Running towards health: the association of running volume with running-related injuries

• Ano de defesa: 2020
• Autor(a) principal: Nakaoka, Gustavo Bezerra
• Orientador(a): Hespanhol Junior, Luiz Carlos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Cidade de São Paulo
• Programa de Pós-Graduação: Programa de Pós-Graduação de Mestrado em Fisioterapia
• Departamento: Pós-Graduação
• País: Brasil
• Palavras-chave em Português: Running; Sports injuries; Decision support techniques; Artificial intelligence; Neural networks (Computer); Machine learning; Athletic injuries; Sports; Risk management; Endurance training
• Área do conhecimento CNPq: CNPQ::CIENCIAS DA SAUDE::FISIOTERAPIA E TERAPIA OCUPACIONAL
• Link de acesso: https://repositorio.cruzeirodosul.edu.br/handle/123456789/881

Objectives: To develop an artificial intelligence (AI) algorithm in order to identify running-related injury (RRI) risk profiles in recreational runners, and to investigate the internal validity of such algorithm. Methods: This was a 3-step AI study using data from a prospective cohort study. In step 1, variable selection and exploratory analyses were conducted in the original (n=191) and simulated data (n=5000). In step 2, the AI algorithm was developed using machine learning techniques (selforganising maps, k-means and probabilistic neural network). The algorithm was trained in 80% (n=4000) of the simulated data, and the internal validity was investigated applying the algorithm in the remaining 20% (n=1000). The characterisation of RRI risk profiles was performed in step 3. Results: Four out of eight variables included in the algorithm were considered the main classification features: sex; running intensity; history of RRIs; and current musculoskeletal complaints or discomfort related to running Five groups were suggested by the AI algorithm. Male runners reporting previous RRIs and running in low-to-moderate intensities (>6 min/km) were at the highest risk of RRIs. Male runners reporting previous RRIs and running in high intensities (3 to 5 min/km) in about 32.1% of the time were at the lowest risk of RRIs. The accuracy of the RRI risk algorithm presented a median of 99.6% (25% to 75% interquartile range 99.5% to 99.8%). Conclusions: An AI algorithm was successfully developed and was able to correctly classify more than 99% of the runners in five RRI risk profiles.