[Predictive Model for the Number of Patients Presenting to the Emergency Department of a University Hospital in Santiago, Chile].

UNLABELLED: Emergency department (ED) overcrowding is a critical global challenge, affecting the quality and timeliness of medical care. Chile is no exception: In 2024, over 18 million ED visits were recorded in the public health system, reflecting high demand and complicating efficient resource planning and allocation.

AIM: To develop a predictive model capable of accurately estimating the weekly number of consultations in the emergency department of a private university hospital in Santiago, Chile. The model was expected to generate a management impact if it could predict ED visits 14 days in advance with a margin of error below 5%.

METHODS

A regression model was developed using supervised machine learning based on time series data. The methodology followed the Knowledge Discovery in Databases (KDD) process, utilizing historical data from 2024. Relevant variables were selected to train and validate the model, with a focus on accurately forecasting weekly consultation volume.

RESULTS

The model successfully predicted the number of weekly ED consultations 14 days in advance, achieving an average error rate of 3.3%. This represents a significant improvement over traditional methods based on historical averages and enables better anticipation of patient demand and improved resource management.

CONCLUSIONS

Weekly ED demand can be accurately predicted with a margin of error below 5% using supervised learning-based predictive models. This type of tool can contribute significantly to improving internal management and strategic planning in university hospitals and other emergency care centers across the country.