In an effort to overcome this, one method is using tools that search the text in the electronic medical record for specific terms, regardless of who saw the patient. For lesions such as kidney masses, often diagnosed on imaging, these “crawlers” or “datamining” programs can search radiology imaging reports to identify all patients for specific terms – regardless of who they were seen and managed by. This could eliminate these biases and benefit patient care, medical research, and trial recruitment.
While prior studies have already used this technology, this particular abstract derived and internally validated text search algorithms by using one such program to identify four common urological lesions (solid renal masses, complex renal cysts, adrenal masses, and simple renal cysts) using radiology text reports.
To accomplish this, they utilized a random sample of reports from 10,000 ultrasound and computed tomography studies of the abdomen from their institution’s data warehouse. Reports were also manually reviewed to determine the true status of the four lesions. Using commonly available software, they created logistic regression models having as predictors the presence of a priori selected text terms in the report. As a validation, they utilized a second external random sample of 2,855 reports, stratified by the number of points for each abnormality, to measure the accuracy of each lesion’s point system.
The protocol is diagrammed below:
In the original dataset, the prevalence of solid renal mass, complex renal cyst, adrenal mass and simple renal cyst, was 2.0%, 1.7%, 3.2%, and 20.0%, respectively. The table below demonstrates the increasing OR with higher points – and the point scoring system itself.
Each model they generated for each of the 4 terms contained between 1 and 5 text terms with c-statistics ranging between 0.66 and 0.90. Using the second dataset, in the external validation, the scoring systems accurately predicted the probability that a text report cited the four lesions.
The table below demonstrates the prevalence and probability of the lesion scores in both datasets.
Based on these results, the authors conclude that textual radiology reports can be analyzed using common statistical software to accurately determine the probability that important abnormalities of the kidneys or adrenal glands exist. These methods can be used for case identification or epidemiological studies. In fact, these have already been used to good effect!
Presented by: James Ross, MD, University of Ottawa, Ottawa, Canada
Written by: Thenappan Chandrasekar, MD, Clinical Instructor, Thomas Jefferson University, @tchandra_uromd, @JEFFUrology at American Urological Association's 2019 Annual Meeting (AUA 2019), May 3 – 6, 2019 in Chicago, Illinois