Natural language processing and the oncologic history: is there a match?

Area of Science:

• Health Informatics
• Oncology
• Natural Language Processing

Background:

• Electronic health records (EHRs) contain valuable cancer patient data, but much is unstructured free text.
• Translating narrative oncologic histories into treatment summaries is difficult and time-consuming.
• Incentives increasingly require treatment summaries for patient use and care transitions.

Purpose of the Study:

• To assess provider perspectives on cancer treatment summaries.
• To evaluate the feasibility of using natural language processing (NLP) to extract information from EHRs for treatment summaries.

Main Methods:

• Surveyed providers at Beth Israel Deaconess Medical Center on treatment summary importance and use.
• Developed and applied an NLP program to extract information from narrative oncologic histories in EHRs.

Main Results:

• 86% of providers found treatment summaries important, but only 11% implemented them.
• Lack of time and EHR tools were key obstacles.
• Automated NLP demonstrated reasonable accuracy in extracting medical concepts from oncologic histories.

Conclusions:

• NLP technology shows promise for structuring free-text oncologic histories into compact treatment summaries.
• This can improve communication between providers and between providers and patients.
• Automated summarization can address current barriers to treatment summary formulation.