Better offer screening
A pilot project tests ways to avoid unwanted offers, speed up organ allocation and transplant more organs.
To identify a program’s recommended filters, the model first finds the most effective offer filter for the kidney program. The most effective filter is the one that screens off the most donors.
Some transplant programs field many thousands of offers in a year, but might accept only a few hundred for their patients. In June, UNOS launched the first phase of a pilot project to develop an innovative tool that will help change that by altering how offers are filtered. The goal? Increase kidney utilization by helping OPOs find an accepting candidate more quickly.
"We want programs to understand what their practice is and to refine it—so organs are directed to centers that will take them, faster.” - Rob McTier, UNOS Business Architect
The offer filters tool will allow transplant hospitals to filter out donor offers they do NOT want to receive by creating custom-designed, multi-criteria filters, says UNOS Business Architect Rob McTier, who helms the project along with UNOS principal research scientists John Rosendale and Darren Stewart and UNOS Data Scientist Harris McGehee. Centers will also be able to select model-derived filters UNOS generated by applying data science to identify consistent patterns of offer refusal. “You don’t have to worry about there being overlap between any filters,” explains McTier, “if it meets one of your filters or all of your filters, you won’t get any offers for the organ.” Reducing unwanted organ offers reduces cold ischemic time and promotes increased utilization overall.
OPOs will also play a part in the pilot by reporting cross-clamp date and time, warm ischemic time, and kidney biopsy information, when making electronic organ offers on kidney matches.
The pilot is a collaborative effort among the transplant community and UNOS and has been years in the making. In preparation, a work group comprised of representatives from transplant centers and OPOs around the country was formed and staff from the 29 participating transplant hospitals reviewed modeling data to set their offer filters.
“That modeling data is a new way of understanding acceptance practices,” says McTier, adding that dialogue between UNOS researchers and participants about filters was an effective way for transplant programs to understand their decisions.
Filters chosen, evaluated, but remain off
On June 17, UNOS launched the first phase of the pilot with 29 kidney programs from across the country. Participants are using a tool called Offer Filters Explorer to select criteria to screen out kidney donor offers they do not want to receive. Though the selected filters will not actually screen offers during this phase, UNOS IT is tracking whether each offer would have been screened if the filter was turned on. The filter information will be displayed when a program receives a kidney offer. The filters are set at the center level so they apply to all candidates at the kidney program.
Chosen filters turned on to screen offers
The project is iterative and will include more participants and options as time goes on. In phase two, transplant centers will be able to turn their offer filters on and off. When a filter is active, the filter criteria are compared with donor data before electronic offers are sent. When the OPO initiates the electronic organ offer notification process, a bypass code is applied to all candidate offers where the donor data meets the filter criteria and the filter is active.
Candidate-level factors added
Phase three will give transplant centers the ability to customize filters and exclude candidates from filters based upon criteria such as CPRA (calculated panel reactor antibody), candidate age and candidate ABO. Selected candidate criteria will apply to all candidates on that center’s waiting list.
The model uses a data science method called recursive partitioning. It evaluates all combinations of donor factors and applies a kidney program’s last two years of offers to find potentially effective offer filters. These combinations are like the branches of a tree.
Originally published June 28, 2019