Tadese Zinabu Bekele, Arage Fetlework Gubena, Tsegaw Tigist Kifle, Alemu Eyob Akalewold, Abate Tsegasilassie Gebremariam, Taye Eliyas Addisu
Department of Health Informatics, College of Medicine and Health Sciences, Samara University, Semera, Ethiopia.
Department of Public Health Officer, Institute of Public Health, College of Medicine and Health Sciences, University of Gondar, Gondar, Ethiopia.
BMC Infect Dis. 2025 Jul 1;25(1):870. doi: 10.1186/s12879-025-11228-3.
Tuberculosis (TB) is a preventable and treatable disease caused by Mycobacterium tuberculosis, which most often affects lungs and remains the second leading cause of death from infectious diseases worldwide. The National End TB Strategy aims to eliminate the TB epidemic by reducing TB-related deaths by 95% and decreasing incident TB cases by 90% by 2030, using 2015 as the baseline. Tuberculosis is the primary cause of morbidity, ranks third in hospital admissions, and is the second leading cause of death in Ethiopia, following malaria. Hence, this analysis aims to forecast and provide evidence that supports the combined intervention to monitor TB incidence in Ethiopia's progress toward the Sustainable Development Goals.
Study employed secondary data analysis from the Global Burden of Disease database (1990-2021) to forecast tuberculosis incidence in Ethiopia. LSTM-based models, including multistep LSTM and hybrid ARIMA + LSTM, were implemented for prediction in TensorFlow frameworks while ARIMA model was built using the statsmodels and pmdarima libraries using the Python programming language. The statistical significance level was set at 0.05 to check data stationarity. Model performance was evaluated using Root Mean Squared Error, Mean Absolute Error, Mean Absolute Percentage Error, and Symmetric Mean Absolute Percentage Error. Finally, the best model was used to forecast the next 9 years from 2021 to 2030.
According to GBD data, the incidence of TB in Ethiopia shows a long-term downward trend, decreasing from 466.93 cases per 100,000 in 1990 to 185.53 by 2021. The analysis result revealed that multistep LSTM model outperformed all achieving MAE: 5.53, RMSE: 6.74, MAPE: 2.72% and sMAPE:2.76%. The incidence of tuberculosis in Ethiopia is projected to decline slightly through 2030, according to a multi-step LSTM model. The forecast estimates that the TB incidence will be 189 cases per 100,000 people by 2025, decreasing further to 179 by 2030.
Overall, the analysis indicates that Ethiopia is still falling short of the national "END TB strategy" goal of 90% reduction in TB incidence cases per 100,000 population by 2030. It highlights the necessity for Ethiopia's TB control strategies to improve access to prevention, early diagnosis, and treatment, focusing on high-risk groups and vulnerable populations.
结核病(TB)是一种由结核分枝杆菌引起的可预防和可治疗的疾病,最常影响肺部,仍然是全球传染病死亡的第二大主要原因。《国家终结结核病战略》旨在以2015年为基线,到2030年将与结核病相关的死亡减少95%,并将结核病发病病例减少90%,从而消除结核病流行。结核病是发病的主要原因,在住院患者中排名第三,在埃塞俄比亚是仅次于疟疾的第二大主要死因。因此,本分析旨在预测并提供证据,支持在埃塞俄比亚实现可持续发展目标的进程中监测结核病发病率的联合干预措施。
本研究采用全球疾病负担数据库(1990 - 2021年)的二手数据分析来预测埃塞俄比亚的结核病发病率。在TensorFlow框架中实施了基于长短期记忆网络(LSTM)的模型,包括多步LSTM和混合自回归积分移动平均(ARIMA)+LSTM进行预测,而使用Python编程语言通过statsmodels和pmdarima库构建ARIMA模型。设定统计显著性水平为0.05以检验数据平稳性。使用均方根误差、平均绝对误差、平均绝对百分比误差和对称平均绝对百分比误差评估模型性能。最后,使用最佳模型预测2021年至2030年的未来9年。
根据全球疾病负担数据,埃塞俄比亚的结核病发病率呈长期下降趋势,从1990年的每10万人466.93例降至2021年的185.53例。分析结果显示,多步LSTM模型表现最佳,平均绝对误差为5.53,均方根误差为6.74,平均绝对百分比误差为2.72%,对称平均绝对百分比误差为2.76%。根据多步LSTM模型,预计到2030年埃塞俄比亚的结核病发病率将略有下降。预测估计到2025年结核病发病率将为每10万人189例,到2030年进一步降至179例。
总体而言,分析表明埃塞俄比亚仍未达到到2030年将每10万人口中结核病发病病例减少90%的国家“终结结核病战略”目标。这凸显了埃塞俄比亚结核病控制策略改善预防、早期诊断和治疗可及性的必要性,重点关注高危人群和弱势群体。