The Timeliness Audit, Missing Data Items, and Nine Refusal Charts
A three-hospital CoC-accredited program accessions about 1,100 analytic cases a year with four CTRs and two non-credentialed abstractors. The cancer committee's quarterly data quality report shows that only 68% of analytic cases were completed within 120 days of diagnosis against an internal target of 90%, and 21% of records are missing at least one required site-specific data item. Nine charts show the treatment field coded as unknown even though the physician notes document that the patient declined chemotherapy or chose hospice.
Informal review suggests three patterns: outside hospital records arriving weeks after the patient is accessioned, physician queries sitting unanswered for two to three weeks, and abstractors prioritizing easy breast and prostate cases over complex pancreas and lung cases. A CoC survey is scheduled in ten months, and the cancer committee chair wants a corrective action plan it can approve at the next meeting.
1. Build a corrective action plan for the timeliness and completeness problems. Identify root causes, interventions, owners, and measurable targets.
Read the worked answer
Start by separating the two problems, because they have different causes and different fixes. Timeliness is a workflow problem; missing site-specific data items are largely an information-arrival and knowledge problem.
The outside-records lag points to a casefinding and retrieval process that accessions before records are complete, then waits passively. Fix it with a defined escalation: request outside records within two business days of identifying the case, escalate at 14 days to the physician liaison or health information management, and track each request in a shared log. The query backlog points to the absence of a service-level expectation; introduce standardized query templates with a five-business-day turnaround, route unanswered queries to the medical director, and report physician response rates to the cancer committee.
The cherry-picking pattern indicates a workload design flaw, not a character flaw. Assign cases by site or by diagnosis month rather than by preference, balance complex cases across abstractors, and give the non-credentialed staff structured work such as casefinding and data entry with a CTR reviewing their abstractions.
Measure monthly: percentage abstracted within 120 days, percentage with complete site-specific data items, and query turnaround. Re-audit a random sample of 30 charts per month against a 90% timeliness and 95% completeness target. Assign an owner and due date to each action and present a run chart, not just a percentage, to the committee.
2. How should the nine refusal charts be corrected, and why does the distinction between unknown and patient refusal matter beyond the individual abstract?
Read the worked answer
Read each note and code what actually happened. If the physician recommended chemotherapy and the patient declined, the correct entry is that treatment was not administered because the patient or family refused, not unknown. If the patient elected hospice and the oncologist documented that treatment was not recommended given performance status, that is a different reason for no treatment, and it should be coded as not recommended rather than as a refusal. The two are not interchangeable.
Unknown is reserved for cases where the record genuinely does not document what happened. Using it when the answer is in the note converts a known clinical decision into missing data and cannot be rescued later, because abstractors and analysts downstream will not reopen the chart.
The consequences go beyond nine records. Treatment and reason-for-no-treatment fields feed quality measures, treatment-pattern analyses, and public reporting. Inflating the unknown category makes the hospital look like it has a data quality problem when it has a coding problem, and it can distort measures that exclude or include untreated patients. It may also mask a real pattern, such as a high refusal rate among a specific population that the cancer committee should be examining.
The correction is straightforward: re-abstract the nine charts, document the source note for each, and add a brief chart audit of all cases with unknown treatment fields that quarter to see whether the error is isolated or systemic.
3. The medical oncologist asks the registry to remove the nine refusal patients from the quality measure denominator so the reported rate improves. How should the registry respond?
Read the worked answer
Decline, and explain why in terms the physician will accept. Measure denominators are defined by eligibility criteria, not by outcomes or by how a case affects the score. Removing patients because they did not receive treatment is a form of selection that makes the remaining group look better than the population the hospital actually serves, and it would fail an audit immediately because the exclusion cannot be justified against the measure specification.
The correct handling is already covered by the measure: patients who were recommended treatment and refused remain in the denominator, with the reason for no treatment documented. That is exactly the information the measure is designed to capture. If the physician believes a specific patient was never eligible for the measured therapy, the route is to verify eligibility against the specification and document it, not to delete the case.
The registry should also point out the governance issue. Changes to measure denominators should be approved through the cancer committee and documented, so that there is a record of who decided what and why. A verbal request from a physician to remove cases is precisely the situation that produces survey findings and credibility damage. Offer instead to review the measure definition together and to present the corrected data at the next committee meeting.