Spatial Patterns of Tuberculosis and Diabetes Mellitus in Los Angeles County, California

Background Tuberculosis remains a public health problem that disproportionately affects vulnerable populations. The objective of the study was to apply Geographic Information Systems (GIS) methods to identify statistically significant hot spots of tuberculosis (TB) and diabetes mellitus (DM) and identify areas and populations that are disproportionately burdened by TB and DM. Methods Verified cases of TB reported in Los Angeles County (LAC), California between 01/01/2015 and 06/30/2017 were identified from the LAC TB Control Program Surveillance Registry. The addresses for patients residing in LAC at the time of TB diagnosis were geocoded and mapped. Hot spot analyses were performed utilizing the Getis-Ord Gi* statistic to identify statistically significant hot spots of TB and DM. Results Among 389 TB cases with DM, 43% were Hispanic and 50% were Asian. Geographic variations for Hispanic and Asian TB cases were found (p<0.05). Hot spots of TB and DM were identified among Asians residing in the southwestern and southern regions of LAC and among Hispanics residing in south central and northwestern LAC. Conclusions GIS methods are important epidemiological tools for identifying and assessing geographic variations in disease morbidity. These findings highlight opportunities for public health interventions aimed at reducing health disparities in underserved communities.

Registry. The addresses for patients residing in LAC at the time of TB diagnosis were geocoded and mapped. Hot spot analyses were performed utilizing the Getis-Ord Gi* statistic to identify statistically significant hot spots of TB and DM.
Results Among 389 TB cases with DM, 43% were Hispanic and 50% were Asian. Geographic variations for Hispanic and Asian TB cases were found (p<0.05). Hot spots of TB and DM were identified among Asians residing in the southwestern and southern regions of LAC and among Hispanics residing in south central and northwestern LAC.
Conclusions GIS methods are important epidemiological tools for identifying and assessing geographic variations in disease morbidity. These findings highlight opportunities for public health interventions aimed at reducing health disparities in underserved communities.

Background
Despite a decline in active tuberculosis (TB) disease rates in recent decades, both globally 1 and in the United States, 2 TB remains a public health concern. TB affects many communities disproportionately, with disparities in TB incidence across racial/ethnic groups, socioeconomic status, and nativity. 3 In Los Angeles County (LAC), California, 4 the 3 TB rate is higher than the rates in the state of California 5 and in the United States, 2 with the highest rates observed among Asian and Hispanic racial/ethnic groups.
The growing prevalence of diabetes mellitus (DM) is now a global health concern. 6 In the past 3 decades, the global DM prevalence has increased from 4.7% to 8.5%. The DM epidemic could contribute to an increase in TB burden and consequently pose a challenge to the World Health Organization's (WHO) "End TB Strategy" efforts 7 to end TB. 1, 8 The cooccurrence of DM is a TB risk factor that can affect TB disease presentation and treatment response. 9 Treatment of persons with DM poses a challenge for TB control programs since having DM increases the risk of progression from TB infection to active TB disease 10 and contributes to poor outcomes. [10][11][12] In LAC, 10% of residents have DM and 44% have prediabetes, 13 similar to statewide estimates (9% and 46%, respectively). 14 As the prevalence of DM increases, the number of people with TB and DM is also expected to rise. Furthermore, DM disproportionately affects racial/ethnic groups (e.g. Asian, Hispanic) in LAC and in California, 14 groups that are also more likely to be faced with the cooccurrence of TB and DM. 4 This is a concern in LAC, where in recent years 25%-30% of persons diagnosed with TB were reported to have DM. 4 Targeted efforts to prevent TB in at-risk groups are urgently needed. However, initial steps must be taken to identify the areas and populations highly burdened by these diseases. The application of Geographic Information Systems (GIS) methods to public health research can elucidate the underlying geography of health disparities, which are not evident with traditional statistical analyses or statistical packages. GIS spatial analysis tools use geographic data to better understand risk factor-disease relationships 15 and identify targets for public heath prevention and intervention. 15,16 A better understanding of the epidemiology of TB and DM co-occurrence and the geographical context in which it occurs can aid our work toward achieving the goal of TB elimination.
Thus, the objective of this study was to apply GIS methodology to identify geographic areas and populations that are disproportionately burdened by the co-occurrence of TB and DM in LAC, California.

Methods
We conducted a retrospective analysis in a cohort of verified cases of TB reported between 01/01/2015 and 06/30/2017 in LAC. TB cases were identified from the LAC TB Control Program Surveillance Registry. The TB Registry collects information on cooccurring medical conditions, including DM. This information is gathered through a variety of methods, including provider report, laboratory confirmation, and/or patient self-report.
TB cases were grouped based on DM status (with DM or without DM). Descriptive statistics (e.g. frequencies, percentages) summarize the characteristics of TB cases. Demographic and clinical characteristics were compared between TB cases with DM and TB cases without DM. Chi-square tests were used to detect statistically significant (p<0.05) differences in demographic and clinical characteristics among TB cases according to DM status. Descriptive statistics and chi-square test analyses were conducted in SAS Enterprise Guide 7.11. Spatial analyses were conducted using ArcGIS 10.3.1. Addresses of incident TB cases residing in LAC at the time of TB diagnosis were successfully geocoded to obtain longitude and latitude coordinates and mapped using the cartographic base map of LAC, LAC 2010 census tracts map, and LAC health district boundaries map. Since LAC is comprised of 24 health districts that are used to plan and manage health service delivery, geocoded addresses were first spatially joined to their respective health district and summarized as choropleth maps to better assess distribution of TB and DM across LAC. Next, hot spot 5 analyses using the Getis-Ord Gi* statistic (p<0.05) were used to geographically assess the burden of TB and DM by census tracts. The Optimized Hot Spot Analysis tool in ArcMap 10.3 identifies statistically significant spatial clusters of high values (hot spots) and low values (cold spots). Areas with elevated burden of TB and DM were considered hot spots when confidence intervals ranged from 90% to 99%. The map of LAC health district boundaries was overlaid on the LAC maps to determine which health districts were located within the hot spots.
The study was considered exempt by the LAC Department of Public Health Institutional Review Board. Informed consent was waived given that the data used were retrospectively collected and the work pertained to LAC TB Control Program's federal requirement to conduct active TB surveillance.

Discussion
The co-occurrence of TB and DM has been recognized for centuries. 9 In LAC, TB and DM disproportionately affect Asians and Hispanics. Furthermore, a majority (91%) of TB cases with DM were born outside the U.S. Spatial analyses identified hot spots of TB and DM among Asian TB cases residing in the southwestern and southern regions of LAC. These regions correspond to areas in LAC with large Asian populations, 14 with the southern region being an area where low English language proficiency is reported. 14 TB and DM burden was identified among Hispanic TB cases residing in southcentral and northwestern LAC, regions with large concentrations of Hispanics with low English language proficiency. 14 The elevated TB and DM burden delineated in the maps presented here are in line with reports and prior research indicating that TB and DM disproportionately and unequally 7 affect racial/ethnic minorities and non-U.S. born persons. 2, 5, 10, 17, 18 The co-occurrence of TB and DM should be closely examined by race/ethnicity given the increasing prevalence of DM, 13 the growing non-U.S. born population in LAC, 19  While the strength of this paper is the application of GIS methodology to identify geographical areas and populations in LAC burdened by TB and DM, there are some limitations to these findings. Assessment of DM status was partly based on patient selfreport. However, it should be noted that national estimates of DM incidence and prevalence are also based on self-report data. 25 Furthermore, self-reported DM status is reliable and has high sensitivity and specificity. 26,27,28 The data presented here are cross-sectional and do not indicate temporal order of onset of DM or TB. The TB and DM surveillance data do not provide information on severity or duration of DM among TB cases. Also, we did not examine the data by nativity to assess disease burden in areas where people from common nationalities reside. Despite these limitations, to our knowledge, this is one of the first set of analyses to examine spatial patterns of TB and DM burden.

Conclusions
In summary, our findings demonstrate the potential of GIS methods for identifying areas of elevated disease burden. Contextualizing disease burden geographically facilitates focusing and implementing community-based outreach initiatives, such as educational programs targeting both patients and providers, in areas at high risk for disease burden.

Ethics approval and consent to participate
The study was considered exempt by the LAC Department of Public Health Institutional Review Board. Informed consent was waived given that the data used were retrospectively collected and the work pertained to LAC TB Control Program's federal requirement to conduct active TB surveillance.

Consent for publication
Not applicable.

Availability of data and material
The datasets generated and/or analyzed for the current study are not publicly available in its entirety. However, reports that incorporate these data are available for download on the website (http://ph.lacounty.gov/tb/reports.htm). Also, a special data request can be submitted to request specific data presented here. editing, and write-up of this article. 14 c Includes TNF antagonist therapy, post-organ transplantation, immunosuppression (not HIV). *χ 2 comparing cases with DM and without DM.