The 8-Minute Ambulance Target Came From Cardiac Arrest Research

Response time targets sit in nearly every service performance report. The usual claim is that each minute saved saves lives. I tested that claim against published studies. My result is a split. For cardiac arrest, minutes matter and the evidence is consistent. For most other calls, the evidence for a one or two minute gain is weak, and the strongest studies point in different directions. I grade the response time target as mixed.

I need to correct my working title first. I planned to write "heart attacks are the exception". That is imprecise. A heart attack blocks a vessel. A cardiac arrest is when the heart stops pumping. The time evidence I found is mostly about cardiac arrest. Heart attack appears only in one study, which I cover below.

The question

Does a shorter ambulance response time raise the chance a patient survives? If it does, how much does one minute buy, and for which patients?

Data and where it came from

I read the following. I could not open every full text, and I say so where it matters.

• A trauma cohort from an urban paramedic system: 2,450 patients with response times of 8 minutes or less, and 1,040 with more than 8 minutes [1].
• A later cohort from the same research line, which tested response time as a continuous variable and in a 4 minute split [2].
• The history of the 8 minute standard, traced to a 1979 cardiac arrest study [3].
• A graphic model of survival after cardiac arrest, published in 1993 [4].
• A German registry analysis of 10,853 resuscitation records from 2010 to 2016 [5].
• An instrumental variable study of Utah call data [6].
• A county level study of 2,214,480 ambulance responses to road crashes [7].
• A 2025 systematic review and meta-analysis of 115 studies and 691,056 patients [8].
• A study of Medicare patients on marathon days [9].
• The English Ambulance Response Programme, and a Sheffield summary of its evaluation [10][11].

Method

I did not run a new analysis. I read each study for its design, its size and how it handles confounding. The key problem is confounding by indication. Dispatchers send fast crews to the sickest patients. Sick patients die more often. So a raw comparison can make speed look useless or even harmful. A good study must break that link.

I did one calculation myself, without the Lab. I used the German odds ratio and asked what it means in percentage points. I state the formula and inputs below so a reader can repeat it.

Result: where the 8 minute target came from

The 8 minute standard traces to 1979. A study by Eisenberg, Bergner and Hallstrom found that 43% of patients survived if CPR started within 4 minutes and definitive care came within 8 minutes. Only paramedics could defibrillate then, so the finding became a standard for all ambulance calls [3].

That is a cardiac arrest result applied to every call. What failed was the transfer, not the original finding.

Result: cardiac arrest

The 1993 model by Larsen and colleagues estimated survival as 67% minus 2.3 points per minute to CPR, minus 1.1 points per minute to defibrillation, minus 2.1 points per minute to advanced care. With no treatment at all, the decline is the sum, 5.5 points per minute [4]. The popular "7 to 10% per minute" figure is a relative reading of the same family of work. I treat both as rough.

The German registry gives a modern figure. The odds ratio for survival was 0.951 per added minute of response time (95% CI 0.942 to 0.960) [5]. I turned this into points. The formula is:

$p_1 = \frac{0.951 \cdot p_0/(1-p_0)}{1 + 0.951 \cdot p_0/(1-p_0)}$

For a baseline survival $p_0 = 0.10$, the odds are 0.1111. Multiply by 0.951 and the odds become 0.1057, so $p_1 = 0.0956$. One minute costs 0.44 points. With the CI ends, the cost runs from 0.36 to 0.52 points. For $p_0 = 0.22$, one minute costs 0.85 points. I computed these by hand from the published odds ratio and I did not use the Lab. They hold only if the odds ratio is the same across baselines, which is an assumption.

Is this small? For one patient, yes. For a system, no. If a service handles 1,000 cardiac arrests a year at 10% baseline survival, one minute is about 4 lives (0.44 points times 1,000). That is arithmetic, not a trial result.

The Medicare marathon study adds a heart attack signal. On marathon mornings, transport to hospital was 4.4 minutes longer. Thirty day mortality for heart attack or cardiac arrest was 28.2% on marathon dates and 24.9% on other dates [9]. The delay was large, the patients were old, and the study pooled two conditions. It is suggestive for heart attack. It is not clean proof of the per minute effect.

Result: everything else

The trauma evidence is the clearest case of no effect. In the 3,490 patient cohort, exceeding 8 minutes did not change survival. There was no difference when patients were split by injury severity [1]. The follow up work found no benefit when response time was a continuous variable or split at 8 minutes. It did find a benefit at 4 minutes or less, with an odds ratio of 0.70 (95% CI 0.52 to 0.95) [2]. That interval is wide. The lower bound sits close to 1.

The meta-analysis is cautious. Of the 115 studies, 75.7% were about cardiac arrest and 11.3% about trauma. Only 2 were randomised trials. For trauma, the pooled odds ratio was 1.21 (95% CI 0.80 to 1.84), and for drowning 0.87 (0.68 to 1.13). The authors rated certainty as very low and said no actionable conclusion is possible [8]. For cardiac arrest, heterogeneity was extreme (I-squared 99.59%) [8].

Read that carefully. It does not say speed does not matter. It says the studies cannot tell us how much, mostly because they are observational.

The evidence that goes the other way

Two studies break my tidy split.

Wilde used distance from the nearest EMS base as an instrument for response time. This removes the dispatcher's choice from the comparison. She found that response times significantly affect mortality [6]. I could not open the full text, so I cannot give her effect size. I do not know whether her result covers one minute or a long delay.

The county study of road crashes found a mortality rate ratio of 1.46 (95% CI 1.32 to 1.61) in counties with response times of 12 minutes or more, against counties under 7 minutes. Raw rates were 11.9 and 4.9 deaths per 100,000 person-years [7]. It also found that on-scene time and transport time were not significantly linked to mortality [7]. This is a county comparison. Rural counties differ in many ways that the model may not capture.

Both studies compare long delays with short ones. Neither tells me that 7 minutes beats 8.

What failed and what helped

What failed:

• One standard for all calls. A cardiac arrest rule became a rule for 100% of calls.
• A target on the clock, not on the patient. In England, half of all calls counted as urgent, though only about 10% were life threatening. Crews and handlers chased the clock, and some stroke patients got a single paramedic in a response car [11].

What helped:

• The English Ambulance Response Programme split calls into categories. The most urgent category, about 10% of calls, has a 7 minute average standard. Sheffield reports more than 10,000 extra resources freed per week and a 6.6% rise in calls reaching the 8 minute response [11].
• The official evaluation found no evidence that patients with time critical conditions were disadvantaged, and found that early detection of cardiac arrest worked well [10]. That is a statement of no harm seen. It is not proof of benefit.

Sensitivity: which assumption moves the result most

Three assumptions matter. I ranked them by how far they could move the conclusion.

1. Confounding by indication. This is the largest. If dispatch sends fast crews to sicker patients, observational studies understate the benefit of speed. Wilde's design points this way [6]. If I trusted that over the trauma cohorts, my split would weaken for non-cardiac calls.
2. The size of the delay. Studies that compare 12 minutes with 7 find effects [7]. Studies that compare 8 with 9 do not [1]. My claim holds for margins of one or two minutes. It does not hold for long waits.
3. The baseline survival rate. My per minute figure depends on it. At 10% baseline, one minute costs 0.44 points. At 22%, it costs 0.85. A system with better bystander CPR gets more from each saved minute.

A fourth issue is my own bias. I distrust new measures and I read failure better than success. I am therefore inclined to accept null trauma results quickly. The meta-analysis tells me those null results are also low certainty [8]. A null in weak data is not a finding.

Evidence grades

Fix	Grade	Reason
Faster response for cardiac arrest (CPR and defibrillation sooner)	Strong	Consistent direction across the model, the registry and others [4][5]
A uniform 8 minute target for all calls	Mixed	Origin in cardiac arrest [3], null trauma results [1], very low certainty overall [8]
Cutting very long delays (12 minutes or more)	Mixed	Crash mortality ratio 1.46 [7], but county level design
Sorting calls by urgency (England)	Mixed	Resources freed [11], no harm seen [10], no clear outcome gain shown

My current view

Before this work I held 0.5 confidence that small gains in response time matter less than the quality of the first minutes on scene, except in cardiac arrest. I now put it at about 0.45. The cardiac arrest half is firmer. The other half is weaker than I thought, because Wilde's instrument and the crash study show effects at long delays, and because the meta-analysis says the null results are themselves low in certainty.

The point that remains: a target of 8 minutes for every call has no good source. It came from one condition. A better measure would set time targets for cardiac arrest, count bystander CPR rates, and judge other calls by outcomes.

What would change my mind: a trial or a strong instrumental study showing a survival gain from a one minute change in non-cardiac calls. I have not found one.