Trends in All-Cause Mortality Among US Veterans With Sarcoidosis, 2004-2022

Abstract

Background

Sarcoidosis is an idiopathic multiorgan disease with variable clinical outcomes. Comprehensive analysis of sarcoidosis mortality in US veterans is lacking.

Research Question

What are the trends in all-cause mortality among US veterans with sarcoidosis, and how are these trends influenced by demographics, Black vs White racial disparities, and geographic variability in relationship to mortality?

Study Design and Methods

Using Veterans Health Administration (VHA) electronic health records (EHRs), we conducted a population-based retrospective cohort study of adjusted all-cause mortality from 2004 through 2022 among veterans with a diagnosis of sarcoidosis who received care through the VHA. Demographics, region of residence, service branch, tobacco use, and comorbidities were extracted from the EHR. Annual trends in all-cause mortality and patient-level characteristics associated with mortality were examined with multivariable ungrouped Poisson regression. We visualized trends and analyzed state-by-state mortality using the marginal means procedure. In subgroup analysis (2015-2022), we considered the impact of neighborhood-level socioeconomic disparities using the Area Deprivation Index (ADI).

Results

In all, 23,745 veterans received a diagnosis of sarcoidosis between 2004 and 2019 and were followed up through 2022. After adjustment, including age and sex, all-cause mortality increased annually by 4.7% (P < .0001) and was 6.4% higher in Black than White veterans (mortality rate ratio, 1.064; P = .02). A subgroup analysis comparing models with and without ADI adjustment showed no meaningful change in mortality trends. Risk factors for increased all-cause mortality included older age, male sex, Black race, Northeast residence, and lower risk with other service branches. Despite distinct geographical variations in mortality rates, no clear patterns emerged.

Interpretation

Mortality among veterans with sarcoidosis is rising. Differences identified by service branch and higher risk among male Veterans raise questions about differences in environmental exposures. The narrower racial disparities and smaller impact of ADI than in other studies may highlight the role of universal health care access in achieving equitable outcomes.

FOR EDITORIAL COMMENT, SEE PAGE 1263

Take-Home Points.

Study Question: What are the trends in all-cause mortality among veterans with sarcoidosis, and how do these trends vary by demographics and other service- and health-related characteristics?

Results: Mortality among veterans with sarcoidosis is rising, with higher risk observed among male and Black Veterans, differences identified by service branch, and persistent but narrower racial disparities. No significant impact of neighborhood socioeconomic status was detected.

Interpretation: The findings suggest that differences in environmental exposures may contribute to the observed mortality trends among veterans, and the narrower racial disparities and smaller impact of neighborhood socioeconomic status compared with other studies may highlight the benefits of universal health care access provided through the Veterans Health Administration.

Sarcoidosis, a heterogeneous, multiorgan disease of unknown cause, is characterized by nonnecrotizing granulomas in affected organs, most commonly the lungs.¹ Clinical outcomes vary. The all-cause mortality rate has been reported as 11 per 1,000 person-years (PYs) in Sweden.² A US civilian population study between 1988 and 2007 showed an age- and sex-standardized sarcoidosis-related mortality rate of 4.3 per 1,000,000 people,³ similar to that in the English population.⁴ However, the mortality rate has increased, especially among Black women 55 years of age or older.³ It is unclear whether these results apply to veterans who are older, predominantly male, and often of White, non-Hispanic ethnicity.^5,6 Although veterans experience racial disparities in socioeconomic status and health outcomes, these differences tend to be smaller than in the general population.⁷

Veterans have higher mortality rates and shorter life expectancies than the general population.⁸ Elevated all-cause mortality may be associated with specific exposures (eg, Agent Orange⁹ and inorganic dust¹⁰). A recent study found reduced cardiovascular disease mortality in civilian women, but not in female veterans.¹¹ Analysis of Veterans Health Administration (VHA) data showed that the incidence of sarcoidosis was 5-fold higher than in civilians and 3 times more prevalent¹² and varied by branch of service, raising questions about service-related toxic exposures. However, comprehensive analyses of disease mortality in veterans are lacking.

We analyzed data from the electronic health records (EHRs) of the VHA, the largest integrated health care system in the United States, which offers comprehensive longitudinal care to veterans.¹³ Our goals were to analyze trends in annual all-cause mortality, to examine Black vs White racial disparities and geographic variability in those trends, and to identify the association of demographics, including service branch, and all-cause mortality in a national cohort of US veterans with sarcoidosis.

Study Design and Methods

Study Design and Data Source

This was a population-based and historical cohort study among veterans who received care through the VHA. We analyzed data extracted from VHA EHRs to the central Corporate Data Warehouse, which contains records for approximately 24 million veterans seen at a facility after October 1999, and analyzed with the Department of Veterans Affairs (VA) Informatics and Computing Infrastructure.¹⁴ The Corporate Data Warehouse is linked with VA sources that provide more complete mortality data than the National Death Index.¹⁵ All-cause mortality was used to avoid bias from incorrect adjudications about the cause of death.¹⁶ The study was approved by the University of California San Francisco Institutional Review Board (Identifier: 15-16660) and the San Francisco Veterans Affairs Healthcare System Research and Development Committee (Identifier: ARJM-0009). A Health Insurance Portability and Accountability Act waiver was obtained to use identifiable data (eg, diagnosis dates).

Defining the Sarcoidosis Population

As described elsewhere,¹² patients were defined as those who underwent either 1 inpatient visit or 2 outpatient visits at least 1 month apart at which the International Classification of Diseases codes for sarcoidosis were documented in the EHR. Our definition further required an absence of International Classification of Diseases codes for alternative diagnoses¹⁷ within 2 years before and up to 1 year after, unless the patient died within that year, in which case they were included as long as no alternative diagnoses were recorded before death.⁴ Incident patients required the absence of sarcoidosis-specific diagnostic codes within 2 years before the index date.¹²

We enrolled all patients with sarcoidosis who received a diagnosis between January 1, 2004, and December 31, 2019, using EHR data from veterans who received care at a VA facility or non-VA facility paid for by the VA. Data were analyzed annually until death or the end of follow-up (time from the index date [first sarcoidosis-specific diagnostic code date] to death or the end of the study [December 31, 2022]).

Demographic and Disease-Related Variables

In all cases, we extracted demographic data (age, sex, race, ethnicity), residence region,¹⁸ and rurality.¹⁹ Race was coded as Black, White, or other (eg, Asian, mixed race). Service in the military included Army, Air Force, Navy, Marines, or other (eg, Coast Guard, National Guard). Tobacco use was defined dichotomously by any positive tobacco use determined from the health factor, dental, or Systematized Nomenclature of Medicine Clinical Terms domains or any claims with procedural codes of tobacco use disorder.^12,20

Comorbidities and Neighborhood Socioeconomic Status

Patients were considered to have previous comorbid conditions if they had International Classification of Diseases codes for the following leading causes of death in the United States: ischemic heart disease, heart failure, chronic obstructive lung disease, pneumonia or influenza A or B, stroke,²¹ or any of the cancers most prevalent among veterans (melanoma or lung, prostate, colorectal, or bladder cancer).²² Posttraumatic stress disorder and depression were considered comorbid conditions because they increase the risk of intentional or unintentional injury.²³

The Area Deprivation Index (ADI), which aggregates 17 census variables, was used to examine the impact of neighborhood-level socioeconomic disparities. The national ADI is reported as a percentile score from 1 (lowest disadvantage) to 100 (highest disadvantage).²³

Statistical Analyses

R version 4.3.2 software (R Foundation for Statistical Computing) was used for statistical analyses. Descriptive statistics are presented as PYs of follow-up, number of deaths, and raw (unmodeled) mortality rates (RMRs) by covariate category; RMRs were calculated by dividing new deaths by total PYs standardized per 1,000 PYs. Missing data for categorical variables were coded as unknown.

To examine annual trends in adjusted all-cause mortality rates, we conducted multivariable Poisson regression analyses of ungrouped data, a method equivalent to the risk-set approach in Cox proportional hazards regression, but with the advantage of directly estimating mortality rate ratios (MRRs) while accounting for follow-up time variability.²⁴ Additionally, it supported adjustments for many covariates, offering analytical benefits over methods that may not handle such complexities as effectively. For this approach, the data set was reshaped to include the unique identifier for each veteran and the year of observation. Each veteran contributed multiple observations, and the total number of observations equaled the total PY from diagnosis to death or the last day of follow-up. All entries after the year of death were removed; thus, each unit of person-time at risk was a distinct observation.²⁴

To calculate unadjusted MRRs, we fitted a Poisson regression model of our reshaped data, where the unit of observation was an individual PY. To calculate adjusted all-cause MRRs, we incrementally adjusted for the following covariate sets, selected based on the directed acyclic graph informed by prior literature^12,25, 26, 27: demographics (age, sex, race, ethnicity), rurality and region, branch of service, and tobacco use. Atop the previously adjusted multivariable model, we also adjusted for common causes of death in the general population. We reran our primary analysis using a complete case approach as a sensitivity analysis. To ensure an unbiased assessment of the effects of neighborhood-level socioeconomic disparities, we conducted a subgroup analysis of a period for which complete ADI data were available (2015-2022); the most recent ADI score was included as a continuous variable. We aimed to determine whether these covariates accounted for the observed trends in all-cause mortality. A P value of < .05 was considered significant; all hypothesis tests were 2-sided.

To investigate racial disparities in annual adjusted all-cause mortality rates, we extracted race coefficients from the ungrouped Poisson regression model, using different adjustment sets to estimate MRRs for Black vs White veterans. To explore MRRs over time, we tested for an interaction term between race and calendar year. To estimate geographic variability in all-cause mortality, we used the same Poisson models, but included the state variable instead of the region variable in all adjustment sets.

To calculate all-cause mortality, to visualize mortality trends (overall and in Black and White veterans), and to examine state-to-state variability in all-cause mortality, we used an estimated marginal means (EMM) procedure,²⁸ which allows comparisons across adjustment models by averaging mortality rates from ungrouped Poisson regression models across all predictor categories. This analysis was used to account for different covariate distributions to ensure meaningful comparisons.

Results

Descriptive Statistics

We identified 23,745 patients with sarcoidosis; 1,971 veterans (8.3%) received non-VA purchased care, and 6,985 veterans (29%) died between 2004 and 2022. The RMR per 1,000 PYs varied across patient characteristics (Table 1). RMRs increased with age and were highest in those aged > 70 years and were higher in male than female veterans and in White veterans than in Black and Asian veterans. RMRs were highest in the Northeast, lowest in the South, and similar in rural and urban areas. RMRs also were similar in Air Force, Army, Marine, and Navy veterans and lowest in veterans of other branches. Those who had ever used tobacco showed a slightly lower RMR than those who never used tobacco. Among veterans with comorbidities, RMRs were highest in those with a history of lung cancer and lowest in those with a history of posttraumatic stress disorder.

Table 1.

Characteristics and Raw Mortality Rates of US Veterans With Sarcoidosis

Characteristic	Veterans With Sarcoidosis (N = 23,745)
	No. (%)	Total No. of Deaths	Total Person-Years	Raw Mortality Rate Per 1,000 Person-Years
Crude	23,745	6,985	225,196	31
Age, y
18-30	480 (2)	29	4,890	5.9
31-50	8,011 (34)	1,226	89,232	13.7
51-70	12,838 (54)	4,176	117,726	35.5
> 70	2,416 (10)	1,554	13,348	116.4
Sex
Female	3,202 (13.5)	475	32,781	14.5
Male	20,543 (86.5)	6,510	192,415	33.8
Race
White	9,677 (41)	2,944	86,058	34.2
Black	12,378 (52)	3,341	124,384	26.9
Asian and other	261 (1)	70	2,432	28.8
Unknown	1,429 (6)	630	12,322	51.1
Ethnicity
Hispanic	680 (3)	171	6,240	27.4
Non-Hispanic	22,286 (94)	6,424	212,963	30.2
Unknown	779 (3)	390	5,993	65.1
Residence by region
South	12,870 (54)	3,488	125,737	27.7
Northeast	3,271 (14)	1,140	30,096	37.9
Midwest	4,251 (18)	1,289	38,611	33.4
West	3,145 (13.2)	975	29,021	33.6
US territories	127 (0.5)	47	1,111	42.3
Unknown	81 (0.3)	46	620	74.2
Rural residence
Rural	6,439 (27.3)	1,864	61,314	30.4
Urban	17,131 (72.2)	5,085	162,713	31.3
Unknown	121 (0.5)	36	1,169	30.8
Branch of service
Air Force	3,129 (13.3)	1,012	30,604	33.1
Army	11,801 (49.7)	3,737	117,689	31.8
Marine	1,926 (8.1)	606	18,511	32.7
Navy	3,494 (14.7)	1,073	33,982	31.6
Other	2,213 (9.3)	255	12,353	20.6
Multiple	1,130 (4.7)	297	11,701	25.4
Unknown	52 (0.2)	5	356	14
Ever tobacco use
No	6,115 (25.7)	1,780	54,146	32.9
Yes	17,630 (74.3)	5,205	171,050	30.4
Ever PTSD
No	16,257 (68.5)	5,460	151,338	36.1
Yes	7,488 (31.5)	1,525	73,858	20.6
Ever depression
No	10,287 (43.4)	3,515	91,575	38.4
Yes	13,458 (56.7)	3,470	133,621	26
Ever pneumonia or influenza
No	16,157 (68)	3,788	152,247	24.9
Yes	7,588 (32)	3,197	72,949	43.8
Ever COPD
No	16,160 (68.1)	4,668	147,940	31.6
Yes	7,585 (31.9)	2,317	77,256	30
Ever ischemic heart disease
No	15,246 (64)	3,414	144,660	23.6
Yes	8,499 (36)	3,571	80,536	44.3
Ever heart failure
No	17,636 (74)	3,881	170,703	22.7
Yes	6,109 (26)	3,104	54,493	57
Ever stroke
No	19,763 (83)	5,171	186,828	27.7
Yes	3,982 (17)	1,814	38,368	47.3
Ever lung cancer
No	22,246 (93.7)	6,099	212,705	28.7
Yes	1,499 (6.3)	886	12,491	70.9
Ever colorectal cancer
No	22,713 (95.7)	6,652	214,880	31
Yes	1,032 (4.3)	333	10,316	32.3
Ever prostate cancer
No	21,327 (89.8)	6,136	200,554	30.6
Yes	2,418 (10.2)	849	24,642	34.5
Ever bladder cancer
No	23,333 (98.3)	6,807	221,363	30.8
Yes	412 (1.7)	178	3,833	46.4
Ever melanoma
No	23,422 (98.6)	6,881	221,937	31
Yes	323 (1.4)	104	3,259	31.9

PTSD = posttraumatic stress disorder.

Adjusted All-Cause Mortality Rates and Trends

All-cause mortality rates increased annually (Table 2) and rose with each adjustment set. After adjustment for demographics, the mortality rate increased by 2.6% annually. When comorbidity history was included in the fully adjusted model, the annual increase was 4.7%, and the EMM of mortality rates ranged from 11.1 to 25.5 patients per 1,000 PYs (Fig 1). In a subgroup analysis of 2015 through 2022 data, additional adjustment with the ADI minimally impacted annual all-cause mortality from 6.7% to 6.8% (95% CI, 1.03-1.10; P < .001 in both models).

Table 2.

Mortality Rate Ratio Determined by Ungrouped Poisson Regression per Calendar Year and for Black vs White Veterans Adjusted for the Different Sets of Covariates Among US Veterans With Sarcoidosis

Model (Adjustment Set)	Year MMR (95% CI)	Black Race MMR (95% CI)
Unadjusted	1.021 (1.016-1.026)	0.79 (0.76-0.84)
Plus demographic characteristicsa	1.026 (1.022-1.031)	1.076 (1.02-1.13)
Plus rurality and regions	1.027 (1.022-1.032)	1.070 (1.01-1.13)
Plus branch of service	1.031 (1.026-1.036)	1.057 (1.001-1.12)
Plus tobacco use	1.031 (1.026-1.036)	1.058 (1.002-1.12)
Plus comorbid conditions	1.047 (1.042-1.053)	1.064 (1.01-1.13)

MMR = mortality rate ratio.

^aInclude age, sex, race, and ethnicity.

Figure 1.

A, Graphs showing annual estimated marginal means of the mortality rate trends between 2004 and 2022 adjusted for the different sets of covariates among US veterans with sarcoidosis. B, Graphs showing the annual estimated marginal means of the mortality rate trends by Black (gray line) vs White (blue line) veterans between 2004 and 2022, adjusted for the different sets of covariates among US veterans with sarcoidosis.

Trends by Race

In an ungrouped Poisson regression model examining the association between race and mortality rate adjusting for the calendar year, the MRR of Black to White veterans was 0.79 (95% CI, 0.76-0.84; P < .001); however, after adjustment for other demographic factors, it was 1.076 (95% CI, 1.02-1.13; P = .01). The MRR of Black veterans in the full model was 1.064 (95% CI, 1.01-1.13; P = .02), indicating a 6.4% higher mortality rate among Black compared with White veterans (Table 2). No interaction between race and calendar year was found. Reassuringly, sensitivity analysis with complete cases yielded highly similar results (e-Table 1), and conclusions remained internally consistent. Figure 1 shows different EMM of mortality rate trends by race.

Geographic Variability

The geographic distribution of adjusted all-cause mortality varied considerably across different models (Fig 2). The EMMs of mortality rates were highest in the District of Columbia, New York, New Jersey, Pennsylvania, Oklahoma, Mississippi, Alabama, Nevada, and Wyoming. The comprehensive model, including all 5 sets of covariates and the state covariate, yielded EMMs ranging from 10.5 patients per 1,000 PYs in Vermont to 27.2 patients per 1,000 PYs in Wyoming. Although all other models showed that the EMM was highest in the District of Columbia, this pattern was altered in the full model. Overall, we found state-level differences in adjusted all-cause mortality, but no distinct regional differences.

Figure 2.

A-E, Maps showing the geographic distribution of estimated marginal means of the mortality rate between 2004 and 2022, adjusted for the different sets of covariates among US veterans with sarcoidosis: demographics (A), rurality and region (B), ever tobacco use (C), branch of service (D), and comorbidities (E).

Risk Factors

Several risk factors—age group of 51 to 70 years and older than 70 years at diagnosis, male sex, Black race, Hispanic ethnicity, living in the South, and Army or Air Force service—were associated significantly (P < .05) with higher all-cause mortality rates in the fully adjusted model (Fig 3). Mortality risk was 2-fold greater in veterans 51 to 70 years of age and 5-fold greater in those older than 70 years than in those younger than 30 years. Mortality risk was 73% higher in male than female veterans, 6.4% higher in Black than White veterans, and 9.3% higher among those who lived in the Northeast (vs West). Taking service in the Navy as a referent, serving in other branches was associated with a 35% lower risk.

Figure 3.

Multivariable Poisson regression modeling of risk factors associated with all-cause mortality among US veterans with sarcoidosis between 2004 and 2022. Values are mortality rate ratios; error bars indicate the 95% CI. Red bars indicate a significantly associated increased risk; blue bars indicate a significantly associated decreased risk; gray bars indicate a nonsignificant association. BOS = branch of service

Discussion

This study examined adjusted annual all-cause mortality rates of 23,745 US veterans with sarcoidosis between 2004 and 2019 and followed up through 2022. It showed that mortality rates increased by 4.7% annually, reached 25 deaths per 1,000 veterans by the end of the study and were slightly higher in a subgroup analysis that included ADI. Racial disparities in all-cause mortality were smaller than in previous studies. Mortality rates varied by service branch and geographically, but no regional patterns emerged. A better understanding of trends in all-cause mortality for sarcoidosis may help policymakers to develop effective systems for comprehensive, personalized management of sarcoidosis.

The rising mortality rates are consistent with multiple studies across the world, including the United States.^2,3,25,27,29 However, a study of sarcoidosis-specific mortality in the general population found a trend toward stable longitudinal mortality.³⁰ This difference was unsurprising because that study adjusted for age only and used sarcoidosis-specific mortality, whereas we adjusted for many possible confounders and used all-cause mortality to avoid ascertainment bias. As in other studies, Black race was a mortality risk.³ In contrast to the general population,³ we found male Veterans to be at higher risk of mortality than female veterans. Although the exact meaning of this finding deserves further research to clarify, one strong possibility is exposure related. Hispanic ethnicity also increased mortality risk, a finding that will require follow-up in dedicated studies, given the small sample of Hispanic veterans in this cohort.

Another consideration is whether disease phenotypes vary intrinsically between civilians and veterans. Mortality decreased among veterans in other branches, and a similar pattern was found previously,¹² suggesting that service-related exposures influence the natural history of sarcoidosis. Rising mortality rates have been attributed to the increasing age at diagnosis among patients with sarcoidosis, including veterans.^12,31 Although age and comorbid diseases of senescence were important risk factors, neither wholly explained the increasing mortality. For example, when we controlled for coronary heart disease, excess mortality increased. A meaningful fraction of these excess deaths are likely related to sarcoidosis.^2,3,31,32 Methodologic differences between studies preclude direct comparison, but the high average all-cause mortality rate in our study suggests differences in military vs civilian mortality rates as a focus of future studies.²

Given historic racial disparities in sarcoidosis outcomes, we looked for racial differences in mortality. Because differences in disease phenotype may represent genetic, environmental, and social factors, some have proposed an implied degree of inevitability in these disparities. Our results suggest that with a sufficiently robust, adaptive health care system, even baseline differences in disease severity that may have a biological origin do not necessarily result in similarly disparate treatment outcomes. For example, racial disparities in screening and disease-specific survival in patients with prostate cancer were reduced or reversed within the VHA.³³ In another study, Black veterans with lung cancer received equitable access to curative resection via lobectomy, which has not been replicated in the broader American health care system.³⁴ Because the military was desegregated in 1948,³⁵ the greater longitudinal experience with interracial collaboration may have contributed to these results. Features related to the VHA’s universal health care model (e., low cost and ease of access to specialty care) also may have contributed. More work is needed to identify and disseminate the features of the VHA that contribute to racially equitable care.

Our results also suggest progress in health equity regarding socioeconomic disparities. Among civilians with sarcoidosis, living in lower-resourced neighborhoods is associated with accelerated loss of lung function.³⁶ Such neighborhoods have worse outcomes after hospitalization across a range of respiratory diseases.³⁷ However, in our subgroup analysis, including ADI minimally affected the observed mortality trends, suggesting that increased morbidity from neighborhood-level socioeconomic disparities does not translate to increased mortality or that the VHA’s unique design helps to obviate some structural disadvantages of these patients.

We did not directly measure psychosocial challenges, another important nonbiological consideration. Reintegration into civilian life is a known challenge, and psychosocial issues related to military service and the stigma surrounding them are barriers to treatment.³⁸ Military service is associated with increased odds of mental health conditions, and psychological symptoms are associated with a greater likelihood of hospitalization.³⁹ A holistic approach considering the entire range of social determinants likely will be necessary to understand or address disparities among veterans with sarcoidosis.

Despite conflicting evidence on survival among veterans vs civilians, veterans on active duty during wartime may lose the survival benefit thought to come with military service.⁴⁰ Of particular concern are service-related exposures that could influence sarcoidosis incidence or mortality. In contrast to civilian populations,³ men were at higher mortality risk than women, possibly reflecting different exposure profiles between the two groups, much as we hypothesize between civilians vs veterans. Combat roles were not fully open to female veterans until 2016,⁴¹ and given the demographics of the present cohort, their combat deployments likely reflect this prior difference in role assignment. We postulate that male veterans having a higher likelihood of combat deployment could have more toxic exposure that yielded higher mortality. Branch of service may be a proxy for service-related exposures linked to risk of sarcoidosis¹² and similarly might influence mortality rates. In our analysis, serving in other branches was associated with a lower risk of all-cause mortality. Those veterans are less likely to be deployed into combat operations, which decreases their likelihood for service-related toxic exposure. However, providers and veterans may be unaware that complex exposure patterns may influence the risk of sarcoidosis and similar conditions.⁴²

Exposures that may differ across service branches and require further exploration include inorganic metal dust and burn pits.⁴³ This mix of influences and unidentified factors may help to explain the stark contrast to the decreasing mortality resulting from other diseases.⁴⁴ An urgent need exists to disentangle these contributors to guide interventions better and improve patient outcomes.

In a study of state health rankings and sarcoidosis-related mortality in civilians,⁴⁵ some of the lowest mortality rates were in Vermont and some of the highest were in Louisiana and Mississippi, as in our study. These states were at the high and low extremes of the health ranking used as a primary predictor in that study. The authors suggested that sarcoidosis-related mortality reflects differences in local policies that shape the social safety net, racial equality, and public health resources, as in other mortality analyses in marginalized populations. In a study of female mortality in the general population, state-level factors (eg, affordable housing and education spending per capita) were associated with mortality rates and were more than twice as important as individual-level factors in explaining the data variance.⁴⁶ The role of the state-specific or region-specific exposures remains to be explored, although areas where mortality is highest do not overlap with those with the greatest number of sarcoidosis cases.¹²

Our findings raise issues for future inquiry. More work is needed to understand how social context affects care for patients with sarcoidosis and the impact of mental illness and psychosocial support. Work is also needed to understand how comorbidities and their interactive effects impact sarcoidosis. These analyses will be incomplete without a better understanding of the influence of exposure history. Differences in exposures may help to explain why our state-based analysis is discordant with our understanding of the effects of state-level social services or other social determinants. Also, deeper inquiry and analysis into place-based relationships in sarcoidosis mortality could reveal important association. Similarly, the next step in understanding the higher mortality among male veterans and differences between service branches is to identify and characterize differences in service-related exposures, including burn pits, inorganic dust, beryllium, and other substances not yet identified. A comprehensive strategy to decrease sarcoidosis mortality requires a better understanding of how service-related exposures and social conditions influence the course of the disease.

Strengths and Limitations

The primary strengths of this study are its large sample size and the long follow-up time. The VHA is notable for its integration, low cost, and focus on the care of veterans, whose exposure histories differ from those of the general population. However, our study has limitations. Although the VHA replaced the Vital Status File, a registry for mortality ascertainment since 2006, with the Death Ascertainment File in August 2023, around 99% of the death information is shared between the two files.¹⁵ Although privately paid community care is known to occur among veterans, we are unaware of any database that would allow us to capture this for inclusion in our analysis. Although we adjusted for some comorbidities and tobacco use, complex causal pathways meant that their point estimates in this analysis were at risk of being confounded or mediated; dedicated work is required to understand their true effect. The comorbidities considered were not comprehensive, and we could not assess disease severity in individual cases, which may have differed by race. In selecting major causes of death among all veterans, we could not compensate fully for racial differences in the leading causes of death in the relevant age groups.⁴⁷ Patterns of nonrandom missingness limited our ability to assess the impact of socioeconomic differences for the entire study period, so we may have underestimated the impact of ADI on mortality. Finally, psychosocial factors are not well captured by structured VHA data and could not be included.

Interpretation

In our retrospective examination of veterans with sarcoidosis, we found that adjusted all-cause mortality is rising, with higher risk among male veterans and significant geographic and service-related variability, which may reflect divergent exposure patterns. Racial disparities were persistent but narrower than in previous reports, and neighborhood socioeconomic status played a smaller role than expected. Both suggest the potential benefit of a universal health care system as offered through the VHA. To develop a truly effective strategy to address sarcoidosis, we must capture the complete range of environmental exposures, social conditions, and individual-level health challenges in patients with sarcoidosis. The complexities we found emphasize the need for in-depth future research.

Funding/Support

This work was supported by the National Center for Advancing Translational Science, National Institute of Health, through the University of California, San Francisco, Clinical Research Informatics Postdoctoral Fellowship Award [Grant TL1-5TL1TR001871-05 to M. I. S.]; a University of California, San Francisco, Academic Senate Committee on Research-Resource Allocation Program grant to M. I. S.; the National Heart, Lung, and Blood Institute of the National Institutes of Health [Diversity Supplement Award 3R01HL157533-02S1 to M. I. S. and Grant R01HL157533 to L. L. K.]; the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development [M. I. S.]; the Oscar Auerbach Visiting Scholar Program, VA Airborne Hazards and Burn Pits Center of Excellence [M. I. S.]; the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development Quality Enhancement Research Initiative [Grant I50-HX002756 to M. A. W.]; the California Tobacco-Related Disease Research Program [Grant T29IR0715 to M. A.]; and the Flight Attendant Medical Research Institute [Grant CIA190001 to M. A.].

Financial/Nonfinancial Disclosures

The authors have reported to CHEST the following: M. A. reports grants from the Departments of Defense [Grant W81XWH-20-1-0158], Veterans Affairs [Grant CXV-00125], the Flight Attendant Medical Research Institute [Grants 012500WG and CIA190001], and the California Tobacco-related Disease Research Program [Grant T29IR0715], and financial support from Guardant Health and Genentech during the conduct of the study. M. A. W. reports financial support from the US Veterans Health Administration, the National Institute of Health—NCATS, and McGraw Hill Education. None declared (M. I. S., M. T. A., A. D. B., W. F. G., S. M. N., K. F. G., C. E. M., L. L. K.).