JMIR Publications

ABSTRACT

Background:

NLP techniques are widely used to support information extraction from unstructured documents in number of industries. Although there are some demonstrated use cases, the adoption of NLP techniques for extracting patient information from electronic medical record data for population health management and measurement has been slow.

Objective:

To outline practical considerations for clinicians and scientists that seek to translate their research NLP systems into health care operations and evaluate their potential benefit to patient care.

Methods:

Based on collective experiences developing NLP systems to support over 20 different EMR products, we enumerate a set of practical considerations for future system developers. Although they are often overlooked, thier consideration before key system decisions can help maximize the relevance of a new NLP system for real-world population health managment and measurement.

Results:

Practical considerations for a new NLP system include (1) determining the readiness of the data and compute resources for NLP, (2) the organizational incentives to use and maintain the NLP systems and (3) the feasibility of implementation and evaluation.

Conclusions:

In addition to demonstrating the ability to extract information from unstructured patient data accurately, minimum criteria are needed to establish the foundation of a successful translational NLP system. These are independent of the medical specialty, and relevant to new NLP systems for the US healthcare system, and globally.