Three Different Heart Failure Readmission Rates and a Request to Challenge the State
Your 340-bed hospital publicly reports a 30-day all-cause readmission rate for heart failure. Three numbers are circulating: the internal EHR report shows 21.4% (n=312 discharges), the claims-based vendor measure shows 24.8% (n=298), and the state's public report, released last week, shows 27.1% and places the hospital in the bottom decile. The CFO wants to publicly challenge the state's methodology before the next board meeting. The quality analyst notes that the EHR report counts only readmissions to your hospital, excludes observation stays, and does not apply the planned-readmission algorithm. A re-abstraction of 20 charts found 4 discrepancies against the current specification. The vendor measure includes Medicare Advantage but not VA or out-of-state claims. The state report's confidence interval is 24.2%–30.4%. The board quality committee meets in 10 days and the hospital's community benefit report is due in six weeks.
1. Reconcile the three rates. What explains the gaps, and what must be done before any external response?
Read the worked answer
They are three different measures wearing the same name. The EHR rate is internal-only: readmissions to your hospital, observation stays excluded, no planned-readmission algorithm, and dependent on discharge-disposition documentation - each of those pushes it downward. The vendor measure adds outside hospitals and Medicare Advantage but misses VA and out-of-state claims, so it is closer to the state figure without being identical. The state measure presumably follows a defined specification: index heart failure cohort, 30-day all-cause, planned readmissions excluded, risk adjusted, with defined exclusions such as hospice and transfers.
Four discrepancies in 20 charts is a 20% disagreement rate against your own specification. That is a measurement-system failure, not a rounding issue. Before responding, commission a proper re-abstraction - larger sample, independent reviewer, inter-rater reliability calculated - and reconcile chart by chart against the state's published specification. Verify data submission integrity: were all eligible discharges submitted, how were Medicare Advantage and VA patients handled, were diagnosis mappings correct.
Check statistical inference too. The point estimate is 27.1% with a confidence interval of 24.2% to 30.4%; determine whether the state's method flags your hospital as a statistical outlier or merely above average before anyone uses the word outlier in public.
2. What do you tell the board, and do you file a public challenge?
Read the worked answer
Appeal only on a documented specification or data error. "Our internal number is lower" is not grounds, and a public challenge that fails costs credibility with the board, patients, and the state - credibility you will need the next time a number is genuinely wrong.
Board presentation in 10 days: one slide with the three numbers, why they differ, the validation result and what it did or did not find, the confidence interval in plain language, the appeal decision with its rationale, and a 90-day action plan with named owners. Be explicit that the internal measure was incomplete and is being retired.
Externally, coordinate with communications and legal on a single statement focused on patients: what the measure means, what the hospital is doing about it, and when the next result will be available. Do not attack the state's methodology without evidence. If the re-abstraction does reveal a submission or specification error, file a factual correction request through the state's published process rather than a press release.
Governance fix, and this is the durable answer: designate one official readmission specification with a written definition, a validation schedule, and a single owner, and stop circulating competing internal numbers. For the community benefit report due in six weeks, use the validated figure with context and the improvement plan, not the most flattering number.