First Real-Time Artificial Intelligence (AI) flu prediction model appeared

Recently, E-BioMedicine, a sub-journal of The Lancet, published a study of Chinese scientists using adaptive  Artificial Intelligence (AI) models and multi-source data to predict the flu activity in Chongqing. The influenza real-time prediction model is also a very influential result of AI in the field of infectious disease prediction.

(Source: Pixabay)

The research results were jointly completed by Ping Technology, Ping Smart City and Chongqing Center for Disease Control and Prevention, Army Military Medical University and Tsinghua University. As we all know, influenza continues to threaten global public health security. Seasonal flu causes hundreds of millions of people worldwide each year and hundreds of thousands of people die. In addition, influenza virus genes continue to mutate, and the risk of a global influenza pandemic persists. Monitoring and forecasting influenza activities and making appropriate prevention and control preparations in time are essential for the prevention and control of seasonal influenza and pandemic influenza. However, traditional influenza surveillance requires time for reporting, collating and statistical processing of data, and indicators related to influenza activity levels are often delayed by one to two weeks to obtain, which is very unfavorable for the prevention and control of influenza, an acute respiratory infection Will lead to miss the best time for epidemic control.

In order to solve the lag problem of influenza activity index data, in the past decade, people have conducted a lot of exploration on the prediction of influenza activity. Among them, Google Flu Trends (Google Flu Trends) created a new pattern of using Google search data to predict the level of flu activity. Since then, multi-source electronic data, including Internet search data, influenza surveillance data, influenza-related posts on Twitter, Wikipedia access logs, and electronic health records, combined with mathematical models, have very good disease tracking and prediction effects. However, many of these studies are conducted at the national level, and their predictions are difficult to translate into actionable information that local health officials can make better decisions. In some areas, such as a city, the flu activity is not equivalent to the situation at the national level. Because of the influence of factors such as specific weather changes in the region, economic and social activities, population immunization and personal habits, flu epidemics in some regions often show a more diverse epidemic pattern.

Today, some studies have established influenza prediction models in cities such as New York, Melbourne, and Hong Kong, and have done many valuable explorations and attempts. However, the current method for predicting influenza activity in urban areas with irregular influenza activity still lacks an effective method with higher accuracy. Therefore, a more scientific and smarter method of influenza prediction has become an urgent need to implement influenza prevention and control.

Theoretically: (1) Combining multi-source big data and artificial intelligence methods to predict influenza activities one week in advance, overcoming the problem of relatively lagging results brought by traditional influenza surveillance "timed sampling, weekly summary" method; (2) An artificial intelligence algorithm establishes an adaptive model and dynamically adjusts parameters, which can accurately capture the irregular seasonal trends of influenza, and provides an effective method for influenza prediction in such areas; (3) Analysis of the importance of the model to the characteristics of the influenza epidemic, Revealed some relevant factors of the influenza epidemic, and provided new ideas for public health workers to study influenza epidemic.

Practical application: (1) Timely and reliable influenza prediction can help government agencies make the scientific and prospective allocation of public health resources, and enable medical and health institutions to prepare in advance during the influenza season and respond to patient flow reasonably; (2 ) In the long run, the release of influenza forecasts can also increase public awareness of the prevention and control of infectious diseases such as influenza. Implementing effective influenza prediction is an inevitable way to move the disease prevention and control forward.

Influenza prediction models are not only theoretical innovations in the prevention and control of infectious diseases but also play a role in actual prevention and control. We developed an influenza prediction system based on the research results. From the 12th week of 2018, we have implemented it in the Chongqing CDC, predicting influenza activities a week in advance, and made early warnings for influenza prevention and control, helping health departments to carry out influenza activities. Online monitoring and early response. The establishment of the disease prediction model and the launch of the intelligent infectious disease prediction system can help the health department to carry out a predictive and early warning more scientifically and accurately, and help government departments to improve efficiency in the prevention and control of related diseases and reduce the cost of disease prevention and control. For ordinary people, the epidemic index of infectious diseases released regularly also improves the people's understanding of the disease and the awareness of health and safety, so that it becomes passive to deal with after the event to take the initiative to prevent in advance. Therefore, our disease prediction model not only achieves a leap in medical and health technologies, improves the prevention effect of the people, but also has a positive impact on the people's health awareness, and realizes the real advancement of disease prevention and control.