Hospital Readmissions Reduction Program

Overview

The Hospital Readmissions Reduction Program (HRRP) dataset is published by CMS as part of the Hospital Compare provider data files on data.cms.gov. It contains hospital-level Excess Readmission Ratios (ERRs) for six condition/procedure cohorts: acute myocardial infarction (AMI), heart failure (HF), pneumonia, chronic obstructive pulmonary disease (COPD), elective primary total hip/knee arthroplasty (THA/TKA), and coronary artery bypass graft surgery (CABG). Each row represents one measure for one hospital, with fields for the ERR, predicted and expected readmission rates, number of discharges, and number of readmissions. The current file covers FY2026, which uses a 3-year measurement window of approximately July 2019 through June 2022.

This dataset answers questions such as: which hospitals have higher- or lower-than-expected 30-day readmission rates for a given condition, how a hospital's readmission performance compares to the national expected rate, and which hospitals face payment reductions under the HRRP. The ERR is the key metric — a value above 1.0 means the hospital's readmission rate exceeds the risk-adjusted expected rate; below 1.0 means it performs better than expected.

Join Strategy

This dataset joins to hospital entity pages on CareGraph using the Facility ID field, which contains the CMS Certification Number (CCN) as a 6-digit zero-padded string. During ETL, the normalize_ccn() function strips whitespace and zero-pads values shorter than 6 characters to ensure consistent matching. Each source row is matched to a hospital entity page by CCN; rows with CCNs that do not match an existing hospital manifest are skipped. Because each hospital has multiple rows (one per condition measure), the ETL groups rows by CCN and attaches the full set of condition-level measures to the hospital's JSON manifest as an array. The join is a left join from the hospital manifest to the HRRP data — hospitals without HRRP records display missing data indicators rather than being excluded from the site.

Known Limitations

• Suppression for low volume. Hospitals with fewer than 25 eligible discharges for a given condition are excluded from that condition's ERR calculation and show "Too Few to Report." This is suppression, not an indicator of zero readmissions. Approximately 20–30% of measure-hospital combinations are suppressed, disproportionately affecting small and rural hospitals.
• No socioeconomic adjustment. ERRs are risk-adjusted using CMS's hierarchical logistic regression model, which adjusts for patient age, comorbidities, and clinical history but does not adjust for socioeconomic status. Safety-net hospitals serving high-poverty populations face systematically higher readmission rates that the model attributes to hospital performance rather than social determinants of health. The dual-eligible adjustment factor added in FY2019 partially mitigates this but does not fully account for social risk.
• Medicare fee-for-service only. The HRRP covers only Original Medicare (fee-for-service) discharges. Medicare Advantage enrollees, Medicaid-only patients, commercially insured patients, and uninsured patients are excluded. In markets with high MA penetration, the ERR is calculated on a smaller, potentially non-representative subset of a hospital's total discharges.
• IPPS hospitals only for penalties. HRRP payment reductions apply only to hospitals paid under the Inpatient Prospective Payment System. Critical Access Hospitals, Maryland waiver hospitals, and hospitals in US territories are exempt from penalties but may still appear in the data with ERR values, which can be misleading when comparing penalty status across facilities.
• 3-year measurement lag. The FY2026 data uses discharges from approximately July 2019 through June 2022. This means the data is a lagging indicator — recent quality improvements or declines at a hospital are not reflected in the current ERR values. The measurement window also overlaps with the COVID-19 public health emergency, during which CMS excluded certain discharge periods.

Data Quality Notes

• ERR fields stored as strings. The source CSV encodes Excess Readmission Ratio, Predicted Readmission Rate, and Expected Readmission Rate as string values. Suppressed rows contain "Not Available" or "Too Few to Report" instead of numeric values. The ETL parses these with _try_float(), converting non-numeric entries to null in the JSON manifest.
• Count fields with suppression markers. Number of Discharges and Number of Readmissions appear as string-encoded integers in the source CSV but contain "Too Few to Report" for suppressed rows. The ETL parses these with _try_int(), setting non-numeric values to null. When both count fields are null for a row, the corresponding ERR is also null.
• Column name variation across vintages. CMS has changed column header casing and naming between file releases (e.g., "Excess Readmission Ratio" vs. "Excess readmission ratio", "Facility ID" vs. "Facility Id"). The ETL uses a candidate-list column matching strategy to handle these variations without manual mapping updates.
• Missing value encoding inconsistency. The source data uses "Not Available", "N/A", and empty strings interchangeably for missing values depending on the field. The ETL normalizes all of these to null. Date fields, where present, use varying formats (MM/DD/YYYY, YYYY-MM-DD) and are standardized to ISO 8601 during processing.