**4. Discussion**

ESDA identified statistically significant clusters in TB treatment completion rates among some countries of Africa. Most African countries except those in the Southern region had significant clusters (**Figures 1–4**), predominantly Low-Low clusters (14 countries) and High-Low clusters (10 countries). Low-Low clusters suggest that TB treatment completion rates between 2000 and 2015 were low among identified countries and were surrounded by countries with similar low changes. High-Low clusters suggest high changes in the core countries versus low changes in their surrounding countries (18, 19). In year 2000, Algeria had a "High-Low" cluster tract (**Figure 1**). This suggests that the TB treatment completion rates for Algeria in 2000 was high, surrounded by countries with low rates. DRC had a "Low-Low" cluster tract in year 2010; suggesting that TB treatment completion rates was low for DRC, surrounded by countries with similar low changes (**Figure 3**). These findings could inform the formation of health policies for TB management strategies including resource allocation frameworks in countries. Study results also indicated that only a few countries had complete treatments for TB across Africa within the period of this study analyses. 13 countries out of 53 (**Figure 1**) had treatment completion rates in the year 2000, with all clusters being either outliers or cold spots. Nine countries in 2005 (**Figure 2**), eight countries in 2010 (**Figure 3**), and eight countries in 2015 (**Figure 4**). These suggest that poor

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

compliance and nonadherence could impact treatment completion rates for which the mHealth technology could be harnessed to address issues related to access, utilization, and attrition [8, 14].

Differential Moran's I cluster maps identified hot spots, cold spots, and outliers among African countries. The base-case analysis identified three countries with significant clusters including Algeria, Burkina Faso, and Senegal. Cluster patterns for Burkina Faso and Senegal were Low-Low cluster tracts, which suggest low TB treatment completion rates surrounded by countries with similar low changes. Conversely, Algeria had High-Low clusters suggesting high changes in TB treatment completion rates versus low changes in surrounding countries (Supplementary Material—C, Figure 6). These findings could inform planning and suggest that the mHealth technology have opportunities to improve outcomes by facilitating fast, reliable, and updated health information.

While the time-period 2 analysis identified 12 countries with High-High clusters (Supplementary Material—C, Figure 7); the time-period 3 analysis identified eight countries with significant clusters (Supplementary Material—C, Figure 8). Altogether, two countries including Algeria and Senegal had significant clusters across the three time-periods of study evaluation (Supplementary Material—C). Nonetheless, of note is the fact that only three countries had significant clusters from LDMI analytics in year 2005 (Supplementary Material—C, Figure 6), compared to 2010 and 2015 where 12 countries (Supplementary Material—C, Figure 7) and seven countries (Supplementary Material—C, Figure 8) had significant clusters respectively. A possible explanation could be that 5-year period may not be enough time to appreciate significant changes in TB treatment rates (23). Conversely, the 15-year period may be too long to observe this trend. Thus, it appears the ideal time for this evaluation should be around 10-year time-periods. This gives credence to treatment guideline policy change introduced in 2009 by WHO that advocates following patients post treatment for longer periods irrespective of the form of TB [23]. In view of this, future studies should aim to do a sensitivity analysis to determine the precise time to explore TB treatment completion rates among African health systems.

It must be noted that identifying significant clusters in TB treatment completion rates in any country does not translate to TB-free nations. South Africa is one of the nations with good TB programs in Africa [22], corroborating this study findings (**Tables 2–4**). South Africa had significant clusters for TB treatment completion rates (Supplementary Material—C); which could possibly be attributed to the coordinated TB control measures introduced by the South African government [7]. However, despite government efforts to curb the incidence of TB in South Africa, recent study by the WHO identified South Africa as one of the seven countries that accounted for 64% of global new cases of TB [24]. Thus, notwithstanding significant TB treatment completion clusters identified by this study, the burden of TB in South Africa remains high.

Exploratory spatial data analyses identified significant association between TB treatment completion rates and mobile phone use rates. Countries with higher rates of mobile phone use, showed higher TB treatment completion rates suggesting enhanced program uptake. Dissecting this association with local level geographical data revealed differing cluster patterns suggesting that the diffusion was not consistent across the region.

High-High clusters indicate countries with high TB treatment completion rates surrounded by countries with similar high mobile phone use rates. Tunisia, Sierra Leone, Eritrea, Kenya, Rwanda, Tanzania, and Mozambique had High-High clustering of these two attributes (**Figure 5**). This suggests that the use of mobile phones may be facilitating TB treatment completion rates; and gives credence to

findings by Chadha et al. which demonstrated how the mobile phone technology strengthened and optimized TB health outcomes [11].

Low-Low clusters indicate countries with low TB completion rates surrounded by countries with low mobile phone use rates. Mauritania, Nigeria, Cameroon, Equatorial Guinea, Gabon, Angola, and South Africa had significant Low-Low clustering of these two attributes (**Figure 5**). Such low uptakes may be contributory to the high burden of TB in Nigeria and South Africa and lends credence to reports by WHO that listed them among the six countries that contributed to the high burden of global new cases of TB in 2017 [24].

Spatial outliers indicate countries with either high or low TB completion rates surrounded by countries with either low or high mobile phone use rates. Identified Low-High outliers include Zimbabwe, Somalia, Ethiopia, Libya, Sudan, and Uganda (**Figure 5**). These are potential moderators and mediators for this study; and could possibly be related to factors that impede the use of mobile phones for health informatics including poverty, ignorance, and poor access to mHealth technologies [14].

This study had some limitations. It did not control for the presence or absence of factors that could influence access and utilization of services, which could impact the robustness of study findings. In addition, there were some limitations in the datasets used for this study. Geographical data for TB cases from the World Bank were at the country level only– granular spatial relationships could have been used and would have revealed cases at a finer resolution. More so, this study is exploratory in nature. It assesses correlation and not causation and becomes the first step in assessing the relationship between TB health outcomes and potential impacts of ICT tools such as mobile phone use on TB programs among African health systems.

#### **4.1 Policy implications**

Evidence-based decision making, monitoring of health status, tracking of expenditures and outputs for improving efficiency of investments are hallmarks of successful health programs. In recent times, public health intelligence platforms, such as the WHO's Epidemic Intelligence from open sources (EIOS), utilize strong data systems facilitated by digital technologies for health emergency preparedness and disease surveillance. Innovative digital technologies including mHealth have potential to offer new insights and tools to improve clinical decision making, data security and predictive analytics. Realizing the potential of mHealth to achieve development goals will require collaboratively addressing multiple challenges, including strengthening underlying digital infrastructure and digital health systems, working to improve data quality, responsible management and sharing of patient data, and eventually trust, security and utilization of data to inform strategic planning of health interventions. Thus, it is recommended that African health sector leaderships convene roundtable discussions to discuss critical policy dimensions of strengthening mHealth ecosystems, and lay the foundation for responsibly developing and adopting new innovations like mHealth-facilitated Artificial Intelligence (AI) applications in addressing privacy related issues. Such discussions should include but not limited to discussions on adopting latest advances in mHealth technologies including AI tools and telemedicine techniques in advancing population health. They should identify concrete actions for shaping policy and enabling environments to foster mHealth technologies for accelerated improvements in health and related development programs; showcase AI-driven innovative solutions for health, and share best practices from countries, including possibilities for customizations. Such approach could foster a deeper interest on the use of mHealth *Geospatial Clustering of Mobile Phone Use and Tuberculosis Health Outcomes among African… DOI: http://dx.doi.org/10.5772/intechopen.98528*

technologies to strengthen service delivery, improve the collection of quality data and promote data protection rights among African economies.
