19 20

Statistical analyses

Baseline participant characteristics were reported by diabetes status in the total population and by quartiles of SDOH burden in adults with diabetes. We also reported the distribution of individual SDOH in adults with and without diabetes. Weighted proportions were generated to report nationally representative estimates. Statistically significant differences in participant characteristics were assessed using χ2 tests. Poisson regression was used to generate age-adjusted mortality rates (AAMR) per 100 000 person-years (PYs) and 95% CI for each SDOH quartile, overall and by subgroups of interest, including sex, race/ethnicity, and comorbidity burden in adults with diabetes. For comparison, we also presented AAMR in the population without diabetes (online supplemental eTable 2).

Survival time was modeled as the number of years of follow-up from baseline (index quarter/year of survey participation) to death or end of study period (December 31, 2019), whichever came first, with a maximum follow-up time of 6 years. The proportional hazards assumption was checked by inspecting the log-log plots of survival (plots of ln(−ln[S(t)]) [log(−log) survival function] against survival time) for the primary exposure variable (SDOH quartiles), adjusted for study covariates. Plots for each quartile were found to be approximately parallel, thus satisfying the assumption of proportionality.

Kaplan-Meier survival curves were generated to estimate survival probability by SDOH quartiles in the total population and in adults with and without diabetes (online supplemental eFigure 1). Multivariable Cox proportional hazards regression models were used to derive adjusted HR (aHR) and 95% CI for all-cause mortality, overall and by sex and race/ethnicity in adults with diabetes. Four models were tested: model 1 adjusted for age; model 2 adjusted for age, sex, and race/ethnicity; model 3 adjusted for age, sex, race/ethnicity, and CRFs; model 4 adjusted for age, sex, race/ethnicity, CRFs, and comorbidities.


Descriptive characteristics

The final analytical sample comprised 182 445 adults ≥18 years of age, of whom 20 079 (11%) had a diagnosis of diabetes. Table 1 highlights the relevant demographic and clinical characteristics of the study population. Participants with diabetes were older (mean age in years (SD): 61 (16.3) vs 46 (10.2); 42.4% over 65 vs 16.6%) and more likely to be male (50.5% vs 47.9%) and NHB (15.5% vs 11.7%) than their counterparts without diabetes. Similarly, those with diabetes had worse CRF profiles (poor CRF: 49.0% vs 16.3%) and higher comorbidity burden (≥2 comorbidities: 28.9% vs 7.7%).

Table 1

Descriptive characteristics by diabetes status from the National Health Interview Survey 2013–2018

Among adults with diabetes, young and middle-aged, female, NHB, and Hispanic individuals were more likely to experience high levels of social disadvantage (SDOH-Q3/Q4), whereas NHW, male, and elderly individuals reported relatively favorable social profiles (SDOH-Q1/Q2) (table 2). Similarly, those with worse CRF profiles and higher comorbidity burden reported higher burden of unfavorable SDOH, relative to those with favorable clinical profiles.

Table 2

Descriptive characteristics by SDOH quartiles in adults with diabetes from the National Health Interview Survey 2013–2018

The distribution of individual SDOH characteristics by diabetes status is presented in online supplemental eTable 1. Individuals with diabetes had a higher burden of unfavorable SDOH across nearly all domains. Adults with diabetes were more likely to experience lower educational attainment and household income, poor access to healthcare, food insecurity, and live in disadvantaged neighborhoods compared with those without diabetes (figure 1). Kaplan-Meier survival plots showed decreasing probability of survival with increasing levels of social disadvantage (online supplemental eFigure 1). Adults with diabetes had lower probability of survival than their counterparts without diabetes at all levels of SDOH burden.

Figure 1

Distribution of individual SDOH characteristics by diabetes status. While people with diabetes were more likely to be insured than those without diabetes, they were also more likely to experience the other adverse conditions, such as limited transportation, food insecurity, and low-income status. All differences listed were statistically significant, with p values <0.001. Not all 47 SDOH are shown (see online supplemental eTable 1 for the full list). SDOH, social determinants of health.

Age-adjusted mortality rates

In individuals with diabetes, AAMR increased significantly with higher SDOH burden, from 1052.7 (95% CI 884.9, 1220.5) per 100 000 PYs in SDOH-Q1 to 2073.1 (95% CI 1827.2, 2319.0) per 100 000 PYs in SDOH-Q4. Similar patterns were observed across demographic and clinical subgroups (figure 2). Overall, mortality rates were higher for male, NHW participants, and those with poor CRF profiles or higher comorbidity burden, relative to female, Hispanic participants, and those with favorable CRF and comorbidity profiles, respectively. For nearly each sociodemographic and clinical subgroup, AAMR increased by 1.5-fold to 2-fold from SDOH-Q1 to SDOH-Q4.

Figure 2

Age-adjusted mortality rates in adults with diabetes. (A)AAMR, total and by sex in adults with diabetes. (B)AAMR by race/ethnicity in adults with diabetes. In individuals with diabetes, AAMR increased with higher SDOH burden, with significant differences between Q1 and Q4 for various demographics. (A) AAMR nearly doubled for both male and female participants in Q4 compared with their counterparts in Q1. (B) Differences in AAMR by race/ethnicity. While Hispanic and non-Hispanic black individuals had lower overall mortality rates, the association between higher SDOH burden and increased mortality risk was still present. AAMR, age-adjusted mortality rate; PY, person-years; Q1, first quartile; Q4, fourth quartile; SDOH, social determinants of health.

In the supplementary analyses, we found that mortality rates were generally higher for participants with diabetes compared with those without diabetes, overall and across sociodemographic and clinical strata. Patterns of mortality across SDOH quartiles were observed similarly for both groups, with people with diabetes facing sharper increases in mortality risk at higher levels of SDOH burden (online supplemental eTable 2).

Multivariable regression

Cox proportional hazard regression models demonstrated a consistent association between SDOH burden and mortality risk, independent of demographic and clinical covariates (table 3). In adults with diabetes overall, individuals in SDOH-Q4 experienced up to twofold higher risk of mortality, relative to those in SDOH-Q1 (aHR, Q4=2.37, 95% CI 1.94, 2.90). While the observed association was somewhat attenuated by clinical risk factors (models 3 and 4), higher social disadvantage was consistently associated with higher mortality risk, with risk increasing in a stepwise manner across increasing quartiles of SDOH burden.

Table 3

SDOH and risk of mortality in adults with diabetes from the National Health Interview Survey 2013–2018

Higher social disadvantage was associated with higher hazards of mortality for both male and female participants and for each racial/ethnic subgroup. Being in SDOH-Q4 was associated with 2-fold to nearly 2.5-fold higher risk of mortality for both male and female participants compared with those in Q1. Unfavorable SDOH burden was significantly associated with higher risk of mortality for only NHW participants. We found a significant interaction effect between race and SDOH (p for interaction=0.04); in models adjusted for age, sex, race/ethnicity, and CRFs (model 3), NHW individuals in SDOH-Q4 experienced over 2.5-fold higher risk of mortality (aHR=2.72, 95% CI 2.13, 3.47), whereas NHB and Hispanic adults in the same quartile (Q4) experienced approximately 1.5-fold higher hazard of mortality, relative to those in Q1; however, the estimates for the latter did not achieve statistical significance (NHB: aHR=1.43, 95% CI 0.88, 2.32; Hispanic: aHR=1.72, 95% CI 0.75, 3.97).


In this nationally representative study, we found that adults with diabetes generally experience greater burden of unfavorable SDOH compared with their counterparts without diabetes, and that higher SDOH burden is an independent risk factor for all-cause mortality in this population. Despite the known association between diabetes and mortality, as well as increasing evidence of the role of SDOH in explaining disparities in diabetes, relatively little is known about the effect of SDOH on all-cause mortality in the diabetes population.8 This is the first large-scale, nationally representative study, to our knowledge, to examine the role of cumulative social disadvantage—experienced across multiple SDOH domains—in determining mortality risk in adults with diabetes.23

Previous research has reported the association between adverse SDOH and diabetes risk, while protective factors such as social support have been shown to mitigate this risk somewhat.24 Diabetes is also linked with low socioeconomic status, environmental risk factors, poor access to healthcare, and food insecurity.8 However, prior studies have not assessed the association between cumulative social disadvantage and mortality in a national sample of adults with diabetes in the USA. Furthermore, prior work assessing aggregate SDOH burden and clinical outcomes is primarily based on a relatively small number of SDOH, is focused on the general population or defined clinical subgroups other than diabetes, or assesses non-mortality outcomes.25 26 A comprehensive SDOH index, as used in this study, may inform future development of holistic social risk assessment approaches and evidence-based, individualized social support interventions.27 In contrast to prior reports, ours is the first nationally representative study to comprehensively describe the burden of social disadvantage experienced by adults with diabetes and the extent to which it predicts mortality in this population. We found that, at each SDOH quartile, people with diabetes faced 1.5-fold to 2-fold higher AAMR than people without diabetes. Furthermore, higher SDOH burden was associated with over twofold increased risk of mortality in adults with diabetes, with the highest risk attributable to the highest degree of social disadvantage (SDOH-Q4). This pattern was observed similarly across race and sex, although with a stronger association for male and NHW adults.

Various pathways may explain the association between SDOH and diabetes observed in this study. Limited green space and exposure to environmental risk factors can increase the risk of diabetes and subsequent mortality.8 It is thought that low health literacy, which is often associated with low educational attainment, may further contribute to the link between low socioeconomic status and poor diabetes outcomes.28 Low income status and limited availability of nutritious food can also promote food insecurity, which is a significant risk factor for developing diabetes and experiencing associated complications like hypertension and cardiovascular disease.29 Additionally, low socioeconomic status is associated with poor glycemic control, which may further contribute to the observed link between adverse SDOH and increased mortality risk among people with diabetes.30 Future work can elucidate how various SDOH influence risk of diabetes, subsequent mortality, or both.

We found a relatively weaker SDOH effect on mortality for Hispanic and non-Hispanic black participants, which merits additional study. This finding may be explained by the higher burden of diabetes in Hispanic and NHB participants (relative to NHW participants), potentially attenuating the SDOH effect on mortality to a greater extent for Hispanic and NHB subgroups, relative to NHW subgroups. The weaker association for Hispanic individuals may also be attributed to lower overall all-cause mortality rates in this population, possibly contributing to relatively lower power for mortality assessment, especially given that mortality was assessed across four separate SDOH levels (Q1–Q4).31 While prior studies have pointed to potential health benefits of community support systems in this population, additional study is needed to fully understand potential racial/ethnic variation in community support systems, social networks, and neighborhood-level factors, with implications for coping with adverse SDOH and affecting downstream mortality risk.32 We also found a relatively stronger SDOH–mortality association for male compared with female individuals, which may be attributable to the higher overall mortality rate and higher prevalence of diagnosed and undiagnosed diabetes in the former.33 Additional research is warranted to fully understand the correlates and potential mediators of the “SDOH effect” on mortality in diverse gender and racial/ethnic subgroups.

While SDOH play large roles in determining the risk of diabetes, the reverse relationship is also worth noting. Indeed, individuals with diabetes are more likely to experience financial toxicity—defined as the negative financial consequences associated with disease—due to increased medical costs from clinical visits, medication, and treatment equipment.34 35 This bidirectional relationship may explain the greater burden of adverse SDOH on people with diabetes. Thus, our study underscores the importance of addressing socioeconomic barriers to diabetes prevention and treatment, whether it is through promotion of affordable insulin programs or reduction of insulin costs.36

Strengths and limitations

This study’s strengths lie in its large, nationally representative sample size and the application of a comprehensive SDOH framework comprising over 40 SDOH variables across six established domains to capture social disadvantage. Additionally, use of data from the NHIS and NDI—the principal sources of health and mortality information in the USA—enables generalizability of our findings to the adult US population with diabetes. We used multiple multivariate models to adjust for traditional risk factors of diabetes and/or cardiovascular disease, as well as established clinical predictors of mortality (such as cancer and ASCVD), in order to account for their potential confounding effect on the SDOH–mortality association. However, NHIS data are cross-sectional, which precludes assessment of potential temporal variation or change in SDOH burden. Future studies should consider replicating our methodology in longitudinal data sets to potentially capture temporal variation in SDOH. Furthermore, non-Hispanic Asian and American Indian/other subpopulations were not included in this study due to low sample size, which when divided among SDOH quartiles for mortality assessment would have yielded potentially unstable estimates due to low power. Another potential limitation lies in the self-reported nature of NHIS data, including diabetes, as well as lack of information about diabetes type. While prior reports have shown good correlation between NHIS data and clinically ascertained measures,37 the latter may reduce potential biases associated with self-report. Increased efforts should be made to enable cross-talks between survey-based and clinically measured data. Similar efforts should be made to capture diabetes type in population-based survey data, given prior evidence showing that patients with type 1 versus type 2 diabetes may have different SDOH profiles, with implications for downstream mortality risk.38 Uncovering these incompletely understood differences and the extent to which they affect the diabetes–mortality association could inform risk stratification and care pathways for adults with type 1 or type 2 diabetes.


Our work may provide further impetus to develop robust polysocial risk scores for mortality prediction in individuals with diabetes, as we reported previously for the ASCVD population.39 Contemporary risk prediction models are primarily reliant on clinical predictors and often ignore SDOH or consider a small subset of socioeconomic factors.40 Available population health databases provide unique opportunities to develop such indices and assess the effects of SDOH burden on mortality risk in patients with varying cardiovascular risk profiles. While assessing over 40 SDOH such as in this study may not always be feasible, particularly in clinical settings, polysocial risk scores provide a parsimonious prediction model which may lower the burden of screening. Similarly, there are increasing opportunities to integrate available validated indices such as the Social Vulnerability Index and the PRAPARE (Protocol for Responding to and Assessing Patients’ Assets, Risks, and Experiences) screening tool into clinical workflows via geocodes and electronic health record plug-ins.

In turn, screening patients for adverse socioeconomic conditions that may impact the risk of diabetes as well as mortality can help improve risk stratification and guide clinical care. For instance, efficient SDOH screening may highlight important barriers to diabetes care, such as food insecurity, pharmacy deserts, transportation barriers, or prohibitively high prescription costs.41 This may help develop critical partnerships between healthcare systems and community stakeholders to address unfavorable SDOH and mitigate their burden on adults with diabetes,42 with the goal of improving life expectancy and reducing mortality in this high-risk population.


The results of this national study suggest that US adults with diabetes experience high burden of unfavorable SDOH, which is associated with significantly elevated risk of all-cause mortality independent of established clinical predictors. Our findings highlight the importance of carefully screening for unfavorable SDOH and finding novel solutions to addressing SDOH to mitigate mortality risk in this clinically high-risk population.

Data availability statement

Data are available in a public, open access repository. All NHIS data are de-identified and publicly available at

Ethics statements

Patient consent for publication

Ethics approval

Given the publicly available and de-identified nature of the NHIS-NDI data, this study was exempt from purview of the Houston Methodist Institutional Review Board.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors RC: conceptualization, data interpretation, writing (first draft and revisions), and formatting. JP: writing (first draft and revisions) and final submission. UJ, AT: literature review, writing (revisions), and formatting. SKG: writing (revisions) and formatting. HK: critical review and formatting. RY: literature review, writing (revisions), and formatting. AAH, EM: critical review and revisions. KN: conceptualization, study design and methods, critical review, and revisions. ZJ: conceptualization, study design and methods, statistical analysis, writing (first draft and revisions), and critical review. Guarantors: ZJ and KN.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests KN is on the advisory boards of Amgen and Novartis and his research is partly supported by the Jerold B Katz Academy of Translational Research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.