Assessing the age specificity of infection fatality rates for COVID-19: systematic review, meta-analysis, and public policy implications

Determine age-specific infection fatality rates for COVID-19 to inform public health policies and communications that help protect vulnerable age groups. Studies of COVID-19 prevalence were collected by conducting an online search of published articles, preprints, and government reports that were publicly disseminated prior to 18 September 2020. The systematic review encompassed 113 studies, of which 27 studies (covering 34 geographical locations) satisfied the inclusion criteria and were included in the meta-analysis. Age-specific IFRs were computed using the prevalence data in conjunction with reported fatalities 4 weeks after the midpoint date of the study, reflecting typical lags in fatalities and reporting. Meta-regression procedures in Stata were used to analyze the infection fatality rate (IFR) by age. Our analysis finds a exponential relationship between age and IFR for COVID-19. The estimated age-specific IFR is very low for children and younger adults (e.g., 0.002% at age 10 and 0.01% at age 25) but increases progressively to 0.4% at age 55, 1.4% at age 65, 4.6% at age 75, and 15% at age 85. Moreover, our results indicate that about 90% of the variation in population IFR across geographical locations reflects differences in the age composition of the population and the extent to which relatively vulnerable age groups were exposed to the virus. These results indicate that COVID-19 is hazardous not only for the elderly but also for middle-aged adults, for whom the infection fatality rate is two orders of magnitude greater than the annualized risk of a fatal automobile accident and far more dangerous than seasonal influenza. Moreover, the overall IFR for COVID-19 should not be viewed as a fixed parameter but as intrinsically linked to the age-specific pattern of infections. Consequently, public health measures to mitigate infections in older adults could substantially decrease total deaths. Electronic supplementary material The online version of this article (10.1007/s10654-020-00698-1) contains supplementary material, which is available to authorized users.


Supplementary Appendix B: Meta-analysis search procedure
To perform the present meta-analysis, we collected published papers and preprints regarding the seroprevalence and/or infection fatality rate of COVID-19. To identify these studies, we systematically performed online searches in MedRxiv, Medline, PubMed, and Google Scholar using the criterion (("infection fatality rate" or "IFR" or "seroprevalence" or "antibodies") and ("COVID- 19" or "SARS-Cov-2")). We also used a search tool created by the University of Zurich for searching EMBASE using the same search criterion. [2] We identified other studies listed in reports by government institutions such as the U.K. Parliament Office. [3] Finally, we confirmed the coverage of our search by referring to two recent meta-analysis studies of the overall IFR for COVID-19, a recent meta-analysis of the ratio of measured seroprevalence to reported cases, and the SeroTracker global dashboard of SARS-CoV-2 seroprevalence studies. [4][5][6][7] Our search encompassed studies that were publicly disseminated prior to 25 September 2020. For cases in which a study was identified by the aforementioned search but age-specific seroprevalence was not found, an expanded search was performed to obtain those details using additional keywords (e.g., the location of the study). Data was extracted from studies by three authors and verified prior to inclusion.
If the sensitivity and specificity of the test are assumed to be known with certainty, then the test-adjusted prevalence can be computed using the Gladen-Rogan formula [9] as follows: Likewise, the test-adjusted confidence interval can be computed by multiplying the raw confidence interval by the factor 1/(sensitivity + specificity -1).
Nonetheless, the sensitivity and specificity of COVID-19 antibody test kits should generally not be treated as parameters known with certainty. Indeed, the U.S. Food and Drug Administration reports 95% confidence intervals for each of these test properties in its information sheet for all EUA test kits. [21] In particular, following the approach of Gelman and Carpenter (2020), we use a Bayesian procedure to compute seroprevalence estimates and confidence intervals that incorporate uncertainty about the sensitivity and specificity of the test kit used in each of these seroprevalence studies. [22,23] As shown in Table C1, the contrast between the Gladen-Rogan vs. Bayesian results are most striking for study observations obtained using a relatively small sample size in a context of low prevalence. In some cases, the Bayesian confidence interval indicates that the level of seroprevalence cannot be distinguished from zero, and hence those observations are excluded from our metaregression (see Appendix X).   Zealand  1·1  1862  Australia  6·7  1054  South Korea  10·8  576  Lithuania  1·4  415  Iceland  1·8  321  Slovakia  1·4  194  Latvia  0·8  191  Austria  15·4  115  Slovenia  1·4  112  Czech Republic  7·6  104  Greece  2·6  95  Denmark  9·0  94  Estonia  1·7  70  Luxembourg  3·8  68  Israel  15·8  57  Norway  7·7  47  Poland  14·0  37  Hungary  2·8  36  Portugal  24·7  35  Belgium  49·9  32  Germany  159·1  31  Finland  4·9  31  Switzerland  29·3  27  Spain  215·2  27  Japan  14·1  25  Italy  203·6  19  Colombia  6·2  16  Canada  51·6  15  Ireland  20·3  13  Turkey  117·6  13  Chile  14·9  12 Note: Age-specific fatality data for Lithuania was published as of 01 June 2020, at which point there was a total of 70 reported fatalities; thus, the six subsequent fatalities through 22 June 2020 were assumed to have the same age distribution as the fatalities through 01 June 2020.

Supplementary Appendix G: Metaregression methodology
To analyze IFR by age, we use meta-regression with random effects, using the meta regress procedure in Stata v16. [64,65] We used a random-effects procedures to allow for residual heterogeneity between studies and across age groups by assuming that these divergences are drawn from a Gaussian distribution. The procedure provides reasonable results even if the errors are not strictly normal but may be unsatisfactory if the sample includes large outliers or the distribution of groups is not unimodal. In analytical terms, this framework can be expressed as follows: where ~ (0, 2 ) and ~ �0, 2 � In this specification, is the estimated IFR in study i for age group j, denotes the median age of that group, denotes the source of idiosyncratic variations for that particular location and age group, and denotes the random effects that characterize any systematic deviations in outcomes across locations and age groups. Under the maintained assumption that each idiosyncratic term has a normal distribution, the idiosyncratic variance is 2 = (( − )/3.96) 2 , where and denote the upper and lower bounds of the 95% confidence interval for that study-age group. The random effects are assumed to be drawn from a homogeneous distribution with zero mean and variance 2 . The null hypothesis of 2 = 0 characterizes the case in which there are no systematic deviations across studies or age groups. If that null hypothesis is rejected, then the estimated value of 2 encapsulates the magnitude of those systematic deviations.
Under our baseline specification, the infection fatality rate increases exponentially with age-a pattern that has been evident in prior studies of age-specific case fatality rates.[66, 67] Consequently, our meta-regression is specified in logarithmic terms, with the slope coefficient encapsulating the impact of higher age on log(IFR). Consequently, the null hypothesis that IFR is unrelated to age can be evaluated by testing whether the value of is significantly different from zero. If that null hypothesis is rejected, then the estimated values of and characterize the estimated relationship between log(IFR) and age. Consequently, the predicted relationship between IFR and age can be expressed as follows: The 95% confidence interval for this prediction can obtained using the delta method. In particular, let denote the infection fatality rate for age a, and let denote the standard error of the meta-regression estimate of log( ). If has a non-zero value, then the delta method indicates that its standard error equals / , and this standard error is used to construct the confidence interval for at each age a. Likewise, the prediction interval for log( ) is computed using a standard error of + that incorporates the systematic variation in the random effects across studies and age groups, and hence the corresponding prediction interval for is computed using a standard error of ( + )/ .
In estimating this metaregression, we exclude observations for which the lower bound of the 95% confidence interval for seroprevalence of that particular age group equals zero, and hence the upper bound of that age-specific IFR is not well defined. Similarly, we exclude observations for which the lower bound of the 95% confidence interval for seroprevalence is less than the observed COVID-19 mortality rate for that age group, since such observations would imply an upper bound for the IFR that exceeds 100%. Finally, we exclude observations for which no COVID-19 fatalities were recorded for a given age group and hence the implied value of the infection fatality rate is at its lower bound of zero and the corresponding confidence interval cannot be precisely determined. On July 30, Alberta's chief medical officer announced the results of a serology study of 9400 blood specimens collected for other purposes and indicated that "less than one percent" were positive.

S15
No further details were available as of September 17, 2020.
Ariano Irpino, Italy [69] This seroprevalence study collected specimens in late May from 13444 individuals (about 75% of municipality residents) and found a raw prevalence of 4·83%.
No age-specific results were reported.
Australia [70] This study analyzed specimens from 2991 individuals undergoing elective surgery during May-June 2020 and found test-adjusted seroprevalence of 0·28% (CI: 0-0·71%). No age-specific seroprevalence results were reported. Although the sample of patients may not be fully representative of the Australian population, these results confirm that prevalence in Australia was indistinguishable from zero at the time of the study, as expected for a comprehensive tracing program (see Supplementary Appendix C).
Bad Feilnbach, Germany [72] This study collected specimens from a random sample of 2153 adults between 23 June and 4 July 2020 and found raw seroprevalence of 6·0%. No age-specific results were reported as of September 17.
Baton Rouge, Louisiana, USA [73] This study analyzed specimens from a random sample of 2138 individuals between July 15-31 and found seroprevalence of 3·6%. No age-specific seroprevalence results were reported.
Blaine County, Idaho, USA [74] This study collected specimens from 972 individuals on May 4-19 and found an IgG prevalence of 22·7% (CI: 20-25·5%). The authors concluded that "the small number of county deaths (n=5) makes estimating the infection fatality rate unreliable." No age-specific fatality data is publicly available for this county.
Bonn, Germany [76] This study analyzed specimens from 4771 individuals participating in the Rhineland Study, an ongoing community-based prospective cohort study of Bonn residents ages 30 years and older. Specimens were collected between 24 April and 30 June 2020 and analyzed using a combination of ELISA, RIA, and PRNT methods. The ELISA assay found 46 positive results and indicated seroprevalence of 0·97% (CI 0·72-1·30%). By contrast, the PRNT assay found only 17 positive results and indicated a markedly lower seroprevalence of 0·36% (CI: 0·21-0·61%). No age-specific seroprevalence results were reported.
British Columbia, Canada [77] This study analyzed 885 laboratory specimens from outpatient clinics for the period May 15-27 and found only four positive cases (0·6%).
No age-specific seroprevalence was reported.

Burlington, Vermont, USA[78]
This study analyzed specimens from 454 primary care patients at a Level 1 medical center in Burlington on 25-28 June 2020 and found 10 positive results using a two-step serologic assay and found raw prevalence of 2·2 percent (CI: 0·8-3·6%). The median age of individuals with positive vs. negative results were nearly identical (51·9 vs. 51·4 years).
Caldari Ortona, Italy [79] This study collected specimens from 640 residents on April 18-19 and found raw prevalence of 12%. No age-specific results were reported.
Chelsea, Massachusetts, USA [80] This study collected specimens from 200 pedestrians and found a raw seroprevalence of 22·5% using an IgG LFA. No age-specific results were reported.
Connecticut, USA [81] This study analyzed specimens from a random sample of 505 adults residing in non-congregate settings. The sample design reflected the assumption of statewide prevalence of 10% (roughly similar to that of the neighboring state of New York) with the aim of obtaining prevalence estimates with precision of 2% at a confidence level of 90%. However, the study obtained a much lower estimated prevalence of 3·1% (95% CI: 1·1-5·1%). Consequently, the sample size proved to be insufficient to provide reliable age-specific results; the margin of error exceeds the estimated prevalence for all age groups reported in the study.
Czech Republic [82] The Czech Ministry of Health conducted a large-scale seroprevalence survey on April 23-May 1, collecting specimens from a random sample of 22316 residents and testing for IgG antibodies using the Wantai test kit. Only 107 positive cases were identified (raw prevalence = 0·4%), and hence the testadjusted confidence intervals include the lower bound of zero prevalence. That result is consistent with the very low number of reported cases in the Czech Republic as of early May; for example, Prague had only 1,638 reported cases for a population of 1·3 million.
Denmark [83] This study analyzed specimens from a random sample of 2427 individuals in early June and identified 34 positive cases, yielding a test-adjusted prevalence of 1·2% (CI: 0·7-1·7%). Age-specific estimates were not reported as of September 17.

Faroe Islands
Denmark [84] This study analyzed specimens from a random sample of 1075 participants during late April and obtained 6 positive results; the test-adjusted prevalence was 0·7% (CI: 0·3-1·3%). No age-specific results were reported.
Finland [85] Finland National Institute for Health and Welfare has been conducting an ongoing study of seroprevalence using random sampling of the population. Each specimen is initially screened for antibodies using a rapid test, and all specimens with positive screening results are analyzed using a microneutralization test (MNT) with confirmed specificity of 100%. As of August 8, this process screened 3155 specimens and obtained 8 positive MNT results (0.25%). No age-specific results were reported as of September 17.
Gangelt, Germany [86] This study analyzed specimens from a random sample of 919 participants from the municipality of Gangelt (population 12,597) on March 31 to April 6 and obtained a test-adjusted prevalence of 15·5% (CI: 12·3-19·0%). Official government reports indicate that Gangelt had 7 COVID-19 fatalities at the time of the study but the death toll rose to 12 by late June, indicating an overall IFR of about 0·6%, similar to the IFR for Geneva. Age-specific fatalities have not been reported for Gangelt.
Hibino, Japan [88] This study analyzed specimens from 615 employees of a Japanese company in Tokyo, including 350 individuals who were tested on two separate occasions. Among the 95 individuals with positive IgG results on the first occasion, there were 12 with negative IgG results in their second test. Among the 252 with an initial negative result, 54 had positive IgG results in their second test.
No age-specific test results were reported.
Hideaway, Texas USA [89] This study analyzed specimens from 293 adults and found 7 positive results.
Age-specific results were not reported.
Ischgl, Austria [90] This study analyzed specimens from 184 adults in Ischgl (an Austrian municipality of 1,604 residents) and obtained 85 positive results, i.e., prevalence of 46·2%. The study reported the fraction of positive results for specific age groups (4 out of 11 adults 55-64 years, 2 out of 8 adults 65-74 years, and 1 out of 2 adults ages 75+) but did not report test-adjusted estimates or confidence intervals by age group. Ischgl had only 2 reported COVID-19 fatalities as of July 1.

Israel[91]
Israel Health Ministry initiated a large-scale seroprevalence study in May. Subsequent media reports indicated that initial tests of 70000 Israelis indicated that prevalence varied significantly across regions and health organizations. No age-specific results had been released as of September 17.
Japanese Evacuees [92] This study performed PCR tests on 565 Japanese citizens expatriated from Wuhan, China. There were eight positive tests, indicating a raw prevalence of 1·4%, but assessment of age-specific prevalence or IFRs is not feasible given the small sample, low prevalence, and lack of data on case outcomes.
Jersey, United Kingdom [93] This study collected samples from 629 households comprising 1,062 individuals and estimated seroprevalence at 4·2% (CI 2·9 to 5·5%), indicating that about 3,300 Jersey residents have been infected. Jersey has had 30 COVID-19 fatalities (as of July 15), and hence the overall IFR is about 1% (similar to that of NYC). However, the seroprevalence sample is too small to facilitate accurate assessments of age-specific IFRs; for ages 55+, there were 258 samples and 12 positive cases,

S18
Slovenia [111] Researchers at the University of Ljubljana assessed seroprevalence using an IgG ELISA test for a random sample of 1,318 participants on April 20 to May 3. Test-adjusted prevalence was 0·9% (CI: 0 to 2·1%), indicating that the sample may have included only 10 infected individuals; no age-specific results were reported.
South-East England [112] This study collected samples from 481 participants of the TwinsUK cohort and obtained 51 positive results (raw prevalence of 12%). No age-specific results were reported. Stockholm, Sweden [113] This study did not directly assess prevalence but produced estimates of IFR for two age groups (ages 0-69 and 70+) using a novel methodology linking live virus tests, reported cases, and mortality outcomes.
Stockholm Districts, Sweden [114] This study analyzed samples from 213 randomly selected individuals in two residential areas of Stockholm on June 17-18 and found markedly different seroprevalence rates of 4·1% and 30%, respectively. No age-specific results were reported.
Stockholm Region, Sweden [115] Stockholm County began offering antibody testing on a free walk-up basis. As of July 20, 166,431 antibody tests had been performed, of which 17·7% were positive. No demographic data or test-adjusted seroprevalence results had been reported as of September 17.
Miyagi, Osaka, and Tokyo, Japan [116] This study collected samples from randomly-selected residents of three cities on June 1-7 and used two IgG test kits (Abbott and Roche); results were deemed "positive" only if confirmed by both tests Estimated seroprevalence was 0·1% in Tokyo (2 positive results from 1,971 specimens), 0·17% in Osaka (5 positive results from 2,970 specimens), and 0·03% in Miyagi (1 positive result from 3,009 specimens). No age-specific prevalence estimates were reported.
United States [117] Seroprevalence estimates are reported in the U.S. CDC's weekly COVID-19 surveillance summary using data collected by 85 state and local public health laboratories. These reports include age-specific seroprevalence but no details regarding sample selection, test characteristics, or confidence intervals and hence could not be used in our metaregression.
Utsunomiya, Japan [118] This study tested a random sample of 742 participants and found 3 confirmed positive results among 463 adults ages 18 to 65 years; the test-adjusted prevalence for that age group was 0·65% (CI: 0·13-1·8%). No positive results were obtained for the sample of 181 adults ages 65+ years.
Virginia, USA [119] Virginia Department of Health collected specimens from a random sample of 3113 participants ages 16+ during early June and estimated prevalence of 2·4%. No confidence intervals or age-specific results had been released as of September 17.
Vo, Italy [120] Vo' is a municipality of 3,300 people, nearly all of whom (87%) participated in an infection survey in late February. However, there were only 54 infections among people ages 50+, so assessing age-specific IFRs is not feasible.
Washoe County, Nevada, USA [121] This study collected samples from 234 individuals on June 9-10 and obtained 5 positive IgG results. No age-specific results were reported.
Winston-Salem, North Carolina, USA [122] This ongoing study has been collecting specimens from a representative sample of area residents since mid-April, and raw prevalence was characterized as "about 10%." On July 28 the researchers reported that the test was not sufficiently sensitive and that a new test would be deployed henceforth.
Zurich, Switzerland [123] This study analyzed specimens from 578 individuals, including 90 with prior confirmed COVID-19 infections, 177 with positive patient contacts, and 311 who were randomly selected residents of Zurich. Seroprevalence in the randomly-selected group was estimated at 3·9% using the optimized test method.

Location Description
Luxembourg [125] Of the 35 participants who tested positive, 19 had previously interacted with a person who was known to be infected or had a prior test for SARS-CoV-2.
Boise, Idaho [126] This study was promoted during a "Crush the Curve" publicity campaign and required participants to sign up for a test.
Santa Clara, California, USA [127] Participants were recruited via social media and needed to drive to the testing site. Stanford Medicine subsequently released a statement indicating that the study was under review due to concerns about potential biases. [128] Frankfurt, Germany [129] This study was conducted at an industrial worksite. Among the 5 seropositive participants, 3 had prior positive tests or direct contact with a known positive case.

Location Description
Brooklyn, New York, USA [130] This study used samples from an outpatient clinic and yielded a much higher infection rate than other seroprevalence studies of the New York metropolitan area.
Kobe, Japan [131] This study tested for IgG antibodies in 1,000 specimens from an outpatient clinic and found 33 positive cases. However, the study did not screen out samples from patients who were seeking treatment for COVID-related symptoms. Moreover, the study reported raw prevalence and confidence interval but did not report statistics adjusted for test characteristics. The manufacturer (ADS Biotec / Kurabo Japan) has indicated that this test has specificity of 100%, based on a sample of 14 pre-COVID specimens, but that specificity has not been evaluated by any independent study. The authors concluded by noting the selection bias and recommended that "further serological studies targeting randomly selected people in Kobe City could clarify this potential limitation." Tokyo, Japan [132,133] The authors of this study specifically cautioned against interpreting their results as representative of the general population. In particular, the sample of 1,071 participants included 175 healthcare workers, 332 individuals who had experienced a fever in the past four months, 45 individuals who had previously taken a PCR test, and 9 people living with a COVID-positive cohabitant. The study obtained a raw infection rate of 3·8%, but the rate is only 0·8% if those subgroups are excluded.
United States [134] This study analyzed residual plasma of 28503 randomly selected adult patients receiving dialysis in July 2020. Seroprevalence for the tested sample was 8·0% (CI: 7·7-8·4%), However, prevalence of dialysis patients can diverge markedly from that of the general population; for example, this sudy finds a seropositive rate of 34% for patients residing in New York state, nearly three times higher than the 14% seroprevalence rate obtained using a random sample of that state. [11] Zurich, Switzerland [135] This study analyzed two distinct set of samples: (i) blood donors and (ii) hospital patients. Nearly all blood donors were ages 20 to 55, so that sample is not useful for assessing age-specific IFRs for older adults. The sample of hospital patients was not screened to eliminate cases directly related to COVID-19, so that sample may not be representative of the broader population, e.g., inhabitants of the city of Zurich constituted a relatively large fraction of seropositive results compared to residents from the rest of the canton of Zurich. The study found an overall IFR of 0·5% similar to that of Geneva.

Location Description
Apulia, Italy [136] This study assessed specimens from a sample of 904 healthy blood donors at a transfusion center in southeastern Italy and obtained 9 positive results (0·99%).
Denmark [138] This study assessed specimens from a sample of 20640 Danish blood donors and calculated a test-adjusted prevalence of 1·9% (CI:0·8-2·3). Unfortunately, the antibody test used in this study was subsequently identified as unreliable, and the Danish government returned all remaining test kits to the manufacturer. [139] England [140] Public Health England has conducted ongoing surveillance of seroprevalence using specimens from healthy adult blood donors. For example, in 7694 samples tested during May (weeks [18][19][20][21], the testadjusted prevalence was 8·5% (CI: 6·9-10%).
Germany [141] This study assessed residue sera from 3186 regular blood donors collected during March 9-June 3 and obtained 29 positive results (raw prevalence 0·9%). The authors stated: "It should be emphasized that the preselection of blood donors as a study cohort is accompanied by limitations regarding representation of population." Lombardy, Italy [142] This study assessed specimens from 390 blood donors residing in the Lodi red zone collected on April 6 and found a raw seroprevalence rate of 23%.
Milan, Italy [143] This study assessed specimens from a random sample of 789 blood donors over the period from February 24 (at the start of the outbreak) to April 8.
Netherlands [144] This study assessed specimens from 7361 adult blood donors collected on April 1-15 and found seroprevalence of 2·7%.
Rhode Island, USA [145] This study assessed specimens from 2008 blood donors collected during April 27-May 11 and found seroprevalence of 0·6%.
Scotland [146] This study assessed specimens from 3500 blood donors collected between This study assessed specimens from 1000 blood donors that were collected during March and found one positive result (raw prevalence 0·1%).
United States [148] This study analyzed residual sera from 252882 U.S. blood donors obtained between June 1 and July 31 and found an overall seroprevalence of 1·83%.

Location Description
Oisie, France [150] This sample of 1,340 participants included elementary school teachers, pupils, and their families. Only two individuals in the sample were ages 65 years and above.
Saxony, Germany [151] This study analyzed specimen samples from students and teachers at thirteen secondary schools in eastern Saxony and found very low seroprevalence (0·6%).

Location Description
United States [153] This study analyzed specimens for 50130 consecutive life insurance applicants whose blood samples were collected for insurance underwriting purposes between 12 May and 25 June 2020. The study found 1520 positive results, that is, raw seroprevalence of 3·0%. The study did not find significant differences in prevalence across three age groups (18-40, 41-60, and 61-85 years).

Location Description
Lombardy, Italy [154] This study used a database of 62881 contacts of COVID-19 cases and conducted RT-PCR tests and antibody screening on 5484 individuals. The study reported that 2824 individuals had positive tests (51·5% of the sample), of which 62 individuals subsequently died with a COVID-19 diagnosis.

H.4 Exclusion of observations with seroprevalence indistinguishable from zero
Note: The metaregression analysis excludes observations for which either (a) the lower bound of the 95% confidence interval equals zero, and hence the upper bound of the IFR is not well defined; or (b) the lower bound of the 95% confidence interval is less than the observed COVID-19 mortality rate for that age group, implying an upper bound for the IFR that exceeds 100%.

Study Description
Castiglione d'Adda, Italy [163] This study assessed seroprevalence in a random sample of 509 residents of the municipality of Castiglione d'Adda, the location of the first COVID-related fatality in Italy. Specimens were collected on May 18-25. This study is included in our meta-analysis but not in our metaregression because this municipality is covered by a nationwide seroprevalence study of Italy. [16] Diamond Princess Cruise Ship [164] This ship was carrying 3,711 passengers and crew. RT-PCR tests indicated that 619 individuals had been infected prior to the ship's debarkation on March 7, and 14 individuals subsequently died due to COVID-related causes.
France Laboratories [162] Santé Publique France analyzed 3084 specimens of residual sera from two French laboratories (Cerba and Eurofins Bimnis) obtained during the week of 6-12 April, using the LuLISA-N test developed at the Institut Pasteur. Our metaregression includes a larger study of a representative sample of the French population in three regions (Grand Est, Ile de France, and Nouvelle Aquitaine) that accounted for about two-thirds of COVID-19 cases and fatalities during spring 2020. Consequently, this study is included in our meta-analysis but not in our metaregression to avoid pitfalls of nested or overlapping samples.
U.K. Biobank [165] This study of Great Britain assessed seroprevalence using specimens collected from a demographically balanced panel of 17,776 participants on May 27 to July 6. Our metaregression includes a much larger seroprevalence study of the English population. [15] Consequently, this study is included in our metaanalysis but not in our metaregression to avoid pitfalls of nested or overlapping samples.
U.K. Office of National Statistics [13] The U.K. Office for National Statistics (ONS) regularly reports estimates of seroprevalence from specimens provided for routine testing using an IgG ELISA test conducted by research staff at the University of Oxford. On August 18 the ONS reported age-specific results for the cumulative sample of 4840 specimens received from 26 April to 26 July and indicated that these results were broadly consistent with the findings of the UK REACT-2 study (which utilized a much larger sample).
Utah, USA [10] This study analyzed commercial lab specimens from 1132 individuals collected during April 20-May 3. This study is not included in our meta-analysis because a subsequent study analyzed a much larger randomized sample of 6527 residents of the Salt Lake City metropolitan area during May 4-June 10. [160] As of May, that metro area accounted for nearly 90% of COVID-related fatalities in Utah.

Supplementary Appendix S: Comorbidities and Demographic Factors
While age and fatality risk are closely related, differences in the age structure of the population and age-specific infection rates surely cannot explain all deviations in IFR across regions and populations. Consequently, the role of co-morbidities and other demographic and socioeconomic factors merits further research that carefully distinguishes between infection risk and IFR.
A recent U.K. study has shown that COVID-19 mortality outcomes are strongly linked to comorbidities such as chronic pulmonary disease, diabetes, and obesity. [172] However, that study specifically warns against drawing causal conclusions from those findings, which may reflect a higher incidence of COVID-19 rather than a higher IFR for individuals with those comorbidities. Indeed, as shown in Table S1, a study of hospitalized U.K. COVID-19 patients found that patient age was far more important than any specific comorbidity in determining mortality risk. [67] For example, the COVID-19 fatality risk for an obese 40-year-old hospital patient was found to be moderately higher than for a non-obese individual of the same cohort but only one-tenth the fatality risk for a nonobese 75-year-old hospital patient.
The high prevalence of comorbidities among COVID-19 patients has been well documented but not compared systematically to the prevalence of such comorbidities in the general population. For example, one recent study of hospitalized COVID-19 patients in New York City (NYC) reported that 94% of those patients had at least one chronic health condition. [173] However, as shown in Table S2, that finding is not particularly surprising given the prevalence of comorbidities among middle-aged and elderly NYC residents. For example, nearly 30% of older NYC adults (ages 60+) are diabetic, while 23% have cardiovascular disease (including hypertension), and 8% have chronic pulmonary diseases-practically identical to the incidence of those comorbidities in the sample of hospitalized COVID-19 patients. Indeed, obesity was the only comorbidity that was much more prevalent among hospitalized COVID-19 patients than in the general population of older NYC adults. Nonetheless, obesity is also much more prevalent among lower-income groups who are more likely to live in high-density neighborhoods and work in high-exposure jobs, and hence such data clearly cannot be used to distinguish prevalence vs. severity of COVID-19.
Our meta-analysis has not directly considered the extent to which IFRs may vary with other demographic factors, including race and ethnicity. Fortunately, valuable insights can be garnered from other recent studies. In particular, one recent seroprevalence study of residents of two urban locations in Louisiana found no significant difference in IFRs between whites and Blacks. [99] Nonetheless, the incidence of COVID-19 mortality among people of color is extraordinarily high due to markedly different infection rates that reflect systematic racial and ethnic disparities in housing and employment. For example, a recent infection study of a San Francisco neighborhood found that 80% of positive cases were Latinx -far higher than the proportion of Latinx residents in that neighborhood. [109] That study concluded as follows: "Risk factors for recent infection were Latinx ethnicity, inability to shelter-in-place and maintain income, frontline service work, unemployment, and household income less than $50,000 per year." Other researchers have reached similar conclusions, attributing elevated infection rates among Blacks and Hispanics to dense housing of multi-generational families, increased employment in high-contact service jobs, high incidence of chronic health conditions, and lower quality of health care. [174] In summary, while our meta-analysis has investigated the effects of age on IFR for COVID-19, further research needs to be done on how infection and fatality rates for this disease are affected by comorbidities as well as demographic and socioeconomic factors.  Figure 5.

Supplementary Appendix T: Metaregression Predictions for Manaus, Brazil
A recent seroprevalence study of Manaus, Brazil found prevalence of about 66% as of August 2020; moreover, that prevalence was roughly similar across age groups. [175] As of October 29, the Brazil Ministry of Health reported 2853 confirmed COVID-19 deaths in Manaus. [176] Assuming a uniform prevalence of 66%, the following table shows the age-specific predictions of our metaregression based on the age structure of the Manaus population. The predicted death toll of 3108 is well aligned with the actual number of confirmed COVID-19 deaths, and the predicted population IFR of 0·22% is well aligned with the observed population IFR of 0·2%.