John Murphy, CEO The COVID-19 Long haul Foundation
Executive finding
The available evidence supports three conclusions simultaneously:
- COVID itself caused substantial mortality, particularly among older and medically vulnerable patients.
- Early treatment practices almost certainly contributed to some deaths and complications, and management changed substantially during 2020.
- There is no credible national dataset that allows us to calculate how many U.S. deaths were caused by inappropriate ventilation rather than COVID itself.
There is, however, strong evidence that hospital mortality fell dramatically during 2020 even after adjustment for age, comorbidities and disease severity, while the use of invasive mechanical ventilation fell and use of noninvasive respiratory support, corticosteroids and remdesivir increased.
That is probably the most important finding.
1. The strongest U.S. evidence: mortality fell by roughly 40% during 2020
A large U.S. analysis of 20,736 hospitalized COVID patients examined outcomes from March through November 2020.
The observed mortality was:
| Admission period | Mortality |
|---|---|
| March–April 2020 | 19.1% |
| May–June | 11.9% |
| July–August | 11.0% |
| September–November | 10.8% |
Thus, mortality fell from 19.1% to 10.8%, a reduction of approximately 43%.
More importantly, the investigators adjusted for:
- age;
- sex;
- comorbidities; and
- disease severity.
The reduction remained statistically significant.
Compared with March–April, adjusted odds of death were:
- May–June: OR 0.66
- July–August: OR 0.58
- September–November: OR 0.59
The authors concluded that changing patient characteristics did not fully explain the decline.
That is powerful evidence that something about clinical management and/or the disease environment changed.
2. Respiratory management changed dramatically
The CDC’s COVID-NET surveillance system provides particularly useful evidence.
Among hospitalized U.S. adults:
ICU admission
March 2020: 37.8%
December 2020: 20.5%
Mechanical ventilation
March: 27.8%
December: 12.3%
Vasopressors
March: 22.7%
December: 12.8%
At the same time, use of:
remdesivir: 1.7% → 53.8%
systemic corticosteroids: 18.9% → 74.2%
And noninvasive respiratory support increased.
This is an extremely important observation.
The medical profession did not simply stop treating COVID aggressively.
Rather, the treatment algorithm changed:
less invasive mechanical ventilation
and
more noninvasive respiratory support + steroids + antiviral treatment + improved critical-care management.
3. Was early intubation actually harmful?
Here the scientific literature is genuinely conflicting.
One large U.S. study examined 7,597 COVID patients at 11 New York City municipal hospitals between March and December 2020.
After propensity-score matching, the investigators found:
intubation within 48 hours → HR 1.30 for mortality
and
intubation at any time → HR 1.62
compared with patients who were not intubated.
At first glance, that appears to support your hypothesis.
But there is a major limitation:
Confounding by indication
Patients who required intubation were substantially sicker.
Even sophisticated statistical adjustment cannot completely eliminate this problem.
Therefore:
Higher mortality among ventilated patients does not prove that ventilation caused the deaths.
It demonstrates an association.
4. Another major U.S. study produced the opposite result
Researchers at Mount Sinai examined 755 mechanically ventilated patients admitted between January 30 and April 30, 2020.
Their median age was 65.
Among these ventilated patients:
68% died
by the end of follow-up.
But the study found something striking:
Each additional day between admission and intubation was associated with a 3% increase in the hazard of death.
Adjusted HR:
1.03 per additional day
The authors therefore concluded that earlier rather than later intubation might improve survival in appropriately selected patients.
That is directly contrary to the simplistic claim that early intubation was inherently harmful.
5. A much larger U.S. multicenter study again favored earlier ventilation
A study involving 68 U.S. hospitals examined 1,879 critically ill adults admitted between March and July 2020.
They compared:
early invasive mechanical ventilation: ICU days 1–2
with
late ventilation: ICU days 3–7.
The raw mortality was:
Early: 42.2%
Late: 51.0%
After adjustment:
HR = 0.77
95% CI:
0.65–0.93
That means the adjusted mortality hazard was approximately 23% lower in the early-intubation group.
Again, this does not prove that early intubation is universally superior. But it directly contradicts the proposition that early ventilation was demonstrably responsible for the majority of early deaths.
6. The 2020 literature therefore does NOT support one simple ventilation narrative
Putting the major U.S. studies together:
| Study | Population | Finding concerning timing |
|---|---|---|
| NYC municipal hospitals | 7,597 | Early intubation associated with higher mortality |
| Mount Sinai | 755 ventilated | Later intubation associated with higher mortality |
| 68-hospital U.S. cohort | 1,879 ICU patients | Later ventilation associated with higher mortality |
| New York Northwell | 1,966 ventilated | 61% 28-day mortality during first wave |
| Atlanta | 217 ICU patients | 35.7% mortality among ventilated patients |
The apparent contradiction is largely explained by differences in:
- patient severity;
- timing definitions;
- hospital capacity;
- oxygenation criteria;
- treatment protocols;
- selection of patients for intubation;
- statistical adjustment;
- and changing clinical practice.
7. The New York experience demonstrates just how sick the ventilated population was
The Northwell Health study examined 1,966 mechanically ventilated patients during March and April 2020.
1,198 died within 28 days — 61%.
That is an extraordinarily high mortality.
But this was the period of the greatest medical uncertainty and one of the largest hospital surges in the United States.
The study also compared traditional ICUs with expanded ICUs, created because conventional ICU capacity had been overwhelmed.
This is important because health-system strain itself can influence mortality.
A patient treated by an experienced ICU team in a conventional ICU is not necessarily equivalent to an equally sick patient treated during a catastrophic surge in an improvised ICU.
8. The 5,700-patient New York study must also be interpreted cautiously
The famous Mount Sinai/New York Presbyterian case series involved 5,700 hospitalized patients.
At the time of analysis:
- 12.2% received invasive mechanical ventilation;
- 21% of patients with known outcomes had died.
Among patients requiring mechanical ventilation, the early reported mortality appeared extraordinarily high.
But there was an enormous statistical problem:
many ventilated patients were still hospitalized.
The April 2020 report showed:
- 3.3% discharged alive;
- 24.5% dead;
- 72.2% still hospitalized.
Therefore, the early apparent mortality among ventilated patients substantially underestimated eventual survivors and could not be interpreted as a final case-fatality rate.
This is an important example of why some of the frightening early ventilator statistics were misleading.
9. Mechanical ventilation itself can injure lungs
There is nevertheless a legitimate physiological basis for your concern.
Mechanical ventilation can cause:
- volutrauma;
- barotrauma;
- atelectrauma;
- oxygen toxicity;
- ventilator-associated pneumonia;
- hemodynamic impairment;
- sedation complications;
- diaphragmatic dysfunction.
The appropriate strategy for COVID-associated ARDS consequently became lung-protective ventilation, using relatively low tidal volumes and limiting plateau pressures.
Clinical recommendations subsequently emphasized low-tidal-volume ventilation, appropriate PEEP, avoidance of routine recruitment maneuvers and prone positioning for appropriate patients.
So the proposition:
“Improper ventilation can contribute to death”
is medically well established.
The proposition:
“Most early COVID deaths were caused by improper ventilation”
is not established.
Those are very different claims.
10. One of the clearest signs of improved treatment: corticosteroids
The RECOVERY trial provides unusually strong evidence because it was randomized.
Among hospitalized patients receiving invasive mechanical ventilation:
Dexamethasone
29.3% mortality
versus
Usual care
41.4% mortality
Rate ratio:
0.64
95% CI:
0.51–0.81
That’s an approximately 36% relative reduction in mortality among mechanically ventilated patients.
The absolute difference was approximately:
12.1 percentage points.
That is a very large treatment effect.
Therefore, one explanation for the later reduction in COVID mortality is straightforward:
physicians learned how to treat the inflammatory phase of severe COVID more effectively.
11. The timing of dexamethasone matters
Dexamethasone was not beneficial at every stage of COVID.
Among patients receiving no respiratory support:
17.8% died with dexamethasone
versus
14.0% with usual care.
The difference was not statistically significant and trended in the wrong direction.
This is an important lesson.
The treatment that saved lives in severe COVID was not simply “give everyone steroids.”
It was:
give appropriate anti-inflammatory treatment to patients with clinically significant oxygen requirements.
That is precisely the type of knowledge that was unavailable in March 2020.
12. The overall U.S. trajectory strongly supports a treatment-learning effect
A much larger study encompassing:
104,590 hospitalized patients
across
21 U.S. health systems
examined the first 20 months of the pandemic.
Risk-standardized mortality fell from:
16.4% in February–April 2020
to
9.0% in July–September 2021.
Unadjusted mortality fell from:
18.6%
to
7.3%.
That is approximately a:
45% reduction in risk-standardized mortality.
This occurred despite subsequent waves containing enormous numbers of hospitalized patients.
13. What happened to ventilation during this period?
The same general pattern appears repeatedly:
early pandemic → more invasive ventilation
later pandemic → less invasive ventilation
while:
noninvasive oxygenation increased.
The CDC’s national COVID-NET data provide the clearest numerical evidence:
| Treatment indicator | March 2020 | December 2020 |
|---|---|---|
| ICU admission | 37.8% | 20.5% |
| Mechanical ventilation | 27.8% | 12.3% |
| Vasopressors | 22.7% | 12.8% |
| Remdesivir | 1.7% | 53.8% |
| Corticosteroids | 18.9% | 74.2% |
This is probably the strongest evidence supporting your observation that medical management changed substantially after the first wave.
But it doesn’t establish that the ventilators were the principal cause of death.
14. Could inappropriate ventilation explain the entire mortality decline?
No.
The magnitude of the decline is too large and the number of simultaneous changes too great.
The later reduction in mortality likely reflects a combination of:
Treatment
- dexamethasone;
- improved respiratory management;
- anticoagulation strategies;
- remdesivir;
- monoclonal antibodies;
- improved ICU management;
- prone positioning;
- better management of bacterial complications.
Medical knowledge
- better recognition of COVID-associated ARDS;
- better patient selection for intubation;
- better ventilator settings;
- better recognition of thrombosis;
- better fluid management.
Population factors
- prior infection;
- vaccination;
- changing age distribution;
- changing variants.
Health-system factors
- increased ICU experience;
- less extreme capacity constraints;
- better staffing;
- improved protocols.
It would be scientifically incorrect to assign the entire improvement to any one factor.
15. A particularly important finding: mortality declined even after statistical adjustment
This is where the evidence becomes difficult to dismiss.
The 20,736-patient U.S. analysis adjusted for:
- age;
- sex;
- comorbidities;
- disease severity.
Yet mortality still declined from 19.1% to 10.8%.
The investigators specifically concluded that changes in patient characteristics did not fully account for the improvement.
That strongly supports the existence of a period effect—something about the treatment environment and/or the disease itself changed.
16. But the data cannot tell us “X deaths were caused by ventilators”
This is the critical limitation.
There is no national database containing:
COVID death = 1
Ventilator caused death = 0/1
Nor does the NCHS mortality database contain ventilator settings.
Consequently, a calculation such as:
400,000 COVID deaths × 20% = 80,000 ventilator deaths
would be completely speculative.
It would not be an analysis.
It would be an assumption.
17. What we can estimate with reasonable confidence
The evidence supports several levels of certainty.
Very strong evidence
COVID itself caused substantial mortality.
Very strong evidence
Hospital mortality fell substantially during 2020.
Very strong evidence
Clinical management changed substantially during the same period.
Very strong evidence
Use of mechanical ventilation declined.
Very strong evidence
Use of corticosteroids increased dramatically.
Randomized evidence
Dexamethasone reduced mortality among patients requiring oxygen or mechanical ventilation.
Strong physiological evidence
Mechanical ventilation can cause lung injury when improperly applied.
Observational evidence
Timing of intubation was associated with mortality, but studies disagree about whether earlier or later intubation was preferable.
Insufficient evidence
The number of U.S. deaths specifically caused by inappropriate mechanical ventilation.
No credible evidence
That the majority of 2020 COVID deaths were actually caused by ventilators rather than SARS-CoV-2 disease.
18. A more interesting hypothesis emerges
The evidence suggests that the most productive question is not:
“Did ventilators kill COVID patients?”
It is:
“How much of the extraordinary early COVID mortality was preventable with the medical knowledge that subsequently became available?”
That question is answerable.
It can be approached by comparing:
March–April 2020
with
May–December 2020
while controlling for:
- age;
- sex;
- obesity;
- diabetes;
- renal disease;
- cardiovascular disease;
- baseline oxygenation;
- respiratory rate;
- SOFA score;
- D-dimer;
- CRP;
- ICU admission;
- mechanical ventilation;
- vasopressors;
- renal replacement;
- steroid treatment;
- remdesivir;
- secondary infection;
- hospital;
- and ICU capacity.
The Atlanta study, for example, incorporated many of these variables and still found substantial variation in mortality across successive surges.