Artificial intelligence methods for infectious disease outbreak detection: a systematic review

Background

Outbreak detection is essential to prevent the spread of infectious diseases in the population. Artificial intelligence (AI) expands the possibilities of outbreak detection through additional analysis approaches and enabling the use of untapped data sources, such as unstructured data besides structured data. This systematic review identified AI methods used for outbreak detection of infectious diseases in electronic health data (EHD), and assessed which were potentially suitable for the general practice (GP) setting.

Methods

AI methods were defined as machine learning (ML)-based approaches in which the mapping between inputs and outputs is at least partially learned from data through an algorithmic training process. The search was conducted in Medline, Embase, Web of Science and IEEE Xplore, followed by article screening. Extracted data included AI method, data type, healthcare setting and case vs outbreak detection. Quality assessment was performed with Probast-AI. Studies conducted within general practices, investigating outbreak detection, multiple infectious diseases, validation and/or comparison with other models were potentially suitable for the GP setting.

Results

We identified 14,633 articles, of which 36 were included. These were based on structured (n = 16), unstructured (n = 14) or both (n = 6) data types. Preliminary results showed that 7 AI methods were applied to the GP setting, of which 5 used a natural language processing (NLP) method, whereas the rest used an ML method.

Conclusions

We identified AI methods for outbreak detection of infectious diseases based on structured and/or unstructured EHD. AI methods applied to the GP setting consisted of an NLP component or an ML method. Application of potentially suitable AI methods in the GP setting, could lead to improved signalling about an upcoming outbreak, ultimately allowing for timely public health action.

Key messages

• Identification of AI methods for outbreak detection of infectious diseases based on structured and/or unstructured electronic healthcare data.

• Determining which artificial intelligence methods for outbreak detection of infectious diseases are suitable for the general practice setting.