Cancer registrars are a small workforce with a large job: turning sprawling patient records into standardized registry data that supports accreditation, research, and quality measurement. With fewer than 6,000 ODS-certified registrars worldwide and cancer cases climbing, hospitals face a structural capacity gap. Brent Dover, CEO of Carta Healthcare, argues that AI can close that gap - not by replacing registrars, but by taking on the time-consuming retrieval work while credentialed professionals stay accountable for every submitted answer.
Dover's position is that registry abstraction is not a task suited for full automation. "Cancer is not one event in one note. It is a story that builds over months across pathology, imaging, surgical findings, molecular and genomic testing, and treatment that often moves between different facilities," he said.
Registrars must reconcile conflicting information and apply detailed rules for staging and histology. Errors carry weight beyond a single record. "A mistake in staging or histology is not cosmetic," Dover said. "It can shift a hospital's reported outcomes, break a quality measure, or drop a patient out of a research cohort they should have been counted in."
A shortage with consequences
Fewer than 6,000 ODS-certified cancer registrars exist worldwide, and the pipeline is not keeping pace with rising cancer incidence. Certification requires time, and the work cannot be shifted to general clinical staff. "You have more cases arriving, fewer trained people to handle them, and no fast way to close the distance between the two," Dover said. "That is a structural shortage, not a rough patch in hiring."
For hospitals, the stakes include Commission on Cancer accreditation, which depends on complete and timely submissions to the National Cancer Database. Backlogs can affect a cancer program's standing, weaken outcomes research, and limit the ability to identify patients for clinical trials.
Recruiting more registrars will not solve the backlog problem, Dover said. "The one path that matches the size of the problem is making each registrar far more productive, so the handful of credentialed people a hospital already employs can cover many more cases without cutting corners on quality."
The registrar-in-the-loop model
Carta's approach keeps the registrar in the loop. The AI reads the chart, proposes answers to registry questions, and links each answer to the source material. The registrar verifies or corrects the proposal and owns the final submission.
"The registrar stops hunting and starts checking," Dover said. "The AI handles the retrieval and takes the first swing at the answer. The credentialed professional handles the judgment, which means confirming it, correcting it, and owning what finally gets submitted."
Dover draws a clear line between accelerating the mechanics of abstraction and automating the judgment behind it. He said the goal is to reduce searching, copying, and manual cross-checking while preserving professional review. In registries where Carta's model is already running, he said abstraction time has fallen by as much as 66%, costs by half or more, and inter-rater reliability has stayed above 98%.
Hospitals considering AI for registry work should judge it by operational measures they already track: turnaround time, backlog, cost per case, and inter-rater reliability. "The biggest lesson has almost nothing to do with the software," Dover said. Successful projects fit into existing workflows and make clear from the start that AI is meant to increase productivity, not eliminate jobs.
Why this matters for healthcare professionals
For CIOs, CMIOs, and cancer program leaders, the takeaway is that AI can clear a backlog that recruiting alone will not fix. The technology works best when it accelerates the work of the credentialed registrars already on staff, not when it replaces them. In oncology, where a wrong answer can ripple beyond a single patient record, keeping a qualified professional answerable for every submission is the discipline that makes the speed worthwhile.
Your membership also unlocks:
