Transcript Document
Improving Early Detection of Serious Disease Natural Language Processing to increase follow-up on significant incidental findings DISCLAIMER: The views and opinions expressed in this presentation are those of the author and do not necessarily represent official policy or position of HIMSS. Clinical Information: 38 year old female with diabetes, status post MVA, restrained passenger, with no head trauma or loss of consciousness. Assess for skeletal and internal injuries. Current Imaging Procedure Description: CT Chest Abdomen Oral contrast: Oral contrast was administered. IV contrast: 90 mL Chest: with intravenous contrast Abdomen: with intravenous contrast status post MVA … Assess for skeletal and internal injuries. Radiology Comparison study: None. Procedure Findings: Chest Lungs and large airways: Lungs are clear bilaterally. Pleura: Normal. Vessels: Normal. Atherosclerosis: No evidence. Heart: Heart size normal with no pericardial effusion. Mediastinum and hila: Subcentimeter axillary nodes. Chest wall: The bony structures of the thorax are WNL. No acute rib fractures. There is a 1.8 cm breast mass in the right upper quadrant of the right breast appreciated on image 4. Abdomen Liver: Normal. Bile ducts: Normal. Gallbladder: No calcified gallstones. Normal caliber wall. Pancreas: Normal. Spleen: The spleen is enlarged, but homogenous. Adrenals: Normal. Kidneys: Small 8 mm left superior renal cyst. Bowel: Normal caliber. Mesenteric lymph nodes: No enlarged mesenteric lymph nodes. Peritoneum: No ascites or free air; no fluid collection. Abdominal wall: Normal. Impressions: Essentially negative CT scans for acute traumatic injuries. There is a 1.8 cm breast mass in the right upper quadrant … Impressions … negative … for acute traumatic injuries. Incidental Finding Unexpected medical result not related to the reason for an ordered study lung scan clogged artery x-ray for back pain spot on liver cystitis urinalysis trace protein in liver 3 Mean frequency of incidental findings in CT imaging is 31% Lumbreras et al. 2010. Incidental findings in imaging diagnostic tests: a systematic review. The British Journal of Radiology. 4 28% of incidental breast lesions were unsuspected breast cancers Moyle et al. 2010. Incidental breast lesions detected on CT: what is their significance? The British Journal of Radiology. 5 Most concerning findings referred for follow-up in only 49% of cases Munk et al. 2010. Frequency and Follow-up of Incidental Findings on Trauma Computed Tomography Scans: Experience at a Level One Trauma Center. The Journal of Emergency Medicine. 6 Substantial disability and medical cost, cause for malpractice claims Daugherty et al. 2014. Tracking Incidental Findings. Radiology Today. 7 Low follow-up Volume of exams Impressions vs Findings Complexity Priority Omission of non-suspicious Familiarity with finding User interfaces 8 Demonstration Automated Flagging of Significant Incidental Findings LifeCode® Natural Language Processing Optum Enterprise & Professional CAC & CDI 9 Interoperability Communicating non-significant incidentals may: Overwhelm and panic the patient Divert attention from primary conditions Lead to unnecessary tests Lead to additional treatments with risk Selective targeting with scenario driven logic Flag Abdominal Aortic Aneurysm Flag Lung Cancer Scenario Abdominal aorta, unusual size Scenario Weak evidence with prior exam change abdominal aorta diameter > 3.0 cm lung consolidation worsening “The proximal abdominal aorta measures 3.1 centimeters in diameter.” “... there is interval increase in left upper greater than lower lung consolidation …” 1 Selective targeting with natural language processing • Synonymy, paraphrase, word variants lung, pulmonary, pleural, subpleural, midlung, … breast, nipple, retroareolar, subareolar, … nodule, nodules, nodular, nodularity, … • Negation and History “No sonographic evidence of a liver mass.” Evidence negated, does not support flagging “History: History of ovarian cancer, status post surgery and chemotherapy.” Historical reference, does not support flagging 1 Optum Connecting and serving the health system 74 million Consumers Thank you! 1