46 Automated Derivation of Diagnostic Criteria for Lung Cancer using Natural Language Processing on Electronic Health Records: A pilot study

Abstract

Introduction The digitisation of healthcare records has generated vast amounts of unstructured free text and similar data, this presents opportunities for improvements in disease diagnosis when traditional clinical coding falls short, such as in the recording of patient symptoms. We have piloted an approach using natural language processing (NLP) to extract clinical coding concepts from free text which are used to automatically form diagnostic criteria for lung cancer from unstructured secondary care data. Methods Patients aged 40 and above who underwent a chest x-ray (CXR) between 2016-2022 at a large London NHS Trust were included. ICD-10 codes and unstructured data were pulled from their electronic health records (EHRs) from the 12 month period prior to the CXR. The unstructured data were processed using a combination of NLP techniques and ontological methods to make clinical statements computable, with symptoms extracted and mapped to SNOMED-CT expressions. Candidate features were taken forward to the model development phase, with a 'genetic algorithm' employed to identify the most discriminating features. Results 75,002 patients were included, with 1,012 lung cancer diagnoses made within 12 months of the CXR. The best performing model achieved an AUROC of 0.72, outperforming currently used models (Figure 1). An existing 'disorder of the lung' (e.g. pneumonia) and 'cough' increased the probability of a lung cancer diagnosis. 'Anomalies of great vessel', 'disorder of the retroperitoneal compartment' and 'context-dependent findings' (e.g. pain) statistically reduced the risk of lung cancer, making other diagnoses more likely. [Formula presented] Conclusions The proposed methods demonstrated success in leveraging unstructured secondary care data to derive diagnostic criteria for lung cancer, outperforming existing risk tools. If a similar model could be utilised in primary care this could lead to focused CXR requesting improving rates of early lung cancer diagnosis. Disclosure This work was funded by AstraZeneca; of whom MS is an employee. CT is a Director, employee and shareholder at Clinithink.