DURHAM, NC — Bladder cancer is the VA’s fourth-most-diagnosed cancer, and early diagnosis is important because, if the tumor spreads outside the bladder, the five-year survival rate is only about 38%.

With about 3,200 veterans diagnosed with bladder cancer every year, researchers from the Durham, NC, VAMC and Duke Cancer Center sought to develop and validate a natural language processing (NLP) model for automatically identifying patients with muscle-invasive bladder cancer (MIBC).

The study team selected all 76,060 patients with a Current Procedural Terminology code for transurethral resection of bladder tumor (TURBT) from the VA database. A sample of 600 patients (with 2,337 full-text notes) who had TURBT and confirmed pathology results were selected for NLP model development and validation.

The researchers explained that the NLP performance was assessed by calculating the sensitivity, specificity, positive predictive value, negative predictive value, F1 score and overall accuracy at the individual note and patient levels.

Results from the study should that the NLP model can speedily and accurately pull essential information from databases with an overall accuracy of 92%. The study team advised that the NLP model identified invasion status for 96% of patients with TURBT at the population level, while also spotlighting 98% of patients with non-MIBC (NMIBC) using a range of clinical information, notes and pathology reports.

“Considering the very good performance in classifying patients with bladder cancer using a wide range of note types, our NLP model may be a practical and accurate tool for rapid identification of patients on the basis of invasion status, thus aiding in population-based research,” according to the JCO Clinical Cancer Informatics study.1

It added that, “applying the model to 71,200 patients VA-wide, the model classified 13,642 (19%) as having MIBC and 47,595 (66%) as NMIBC and was able to identify invasion status for 96% of patients with TURBT at the population level. Inherent limitations include a relatively small training set, given the size of the VA population.”

The authors suggested that the NLP model could be a useful tool for efficiently identifying BC invasion status and help in research.

 

  1. Yang R, Zhu D, Howard LE, De Hoedt A, Schroeck FR, Klaassen Z, Freedland SJ, Williams SB. Context-Based Identification of Muscle Invasion Status in Patients With Bladder Cancer Using Natural Language Processing. JCO Clin Cancer Inform. 2022 Jan;6:e2100097. doi: 10.1200/CCI.21.00097. PMID: 35073149.