Service: SNS24 Scout.AI - Analyse the information recorded within the SNS 24 TAE Service

Responsible organisation: SPMS - Serviços Partilhados do Ministério da Saúde, EPE (Central-Government)

The Portuguese national health service is tax-funded, and otherwise mostly free of charge for the citizens. Regarding telephone triage, Portugal has a national free of charge helpline for emergencies (112) and a separate national health service contact center helpline for non-emergent situations (SNS24), at the cost of a local call. Originally established in 1999, SNS24 evolved into a multichannel contact centre with the mission to facilitate the citizen’s access to the public healthcare services. Whilst providing several different services to callers, its core service is the nurse-led triage that uses protocols developed by the governmental public health agency. The Triage, Counselling and Referral Service (TAE) is a telephone service provided by the Contact Centre of the National Health Service - SNS 24. In 2018, more than 1 million calls with an average duration of 7-8 minutes were answered. Being of national scope, this is a service that promotes equity in the access to health care. Telephone service is provided by nurses and follows pre-defined clinical algorithms. Triage is based on a specific clinical algorithm (out of a set of 59), and the choice of the most appropriate algorithm is extremely important and relevant. The selected clinical algorithm should ensure high safety (not failing to identify situations that require urgent medical contact) and should have high discriminatory capability. In this context, the SNS24 Scout.AI will apply Artificial Intelligence (AI) methodologies, aiming the development of decision support tools with two main objectives: 1. Support the nurse in the selection of the most appropriate clinical algorithm. 2. Provide support to Directorate-General of Health (DGS) in the optimization process of the design of clinical algorithms and their referrals. The first objective will be achieved by identifying the most appropriate algorithms for a given set of symptoms, with adjustment for age and sex. The AI methodology to be applied will be based on a classifier built on automatic learning algorithms on an anonymised data set, obtained from contacts of the SNS 24 in 2017 and 2018. This data represents accumulated experience in around 2 million cases. After the creation of the prediction model, it will be implemented in the SNS 24 TAE Service as decision support, indicating in real time which algorithms are most likely to be used. The second goal is to create a support tool in the process of optimising the design of clinical algorithms and their referrals. Adding to the data already mentioned the referral and its adequacy, as well as the diagnoses established at hospital level, it will be possible to: - assess the clinical performance of each referral, when classified in terms of safety and discriminatory capacity; - predict the impact of potential changes to algorithms in terms of safety and performance; - predict hospital-level diagnoses for each algorithm or set of symptoms. With this analysis, it will also be possible to detect any anomalous patterns and allow the DGS to make more informed decisions supported by real world data. In order to achieve these objectives, Natural Language Processing (NLP) and Machine Learning (ML) techniques developed specifically for the Portuguese language will be used, as well as Knowledge Representation and Reasoning (KRR). The motto of the texts, the extraction of appointed entities and the creation of lists of bigrams and trigrams, as well as the association with international medical classification systems (International Classification of Diseases), using ontology mapping techniques will be carried out. At ML level, automatic classifiers based on Support Vector Machines (SVM) and deep neural networks (e.g. bi-LSTM with attention mechanisms) will be developed. Statistical methodologies will also be applied to detect anomalous situations in the application of clinical algorithms and their referral. Therefore, the SNS24 Scout.AI project will automatically analyse the information recorded within the SNS 24 TAE Service, using a built-in classifier using ML techniques to help nurses select the most appropriate clinical algorithm. It will also support the optimisation of the design in the SNS 24 algorithms by analysing the results of the referral of each case, improving its discriminatory capacity and clinical safety.

Additional information

Source Open Innovation Regione Lombardia
Web site https://www.spms.min-saude.pt/2019/08/spms-recebe-apoio-da-fct-para-projetos-de-ciencia-dos-dados-e-ia/ https://nova-lincs.di.fct.unl.pt/news-single/98
Start/end date 2021 -
Still active?

Related cases