Covid-19: Trump declares intention to “re-open economy” within weeks against experts’ adviceBMJ 2020; 368 doi: https://doi.org/10.1136/bmj.m1217 (Published 25 March 2020) Cite this as: BMJ 2020;368:m1217
All rapid responses
The projection of “deaths in the low millions” described in the article by Dyer was based on an Imperial College Model (ICM) that may have substantially overstated the impact of COVID-19 spread on Great Britain (GB) and the United States (US), according to published data, much of which was available when the model was developed. Two key ICM assumptions, the percentage of hospitalizations that require ICU care for invasive mechanical ventilation (IMV) or extracorporeal membrane oxygenation (ECMO) and the percentage of ICU stays that result in death, were based on informal sources, including personal communication and expert opinion, rather than on available published information. An empirically based analysis would have had several features that differed in important ways from the ICM.
First, the ICM-assumed 30% rate of intensive care unit (ICU) treatment for IMV or ECMO would have been replaced with data published on 28 Feb 2020, from a sample of 1,099 hospitalizations for COVID-19 in China (median inpatient age 47.0 years), in which 2.3% of inpatients received IMV, 0.5% ECMO, and 5.0% ICU care. A reasonable range of alternative empirically supported estimates could use data from Wuhan (10.6%), Lombardy, Italy (16%), or Chinese inpatients with severe disease (19.1%).
Second, the ICM-assumed 50% mortality among ICU patients (i.e., 15% of hospitalized patients) would have been replaced with empirical mortality data (1.4% of hospitalized Chinese inpatients) and an alternate of 8.1% (Chinese inpatients with severe disease). These assumptions would be in line with reports of a 1.4% case fatality among all symptomatic individuals in Wuhan and much lower case-fatality rates in all other mainland Chinese cities, although lower than the fatality rate of 1.8%-3.4% reported for all COVID-19 cases in a preliminary US study in which mortality data were missing for 2,001 of 4,226 patients.
Third, an empirical approach would have considered available data on important sources of regional variation. In Italy, severe outcomes would be expected because of elevated baseline rates of health care-associated infections, pollution-related cardiopulmonary disease deaths, and smoking, in addition to a generally older-age population. Similarly, higher risk of negative impact would have been anticipated in the state of New York, which was already experiencing its highest rate of influenza infections in >20 years at the outset of the COVID-19 pandemic, had the highest hospital bed occupancy rate in the US in 2015, and in New York City, where ozone pollution, a known contributor to lung damage, ranks tenth in the nation. In China, resource shortages might have artificially depressed the service utilization rates we report here, but disease severity was likely exacerbated by elevated rates of smoking (25% in China vs. 11% in the US and 17% in the UK) and baseline pollution-related cardiopulmonary disease deaths per 100,000 (140 in China vs. 24 in the US and 32 in the UK).
Fourth, competing causes of illness and death have not been addressed by any estimates to date. Autopsy studies suggest coronavirus infections are commonly accompanied by other viruses[14,15], making the attribution of death solely to COVID-19 problematic because of detection bias; yet, a popular press report suggests that all Italian patients who die in hospital with a concomitant COVID-19 infection are assumed to have died from COVID-19. Policy decisions should be based on excess (i.e., disease-attributable) mortality, not total mortality.
Data available from other, less formal recent sources suggest that the ICM-assumed hospitalization rate of 4.4% of infected individuals (including both symptomatic and asymptomatic cases) may also have been overstated. For example, a total of 100,000 estimated (but undetected) infections in the state of Ohio was announced on March 12; yet, 15 days later, the state’s total reported number of COVID-19 hospitalizations was just 276, far less than would be expected from a disease with a median incubation period of 5 days and hospitalization rate of 4.4% of infected individuals as assumed in the ICM, even accounting for in-hospital testing delays. Early experience in Ohio and New York does suggest a higher ICU admission rate than in China, a geographic variation that should be monitored over time; it may reflect practice-pattern variation or preferential testing speed for higher-acuity patients.
All these sources of uncertainty and variation should, according to best practices for decision analysis modeling, be addressed by transparent reporting of estimates and presentation of numeric results in ranges. Yet, the ICM report publication includes no specific numerical information about projected infected case counts in either GB or the US, no ICU bed counts in the US, no sources or assumptions (e.g., repurposing of elective resources) for ICU bed counts, and no ranges on outcomes associated with the health care utilization assumptions. These data are essential so that policy makers can validate projections against empirical reality on an ongoing basis and adjust decisions accordingly.
Our purpose is not to advocate for any specific policy. Rather, it is to emphasize the need to use the best possible data instead of informal sources when modeling social policies with crucial implications for individual economic and clinical well-being. Others have similarly called for empirical approaches to COVID-19 policy development.[15,20] To make highest-quality decisions, policy makers need highest-quality evidence.
1. Dyer O. Covid-19: Trump declares intention to “re-open economy” within weeks against experts’ advice. BMJ 202 ; 368. doi: https://doi.org/10.1136/bmj.m1217
2. Ferguson NM, Laydon Dm, Nedjati-Gilani G, et al. Impact of non-pharmaceutical interventions (NPIs) to reduce COVID-19 mortality and healthcare demand. Imperial College COVID-19 Response Team. 16 Mar 2020. Available: https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-... [Accessed 24 Mar 2020]
3. Guan W, Ni Z, Hu Y, et al. Clinical characteristics of coronavirus disease 2019 in China. N Engl J Med 28 Feb 2020. doi: 10.1056/NEJMoa2002032. [Epub ahead of print]
4. Li R, Rivers C Tan Q, et al. The demand for inpatient and ICU beds for COVID-19 in the US: Lessons from Chinese cities. Available: https://dash.harvard.edu/handle/1/42599304 [Accessed 24 Mar 2020].
5. Grasselli G, Pesenti A, Cecconi M. Critical care utilization for the COVID-19 outbreak in Lombardy, Italy. JAMA 2020 13 Mar. doi:10.1001/jama.2020.4031. [Epub ahead of print].
6. Wu JT, Leung K, Bushman M, et al. Estimating clinical severity of COVID-19 from the transmission dynamics in Wuhan, China. Nature Research 13 Mar 2020. doi: 10.21203/rs.3.rs-17453/v1 [Epub ahead of print]
7. CDC COVID-19 Response Team. Severe outcomes among patients with coronavirus 2019 (COVID-19)—United States, February 12-March 16, 2020. MMWR 2020;69. Available: https://www.cdc.gov/mmwr/volumes/69/wr/mm6912e2.htm [Accessed 23 Mar 2020]
8. OECD. Health at a Glance 2019. 2019. Available: https://www.oecd-ilibrary.org/docserver/4dd50c09-en.pdf?expires=15849843... [Accessed 23 Mar 2020].
9. Newey S. why have so many coronavirus patients died in Italy? The Telegraph. 22 Mar 2020. Available: https://www.telegraph.co.uk/global-health/science-and-disease/have-many-... [Accessed 23 Mar 2020].
10. New York Department of Health. Governor Cuomo announces record-breaking total number of flu cases in New York even as number of flu cases decline for second week. 27 Feb 2020. Available: https://www.governor.ny.gov/news/governor-cuomo-announces-record-breakin.... [Accessed 28 Mar 2020].
11. CDC. Table 91. Occupancy rates in community hospitals and average annual percent change, by state: United States, selected years 1980–2015. Available: https://www.cdc.gov/nchs/data/hus/2017/091.pdf. [Accessed 28 Mar 2020].
12. Wang M, Aaron CP, Madrigano J, et al. Association between long-term exposure to ambient air pollution and change in quantitatively assessed emphysema and lung function. JAMA 2019;322:546–556. Abstract available: https://jamanetwork.com/journals/jama/article-abstract/2747669. [Accessed 28 Mar 2020].
13. Nichols A. NYC is among smoggiest cities in nation, report says. Patch 24 Apr 2019. Available: https://patch.com/new-york/new-york-city/nyc-among-smoggiest-cities-nati.... [Accessed 28 Mar 2020].
14. Walsh EE, Shin JH, Falsey AR. Clinical impact of human coronaviruses 229E and OC42 infection in diverse adult populations. J Infect Dis 2013;208(10):1634-1642.
15. Ioannidis JPA. A fiasco in the making? As the coronavirus pandemic takes hold, we are making decisions without reliable data. Stat. 17 Mar 2020. Available: https://www.statnews.com/2020/03/17/a-fiasco-in-the-making-as-the-corona... [Accessed 23 Mar 2020].
16. Sullivan P. Ohio health official estimates 100,000 people in state have coronavirus. The Hill. 12 Mar 2020. Available: https://thehill.com/policy/healthcare/487329-ohio-health-official-estima... [Accessed 23 Mar 2020].
17. ABC6. Coronavirus claims the lives of 19 in Ohio as 1137 infected. Available: https://abc6onyourside.com/news/local/march-27-coronavirus-in-ohio-update [Accessed 28 Mar 2020].
18. Millman J, Shea T. ‘They need us’: Cuomo urges strength as NY cases near 46K, more than 600 dead. 4New York 28 Mar 2020. Available: https://www.nbcnewyork.com/news/local/too-many-people-are-dying-alone-ny.... [Accessed 28 Mar 2020].
19. Weinstein MC, O’Brien B, Hornberger J, et al. Principles of good practice of decision analytic modeling in health care evaluation: report of the ISPOR Task Force on Good Research Practices-Modeling Studies. Value Health 2003;6(1):9-17.
20. Bendavid E, Bhattacharya J. Is the coronavirus as deadly as they say? Wall Street Journal 24 Mar 2020. Available with subscription: https://www.wsj.com/articles/is-the-coronavirus-as-deadly-as-they-say-11... [Accessed 26 Mar 2020]
Competing interests: No competing interests