Identifying Frequent Users of Emergency Medical Services: A Study Using Natural Language Processing

Strategic Insights -

A significant proportion of older adults in the Netherlands experience frequent utilization of emergency medical services (EMS), which can be challenging to identify and provide tailored interventions. A recent study from the University of Groningen employed a natural language processing (NLP) approach to analyze EMS records and identify older patients at risk of frequent usage. The study, published in the Journal of the American Medical Directors Association, explored the use of machine learning algorithms to predict frequent EMS users.

Key Takeaways:

  • The study retrospectively analyzed patient data from EMS records from 2013 to 2019, identifying 9.8% of patients as frequent users, who accounted for 28.6% of EMS use.
  • The authors used an Extreme Gradient Boosting (XGBoost) algorithm to develop a machine learning model that achieved a high c-statistic of 0.97 (95% CI, 0.97-0.97) on the training set.
  • Test set performance was slightly lower, with a c-statistic of 0.92 (95% CI, 0.90-0.92), but the model showed good calibration metrics.
  • The study found that the NLP model could identify word importance and provide example predictions, indicating its potential to improve identification of frequent users.
  • The authors suggest that integrating such tools into prehospital care systems may support earlier identification and tailored intervention strategies for older people frequent in use of acute care.

Statistics:

  • 97,736 patients were included in the study, with a median age of 78 years and an interquartile range of 72-85.
  • 9.8% of patients were identified as frequent users, accounting for 28.6% of EMS use.
  • The machine learning model achieved a c-statistic of 0.97 (95% CI, 0.97-0.97) on the training set, with a recall of 0.93, precision of 0.47, and F1 score of 0.63.
  • Test set performance was slightly lower, with a c-statistic of 0.92 (95% CI, 0.90-0.92), recall of 0.82, precision of 0.41, and F1 score of 0.54.

Sources:

  • Frequent Emergency Medical Services Utilization Among Older Patients: Exploration and Automatic Identification Using Natural Language Processing. Journal of the American Medical Directors Association, 2025:105814.