readmit (version 0.0.1)

hsr_readmission_risks: Compute discharge-level readmission risks in a Hospital-Specific Report (HSR)

Description

Computes the predicted and expected readmission risks for each eligible discharge in the specified cohort.

Note: CMS changed the format of Hospital-Specific Reports (HSRs) for FY2026 (see here). The current HSR functions support formats through FY2025.

Usage

hsr_readmission_risks(file, cohort)

Value

A tibble::tibble() containing the following columns:

• ID Number: The unique discharge identifier (see hsr_discharges())

• Predicted: The predicted readmission risk for the discharge

• Expected: The expected readmission risk for the discharge

Arguments

file

File path to a report

cohort

Cohort to compute readmission risks for. One of c("AMI", "COPD", "HF", "PN", "CABG", "HK")

Details

The readmission measure is what CMS uses to grade performance in the Hospital Readmissions Reduction Program (HRRP).

Individual discharges are assigned an adjusted readmission risk (based on clinical history), which then get aggregated into a hospital-level score and compared to peer groups for penalty determination. Specifically, a random-intercept logistic regression model is built for each cohort (see methodology) which serves as the basis for two (2) readmission risks assigned to each discharge:

• Predicted: Adjusted for patient-specific clinical factors plus the hospital-specific effect (random intercept term)

• Expected: Adjusted for patient-specific clinical factors plus the hospital-average effect

These quantities are then aggregated across all discharges and their ratio is taken to form the Excess Readmission Ratio (ERR), which is then used as the cohort-specific comparison metric. Thus, it is a comparative measure of how likely patients are to be readmitted at your hospital versus the average hospital, given your hospital's clinical characteristics.

Examples

# Access a report
my_report <- hsr_mock_reports("FY2025_HRRP_MockHSR.xlsx")

# Compute readmission risks for HF discharges
hf_risks <- hsr_readmission_risks(my_report, "HF")
hf_risks

# Compute the ERR from scratch
hf_risks |>
 dplyr::summarize(
   Discharges = dplyr::n(),
   Predicted = mean(Predicted),
   Expected = mean(Expected),
   ERR = Predicted / Expected
 )


# Check that this matches the report table
hsr_cohort_summary(my_report) |>
 dplyr::select(
  dplyr::matches(
     paste0(
       "^Measure|",
       "^Number of Eligible Discharges|",
       "^Predicted|",
       "^Expected|",
       "^Excess"
     )
   )
 )