Machine Learning Applications in Smartphone Healthcare Services and Fraud Detection: A Ten-Year Systematic Literature Review

Bangladesh’s government-run health system is heavily subsidized yet remains underfunded. According to WHO (2010), health expenditure accounts for only about 3% of GDP. The government provides 34% of total health expenditure (THE), while individuals pay 66% through out-of-pocket (OOP) payments, resulting in substantial inequity in access to care. Drawing on secondary data, this review examines the strengths and weaknesses of Bangladesh’s health system. Bangladesh has made notable progress on health-related Millennium Development Goals (MDGs), particularly MDGs 4 and 5, but it still lacks a coherent health policy for system development despite a rapidly growing private sector focused on tertiary care. As smartphone ownership has become widespread, mobile applications are increasingly used to simplify daily life, creating demand for smartphone-based healthcare services. However, this digital health market is also vulnerable to fraud, especially medical insurance fraud. Manual detection of healthcare fraud is difficult, so machine learning and data mining have been used to automate the process. Many studies have applied these techniques to healthcare fraud detection, yet further research and more sophisticated machine-learning methods are needed to identify atypical patterns in health insurance usage. This review synthesizes literature from the past ten years on smartphone-based healthcare services and machine-learning-driven fraud detection for improving the healthcare system.