Near Miss Rate

Near Miss Rate in Health System Safety Monitoring

Near miss rate measures how often errors in care are caught before they reach a patient, counted against the volume of care in which such errors could occur. Hospitals and health systems track it next to harm rates because near misses are far more frequent than harmful events and are thought to share their causes. For a health economist the number needs care, since a rise can mean more errors, better interception or more reporting, and none of these shows that harm or cost has fallen. The same words carry a separate meaning in maternal health, where the World Health Organization (WHO) counts women who nearly died per 1,000 live births.

Where the near miss boundary sits

The rate is only as clear as the definition behind its numerator, and the main schemes draw the line in different places. The WHO conceptual framework for the International Classification for Patient Safety, version 1.1 of January 2009, separates four kinds of incident. A reportable circumstance is a situation with significant potential for harm in which no incident occurred, a near miss is an incident that did not reach the patient, a no harm incident reached the patient without discernible harm, and a harmful incident is an adverse event. Its near miss example is a unit of blood connected to the wrong patient's line but detected before the infusion started.

The Common Formats used by the Agency for Healthcare Research and Quality (AHRQ) in the United States keep the same boundary. Its hospital reporting format classifies each report as an incident, a near miss or an unsafe condition, and treats any exposure of the patient as reaching the patient, so a wrong tablet handed to a patient who recognises and refuses it is an incident and not a near miss.

The AHRQ Patient Safety Network (PSNet) glossary uses a wider definition, an event or situation that caused no injury only because of chance, which can include a patient with a penicillin allergy who receives penicillin and has no reaction. Under it some no harm incidents count as near misses, so two hospitals can report different rates for identical practice, and a near miss rate needs a named classification scheme held fixed across any comparison.

Why health systems count events that caused no harm

The case for counting near misses rests on a causal claim. Barach and Small, reviewing 25 non-medical reporting systems in fields such as aviation and nuclear power, argued that the same patterns of failure precede both adverse events and near misses, and that only the presence or absence of recovery mechanisms decides the actual outcome. They reported that near misses occur 3 to 300 times more often than adverse events, enough for quantitative analysis, with fewer barriers to collection and limited liability.

In economic terms a near miss is cheap information. The cost to the system is mostly staff time spent detecting, reporting and reviewing it, while what it shows about weak defences benefits later patients. Frequency matters too, since harmful events on one ward may be too rare to show change within a year while near misses accumulate quickly enough to monitor. Whether staff report them depends in part on blame-free reporting.

Building the rate from events and opportunities

A near miss rate divides a count of qualifying events by a measure of how often such an event could have happened. The denominator should count opportunities for the specific error, not a convenient activity figure, and it must cover the same setting and period as the numerator.

$$R = \frac{N}{E} \times k$$

where $R$ is the near miss rate, $N$ is the number of distinct qualifying near misses recorded in the period, $E$ is the number of eligible opportunities for that error in the same setting and period, and $k$ is a scaling constant such as 1,000.

The numerator counts events rather than reports, so duplicate reports of one intercepted error are merged. The table pairs common near miss types with matching denominators.

These pairings are illustrative, not set by any body. A bed-day denominator is easy to obtain but mixes many processes, so a change in case mix moves the rate with no change in safety. The share of incident reports that are near misses is a different number, describing the reporting system, since its denominator holds only reported events.

What a change in the rate can mean

The PSNet primer on reporting patient safety events states that voluntary reports are subject to selection bias, capture only a fraction of events and supply no denominator of patients at risk. It notes that some institutions celebrate a rise in reports as a sign of reporting culture while others celebrate a fall as fewer events. A simple accounting identity shows why both readings can be defended.

$$R = \lambda \times q \times r \times k$$

where $\lambda$ is the number of errors made per eligible opportunity, $q$ is the proportion of those errors intercepted before they reach a patient, $r$ is the proportion of intercepted errors that are reported and counted, and $k$ is the scaling constant as before.

The identity is a framing device, not an estimate taken from any source. A rise in the observed rate can come from more errors or from better interception or fuller reporting, and voluntary reporting cannot separate these causes. The PSNet primer calls it a passive form of surveillance, in contrast to active methods such as direct observation and chart review with trigger tools, so reading a near miss rate as evidence about safety needs an independent estimate of $r$ and a parallel count of incidents that reached patients.

With a few dozen events per period, counts also vary between months when nothing has changed, as the Poisson distribution predicts. A control chart helps separate that variation from a real shift, and a before and after comparison is better summarised by an incidence rate ratio with its interval than by two raw rates.

Worked example with a barcode scanning programme on one ward

All numbers in this example are illustrative. A ward introduces barcode checking of medicines at the bedside together with a campaign encouraging staff to report near misses, and compares one quarter before with one quarter after.

1. Reported rates. Before, the near miss rate is 24 / 12,000 × 1,000 = 2.0 per 1,000 doses and incidents reaching patients run at 18 / 12,000 × 1,000 = 1.5 per 1,000. After, the figures are 45 / 12,500 × 1,000 = 3.6 and 10 / 12,500 × 1,000 = 0.8.

2. The headline change. The near miss rate has risen by a factor of 3.6 / 2.0 = 1.8, or 80%, which read alone suggests a less safe ward.

3. Reporting completeness. Direct observation of medication rounds suggests that 10% of intercepted errors were reported before and 20% after, and 15% of errors reaching patients before and 20% after.

4. Estimated true frequencies. Intercepted errors were about 2.0 / 0.10 = 20 per 1,000 doses before and 3.6 / 0.20 = 18 after. Errors reaching patients were about 1.5 / 0.15 = 10 per 1,000 before and 0.8 / 0.20 = 4 after.

5. Error and interception rates. All errors total 20 + 10 = 30 per 1,000 doses before and 18 + 4 = 22 after, a fall of about 27% because 22 / 30 = 0.733. The share intercepted rises to 18 / 22 = 0.818 from 20 / 30 = 0.667.

6. Harm averted. Errors reaching patients fell by 6 per 1,000 doses, which over 12,500 doses is 6 / 1,000 × 12,500 = 75 errors per quarter. If 10% of such errors cause harm and each harmful event adds GBP 1,500 of treatment cost, the cost avoided is 75 × 0.10 × 1,500 = GBP 11,250 per quarter. Against a programme cost of GBP 9,000 per quarter, the net saving is GBP 2,250 per quarter before any health gain is counted.

7. Interpretation. The near miss rate rose by 80% while the estimated error rate fell by about 27% and errors reaching patients fell by 60%. The economic case rests on the errors that no longer reach patients, the near miss rate enters none of the valuation steps, and the result rests on the observation estimates in step 3 and on crediting the whole before and after change to the programme.

Maternal near miss ratio in obstetric care

In maternal health the term refers to patients, not intercepted errors. The WHO near-miss approach for maternal health, published in 2011, defines a maternal near-miss case as a woman who nearly died but survived a complication that occurred during pregnancy, childbirth or within 42 days of termination of pregnancy, identified by set clinical criteria. Its main indicator is the maternal near-miss ratio.

$$\text{MNMR} = \frac{\text{MNM}}{\text{LB}} \times 1000$$

where $\text{MNMR}$ is the maternal near-miss ratio, $\text{MNM}$ is the number of maternal near-miss cases and $\text{LB}$ is the number of live births in the same facility or area and period.

WHO pairs it with the severe maternal outcome ratio, which adds maternal deaths to the numerator, and says both estimate the care and resources that would be needed in an area or facility. A higher ratio of near-miss cases to maternal deaths indicates better care, and the mortality index divides maternal deaths by all women with life-threatening conditions. As an illustration, a facility with 4,000 live births, 24 near-miss cases and 3 maternal deaths in a year has a near-miss ratio of 24 / 4,000 × 1,000 = 6.0 per 1,000 live births, a severe maternal outcome ratio of 27 / 4,000 × 1,000 = 6.75, eight near-miss cases for each death and a mortality index of 3 / 27 = 0.111, or 11.1%.

Near miss rate beside neighbouring safety measures

The near miss rate sits earlier in the chain of events than most safety measures. The harm rate counts incidents that reached patients and caused harm, and the sentinel event rate counts events on a named list of the most serious failures. An adverse event rate in a cost-effectiveness model describes harms of a treatment rather than failures of the care process, and adverse event reporting covers the systems that capture all of these events.

Near misses show how often defences caught an error, while no harm incidents and harm rates show how often errors got through. A service with a falling harm rate and a stable or rising near miss rate can be more reassuring than one where both fall together, since a joint fall may reflect less reporting of every kind.

Using near miss data in economic evaluation and performance management

The OECD review by Slawomirski, Auraaen and Klazinga estimated that 15% of hospital expenditure and activity in OECD countries can be attributed to treating safety failures. A near miss adds almost none of that cost, because no patient was reached. Its costs lie in detection, review and rework, such as a dose dispensed again, and its value in what it reveals about the defences that stopped it.

In a cost-effectiveness analysis of a safety intervention, the outcome is the harm averted, valued in health and in avoided treatment cost. A near miss rate is at most an intermediate measure whose link to harm must be estimated, deserving the scrutiny given to a surrogate endpoint. Counting a rise in near misses as a benefit would credit a programme with the effect of its own reporting campaign.

As a performance target the rate creates incentives in either direction. A target to raise reporting can produce volume without learning, and the PSNet primer advises against encouraging reports without a concrete plan for following them up, while a target to reduce near misses rewards silence. Staff time spent on review has an opportunity cost, so a reporting system is better judged by the safety changes it produces than by its number of reports.

Common misreadings of near miss rates

Most errors with near miss rates come from treating a reporting statistic as a direct measure of safety. Four mistakes recur in board reports and evaluations.

1. Reading a higher rate as a less safe service. A rise can reflect better interception or reporting, and the worked example shows a rising rate alongside falling harm.
2. Comparing services that use different definitions. The WHO and Common Formats boundary excludes events that reached the patient, and the PSNet glossary can include them.
3. Dividing by a convenient denominator. Bed-days mix processes, and counting reports instead of events inflates the numerator.
4. Pooling the maternal near-miss ratio with error rates. One counts women with life-threatening complications per 1,000 live births, the other intercepted errors per care opportunity.

Sources

• Agency for Healthcare Research and Quality. NPSD Dashboard Information. Network of Patient Safety Databases; content last reviewed September 2025. https://www.ahrq.gov/npsd/data/dashboard/info.html
• Agency for Healthcare Research and Quality, Patient Safety Network. Near Miss. PSNet glossary; 13 September 2021. https://psnet.ahrq.gov/glossary/near-miss
• Agency for Healthcare Research and Quality, Patient Safety Network. Reporting Patient Safety Events. PSNet Patient Safety Primer; last reviewed 2025. https://psnet.ahrq.gov/primer/reporting-patient-safety-events
• Barach P, Small SD. Reporting and preventing medical mishaps: lessons from non-medical near miss reporting systems. BMJ. 2000;320(7237):759-763. https://doi.org/10.1136/bmj.320.7237.759
• Slawomirski L, Auraaen A, Klazinga N. The economics of patient safety: strengthening a value-based approach to reducing patient harm at national level. OECD Health Working Papers No. 96. Paris: OECD Publishing; 2017. https://doi.org/10.1787/5a9858cd-en
• World Health Organization. Conceptual Framework for the International Classification for Patient Safety. Version 1.1, final technical report. WHO/IER/PSP/2010.2. Geneva: WHO; January 2009. https://www.who.int/publications/i/item/WHO-IER-PSP-2010.2
• World Health Organization. Evaluating the quality of care for severe pregnancy complications: the WHO near-miss approach for maternal health. Geneva: WHO; 2011. https://www.who.int/publications/i/item/9789241502221