Race and nativity are major determinants of tuberculosis in the U.S.: evidence of health disparities in tuberculosis incidence in Michigan, 2004–2012

The social underpinnings of tuberculosis (TB) disease have long been documented both in historical narratives and scientific literature. Yet, disparities in the incidence of TB related to nativity (where a person is born), race, and socio-economic status (SES) continue to persist despite organized TB control efforts. Applying a social determinants of health framework to infectious etiologies, specifically TB, could shed light on more distal social and environmental factors that may be inhibiting our ability to reduce enduring disparities in TB [1]. Such an approach may be able to shift our understanding from delineation of risk factors to a more comprehensive understanding of the processes producing such risk factors [2].

Despite a resurgence of TB in the U.S. from 1986 to1992, incidence is now at its lowest (3.2 per 100,000 in 2012) since routine reporting began in 1953 [3, 4]. That decline, however, has recently stagnated [3, 5], in both urban and rural populations [6, 7] and among foreign-born persons [5]. In addition, incidence of TB in the U.S. is much higher among racial/ethnic minorities, people of lower SES, those with HIV, the homeless, and the incarcerated [5, 8‐11]. Consistent with various studies that have examined the effects of SES on health [12], TB incidence shows an SES gradient, whereby people of lower SES experience greater risk of TB—a gradient much steeper among U.S.-born cases [10, 13, 14].

Many studies have reported disparities in TB incidence in the U.S., particularly racial disparities [15‐19]. However, few studies have contextualized these disparities in the larger framework of social and environmental determinants of health. In the late 1980s and early 1990s, several studies examined the social and geographic context of TB cases as means to understand the increase in TB incidence following the HIV epidemic [14, 20, 21]. More recently, studies in Washington State examined the relationship between neighborhood socioeconomic disadvantage and TB incidence and disease progression finding that residence in an area with greater neighborhood disadvantage is associated with increased TB incidence, accelerated progression of disease, and genotypic clustering, particularly among U.S.-born cases [22‐24]. Recent studies seeking to understand the drivers of these persistent disparities posit that SES and/or an unequal burden of TB risk factors may be confounding our understanding of racial and ethnic disparities in TB in the U.S. [15, 16]. It may be that investigations of the social and geographic context of individuals hold clues for understanding the drivers of TB disparities in different contexts, and for designing context-specific interventions to ameliorate them.

In the state of Michigan TB control has largely been successful; the incidence rate is consistently lower than the national average [15]. Yet, despite Michigan being a low-burden state, TB remains a notable public health issue, particularly among socially vulnerable populations. At present about 75% of TB cases occur in the Detroit Metro Area while only 39% of the Michigan population resides there [15, 25]. Moreover, about half of Michigan TB cases are foreign-born, compared to 63% of the national TB cases in 2012 [4]. In 2010, less than 8% of the Michigan population was foreign-born, compared to 13% of the U.S. population [26]. Understanding the differential impacts of social factors on TB infection and disease, and specifically how such factors differ for recently transmitted versus cases resulting from reactivation of latent TB infection (LTBI), will help reduce risk and improve treatment.

Much of the investigation of TB in the U.S. in recent decades has focused on high-burden settings. However, as we move from strategies aimed at controlling TB to strategies aimed at eliminating TB, it is increasingly important to understand the patterns of TB incidence in low-burden settings. This is critical given the levelling off of TB incidence in the U.S. in recent years [27]. Recent reports have emphasized the importance of local approaches to TB control; approaches that take into account context (including socio-demographic context) when thinking about how to best design TB control interventions [28, 29]. Having a more thorough understanding of TB epidemiology in low burden settings, and particularly understanding the local context can allow the development of more effective TB control strategies to achieve TB elimination.

Using Michigan surveillance data from 2004 to 2012, we examined TB incidence patterns for the entire population of Michigan, and within population subgroups. We used both genotypic and temporal data to investigate trends separately for cases due to recent transmission and those due to reactivation of LTBI. We specifically sought to document disparities in TB incidence by race, nativity, gender, and age in Michigan with the aim of laying the foundation for future studies that can explore the mechanisms underlying such disparities.

Of the 1254 cases in the sample, 473 (38%) fit our criteria for a clustered case, while 781 (62%) were considered non-clustered cases. The 473 clustered cases belonged to 95 unique clusters with the size ranging from 2 to 51 cases.

This study analyzed temporal changes in TB incidence patterns in Michigan during 2004–2012, with particular attention to subgroups by race, nativity, gender, and age. We found the incidence rate of TB to be declining overall in Michigan both for recently transmitted cases and reactivated cases. Our results suggest both ongoing transmission and reactivation of LTBI are actively contributing to the burden of TB disease in Michigan. Notably, we found significant subgroup disparities in the average incidence rate, particularly by race and nativity.

On the whole TB incidence is declining in the U.S. We observed an average yearly percentage decline of 8% in the incidence, greater than the average percentage decline (5%) reported for the U.S. as a whole in this time period [4, 35‐42]. The greater decline in Michigan might be partially explained by a population composition that differs from many of the more populous states contributing to the national TB burden such as California, Texas, New York, and Florida who accounted for nearly half of all TB cases in 2014 [43]. The higher TB incidence rate in these states may be driven by a larger proportion of foreign-born populations in these state as compared with Michigan [44].

Our findings show that both ongoing transmission and reactivation of LTBI are contributing to TB incidence in Michigan. Based on our analysis, an average of 62% of the cases in Michigan resulted from reactivation of LTBI. Studies in Arkansas have reported similar findings, though with smaller proportions of cases attributable to reactivation [7, 30]. Based on our analysis, TB control efforts need to implement additional measures that can both reduce Mycobacterium tuberculosis (MTB) transmission and reduce reactivation of LTBI. Routine contact tracing is a key component of the TB control strategy in Michigan; however, additional routine screening for LTBI and active TB, followed by timely appropriate treatment may be beneficial among certain sub-populations in Michigan.

Although the concept of race is widely recognized as a social construction, it represents a strong predictor of health and disease in the U.S. Racial classifications in the U.S. are based on a social definition of race, rather than a biological or genetic definitions [45]. Racial differences in the U.S. reflect both historic and contemporary differences in power, status, and access to resources [46]. Accordingly, differences by race, and the meaning of those differences, must be assessed when examining health disparities in the U.S. context.

In general, we did not find significant differences in the TB decline rate in our subgroup analyses. However, there was evidence of disparities in the average incidence rate, particularly by race and nativity. Among recently transmitted cases, the highest risk racial group was Blacks while among reactivation of LTBI cases it was Asians. The evidence is mixed regarding the racial distribution of TB. While some studies have also reported the greatest racial disparity in TB incidence between Blacks and Whites [47], the national surveillance data suggests the greatest racial disparities exist between Asians and Whites [43]. Our multivariable models allowed us to explore these disparities while accounting for the effects of nativity. For example, among both recently transmitted cases and reactivation of LTBI cases, we demonstrated that nativity confounded the relationship between race and TB disparities. In some cases, the racial disparities were reduced after accounting for nativity, while in other cases the disparities were amplified after accounting for nativity (results shown in Additional file 2: Table S2). Thus, accounting for nativity is critical to understanding the relationship between race and TB incidence.

Disparities by nativity were evident among both cases resulting from recent transmission as well as those resulting from reactivation of LTBI. Immigrants may be infected in their country of origin and may reactivate sometime later when in the U.S. [48‐50]. However, the disparity observed between foreign-born and U.S.-born persons among the recently transmitted cases suggests they are also at risk for acquiring TB infection while in the U.S. This may be resultant from increased exposure to active TB cases among other recent immigrants or it could it be a reflection of their social and environmental circumstances once in the U.S.

Our study is one of only a few to use a multivariable approach to disentangle the differential impact of socio-demographic factors on disparities in TB incidence in the U.S. Bivariate analyses of the socio-demographic patterning of TB risk miss important relationships that are only uncovered when controlling for other factors. Other studies that have used a multivariable approach have also found that social vulnerability based on minority race/ethnicity, nativity, and income was a better predictor of TB incidence than traditional TB risk factors such as over-crowding living conditions or high population density [51].

During the 19th and early 20th centuries, TB was thought to be endemic in the U.S., particularly among those living in poor, crowded housing without access to basic resources: clean water, sanitation, ample food [52, 53]. Yet, in recent decades the focus in infectious disease control in the U.S. has shifted to individual-level predictors of risk and disease, often ignoring the social and environmental context of individuals that may be a more salient predictor of risk. While we recognize that the same types and magnitudes of resource deprivation that occurred more than a century ago are not the same today, our findings indicate that, despite organized TB control efforts, contemporary disparities are still important in explaining TB patterns today, suggesting that it could be important to again address the social and environmental context in which TB cases are arising. It is not sufficient to simply describe the patterns of TB incidence without regard to how these factors are working together to augment risk in certain populations.

Additionally, despite our knowledge of the highly social nature of TB, control efforts in the U.S. rarely emphasize the social and environmental context as means for intervention. Recently, the World Health Organization has begun to recognize the inextricable links between social and economic determinants of TB incidence globally [52, 54]. Globally, the burden of TB often disproportionately falls on the most vulnerable populations. Hargreaves et al. suggests that global TB control needs to develop interventions that simultaneously incorporate the traditional biomedical approaches as well as approaches that address the social determinants of health [54]. These efforts have mainly been targeted to high-burden TB countries in order to reduce the global incidence of TB. However, we advocate for an extension of this expansive view of TB control to low burden TB settings as TB disparities often persist in these settings. While many studies show social factors ranging from neighborhoods to social cohesion to be meaningful predictors of TB risk, we now need to move to incorporating these findings into TB control in the U.S.

One critical limitation to our study was the availability of more detailed socio-demographic data. While our findings suggest significant social patterning of TB incidence, our ability to infer what may be driving the observed disparities is limited based on the data available. Future studies would benefit from the collection of more detailed socio-demographic data. This could be done by improving the data collected on the RVCT form or research studies with the explicit purpose of collecting such data.

Our incidence rate calculations were based on extrapolations of census data. Updated population denominators were not available for all years, possibly reducing accuracy of the population-level data. However, the resulting bias is likely non-differential, biasing the observed estimates towards the null. Additionally, there are several limitations inherent in the methodology utilized to define a genotypic cluster and thereby proxy recent transmission. Some misclassification of genotypic clustering may have occurred for those cases diagnosed in 2004 and 2012. To be classified as clustered, a case had to share a genotype and a diagnostic date within one year of another case. Thus, there may have been some 2004 cases that were misclassified as non-clustered since we could not link them to cases diagnosed in 2003 because of a lack of data for the 2003 cases. The same would be true of those cases diagnosed in 2012. We performed sensitivity analyses excluding 2004 and 2012 cases. The proportion of cases classified as clustered was not substantially different when excluding 2004 and 2012. Moreover in the multivariable models, the effect estimates were relatively close and were most often an underestimate of the disparity.

Moreover, while we believe that the addition of a time restriction in our genotypic cluster definition increased the specificity of the designation, it is also possible that the use of a one-year time period may systematically underestimate the true proportion of recent transmission. One year is the shortest cut point that has been used to distinguish recent transmission from reactivation disease in previous studies [55, 56]. For example, Borgdorff et al. found a significant portion of secondary cases were diagnosed more than one year after infection from the primary case [57]. In using the one year criterion, we followed a previously published methodology used in the U.S. context so as to allow our analysis to be comparable to previous reports regarding recent transmission trends [7, 30, 58]. We also believe that our use of this classification helps to reduce the effect of the relatively lower discriminatory power of 12-loci MIRU typing on the clustering rate, another methodological limitation that may result in underestimation of recent transmission in our study population.

Future studies should seek to employ multi-faceted methodologies for identifying recent transmission, such as the one described by Yuen et al. [59]. They used a plausible source-case methodology that used data on the genotype, infectiousness status, geographic location and time to classify cases as recent transmission. Accordingly, they found that only 14% of national cases were classified as recent transmission, notably lower than what we, and others, have found.

This is the first study offering a more in-depth analysis of the trends in TB incidence in Michigan. Our findings suggest that both ongoing transmission and reactivation of LTBI are contributing to the incidence of TB in Michigan. This is consistent with trends in Arkansas where studies have documented that ongoing transmission is occurring as well [7, 30]. In low incidence settings, such as Arkansas and Michigan, TB control is not often focused on reducing transmission of MTB. However, these studies suggest that additional strategies addressing ongoing control of TB transmission should be a key tenet of TB control programs in low incidence, as well as high incidence settings. As the U.S. moves towards TB elimination, more local approaches to TB control are needed that can address the specific socio-demographic composition in a given context and tailor strategies accordingly [28].

Additionally, disparities in incidence persist despite ongoing TB control efforts suggesting more targeted TB control is needed to reduce incidence in high-risk groups. The greater disparities by race and nativity demonstrate some of the ways that TB incidence is socially patterned, disproportionately affecting the most vulnerable members of the population. Reducing these disparities will require a multi-faceted approach encompassing the social and environmental contexts of high-risk populations.