With newly available electronic health data and a massive increase in processing power, data-driven personalized medicine is just now becoming possible.1 However, advances to improve health care are inherently limited by data quality. One of the most used sources of data, the patient problem list, is also the greatest source of data inaccuracy. According to recent studies, the patient problem list is often less than 50% accurate in documenting the most critical conditions.2 3 4 5 These errors exacerbate inefficiencies throughout the American health care system from care delivery to quality improvement. Primary care physicians rely on problem lists to develop transitional treatment plans for the 68 million Americans who change providers every year. Errors related to care transitions harm more than 1.5 million people each year in the United States, costing the nation an estimated $3.5 billion annually.6 Population health efforts, a cornerstone of value-based healthcare, rely on problem lists to determine risk levels and deployment of resources. These efforts cannot succeed if the source data produce faulty results. This application seeks to enable better individual patient care, enhanced population health management, and effective downstream analytics by building an automated problem list builder, which provides an accurate and granular account of the patient?s medical conditions. If the program is successful, one of the greatest technical risks in value-based healthcare will be addressed. Phase I exceeded success criteria in proving feasibility of core modules in natural language processing (NLP) and artificial intelligence. Based on Phase I success, implementation pathways are demonstrated through pilots with one of the largest US healthcare systems and one of the largest global biotechnology firms. The team is comprised of commercial and academic leaders in the field of NLP-based products applied to value-based healthcare.

Public Health Relevance

VMT proposes to use advanced semantic technologies and artificial intelligence to enhance the individualized patient problem list, which frequently has 50% or lower accuracy for common conditions such as cancer, smoking, and heart failure. 1 2 3 The problem list represents source data in care delivery, population health, shared-risk contracting, and research. This proposal aims to support more accurate and granular source data to enhance care delivery and value-based healthcare.

Agency
National Institute of Health (NIH)
Institute
National Center for Advancing Translational Sciences (NCATS)
Type
Small Business Innovation Research Grants (SBIR) - Phase II (R44)
Project #
9R44TR002437-02A1
Application #
9622425
Study Section
Special Emphasis Panel (ZRG1)
Program Officer
Colvis, Christine
Project Start
2016-07-01
Project End
2020-08-31
Budget Start
2018-09-01
Budget End
2019-08-31
Support Year
2
Fiscal Year
2018
Total Cost
Indirect Cost
Name
Vmt, Inc.
Department
Type
DUNS #
827623245
City
Menlo Park
State
CA
Country
United States
Zip Code
94025