BACKGROUND: Childhood asthma continues to be a growing medical concern in the United States, affecting > 17 million children in 1998. The mortality rate from asthma in children aged 5 to 14 years has nearly doubled, from 1.7 deaths per million to 3.2 deaths per million between 1980 and 1993. OBJECTIVE: To evaluate the use of artificial neural networks (ANNs) to rate problem-based strategies for asthma management in a defined population of children. METHODS: The participants in our study were recruited from a local inner-city medical facility in Los Angeles. The majority of participants had received the diagnosis of mild-to-moderate-persistent asthma. Each participant was given 10 asthma-based problems and asked to manage them. Each management decision and its order were entered into a database. This database was used to train an artificial neural network (ANN). The trained ANN was then used to cluster the various performances, and outputs were evaluated graphically. RESULTS: Three hundred five performances were analyzed through our trained neural network. Our ANN classified five major clusters representing different approaches to solving an acute asthma case. CONCLUSIONS: ANNs can build rich models of complex phenomena through a training and pattern-recognition process. Such networks can solve classification problems with ill-defined categories in which the patterns are deeply hidden within the data, and models of behavior are not well defined. In our pilot study, we have shown that ANNs can be useful in automating evaluation and improving our understanding of how children manage their asthma.
BACKGROUND: Childhood asthma continues to be a growing medical concern in the United States, affecting > 17 million children in 1998. The mortality rate from asthma in children aged 5 to 14 years has nearly doubled, from 1.7 deaths per million to 3.2 deaths per million between 1980 and 1993. OBJECTIVE: To evaluate the use of artificial neural networks (ANNs) to rate problem-based strategies for asthma management in a defined population of children. METHODS: The participants in our study were recruited from a local inner-city medical facility in Los Angeles. The majority of participants had received the diagnosis of mild-to-moderate-persistent asthma. Each participant was given 10 asthma-based problems and asked to manage them. Each management decision and its order were entered into a database. This database was used to train an artificial neural network (ANN). The trained ANN was then used to cluster the various performances, and outputs were evaluated graphically. RESULTS: Three hundred five performances were analyzed through our trained neural network. Our ANN classified five major clusters representing different approaches to solving an acute asthma case. CONCLUSIONS: ANNs can build rich models of complex phenomena through a training and pattern-recognition process. Such networks can solve classification problems with ill-defined categories in which the patterns are deeply hidden within the data, and models of behavior are not well defined. In our pilot study, we have shown that ANNs can be useful in automating evaluation and improving our understanding of how children manage their asthma.