A robust privacy-preserving online monitoring algorithm for disease surveillance

Early detection and treatment of disease are paramount for our well-being. Regular clinical visits are essential for patients with chronic illnesses to prevent the disease from progressing, and individuals with high-risk health factors, such as family history of a specific disease, also benefit from frequent health checks. Effective health monitoring relies on two essential factors. First, a timely and robust algorithm is indispensable for identifying early warning signs of disease from complex health indicators. Second, privacy protection during data usage is crucial due to the sensitive nature of clinical data. For instance, a seemingly innocuous indicator (e.g. neutrophils in asthma detection) may be associated with stigmatized illnesses (e.g. HIV/AIDS) that cause social isolation and discrimination. To achieve disease surveillance in a timely and privacy-preserving manner, we develop an online monitoring algorithm based on differential privacy. We incorporate an independent, Laplace-distributed random variable into the statistics of control chart. Further, we introduce an adaptive estimation method and a weighting function to reduce the sensitivity loss in the monitoring and robustness loss in the private monitoring algorithm. The proposed method adheres to the differential data privacy model. Simulation results and a real example of personal clinical diagnosis demonstrate the superiority of our design.