Small-Cell Suppression in a Public Quality Report
A regional health information exchange publishes a quarterly public report on 30-day readmission rates by hospital and by condition. One rural critical access hospital has 41 eligible discharges for heart failure in the quarter, with 9 readmissions (22%). The state average is 21%. The HIE's policy suppresses cells with fewer than 11 cases. The hospital's CEO argues that suppression hides the hospital's good performance and that the rate should be published because 41 exceeds the threshold. The HIE analyst notes that the readmission count of 9 is itself small and that a single readmission changes the rate by more than two percentage points. A reporter has requested the underlying counts.
1. Evaluate the statistical and privacy considerations that bear on whether this rate should be published.
Read the worked answer
Two separate concerns apply, and the CEO has conflated them.
Statistical reliability. The suppression threshold of 11 is applied to the wrong cell. The relevant small number is not the 41 discharges but the 9 readmissions and, more importantly, the instability of the estimate. With a denominator of 41, one additional readmission moves the rate from 22.0% to 24.4% - a swing larger than the difference from the state average. A confidence interval around 22% with n=41 is wide, likely spanning the state average. Publishing the rate without an interval invites a comparison the data cannot support. The correct treatment is to publish with a confidence interval and an explicit caution, or to aggregate across quarters.
Privacy. The suppression rule protects against re-identification in small cells. Here the readmission count of 9 is below 11, so the count cell itself is subject to suppression. In a small rural community, a count of 9 readmissions for a named condition at a named facility may be re-identifiable when combined with local knowledge.
So the CEO's argument that 41 exceeds the threshold misreads the rule: the rule applies to the count being disclosed, and the count is 9.
2. The reporter requests the underlying counts. Advise the HIE on its response and on how to communicate the rate if it is released.
Read the worked answer
Response to the reporter:
- Apply the published suppression policy consistently. If the policy suppresses cells under 11, the count of 9 is not released, and the response should state the policy and the reason (statistical reliability and privacy), not simply refuse.
- Offer what can be released: the rate with its confidence interval and a note that the denominator is small, or a multi-quarter aggregate that meets the threshold.
- Consistency matters: releasing the count to a reporter after suppressing it publicly would breach the policy and the trust of the reporting hospitals.
If the rate is released, communicate it responsibly:
- Present the rate with the confidence interval and the denominator.
- State explicitly that the difference from the state average is not statistically meaningful.
- Avoid ranking hospitals with small denominators; present as a distribution or with a band rather than a league table.
- Add a plain-language note explaining why small numbers produce unstable rates.
The governing principle: transparency is served by publishing what the data can support, with the uncertainty attached - not by publishing a point estimate that misleads, and not by suppressing everything to avoid discomfort.