Readmission

Readmission

A hospital readmission is a subsequent inpatient admission following discharge from an earlier admission. It is both an event in a patient's care pathway and, when defined over a specified population and interval, a service-use or quality indicator. This page explains how to count it, how to interpret a readmission rate, and why an observed return to hospital is not automatically avoidable or evidence of poor care.

Define the event and observation window

Start by naming the index admission, the qualifying original hospital stay whose discharge starts follow-up. Specify eligible patients, discharge destination, what counts as an inpatient readmission, the length and start of the window, whether returns to any hospital are captured, and whether planned returns are excluded. A 30-day window is common in US quality programmes, but “readmission” does not inherently mean 30 days or use the same rules across datasets.

Design choice	Possible rule	Why the choice matters
Start of follow-up	Date of live discharge from an eligible index admission.	Counting from admission or using incomplete discharge dates changes exposure time.
Subsequent event	First qualifying inpatient return within 30 days, to the same or another hospital.	Emergency visits without inpatient admission may be a separate outcome.
Planned care	Exclude admissions classified as planned under a prespecified algorithm.	A scheduled cancer treatment need not signal a failure of discharge care.
Transfers	Treat contiguous acute-care stays under a stated transfer rule.	An immediate transfer should not casually be counted as a new failure.
Follow-up completeness	Observe the full window after each index discharge or handle truncation appropriately.	A discharge near a data cutoff otherwise appears to have fewer returns.

The US Centers for Medicare & Medicaid Services (CMS) Hospital Readmissions Reduction Program uses condition- and procedure-specific 30-day risk-standardized unplanned readmission measures after an index admission. Its exact exclusions, eligibility and risk model belong to that programme; a local all-cause rate or an AHRQ descriptive rate need not use identical rules. Record the measure version and population before comparing numbers.

Count episodes, then calculate a rate

For a simple descriptive measure, count eligible index discharges with at least one qualifying readmission within the window and divide by all eligible index discharges with adequate follow-up. This is an admission-level proportion: the same person may contribute more than one eligible index episode under some protocols. State how subsequent episodes and multiple returns are treated, and use the same rule in numerator and denominator.

Suppose a fictional hospital has 1,000 eligible live discharges with complete 30-day follow-up. Eighty have a qualifying unplanned inpatient return within 30 days. The crude proportion is $80/1{,}000=0.08$, or 8%. If a later comparable cohort has 60 such discharges among 1,000, the observed proportion is 6%; the absolute change is $0.06-0.08=-0.02$, or a two-percentage-point decline. The relative decline is $(0.08-0.06)/0.08=0.25$, or 25% of the earlier rate.

Quantity	Spreadsheet expression	Result
Earlier crude rate	=80/1000	8%.
Later crude rate	=60/1000	6%.
Later minus earlier	=60/1000-80/1000	-2 percentage points.
Relative decline from earlier	=(80/1000-60/1000)/(80/1000)	25%.

The illustrative 20 fewer index discharges with a return are not necessarily 20 prevented readmissions or 20 patients benefited. Eligibility, severity, coding and capture of returns outside the hospital might have changed; multiple returns per index episode are not represented by the binary indicator. A percentage-point change is not a percent change, and neither establishes a treatment effect.

Interpret the indicator with clinical context

Some returns reflect complications, medication problems, poor transitions or inadequate access to follow-up. Others reflect progression despite appropriate care, a needed new treatment, patient choice or an unrelated acute illness. “Unplanned” is not synonymous with “preventable.” Review clinical cases and the reasons for return before setting a preventability target or attributing every event to the discharging hospital.

Comparing providers requires attention to baseline illness, case mix, social circumstances, service availability and referral patterns. Risk standardization can improve comparability for measured factors, but it does not erase unmeasured differences or decide which factors ought to be adjusted for. A crude 8% rate should not be compared directly with a CMS risk-standardized rate or with a rate from a different service line. Small volumes create unstable rates, so present counts and uncertainty as well as percentages.

Watch for competing outcomes and perverse incentives

Death before a possible return prevents readmission and can make a low rate look favorable for the wrong reason. Examine mortality alongside readmissions, particularly in populations at substantial risk of death, and use a competing-risk framework if estimating the probability of readmission over time is the analytic aim. Transfers, observation stays and emergency-only visits may shift when payment or quality incentives change; track these outcomes instead of rewarding a rate reduced by relabeling care.

A reduction in returns may also shift costs to families or community services, while an early appropriate readmission might avert more serious harm. Monitor patient-reported health, safety, access and total care use. A readmission metric is a signal for investigation, not a complete assessment of hospital quality.

Use readmissions in economic evaluation

An intervention such as discharge planning, medication reconciliation or post-discharge follow-up has implementation costs and may affect return admissions, other services, mortality and quality of life. Estimate its effect against a relevant comparator using an appropriate design and time horizon. Multiply an estimated causal reduction in qualifying events by applicable incremental resource costs only after checking event definitions and whether costs shifted elsewhere; do not use the difference between two crude rates as if it were automatically attributable to the programme.

For a health economic model, specify whether the input is a 30-day risk, a time-to-event quantity, or a rate per patient-year. Align baseline risk, relative effect, follow-up, population and death assumptions. Include uncertainty and plausible heterogeneity. “Avoided admission cost” is not automatically a net saving to every payer or provider, since payment arrangements and fixed costs differ.

Sources and further reading

• Centers for Medicare & Medicaid Services, Hospital Readmissions Reduction Program, for current programme scope and 30-day risk-standardized unplanned measures: https://www.cms.gov/medicare/quality/value-based-programs/hospital-readmissions
• Agency for Healthcare Research and Quality, HCUP Statistical Brief 304, for an explicit descriptive 30-day all-cause admission-based definition: https://hcup-us.ahrq.gov/reports/statbriefs/sb304-readmissions-2016-2020.jsp
• Agency for Healthcare Research and Quality, “How CMS Measures the 30-Day All Cause Readmission Rate,” for risk-standardization concepts: https://www.ahrq.gov/patient-safety/settings/hospital/red/toolkit/redtool-30day.html