We developed an Automated Case Review Assignment System that enhanced Quality Assurance process efficiencies.
Our client was in dire need of an efficient system that would help them choose cases to be reviewed for QA purposes. As one of the leading radiology groups in the US, they had a huge volume of un-reviewed cases that needed a proper QA analysis – a task that was manually laboring and inefficient. Radiologists had to choose which exams to review, and the QA team would have to manually assign cases to fill in any paucity in coverage. Service level agreements with different medical groups dictated levels of coverage, in terms of modality, interpreter, and site, which became a logistical morass for the team. The system also led to an imbalance in workloads
Problems That Needed To Be Addressed
The methodology that was being followed by the company for assigning peer reviews of cases for the QA team was not scalable. The company itself was growing with dozens of sites and thousands of exams per day.
The Solution We Provided
Our team at Expeed conceived and built a rule-based system that allowed for automated selection of cases to be reviewed, thereby ensuring the sample covered all types of cases and with almost no resource hours lost in the selection process.
The Impact and Long-term Results
With our automated review assignment system in place, the client’s QA team was able to keep up with the ever-expanding growth of the parent company and allowed them to reliably deliver contracted reviews to medical providers with whom they had service level agreements.
Our client was especially happy that the intrinsic work related to each review was accounted for instead of treating every review as equal work.
So, Here’s How We Did It
Rule-based System: We conceived and built a rule-based system that enabled automated assignment of exam reviews. These rules could be applied at different levels. For instance, a rule could be applied as broad as an entire practice or as narrowly as an individual radiologist. Specific categories also included sites and types of employees, for instance, new hires. The categories were dynamically linked to the companies database so that as options changed, for example, a new radiologist is hired, options automatically update. In addition to a selection of radiologists to which a rule applied an amount could be indicated, in either percentage or absolute numbers.
Automated Selection: The set rules were then cumulatively applied to exams in a rolling window. For instance, if 2% of exams were to be reviewed, 2% of the exams for the previous 30 days would be randomly selected for review. The process was set to repeat every day and the exams previously selected would already be accounted for.
This simple concept was applied throughout the system and on all rules using hierarchical layering. This helped minimize the number of exams selected while satisfying all rules that had been created.
Effective Sample Selection: Our process ensured that the exams that were selected for review were relatively recent and evenly distributed in time. The size of the window was adjustable so that stochasticity in the sample could be managed.
Resource Management: Finally, each exam was assigned to a radiologist that was credentialled to review the exam. This process included accounting for any specialities that might be applicable, i.e., MSK, Neurology, etc. Workloads were balanced among all radiologists using the Relative Value Unit (RVU) assigned to each procedure. As new exams were selected for review, it was balanced to perform across all viable peers. Thus, no member of the radiology team felt encumbered because they were assigned more complex exams to review.