Predicting patient acuity from electronic patient records.

Area of Science:

• Health Informatics
• Computational Intelligence
• Natural Language Processing

Background:

• Accurate prediction of patient acuity is crucial for effective healthcare resource allocation.
• Healthcare data, including electronic nursing narratives, can be leveraged using computational intelligence.
• Mathematical optimization models can support tactical decision-making in healthcare management.

Purpose of the Study:

• To develop and evaluate a mathematical model for automated patient acuity score prediction.
• To assess the predictive performance of linguistic and text-modeling techniques on electronic nursing records.
• To determine the optimal combination of data sources for accurate acuity score forecasting.

Main Methods:

• Utilized a dataset of 23,528 electronic patient records with nursing narratives and Oulu Patient Classification (OPCq) acuity scores.
• Employed linguistic pre-processing, vector-space text modeling, and regularized least-squares regression for acuity prediction.
• Evaluated a mathematical model for automated assignment and prediction of patient acuity scores.

Main Results:

• Accurate predictions of next-day patient acuity scores are achievable using previous day's data.
• Same-day predictions yield improved performance by incorporating current day's nursing notes.
• Combining textual nursing notes with previous acuity scores provides the most accurate predictions (concordance index of 0.821).

Conclusions:

• Language technology applied to electronic patient documents enables accurate prediction of future patient acuity scores.
• The developed model demonstrates the potential of utilizing nursing narratives for proactive healthcare management.
• Predictive acuity scoring can enhance resource allocation and operational efficiency in healthcare settings.