9

Accompanying a video on MSN News today (17th jan 023) is the following statement :

Pregnant women who catch Covid are SEVEN times more likely to die, warns new research. They are also at a "significantly elevated" risk of being admitted to intensive care and 23 times more likely to develop pneumonia, according to the study published in BMJ Global Health. In addition, the findings suggest that Covid during pregnancy also increases the risk that the baby will need to be admitted to intensive care. Scientists say their findings add urgency to calls for more women of childbearing age to get vaccinated Lead author Professor Emily Smith, of the George Washington University Milken Institute School of Public Health in the US said: “This study provides the most comprehensive evidence to date suggesting that Covid-19 is a threat during pregnancy.

The only recent BMJ Global Health study I can find on the subject does not make any such claim (that I can find) and states :

Taken together, this analysis of 12 studies including 13 136 pregnant women from 12 countries indicates that SARS-CoV-2 infection at any time during pregnancy increases the risk of maternal mortality, severe maternal morbidities and adverse newborn outcomes. These findings underscore the need for global efforts to prevent COVID-19 during pregnancy through targeted administration of vaccines and non-pharmaceutical interventions.

Is this 'seven times' morbidity claim true ?

Nigel J
  • 756
  • 4
  • 11
  • 5
    The abstract of the paper you quote mentions a relative risk of 7.68 for mortality in the abstract, which fits to the claim in the MSN article. – Mad Scientist Jan 17 '23 at 13:55
  • @MadScientist Thank you. Will you answer that and I accept ? Or shall I just delete my question ? – Nigel J Jan 17 '23 at 14:25
  • 5
    @NigelJ, you could also post that as an answer to your own question. (I don't think you should delete your question.) – Ben Bolker Jan 17 '23 at 14:26
  • 3
    7 times more likely than whom? 7 times more likely than pregnant women who do not get COVID? 7 times more likely than non-pregnant women who get COVID? 7 times more likely than dancing bears? – shoover Jan 18 '23 at 03:55
  • 2
    @shoover as compared with uninfected pregnant women, conclusion being SARS-CoV-2 infection at any time during pregnancy increases the risk of maternal death by 7.68 – Timmetje Jan 18 '23 at 09:58
  • How does this compare with the mortality risk increase for the overall population (or at least 18-40yo non-pregnant women)? If the relative mortality increase is about the same, why are pregnant women singled out? – Cristobol Polychronopolis Jan 19 '23 at 21:16
  • 1
    @CristobolPolychronopolis The mortality of pregnant women in the UK has risen since 2020 by about 2.0%. This is due to Covid-19, as the above study proves. – Nigel J Jan 19 '23 at 21:30
  • @NigelJ Okay. How much has the mortality of non-pregnant women risen? The specificity of the claim implies that they're looking at pregnancy as a risk factor, but without data on a control group under the same analysis, it doesn't mean anything in that respect. – Cristobol Polychronopolis Jan 19 '23 at 21:37
  • 1
    @CristobolPolychronopolis Excess deaths in the general population have risen by about the same amount. – Nigel J Jan 19 '23 at 21:42
  • 1
    I found some statistics that depending on race, in the USA the pregnancy related number of deaths is between 11 and 41 per 100,000 of women per live birth. That's deaths during and up to one year after birth, caused by pregnancy (not pregnant women run over by a bus). Times 7 is 77 to 287 per hundred thousand, but then we'd have to subtract the original numbers, so 66 to 246 per 100,000. That's 0.066% to 0.246%. Pregnant women are usually in an age group with fewer Covid deaths, so this is slightly high. Maybe someone knows if pregnancy increases the death risk when you have Covid. – gnasher729 Jan 20 '23 at 18:11

2 Answers2

12

An estimate of being roughly seven times more likely to die is a reasonable summary of the information in the paper you linked yourself. The sentence in the abstract is

Pregnant women with SARS-CoV-2 infection—as compared with uninfected pregnant women—were at significantly increased risk of maternal mortality (10 studies; n=1490; RR 7.68, 95% CI 1.70 to 34.61)

The RR 7.68 means that the relative risk is 7.68 higher for pregnant women with a SARS-CoV-2 infection compared to pregnant women without infection. The studies contained a total of n=1490 pregnant women (I'm not sure whether this is the number of women with infection or the total). The last bit means a 95% CI convidence interval for the estimated relative risk is somewhere between 1.70 and 34.61. Hence there is fairly high convidence that the risk is increased but by how much the risk is increased is very uncertain.

Another useful number for context is the general maternal mortality during child birth which in the US is around 20 per 100.000 birth (source). This number of deaths is estimated to be increased by a factor of around 7 due to a SARS-CoV-2 infection.

Edit: JonathanReez claims that the answer should be inconclusive so a here a few more words on what the study claims (and is correct in claiming) and what it doesn't.

First one can compute backwards that with a base rate of maternal mortality of around 20 per 100.000 and an RR of 7.68 comes out to exactly two deaths in their sample of n=1490 (equivalent to an observed maternal mortality of around 150 per 100.000 in the sample).

This turns this into a classical probability problem. The null hypothesis is that there is no effect due to corona. So how likely is it to observe 2 deaths in 1490 women if the maternal death rate is 20 per 100.000? It turns out this probability is smaller than 5% (but bigger than 1%). Hence one can conclude with 95% confidence that women with a corona infection suffer a higher maternal mortality than women without (which is what the study says). We couldn't make the same conclusion with 99% confidence.

Trying to estimate how much bigger the maternal mortality rate actually is is very difficult due to the limited data. The 2 observed deaths lead to an estimate of 7.68 higher than normal but whether the actual increase is more around twice as big or more around 30 times as big cannot be concluded from the data (the study computes this as a 95% confidence that the effect is in the interval [1.70, 34.61]).

Additionally the claim is only saying that women with a corona infection have higher maternal mortality. It does not claim that the higher mortality is due to the corona infection. This could be the case but it could also be the case that there is a confounding factor that increases both the risk of a corona infection and maternal mortality (for example race or poverty). This question cannot easily be answered with the limited amount of data and the study doesn't make any claims in this direction.

quarague
  • 1,474
  • 9
  • 13
  • I'm trying to think of a way that your first sentence could be improved to include some mention that the "**SEVEN times**" is only the preliminary number. I feel like people will fixate on that 7x, and if future research shows it as higher or lower people will panic or dismiss the research respectively. Also, does the research do anything to establish cause and effect? I wouldn't be surprised to hear that the increased pregnancy risk is because of risky behaviors that also led them to catch Covid. – Rob Watts Jan 18 '23 at 22:32
  • 1
    @RobWatts I changed the first sentence. Afaik the study only talks about correlation. I think the number of deaths is (fortunately) way too small to take any more detailed conclusions like trying to look into causes. See also the huge convidence interval on how big the effect actually is. – quarague Jan 19 '23 at 06:39
  • "around 20 per 100'000 birth. " This number is on par with Saudi Arabia, Iran and Russia ... with a much more expensive health system https://data.worldbank.org/indicator/SH.XPD.CHEX.GD.ZS?locations=US-SA-RU-IR – EarlGrey Jan 19 '23 at 14:11
  • Perhaps it's just me, but I read the claim in that text as "pregnant women are seven times more likely to die of covid than non-pregnant women who get covid" rather than "pregnant women with covid are 7x more likely to die than pregnant women who don't get covid". – Jack Aidley Jan 20 '23 at 11:08
  • @JackAidley The paper is about the sitation that given a pregnancy how do the risks change when comparing covid infection to no infection. The MSN article could have made that more clear. – quarague Jan 20 '23 at 13:57
  • @quarague: Yes, sorry I was referring to the claim in the MSN article, the paper is clear. – Jack Aidley Jan 20 '23 at 13:59
  • 1
    @EarlGrey I found numbers by race, with worst = 41 per 100,000 for black women, and best = 11 per 100,000 for hispanic women. – gnasher729 Jan 20 '23 at 18:13
  • @JonathanReez Starting with 20 per 100.000 as a base rate, the study found a 7-fold increase so around 150 per 100.000 which corresponds to around 2 for their study size of n=1490. So my guess would be 2 deaths actually happened. Fits with the huge range of estimates for the size of the effect. – quarague Jan 21 '23 at 05:55
  • 1
    @JonathanReez a) Presumably there are not that many more women in the US that were pregnant and had a covid infection and b) The study did their math correctly. With n=1490 and 2 deaths you can exclude the null hypothesis of no effect due to Covid with 95% confidence (but not with say 99% confidence) but the estimated range of the size of the effect is huge (somewhere between 1.7 to 35 times worse with 95% confidence). As this is a meta study I suppose that is all the data there is. – quarague Jan 21 '23 at 10:18
  • @JonathanReez No it's not a joke. If you have an event that you expect to see only in around 1 in 5000 cases and you see 2 in 1500 that suffices to conclude with 95% confidence that your expection is wrong. You can't be 100% certain or even 99% but they didn't claim that. If you expect an event to be extremely rare then observing it twice can be enough to conclude that your expectation is likely wrong. That is why studies use maths and statistics and not intuition which can fail badly in such situations. – quarague Jan 21 '23 at 10:57
  • The paper you linked has the raw maternal mortality numbers in a table and it's 38/368 for COVID positive and 3/1122 for COVID negative controls. I don't know why you're all speculating that there were two observed deaths. The edit based on misinterpreting the abstract makes this answer inaccurate and significantly worse. – CJR Jan 21 '23 at 13:01
  • Thanks for the edit, I've retracted my previous comments. The one missing thing you could consider mentioning is selection bias - we cannot be sure if the women with a corona infection were actually picked at random, especially since people with more severe symptoms were also more likely to get tested. – JonathanReez Jan 21 '23 at 14:14
  • Thanks to @CJR I now understand that as per Table 3 - Maternal Deaths is 38/368 and 3/1122 for COVID and non COVID cohorts respectively. Then it's even more of a head scratcher as they claim a whooping 10% mortality rate from COVID? – JonathanReez Jan 21 '23 at 16:28
  • 1
    Was the risk for pregnant women higher than for non-pregnant women of the same age? Every activity has a risk X, and Covid has a risk Y, and you divide Y by X and say “if you do X then Covid increases your risk by Y/X + 1”. So it could be that if you are pregnant, you should get vaccinated. If you like red wine, you should be vaccinated. If you eat vegan or no vegetables at all. You should get vaccinated. If you usually drive drunk at excessive speed, no need to get vaccinated. – gnasher729 Jan 21 '23 at 17:49
  • @gnasher729 Excellent point! The baseline risk for maternal mortality is 1 in 5000. The baseline risk for a person for the original strain of COVID for a person aged 25-40 is ~1 in 1000. So this alone should be responsible for a 5x increase in risks. – JonathanReez Jan 22 '23 at 02:22
1

The CDC has data on all-cause maternal mortality from 2018, 2019 and 2020:

Year Number of live births Number of maternal deaths Maternal mortality rate
2018 3,791,712 658 17.4
2019 3,747,540 754 20.1
2020 3,613,647 861 23.8

A maternal death is defined by the World Health Organization as, “the death of a woman while pregnant or within 42 days of termination of pregnancy, irrespective of the duration and the site of the pregnancy, from any cause related to or aggravated by the pregnancy or its management, but not from accidental or incidental causes”

As we can see from the data, only 0.02% of pregnancies end with a maternal death, so this is a very rare event to begin with. We also see that 2020 wasn't exceptionally bad, as the increase in mortality from 2018 to 2019 (20.1/17.4 = 15%) was similar to the increase in 2020 (23.8 / 20.1 = 18%). COVID might've have had some impact on mortality during pregnancy but it's not easily visible in mortality statistics.

The other answer links to studies that attempt to quantify the relative risks of having a COVID infection but their cohorts are far too small to meaningfully study an event with a 0.02% probability. So I'd say the answer to this question is inconclusive, as we simply don't have enough data to provide a meaningful answer. Compare this to, say, all-cause mortality statistics for people above the age of 65 where there's a clear increased during the pandemic and little argument over the severity of the virus.

JonathanReez
  • 10,917
  • 11
  • 51
  • 102
  • 2
    If you account for the fact that other industrialized countries have a maternal mortality of around 5, other countries much poorer than the US manage values around 20 and these numbers have been decreasing globally for decades, a 15% increase **is exceptionally bad**, even more so if it happens 2 years in a row. Somewhere the US health care system is failing badly in its maternal care and whatever the reason is, it seems to be getting worse. This is probably unrelated to corona though. – quarague Jan 21 '23 at 11:57
  • 2
    The pre-vaccine COVID infection fatality rate in the childbearing age cohort was approximately 1/1000. 7x higher risk would be 0.7%. If the COVID IFR for pregnant women was 0.2/1000, pregnancy would be protective against COVID. Using the base maternal mortality rate is incorrect as the comparison is COVID deaths in pregnant vs not pregnant. The speculative math in this answer is therefore entirely incorrect. – CJR Jan 21 '23 at 14:16
  • @CJR there is no math in the answer, except showing that the maternal mortality rate increased by 18% in 2020 compared to 2019. What the answer does show is that there wasn't a *massive* increase in maternal fatality during 2020 despite a large number of infections. And 0.2/1000 is the *baseline* mortality rate - COVID would definitely increase it somewhat but we don't have numbers to prove by how much. – JonathanReez Jan 21 '23 at 14:20
  • 2
    You don't have any idea what the number of infections is, COVID infections in 2020 were between 5-10% of the population and we were protecting vulnerable populations with NPIs. There is a reason that you need to do control-matched studies for this and that you can't extrapolate off the base mortality rate. – CJR Jan 21 '23 at 15:37
  • @CJR yes, a control-matched study would be ideal, however no such studies have been performed with a sufficient number of participants, so we don't have a good answer to this question. The other answer talks about a study that extrapolated the mortality rate based on a grand total of 2 deaths. So the real answer is: *inconclusive*. – JonathanReez Jan 21 '23 at 15:50
  • 1
    The other study had 38 deaths in the COVID cohort and 3 deaths in the non COVID cohort. You again misinterpreted the abstract because of your misunderstanding of the proper base rate. – CJR Jan 21 '23 at 15:52
  • @CJR where did you see the 38/3 deaths in that paper? If you provide an exact quote I'll delete my answer - quarague claims that there have only been 2 deaths during the study. – JonathanReez Jan 21 '23 at 15:55
  • 1
    Table 3 - Maternal Deaths is 38/368 and 3/1122 for COVID and non COVID cohorts respectively (a better question would be what is the quality of the underlying studies and if they're stratified but that reanalysis is out of scope here IMO) – CJR Jan 21 '23 at 15:56
  • @CJR Okay, then looks like this metastudy is even worse than I've thought, as the people in the cohort had an insane 10% mortality rate from the virus vs. a 0.1% mortality rate for women under the age of 40. Garbage in, garbage out. – JonathanReez Jan 21 '23 at 16:29
  • 1
    This study calculated relative risk at 7x so it's most likely theres one or two very poor studies with denominator problems. But that's what meta-analysis is for, aggregating studies to get better estimates. There may have been additional exclusion criteria that would be useful but that's out of scope here IMO. – CJR Jan 21 '23 at 16:38
  • @CJR questioning the quality of published studies is surely not out of scope here. We don’t take anything at face value :-) – JonathanReez Jan 21 '23 at 16:43
  • 1
    It's not a "this study is garbage" situation though, it's a "I would probably have made a few relatively minor changes to the methods" situation (I'd have probably excluded on collection date too), and the maternal mortality estimate is a very small part of the result anyway. It's more of a stats opinion and would fit better there. The core question here is supported by this published result. – CJR Jan 21 '23 at 16:46
  • @CJR I’d argue that if the 7x claim is about the *general* population then the meta study doesn’t have any data because it clearly didn’t include studies that sampled people at random, given the high mortality rate. If it’s about women with pre existing conditions then yes, it can be valid data, though we need to dig deeper into what those conditions were. Statistics is only useful once you have a good understanding of the underlying data – JonathanReez Jan 21 '23 at 16:49
  • 1
    That's an argument with no bottom - we can't do studies at massive scale to answer every question, and we certainly can't do RCTs for pregnancy. Statistics is only useful when you have an imperfect understanding of the underlying data, otherwise it's wildly inferior to direct observation. Again, why I think this is a stats question. (This paper also does a good job showing the uncertainty around that 7x estimate IMO) – CJR Jan 21 '23 at 16:55
  • @CJR yes that’s why the answer to some questions is… *unknown*. – JonathanReez Jan 21 '23 at 17:10