**1. Introduction**

Mobile phones are becoming increasingly available and accessible globally. The global mobile phone subscription in 2009 was 68.0 per 100 inhabitants compared to 108 in 2019 corresponding to an overall 96% penetration rates<sup>1</sup> . In the African region, the estimated mobile phone penetration rate was 32.2% in 2008 compared to 85% in 2019 and is projected to rise to over 90% by 2025. The mobile broadband penetration rate also increased from 1.7% in 2008 to 30% in 2019 [1]. Many people who could not access fixed telephones for health informatics now use the mobile phone technology to access health services (mHealth). Compared to the wired information technology, the wireless technology is less expensive, more convenient, and readily accessible for individuals in many developing countries, including African countries [2, 3]. The mobile wireless technology has opportunities to facilitate communication among geographically isolated communities and could be harnessed to improve population health.

Mobile phones brought new opportunities for public health to Sub-Saharan Africa. With most of the African population in rural areas, mobile phone use has facilitated infectious disease management irrespective of geographical barriers [4–6]. Tuberculosis (TB) management is one area in which mobile phone use has shown great success because to effectively treat TB, patients must take four pills of anti-tuberculosis medications five times per week, for a period of 6 months [7]. This may create a high level of non-compliance to prescribed medications. Such protracted adherence could be facilitated by removing barriers to access and utilization through mHealth technologies [8]. In 2002, the South African government introduced the use of the mHealth technology, and computer databases to optimize TB treatment adherence. The database repeatedly lists patients who are due for their medications and an automatic Short-Message-Service (SMS) reminder sent to their mobile phones. This model enhanced treatment adherence and completion rates among sampled patients [7]. Thus, mobile phones provide platforms through which the SMS technology and other treatment-reminder protocols can be harnessed by TB patients to improve efficiency and optimize outcomes [6, 8]. Our study assumed that increased mobile phone use translates to broader access to TB services and represents the potential for impact on health with utilization of mobile phones to send/receive TB-related health information [9].

This study provides insight into the geospatial clustering of TB and mobile phone use among African health systems. Using an Exploratory Spatial Data Analytic (ESDA) approach, this study explored the spatial relationships between TB treatment completion rates and mobile phone use. Geospatial analytics of these concepts have opportunities to inform TB surveillance, intervention mapping and resource allocation. Moonan et al. conducted a geospatial epidemiological TB surveillance among newly diagnosed TB patients at the Tarrant County Health Department, Fort Worth, Dallas. Their model facilitated the identification of TB transmissions not identified during routine contact tracing. Thereby enabling the identification of at-risk populations, with an intervention mapping recommended for screening, treatment, and rehabilitation [10].

Conceptually, mobile phones facilitate information exchange and transfer without spatial barriers at high efficiency and low cost [4, 5]. Chadha et al. evaluated the effectiveness of the ComCare mobile application in coordinating TB referrals among patients in the Khunti District of India. It was discovered that the mobile

<sup>1</sup> Mobile phone penetration rate is the degree of diffusion of mobile phone use.

*Geospatial Clustering of Mobile Phone Use and Tuberculosis Health Outcomes among African… DOI: http://dx.doi.org/10.5772/intechopen.98528*

technology increased provider accountability and led to improved patient referral, retention, and treatment completion rates among network members [11]. Other researchers showed similar successes demonstrating the use of geospatial analytics in TB control, prevention, and management. Mwila and Phiri using geospatial analyses, cloud computing and web technologies modeled TB prevention strategies among developing countries. They explored ways to optimize TB monitoring and tracking protocols using technologies that display geographic distribution of TB cases on an mHealth application, while providing policy reports to inform intervention mapping activities [12]. While Yakam et al. spatially identified smearpositive pulmonary TB clusters among poor neighborhoods in Douala, Cameroon using mHealth technologies; Herrero et al. spatially explored cluster patterns of TB nonadherence and treatment dropouts in the metropolitan area of Buenos Aires, Argentina. Risk areas of nonadherence were characterized by poverty, ignorance, and reduced access to mHealth technologies [13, 14].

Multiple studies have documented the impact of mobile phone use on TB health outcomes for varied settings [11–14]. However, it is not immediately clear of the geospatial clustering patterns of TB treatment completion rates and mobile phone use among African countries. Previous studies have focused on evaluating TB medication access using geospatial disaggregated datasets of population characteristics [13, 15]. Hassarangsee et al. investigated the spatial detection and management of TB using information systems in the Si Sa Ket Province, Thailand [16]. However, the focus of this study is to evaluate the geospatial clustering patterns of TB treatment completion rates and mobile phone use among African health systems. It presents an exploratory spatial analysis of the relationships between TB treatment completion rates and mobile phone use for the countries in Africa. Using an ESDA approach to identify countries with low TB treatment completion rates and reduced mobile phone use could be the first step towards addressing issues related to poor TB outcomes. Thus, this presents an opportunity to identify African countries with limited resources and a high need for a wireless technology intervention.

### **2. Materials and methods**

#### **2.1 Data sources**

Data for TB outcome and mobile phone use for African countries were obtained from the World Bank database [17] for the periods 2000 through 2015 (Supplementary Material—A). This study excluded one country due to incomplete data. In total, 53 countries representative of the African continent were included in this study. De-identified information was collated and aggregated per country and published at the end of each year by the World Bank, qualifying it as Institutional Review Board exempt [18].

#### **2.2 Comparative statistical analyses**

ArcGIS and GeoDa statistical software were used in all geospatial analyses which were performed in three stages and completed in November 2020. Univariate local Moran's I and global Moran's I were run on TB treatment completion rates separately. This was followed by a differential local Moran's I analysis to ascertain differential cluster patterns for different time-periods. Finally, spatial relationships between TB treatment completion rates and mobile phone use for year 2015 was evaluated using ESDA. To investigate treatment completion cluster patterns, spatial and tabular data were uploaded into ArcGIS 10.5.1. Geographically referenced data

for TB treatment completion rates and mobile phone use for four time-periods (2000, 2005, 2010, 2015) were extracted for each country. These were added and joined to the African map by country shapefile and analyzed using an ESDA approach to visualize patterns and trends among geographically referenced data.

The Local Indicator of Spatial Association (LISA) represents the localized equivalent of the global Moran's I [19, 20]. For any location on a map, the LISA statistic measures and statistically tests the similarity of the geographically referenced data for that location (e.g., TB treatment completion rates at the source country) with the values of its corresponding local neighbors in space (surrounding countries). According to standard practice for reporting geospatial analytics, positive spatial autocorrelation is placed into High-High and Low-Low clusters; and negative spatial autocorrelation is placed into High-Low and Low-High outliers. High-High clusters denote above-average values of core countries versus surrounding countries. Low-Low clusters represent below average values of core countries versus surrounding countries. Low-High clusters means small changes among core countries versus high changes in the surrounding countries. Conversely, High-Low clusters means high changes among core countries versus small changes in the surrounding countries [20, 21]. For this study, a randomization of 999 permutations was used prior to result interpretations, and this study only analyzed observations with neighbors [20–22].

The Local Differential Moran's I (LDMI) statistic Eq. (1) measures if a variable change in space over time is related to its neighbors, and is calculated thus:

$$I\_{D,i} = \mathcal{c}\left(\mathcal{y}\_{i,t} - \mathcal{y}\_{i,t-1}\right) \sum\_{j} w\_{ij} \left(\mathcal{y}\_{\ j,t} - \mathcal{y}\_{\ j,t-1}\right) \tag{1}$$

Where *y* represents treatment completion rates for country *i* and neighbor *j*. The differential local Moran statistic *ID* is generated based on change over time and is represented by the difference between *yt* and *yt-1*. The geospatial matrix (*Wij*) is a binary spatial weights matrix. Under the queen first order principle, contiguous geospatial neighbors with common borders and vertex weights equals one. Therefore, observations that share common borders are considered neighbors for the calculation– all other locations are equals to zero [20, 21]. LDMI determines spatial autocorrelation on change over time (yt – yt-1). For this study, differential cluster patterns were evaluated between base time 0 (year 2000) and time 1 (year 2005), time 2 (year 2010) and time 3 (year 2015) respectively. Year 2000 was chosen as the base time because of data availability, which has been consistently captured for the 53 African countries included in this study [17]. Also, notable access to free anti-tuberculosis medications commenced in 2000 among most African health systems [23].

#### **3. Results**

Using geospatial data for 53 African countries for the periods 2000–2015, univariate global Moran's I values and associated pseudo p-values were computed and documented (Supplementary Material—B). In addition, LISA analytics identified different cluster patterns (Low-Low and High-Low) that were significant at different p-values (**Table 1**).

**Figure 1** shows the clusters and significance levels of TB treatment completion rates in year 2000. Thirteen countries had significant clusters at different p-values including:

*Geospatial Clustering of Mobile Phone Use and Tuberculosis Health Outcomes among African… DOI: http://dx.doi.org/10.5772/intechopen.98528*

**Figure 2** represents the clusters and significance levels of TB treatment completion rates in year 2005. Nine countries had significant cluster patterns at different p-values.

The clusters and significance levels of TB treatment completion rates for year 2010 are shown in **Figure 3**. Eight countries had significant clusters at different p-values.

The cluster patterns and significance levels of TB treatment completion rates for year 2015 are shown in **Figure 4**. Eight countries had significant clusters at different p-values.


#### **Table 1.**

*Univariate global Moran's I result for the years 2000–2015.*

#### **Figure 1.**

*Clusters and significance levels of TB treatment completion rates in year 2000.*

#### **Figure 2.**

*Clusters and significance levels of TB treatment completion rates in year 2005.*
