1. BackgroundTuberculosis (TB) is a contagious disease caused by the bacillus Mycobacterium tuberculosis (Mtb). Despite the tremendous efforts and encouraging progress obtained towards the control of the TB epidemic globally, it remains the single infectious disease that takes more lives each year, ranking above human immunodeficiency virus/acquired immunodeficiency syndrome (HIV/AIDS) until the coronavirus (COVID-19) pandemic spread worldwide [1]. In 2020, an estimated 9.9 million people fell ill with TB worldwide. This number shows a slight decline, about 0.87%, compared with 2019, and there were faster declines in the African (18%) and European regions (26%) [1]. HIV also continues to be a major global public health issue. According to UNAIDS estimates, globally, an estimated 37.7 million people were living with HIV in 2020 [2]; of these people, two-thirds of them lived in Sub-Saharan Africa (SSA). In 2020 and 2021, there were 1.5 million newly HIV-infected cases each year globally, and about 680,000 and 650,000 people died from AIDS-related illnesses in 2020 and 2021, respectively, worldwide [3,4]. In 2019 and 2020, there were an estimated respective 815,000 and 787,000 TB cases globally who were people living with HIV [3,5].In 2019 and 2020, the estimated incidence rates for TB were 140 and 132 per 100,000 population, and the mortality rates were 21 and 17 per 100,000 population [1,6], respectively. An estimated 670,000 people in Ethiopia acquired HIV, and an estimated 12,000 people died of AIDS-related illness in 2019 [7]. The Government of Ethiopia has been taking several measures to control TB and HIV; for example, it established the National Tuberculosis and Leprosy Control Program (NTLCP) to harmonize the coordination and management of TB and leprosy at the country level in 1994 [8] and improved access to TB diagnostic facilities through its community-based Health Extension Program. Further, based on a WHO recommendation, the national responses for the integrated TB and HIV collaborative activities were updated in 2012 and aimed to reduce the burden of HIV among TB patients and the burden of TB among people living with HIV (PLHIV). Since then, the integrated TB and HIV service has expanded to the tertiary, secondary, and primary health care levels in Ethiopia [9]. Despite the measures that have been taken, Ethiopia remains among the 30 countries reported with a high burden of TB and HIV, TB/HIV co-infection, and multi-drug resistant (MDR)-TB from 2015 to 2020. The spatial analysis of a disease helps to understand the spatial epidemiology of the disease occurrence (incidence) at different administrative levels [10]. There are several studies in Ethiopia that have investigated the spatial clustering of TB and HIV separately at various levels [11,12,13,14,15,16,17,18,19]. Studies in other countries also reported spatial clustering of tuberculosis, e.g., in Brazil [20], China [21], Nigeria [22,23] and in South Africa [24]. Similar studies are also available for HIV, e.g., in Burundi [25], China [26], Malawi [27], Mozambique [28], South Africa [29] and Uganda [30]. Studies also show that an understanding of the spatial distribution pattern of TB (e.g., [31,32]) and HIV (e.g., [26,33,34,35,36]) helps to design focused interventions, such as facilitating access to TB control, community-based awareness creation and intervention, contact tracing and equitable access to treatment for vulnerable populations. HIV and TB are epidemiologically associated [37]. Observed co-dynamics suggest that the two diseases are directly related at the population level [38] and also within the host [39]. A study in Ethiopia also suggests that TB incidence is high in areas where HIV is highly prevalent [40]. Studies on the prevalence of TB-HIV co-infection have shown that the co-infection varies widely in Ethiopia (see, e.g., [40,41]) and has geographical clustering [42]. Similar observations have been made in other countries (see, e.g., [43,44,45]). There are studies in Ethiopia that have reported the epidemiology of TB-HIV co-infection at the hospital level [46,47]. Except for [42], the other studies include either a single district or region or only public health facilities in a district or a town. However, Alene et al. [42] have reported the spatial distribution of TB-HIV co-infection at the national level using three years of aggregated data. Despite the relevance of TB-HIV co-infection, there are still few publications on its spatial co-clustering, at least using yearly data at a national level. To design the most effective strategies that help to reduce the TB and HIV transmission rates, it is essential to have a more in-depth analysis of the epidemiological patterns of TB-HIV co-infection at the district level because targeting high-risk areas with effective control measures yields good results in controlling the pandemic [48]. The results of such an analysis could be used to inform locally targeted disease-specific or integrated control measures. In addition, knowledge of hot-spot and cold-spot (or high- and low-burden) areas is required for successful surveillance programs and optimal resource allocation [49]. Therefore, the main objective of the current study was to assess the spatial co-clustering of TB and HIV in Ethiopia at the district level for a four-year period, 2015 to 2018. Further, we were interested in assessing the spatial clustering of TB and HIV separately for the same period. These objectives were addressed using the aggregated data, collected from the national Health Management Information System (HMIS), on the number of TB cases enrolled in DOTS and who were tested for HIV, and the number of HIV patients enrolled in HIV care who were screened for TB during their last visit to health care facilities. 3. Statistical AnalysesIn this paper, spatial analyses were conducted to identify geographical clustering of the HIV and TB cases separately and their co-clustering simultaneously in Ethiopia at the district levels for each year from 2015 to 2018. For the spatial analyses, the district geographical boundaries were geo-referenced. They were linked to the district HIV and TB data, and choropleth maps were developed for visualization using GeoDa software version 1.2 [52]. Then, the global pattern analysis to study the global spatial autocorrelation for the prevalence of HIV and TB separately was conducted using the univariate global Moran’s I [53]. The two diseases’ prevalences were investigated simultaneously using the bivariate global Moran’s I tests. The univariate local Moran’s I [54] statistics were applied to identify the local spatial clusters of HIV and TB separately [55], whereas the bivariate local Moran’s I [56] statistics were applied to investigate the simultaneous occurrence of both diseases and hence to identify the co-clustering of HIV and TB in Ethiopia for each year in the study period [44]. The significance of Moran’s I statistics is assessed by the Monte Carlo randomization technique. The spatial relationships of the districts were defined using a spatial weight matrix, and neighbourhoods were defined using Queen’s contiguity, where neighbours’ districts are defined as districts sharing borders or a common vertex with each other. The weight matrices for the local indicators of spatial association (autocorrelation) or for local Moran’s I statistics were defined using the nb2listw function from the spdep [57] R package.

In the choropleth map, for a single disease case, the local indicators of spatial analysis (LISA) classify districts into five clustering categories to show the spatial clusters and spatial outliers: high-high (hot-spot), which are districts with a high density of disease notification compared to the expected cases given a random distribution of disease; low-low, which are districts with a low density of disease notification (high-high and low-low suggest clustering of similar districts); low-high, which are districts with a low notification of disease sharing borders with districts with high notifications; high-low, which are districts with a high number of notified cases sharing borders with districts with low notifications; and not significant clustering. For the bivariate case, the software produces cluster maps in which the significant districts are classified into high-high, and low-low, and to describe the nature of spatial outliers of the two disease distributions at the district level, the districts are also classified as high-low and low-high. In each of the four classifications, the first attribute corresponds to the total number of TB cases, and the second corresponds to the total number of HIV cases in the neighbouring districts.

Ethical Consideration

Permission for the study was obtained from the School of Science Ethics Committee, University of South Africa (ERC Reference Number: 2021/CSET/SOS/045). In addition, permission to use the data for this study was obtained from the Ethiopian Ministry of Health Office. In this study, we have used aggregated district-level data where individual patient information was not available; therefore, informed consent was not obtained from the study participants.

5. DiscussionTuberculosis and HIV still represent the most important infectious diseases around the globe. Despite advances made in their eradication via WHO-coordinated efforts, these diseases still present alarming data on morbidity and mortality [1,2]. In the current paper, we have investigated the spatial clustering of TB and HIV separately and the spatial co-clustering of both diseases in Ethiopia at the district level for a four-year period from 2015 to 2018.The findings of this study show that the occurrences of both diseases were geographically heterogeneous over time. These results are consistent with the findings of [42], whose results indicate that the prevalence of TB among people living with HIV and the prevalence of HIV among TB patients varied in Ethiopia at the district level, but the data that we have used are different from theirs. Several studies conducted in Ethiopia [11,12,14,15,16,18] and in other countries, including Nigeria [22,23], China [21], and South Africa [24], revealed a strong tendency toward the spatial clustering of TB at different geographic scales. Similarly, a number of studies showed the spatial clustering of HIV at different scales [19,25,26,27].

The results from LISA show that the notification of TB was strongly spatially clustered in the districts of the Addis Ababa city administration; the Gondar Zuriya, Bahir Dar Zuriya, and Armacho districts in the Amhara region; the Legehare area in Dire Dawa; the Adama, Burayu, Sebeta Hawas, and Shashemene districts in the Oromiya region and the Wendo Genet district in the SNNP region. However, the spatial pattern of TB showed changes in the eastern (in 2017 and 2018) and northern (in 2018) parts of the country. In 2018, the hot spots were more concentrated in the Amhara region; about 70% of the observed hot spots were in this region.

Generally, the TB hot spots in the country over the study period appeared in more urbanized areas such as Addis Ababa, Adama, Dire Dawa, Bahir Dar, and Shashemene. These findings are supported by the results of [13], which identified urbanization and population density as main risk factors for TB. Additionally, it may be related to the migration of people within districts or across neighboring districts for jobs and better living conditions. Studies in Ethiopia have indicated that the spatial clustering of TB is associated with migration [11,12,13,14,15,17], and ongoing TB transmission is also high in overcrowded and congested urban areas [10,13,14,17]. In addition, other studies in Ethiopia [15,18] and in other countries such as Argentina [58], Brazil [59] and China [60] show that TB incidence rates are associated with poor living conditions and housing. The LISA cluster maps illustrated that there were hot spot areas on the border of North Sudan over the study period (2015 to 2018) and Eritrea in 2018. These findings support the results of the existing literature [17,42]; therefore, they may affirm that there is a relationship between TB transmission and international border or territorial space. Hence, it is necessary to extend the country-level analysis to higher spatial dimensions that include at least neighbouring countries to obtain global solutions and targeted interventions [61,62,63].As with the TB case, the results of the LISA analysis show the spatial clustering of HIV in Ethiopia at the district level over the study period (2015 to 2018). The high-high clustering or hot-spots of HIV significantly and consistently appeared in the districts of the Addis Ababa city administration, in three districts from the Amhara region (Ambasel, Dembia, and Gondar Zuriya), the Legehare area from the Dire Dawa city administration, and Adama from the Oromiya region. In addition to Addis Ababa, Dire Dawa, and Adama, the current study identified that the country’s main cities or towns, such as Woldiya, Humera, Kombolcha, Dese, and Kobo, appeared as HIV hot spots. This may be because such places have a higher number of commercial sex workers [64,65,66], and some of the cities/towns are transport corridors for truck drivers or long-distance vehicle drivers, e.g.,Adama, Woldiya, and Kobo for trucks coming or going from or to the Djibouti port [67]. In addition, cities, towns, or generally urban areas are associated with a high rate of HIV infection [68,69] as residents in urban areas have higher population movement due to labor, migration, and trading. A hot spot was observed at Humera town in 2015, which shares a border with North Sudan and Eritrea. This area is in the Tigray region agricultural center and the gateway to North Sudan and is used by returnees [64]. This HIV hot-spot may be related to agricultural activities; some studies suggest that border areas and agricultural activities are potential risk factors for infectious disease [16,68,70].In the current study, the results revealed that TB was more spatially correlated than HIV in Ethiopia for each year in the study period. However, this finding contradicts the results of Aturinde et al. [44], who found that HIV was more spatially correlated than TB in Uganda for the period of 2015–2017. Possibly, this difference could be due to the high burden of HIV (6.4% prevalence rate) in Uganda compared to 0.9% in Ethiopia. The univariate global Moran’s I statistics for TB and HIV were positive, suggesting that neighbouring districts tend to possess similar characteristics in both disease prevalences. Similarly, the global bivariate Moran’s I statistic was positive for the study period, and this implies that the two diseases were positively influenced by neighbouring districts.The simultaneous TB and HIV spatial co-clustering patterns in Ethiopia at the district level, in most cases, overlapped with the hot spots of TB and HIV. The TB and HIV spatial co-clustering for the study period therefore could be explained by reported findings on TB and HIV in the literature, as discussed in the above paragraphs. Our findings on TB-HIV co-clustering clearly indicated an almost similar trend for the study period; hot-spot areas occurred mostly in the central part of the country (including Addis Ababa) and some part of the northeast and northwest. Several studies conducted at the regional level in Ethiopia support our findings [40,41,46,47]. Additionally, the results of a systematic review and meta-analysis of TB-HIV distribution in Ethiopia from 2007 to 2017 [71] showed regional and small-scale spatial variation in TB-HIV, which agrees with our findings. Similarly, the 2016 National TB and Leprosy control report revealed regional variation in TB, HIV, and TB-HIV prevalence, identifying the Addis Ababa and Somali regions as having high and low prevalence rates of TB-HIV co-infection, respectively [72]. Our results also show that there were low-low TB and HIV co-clusters, or cold spots, in most of Afar and the Somali regions, which consistently appeared for the period 2015–2018. This may be due to very low notifications of both diseases in the regions.Although there are few studies that have investigated the spatial clustering of HIV and TB separately at the national scale in Ethiopia [66] and TB/HIV co-infection, e.g., [42], to the best of our knowledge, this is the first spatial study on the co-clustering of TB and HIV using Ethiopian data. However, there were some limitations that could have affected our findings. First, the data were aggregated at the district level; therefore, the findings of this study cannot be representative of small administrative units of the country or kebele or household or individual levels. Second, since HIV and TB data were collected from the national HMIS electronic surveillance system, the reported HIV and TB cases might not reflect the actual burden of the diseases in a district due to the underreporting or underdetection of cases. For example, symptomatic individuals who did not receive HIV or/and TB diagnosis and treatment might remain unreported. Generally, the use of secondary data could affect the results of this study because of selection bias, information bias, or both. First, the data set includes those cases recorded only at health centers, which were compiled and reported to the districts, the zonal level, the regional level, and then the national level of the ministry of health offices. However, there may be unregistered TB and HIV cases mainly due to limitations of health service access, especially in rural remote areas. This missing data, called selection bias, could be one possible bias affecting the study results. Secondly, there may be missing data, alteration of data or systematic distortions when collecting information at any stage, and these could add another potential bias called information bias, which could affect the study’s findings. Third, the results could be more interesting if they were supported by spatial generalized linear models, specifically using count models to assess potential risk factors for TB, HIV, and both TB and HIV; however, these data were not available in the national HMIS.According to recent studies, the outbreak of COVID-19 has become one of the contributing factors to the increase in morbidity and mortality related to TB cases. For instance, the 2020 WHO estimates using data from 84 countries indicate that the number of TB patients receiving care was reduced by 1.4 million in 2020 compared to 2019 (reduced by 21% from 2019). According to this WHO estimate, these COVID-19-related challenges in access to TB care could cause an additional half a million TB deaths. Findings from other studies [73,74,75] also showed strong association between TB and COVID-19 and between the mortality rates of the first cohort of patients with COVID-19 and TB co-infection, indicating the importance of considering their strong relationship in a future study.

Generally, our study results showed strong spatial co-clustering of TB-HIV at the district level. These spatial heterogeneities or clusterings may be due to different factors that should be studied further in detail. Therefore, based on our findings, we recommend that authorities should strengthen their intervention mechanisms, such as facilitating access to TB and HIV control, providing early screening and treatment, the introduction of community-based awareness creation, and allocating resources to hot-spot areas.