Application of artificial intelligence to identify needs in user messages and support digital navigation in the healthcare system: feasibility study

• Grant number: 23/15574-4
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Start date: August 01, 2024
• End date: July 31, 2025
• Field of knowledge: Health Sciences - Collective Health - Preventive Medicine
• Principal Investigator: Renata Luciria Monteiro
• Grantee: Renata Luciria Monteiro
• Company: AGILE HEALTH.TECH INOVA SIMPLES (I.S.)
• CNAE: Atividades de profissionais da área de saúde, exceto médicos e odontólogos ; Atividades de apoio à gestão de saúde ; Atividades de atenção à saúde humana não especificadas anteriormente
• City: São Bernardo do Campo
• Pesquisadores principais: Rogério Silicani Ribeiro
• Associated researchers: Alexandra Rodrigues dos Santos Silva
• Associated scholarship(s): 24/13349-6 - Artificial intelligence to identify needs in user messages and support navigation digital in the health system, BP.PIPE

Abstract

Patient navigation in the healthcare system improves access, care for chronic conditions, and prevents emergencies and hospitalizations. However, low patient adherence, the scarcity and cost of healthcare professionals limit its implementation. Two years ago, we created Journey, a digital navigation platform integrated with WhatsApp. Users are referred by health insurance companies and receive messages offering multidisciplinary support for any need. The team evaluates all responses and classifies them into: no needs (50%), clinical needs (48%) and administrative needs (2%). People with no needs are instructed to use the service when needed and up to 20% request support. People with needs are served with Journey's decision support system to personalize guidance, education, and referral to professionals and services, combining multiple care lines and integrating network services. In 18 months, we served 11,856 people (average of 2.6 visits per user) with high satisfaction (NPS score 91). The use of emergency room and hospitalization was 10% lower than for people who did not interact. Justification: Automatic message classification can direct human assistance only to people with needs and improve the effectiveness and scalability of the service.Objective: To evaluate the feasibility of using natural language processing to classify user responses and direct only people with needs to human assistance. Methodology: Supervised machine learning algorithms.Resources: We will use the anonymized database of user messages classified by the team, the service platform, cloud services, support teams, data, and technology.Expected results: Develop a model that allows to attend in an automated way at least 10% of the captured users.Impact: The company improves its operational efficiency and offers the market a cost-effective navigation service in the healthcare system to improve clinical outcomes and user experience and optimize resource utilization. (AU)