Our project aims to look into the types of injuries that dancers experience, where these injuries are most likely to happen, how they are happening, and if age can predict these injuries. Our main aim is to investigate when injury prevention education should begin for dancers, so that the number of long-term injuries can be prevented.
We did this by using data from the National Electronic Injury Surveillance System. On their website you are able to submit a query with the defined parameters of data you need, which for this project was dance injuries from the past 10 years. After submitting this query, we were able to download the excel files to our computers and begin the data cleaning and analysis process.
For data cleaning we first took out any unnecessary columns. Then we filtered the narrative column to rule out any rows that did not include the words “dance” and “class” in them, so we could be sure we were only looking into the injuries of long-term dancers. We then imported the data into Jupyter Notebook and concatenated all of them together into one dataframe. The data now included 6 columns and 4911 rows.
Then we completed our Exploratory Data Analysis. From this we were able to find that the average age of the dancers in the sample was 14, and the most common injuries were a strain/sprain. We also looked into the relationships between the months and the frequencies of injuries, where we found that February and October were the months with the most frequent injuries happening. We also visualized the proportion of injuries onto a photo of a dancer in order to show where these injuries were most often occurring, which ended up being the knee and ankle. We also ran a Latent Dirichlet Analysis on the narrative column to see if there was a pattern in the way dancers were getting injured.
We then ran Random Forrest Classifier and SVM Classifier models on our data to first predict whether a dancer experienced a strain or not, and then if they experienced a fracture or not. Both models did fairly well, but the fracture models did better in terms of accuracy. We also found the feature importances for each column in contributing to the models and in both the sprain and fracture models, the body part was the most important feature for prediction.
Through our results, we uncovered several important findings related to injury patterns in dancers. We found that dancers are most vulnerable to injury between the ages of 10 and 15, with the highest incidence occurring at age 14, making this age group particularly prone to serious injuries such as sprains, strains, and fractures. Injury prevention education should begin early, ideally before age 10, as injury rates start to rise at this point. Early education on which body parts are most prone to injury, the types of injuries that can occur, and how seasonal patterns affect injury risk can help dancers make more informed decisions about their training and warm-ups. Additionally, our models confirmed that age is a strong predictor of sprains, strains, and fractures, but body part was the most critical factor in predicting these injuries, highlighting the need for targeted education on specific injury-prone areas like the knee, ankle, and foot. The most common injuries, sprains, strains, and fractures, typically occur in the knee, ankle, and foot, and our analysis revealed that terms like “fell,” “pop,” and “twisted” were commonly associated with these injuries, offering insights into their causes. In conclusion, our findings stress the importance of early injury prevention education focused on vulnerable age groups and injury-prone body parts, helping dancers reduce their injury risk and make safer, smarter choices in their training.