AI Analysis Reveals Tuberculosis Treatment Patterns in Indonesia, Bali Identified as a Unique Region
FORMOSA NEWS – Tuberculosis (TB) treatment outcomes in Indonesia are not evenly distributed across provinces. This finding comes from a study conducted by Mahrani of Universitas Negeri Makassar (UNM) and published in 2026 in the International Journal of Applied and Advanced Multidisciplinary Research (IJAAMR). Using an artificial intelligence-based data analysis method known as Density-Based Spatial Clustering of Applications with Noise (DBSCAN), the research classified Indonesian provinces according to similarities in TB treatment outcomes and identified regions with distinctive patterns that differ from the national trend. The study is particularly relevant because Indonesia remains one of the countries with the highest tuberculosis burden in the world. According to the World Health Organization (WHO), Indonesia accounted for approximately 10 percent of global TB cases in 2024, ranking second only to India. Despite ongoing control efforts, significant challenges remain in case detection, treatment access, patient monitoring, and treatment success. Mahrani’s research highlights that tuberculosis treatment outcomes vary considerably across regions. Factors such as population density, socioeconomic conditions, environmental quality, mobility, and healthcare accessibility can influence how TB is managed and treated in different provinces. Understanding these regional differences is essential for designing more effective public health strategies. The study analyzed secondary tuberculosis data from 34 provinces across Indonesia. The dataset included six key indicators of treatment outcomes: cured patients, completed treatment, deaths, treatment failures, loss to follow-up, and cases that were not evaluated. These variables were processed using data mining techniques, including normalization, dimensionality reduction, and density-based clustering through the DBSCAN algorithm. Unlike conventional clustering methods such as K-Means, which require researchers to determine the number of groups in advance, DBSCAN automatically identifies clusters based on data density. The method is also capable of detecting outliers, referred to as “noise,” making it particularly useful for analyzing complex public health data with irregular patterns. The results showed that the optimal clustering configuration was achieved with an epsilon (ε) value of 0.05 and MinPts of 2. Under these parameters, the model produced two major clusters and one noise region. Cluster validation using the Davies-Bouldin Index (DBI) generated a score of 0.450, indicating a strong and reliable clustering structure.
Two Major Clusters and One Outlier Region
The first cluster consisted of 27 provinces, including Aceh, North Sumatra, West Sumatra, Riau, Riau Islands, Jambi, South Sumatra, Bangka Belitung Islands, Bengkulu, Lampung, Banten, DKI Jakarta, West Java, Central Java, DI Yogyakarta, East Java, West Kalimantan, Central Kalimantan, South Kalimantan, East Kalimantan, North Kalimantan, North Sulawesi, Gorontalo, Central Sulawesi, South Sulawesi, West Sulawesi, and West Nusa Tenggara. These provinces shared relatively similar tuberculosis treatment outcome characteristics, suggesting that many regions across Indonesia face comparable challenges and opportunities in TB management. As a result, similar intervention strategies may be applicable across these provinces, although local adjustments remain necessary. The second cluster contained six provinces: Southeast Sulawesi, East Nusa Tenggara, Maluku, North Maluku, Papua, and West Papua. These provinces demonstrated treatment outcome patterns that differed from those observed in the larger cluster. According to the study, the separation of these provinces may be linked to geographical barriers, dispersed populations, limited healthcare infrastructure, and challenges in accessing medical services. These factors can significantly affect patient monitoring, treatment adherence, and healthcare delivery. One of the most notable findings was the classification of Bali as a noise region, meaning its tuberculosis treatment outcome profile was sufficiently different from all other provinces and could not be grouped into either cluster. In the DBSCAN framework, being categorized as noise does not indicate an error or anomaly in the data. Instead, it signals that Bali possesses unique characteristics that distinguish it from the dominant provincial patterns. These differences may be associated with healthcare performance, reporting systems, treatment success rates, population mobility, or other local factors. The study recommends further investigation to understand the underlying causes of Bali’s distinctive profile.
Implications for Public Health Policy
The findings demonstrate how artificial intelligence and data mining can support evidence-based public health planning. By identifying provinces with similar treatment outcome patterns, policymakers can design more targeted and efficient tuberculosis control programs. For provinces in the largest cluster, policy efforts may focus on strengthening existing TB control measures, improving treatment completion rates, enhancing patient follow-up systems, and monitoring treatment outcomes more effectively. Meanwhile, provinces in the second cluster may require tailored interventions that address region-specific challenges such as geographical isolation, healthcare accessibility, and resource limitations. Understanding why these provinces differ from the national trend could help improve TB management strategies across the country. The identification of Bali as an outlier also provides valuable insight. Rather than overlooking unusual patterns, health authorities can use such findings to investigate potential best practices, unique healthcare conditions, or emerging challenges that may otherwise remain unnoticed. According to Mahrani, clustering analysis offers more than a simple count of tuberculosis cases. It reveals patterns of similarity and difference among regions, enabling policymakers to make more informed decisions and allocate resources where they are needed most. The study further demonstrates the value of AI-driven analytical tools in supporting disease surveillance. By combining statistical methods with modern data mining techniques, researchers can transform large health datasets into actionable information that supports more effective and responsive public health policies.
Author Profile
Mahrani is a researcher and academic affiliated with Universitas Negeri Makassar (UNM), Indonesia. Her areas of expertise include data analysis, statistics, data mining, and the application of computational methods in multidisciplinary research, particularly in public health and spatial epidemiology.
Research Source
Article Title: Density Based Spatial Clustering of Application with Noise on Tuberculosis Cases in Indonesia
Author: Mahrani
Affiliation: Universitas Negeri Makassar
Journal: International Journal of Applied and Advanced Multidisciplinary Research (IJAAMR)
Volume: 4, Issue 5
Year: 2026
.jpg)
0 Komentar