Failure to Rescue

Failure to Rescue in Hospital Quality and Economic Evaluation

Failure to rescue (FTR) judges a hospital by what happens to patients after a complication has already occurred, most often after surgery. It estimates the probability that such a patient dies, which separates the response to a complication from the occurrence of the complication itself. The split matters to health systems because hospitals with similar complication rates can have very different death rates, and the gap tends to track rescue capacity such as nurse staffing, intensive care and escalation systems, all of which cost money. For health economists the same split gives decision models a natural structure, so that a safety intervention is credited only with the deaths it actually prevents.

Origins in surgical outcomes research

Silber and colleagues introduced the term in 1992 in a study of 5,972 Medicare patients undergoing elective cholecystectomy or transurethral prostatectomy. Alongside the death rate they measured an adverse occurrence rate (patients with a complication over all patients) and a failure rate (deaths among patients with a complication over patients with a complication). The adverse occurrence rate was associated mainly with patient characteristics, while failure to rescue was associated more with hospital characteristics and was less influenced by severity of illness on admission. A hospital's own contribution to mortality therefore shows up more clearly after a complication than in the complication count.

Splitting operative mortality into complications and rescue

Overall mortality after an operation is a weighted sum of two conditional death rates. The decomposition is an identity, so it holds in any cohort, and it shows that a hospital can lower deaths by preventing complications or by rescuing more of the patients who have them.

$$M = c \times F + (1-c) \times d_0$$

where $M$ is the overall death rate in the surgical cohort, $c$ is the proportion of patients with at least one qualifying complication, $F$ is the failure-to-rescue rate (deaths among patients with a qualifying complication divided by patients with a qualifying complication), and $d_0$ is the death rate among patients without a recorded qualifying complication.

Ghaferi, Birkmeyer and Dimick applied this split to 84,730 general and vascular surgery patients in the American College of Surgeons quality improvement registry (NSQIP) between 2005 and 2007. Death rates were 3.5% in very-low-mortality hospitals and 6.9% in very-high-mortality hospitals, yet overall complication rates were similar, at 26.9% and 24.6% respectively. Mortality among patients with major complications was 21.4% in the very-high-mortality hospitals against 12.5% in the very-low group. The variation in mortality lay mainly in $F$ and hardly at all in $c$.

Measure definitions and why the denominator matters

Every FTR figure rests on choices about eligible patients, qualifying complications, the deaths that count and the follow-up window. Changing any of them can change which hospitals look best, so a reported rate means little without its specification.

Silber and colleagues compared three definitions in 403,679 general surgical patients across 1,567 hospitals in 1999 and 2000 Medicare data. A nursing-sensitive version (FTR-N) and a version used by the Agency for Healthcare Research and Quality (AHRQ), known as FTR-A, restricted complications and deaths and omitted 49% and 42% of deaths respectively. Their reliability was 0.18 against 0.32 for the original all-deaths measure, and the authors recommended the original definition for analysing hospital quality.

AHRQ adopted a measure of this kind as Patient Safety Indicator 04 (PSI 04) in 2003, according to its Patient Safety Network (PSNet) primer. The v2025 specification counted in-hospital deaths per 1,000 surgical discharges with one of five groups of serious treatable complications, so a rate of 120 per 1,000 is an FTR of 12%.

AHRQ retired PSI 04 in its v2026 release and replaced it with Inpatient Quality Indicator 40 (IQI 40), a beta measure that adopts the failure-to-rescue framework of the Centers for Medicare and Medicaid Services (CMS) across 30 postoperative complication categories. IQI 40 counts in-hospital deaths within 30 days of the first operating room procedure and describes itself as an estimate of the probability of death given a postoperative complication. The PSNet primer reports that the CMS version excludes patients whose only qualifying complication was diagnosed before admission and was adopted for Hospital Inpatient Quality Reporting for the fiscal year 2027 payment determination.

Rescue capacity as a health system resource

Rescue depends on a sequence of actions after a complication begins. The PSNet primer describes rescue failing at recognition of the complication, relay of information or reaction in time, and lists staffing levels, electronic early warning systems, rapid response teams and escalation protocols among the interventions studied. Each is a service with a budget line, which is where FTR enters health economics.

The best-known evidence concerns nursing. In 168 Pennsylvania hospitals, Aiken and colleagues found that among 232,342 surgical patients each additional patient per nurse was associated with a 7% increase in the odds of failure to rescue, after adjustment for patient and hospital characteristics.

In a national business case, Needleman and colleagues found that more nursing hours reduced hospital days, adverse outcomes and deaths for a net rise in hospital costs of 1.5% or less at the staffing levels modelled. They concluded that whether staffing should rise depends on the value patients and payers assign to avoided deaths and complications, which is a cost-effectiveness analysis question.

The cost side is thinly studied. In 451,490 elective colorectal resections in the US National Inpatient Sample between 2016 and 2019, Schultz and colleagues found median admission costs of USD 42,295 for rescued patients and USD 53,182 for failures to rescue, a difference of USD 10,887. A rapid review by Subbe and colleagues found no publications on long-term costs outside hospital and judged that care needs after discharge and lost income for patients and carers are likely to outweigh the direct cost of the admission.

Modelling a rescue intervention in an economic evaluation

The mortality identity maps directly onto a decision tree. The first chance node is whether a qualifying complication occurs, with probability $c$, and the second is whether a patient with a complication dies, with probability $F$. A prevention bundle changes $c$ and a rescue programme changes $F$, so their effects combine multiplicatively and adding the deaths each would avert alone overstates the joint effect.

The complication definitions and death window must be the same in both arms and match the source of $F$. Costs should include the programme and the extra treatment that survivors receive, and survivors accrue QALYs and costs over their remaining lifetime.

Worked example with a rescue programme and a prevention bundle

All numbers in this example are illustrative. A hospital performs 2,000 eligible operations a year, 20% of patients develop a qualifying complication, the failure-to-rescue rate is 15%, and the death rate among patients without a complication is 0.5%.

1. Baseline deaths. Complications affect 0.20 × 2,000 = 400 patients, of whom 0.15 × 400 = 60 die. Of the 1,600 patients without a complication, 0.005 × 1,600 = 8 die. Total deaths are 68, so $M$ is 68 / 2,000 = 0.034, matching 0.20 × 0.15 + 0.80 × 0.005 = 0.034.

2. Rescue programme alone. A rapid response and staffing programme lowers $F$ to 12%. Deaths after complications fall to 0.12 × 400 = 48 and total deaths to 48 + 8 = 56, which is 12 fewer than at baseline.

3. Adding a prevention bundle. A bundle that lowers $c$ to 17% leaves 340 complicated patients. Alone it gives 0.15 × 340 = 51 deaths plus 0.005 × 1,660 = 8.3, avoiding 8.7. With the rescue programme as well, deaths are 0.12 × 340 = 40.8 plus 8.3, a total of 49.1 and 18.9 avoided. Adding the separate gains would give 12 + 8.7 = 20.7, overstating the joint effect by 1.8 because the programme now acts on 60 fewer complicated patients (60 × 0.03 = 1.8).

4. Cost-effectiveness of the rescue programme. The programme costs GBP 600,000 a year and each extra survivor needs GBP 15,000 of further care, so 12 × 15,000 = GBP 180,000 and the incremental cost is GBP 780,000. If each survivor gains 6 discounted QALYs, the gain is 12 × 6 = 72 QALYs and the ICER is 780,000 / 72 = GBP 10,833 per QALY gained. The NICE technology appraisal manual (PMG36) gives an English reference point, treating GBP 25,000 to GBP 35,000 per QALY gained as the usual cost-effectiveness range.

5. A measurement trap. Suppose instead that the hospital begins coding minor complications, so 500 patients are recorded as complicated while deaths stay at 60. Measured FTR falls to 60 / 500 = 0.12, the same figure the programme achieves, with no extra survivor. A model built on this before-and-after comparison would credit 12 lives that were never saved.

Comparing hospitals with failure-to-rescue rates

FTR is used for benchmarking hospitals and for public quality reporting. Fair comparison needs risk adjustment and enough cases for a stable rate. Schwappach, Zwahlen and Havranek applied the PSI 04 definition to Swiss data for all acute hospitals between 2019 and 2023, covering 41,506 surgical inpatients with qualifying complications. A multilevel model attributed 1,045 of 7,114 FTR deaths (14.7%) in hospitals with at least 100 cases to below-average hospital performance.

Such results are usually reported as risk-standardised ratios, built as described under observed-to-expected ratio and standardised mortality ratio. The problems specific to FTR lie in the denominator. Measures built from administrative data depend on how thoroughly complications are coded, a form of detection bias. The PSNet primer notes that length of stay and early discharge to post-acute facilities vary widely across hospitals, which can bias comparisons of in-hospital deaths even after risk adjustment.

Boundaries with neighbouring measures

FTR and the adverse event rate are the two main factors in overall mortality, with deaths among patients without a recorded complication making up the rest, so neither substitutes for the overall death rate. FTR also differs from a broad outcome measure, since it is a conditional rate designed to isolate the response of the care system. Within patient safety it captures the recovery side of harm and says nothing on its own about how many complications were avoidable.

Common misreadings

Most errors with FTR come from forgetting that the rate is conditional on a complication being recorded. Others come from reading a hospital signal as a patient-level verdict.

1. Treating a low rate as proof of good rescue. Recording more minor complications lowers the rate with no change in care, as step 5 shows.
2. Reading every death in the numerator as avoidable. The rate signals where to look but does not show which step failed, and it is not a verdict on an individual clinician.
3. Adding prevention and rescue effects. The two act on the same patients in sequence.

Sources

• Agency for Healthcare Research and Quality. Patient Safety Indicator 04 (PSI 04) Death Rate among Surgical Inpatients with Serious Treatable Complications. AHRQ QI ICD-10-CM/PCS Specification v2025, August 2025. https://qualityindicators.ahrq.gov/Downloads/Modules/PSI/V2025/TechSpecs/PSI_04_Death_Rate_among_Surgical_Inpatients_with_Serious_Treatable_Complications.pdf
• Agency for Healthcare Research and Quality. Inpatient Quality Indicator 40 (IQI 40) Inpatient Mortality Rate Among Surgical Inpatients with Complications, Beta. AHRQ QI ICD-10-CM/PCS Specification v2026. https://qualityindicators.ahrq.gov/Downloads/Modules/IQI/V2026/TechSpecs/IQI_40_Inpatient_Mortality_Rate.pdf
• Agency for Healthcare Research and Quality. IQI technical specifications and v2026 release notes. https://qualityindicators.ahrq.gov/measures/IQI_TechSpec
• Aiken LH, Clarke SP, Sloane DM, Sochalski J, Silber JH. Hospital nurse staffing and patient mortality, nurse burnout, and job dissatisfaction. JAMA. 2002;288(16):1987-1993. https://doi.org/10.1001/jama.288.16.1987
• Ghaferi AA, Birkmeyer JD, Dimick JB. Variation in hospital mortality associated with inpatient surgery. New England Journal of Medicine. 2009;361(14):1368-1375. https://doi.org/10.1056/NEJMsa0903048
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). Published 31 January 2022, last updated 13 October 2026. Section 6.3. https://www.nice.org.uk/process/pmg36
• Needleman J, Buerhaus PI, Stewart M, Zelevinsky K, Mattke S. Nurse staffing in hospitals: is there a business case for quality? Health Affairs. 2006;25(1):204-211. https://doi.org/10.1377/hlthaff.25.1.204
• Schultz KS, Moore MS, Pantel HJ, Mongiu AK, Reddy VB, Schneider EB, Leeds IL. For whom the bell tolls: assessing the incremental costs associated with failure to rescue after elective colorectal surgery. Journal of Gastrointestinal Surgery. 2024;28(11):1812-1818. https://doi.org/10.1016/j.gassur.2024.08.019
• Schwappach D, Zwahlen M, Havranek MM. Between-hospital variation in failure to rescue after major surgery. JAMA Network Open. 2026;9(2):e2555855. https://doi.org/10.1001/jamanetworkopen.2025.55855
• Silber JH, Williams SV, Krakauer H, Schwartz JS. Hospital and patient characteristics associated with death after surgery: a study of adverse occurrence and failure to rescue. Medical Care. 1992;30(7):615-629. https://doi.org/10.1097/00005650-199207000-00004
• Silber JH, Romano PS, Rosen AK, Wang Y, Even-Shoshan O, Volpp KG. Failure-to-rescue: comparing definitions to measure quality of care. Medical Care. 2007;45(10):918-925. https://doi.org/10.1097/MLR.0b013e31812e01cc
• Subbe C, Hughes DA, Lewis S, Holmes EA, Kalkman C, So R, Tranka S, Welch J. Value of improving patient safety: health economic considerations for rapid response systems, a rapid review of the literature and expert round table. BMJ Open. 2023;13(4):e065819. https://doi.org/10.1136/bmjopen-2022-065819
• Tokareva I, Romano P. Failure to rescue. PSNet primer. Agency for Healthcare Research and Quality; updated January 2025. https://psnet.ahrq.gov/primer/failure-rescue