Abstract Breast Imaging Reporting and Data System (BI-RADS) is a standard system used by clinicians to describe mammogram findings and sort the results into categories numbered 0 through 6; this makes accurately communicating about these test results and following up after the tests much easier. For example, the BI-RADS 4 score is given to lesions that carry a risk of malignancy between 2 and 95% such that a majority (69 to 95%) of BI-RADS 4 lesions are biopsied. Our goal is to create a better risk estimate score to stratify patients by developing a more accurate threshold for biopsy. To accomplish this goal, we defined the personal risk factors, relevant descriptors in the mammography and pathology findings for every patient with BI-RADS 3, 4, or 5 risk score. Clinical data were sourced from the Methodist Environment for Translational and Outcomes Research (METEOR) data warehouse at the Houston Methodist Hospital. METEOR data warehouse contains 135,280 unique patient mammogram reports dating from January 1, 2006 to May 30, 2015. Manual abstraction from this large number of free text reports is not feasible due to the constraints on time and labor costs, and risk of human-error during manual extraction. We developed a natural language processing (NLP) tool called Methodist hOspital Text Teaser (MOTTE) to extract defined clinical parameters and derive the desired meaning from the volumes of free text reports automatically. MOTTE is a rule-based software tool programmed in Java, Structured Query Language (SQL) and an open source machine learning toolkit, OpenNLP. MOTTE combines regular expressions and algorithms to identify where in the dictionary keywords and phrases related to a concept are found. Mammography findings from January 2006 to May 2015 were extracted including breast density as well as the presence of mass, calcifications, architectural distortion, asymmetric density, and calcification characteristics. A total of 717 patients with BI-RADS 5 mammograms who underwent breast biopsy within 3 months of the abnormal mammogram report date were identified using MOTTE, which also extracted personal history, family history, and subtype information (ER, PR, and HER2 status) of those patients. Clinical data including age, race, height, weight, and body mass index (BMI) were extracted from the METEOR data warehouse for those 717 patients. The extracted data were verified and validated with 99% accuracy by assessing a random 10% of records against a gold standard of manual chart review, which was completed by the clinical coauthors. MOTTE surpasses the manual method in terms of consistency, time, cost, and data preparation. Our ongoing study is to apply the MOTTE to extract distinct characteristics of BI-RADS 4 scored patients. The finding would help to refine biopsy recommendations and create a predictive model to drive evidence-based biopsy decision-making in breast cancer care. Citation Format: Puppala M, He TC, Ogunti R, Wong STC. Use of natural language processing on mammography and pathology findings to supplement BI-RADS to improve clinical decision making in breast cancer care [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P5-03-08.