For example, missing an early clinical clue suggestive of bladder or renal cancer can delay a diagnosis until months later when disease has progressed. Causes of delays are multifactorial and include, among others, information overload, poor follow-up systems and tracking tools, suboptimal teamwork and communication structures, decreasing access to care, as well as patient-related factors, such as patients failing to attend follow-up visits and testing.
Detecting such delays is difficult and time-consuming. Because of the large amount of resources required to perform non-selective patient records reviews (e.g., reviewing all patient records with an abnormal result), few health care organizations employ ongoing methods to detect delays. Fortunately, the increasing use of electronic health records to store clinical data offers a method to automate the detection process using “triggers.” Triggers are electronic algorithms that scan vast amounts of data to find specific clinical clues suggestive of high-risk situations. Use of triggers could allow health care organizations to focus their resources on just those patients at highest risk of having a delay.
Using expert opinion, we developed criteria for an electronic trigger algorithm that could scan hundreds of thousands of records to look for delays in care (>60 days) after clinically-significant hematuria (>50 red blood cells per high powered field). The algorithm automatically excluded patients deemed to be low-risk, such as those at young age, those who received typical follow-up action (e.g., cystoscopy or abdominal imaging), or those with a known benign cause for the hematuria, such as recent surgery or acute nephrolithiasis. The algorithm also excluded patients where follow-up would not be indicated, such as those with a known cancer diagnosis, prior cystectomy, or other terminal illness.
Of 5,857 patients with clinically-significant hematuria, 495 (8.5%) were flagged by the trigger as being at high risk for a delay. We manually reviewed a sample of 400 of the flagged records and found that the trigger correctly flagged 232 (58%) records. Those incorrectly flagged were commonly due to the presence of free-text notes that described a reason the patient was low-risk (e.g., follow-up at an outside institution), but where the trigger algorithm could not make use of the data. Overall, the trigger algorithm achieved a sensitivity and specificity of 64% and 96%, respectively. A total of 14 patients experiencing a delay were subsequently diagnosed with renal or bladder cancer within two years.
Triggers offer the promise of greatly improved efficiency for detecting delays in bladder cancer diagnosis. In this study, we estimated that triggers would reduce the number of manual reviews needed to detect delays by 91%. While our trigger was tested on retrospective data, future application could be performed in real-time to enable measurement of the direct impact on timeliness of diagnosis. This study provides a promising first step in the testing of triggers to detect delays in urologic cancer diagnoses and provides basis for future application and testing of triggers in this and other high-risk conditions.
Written By: Daniel R. Murphy, M.D., MBA, Michael E. DeBakey Veterans Affairs Medical Center (MEDVAMC), Houston Center for Innovations in Quality, Effectiveness & Safety (IQuESt)
Read the Abstract