The historical claim of this section is that the federal report of February 2020 sorted what it knew from what it did not, named the assumption on which the response acted, and admitted that its measures might not prevent what they were meant to slow; and that it named the conditions of its own revision only in general terms. The section adds to the precaution debate a document that states its ignorance as well as its uncertainty and leaves the costs of its measures unstated; to the debate over declarations, a dated record of the international act that preceded the statement of 11 March; and to the timing debate, a correction, printed in the next issue, to the date of one of its own acts.
5.1What the summary box said
The report opens with three questions and short answers. The first answer is what was already known: "In December 2019, an outbreak of acute respiratory illness caused by a novel coronavirus (2019-nCoV) was detected in mainland China. Cases have been reported in 26 additional locations, including the United States." The second is what the report added: "Nine of the first 11 U.S. 2019-nCoV patients were exposed in Wuhan, China. CDC expects more U.S. cases." The third is what it implied for practice: "CDC, multiple other federal agencies, state and local health departments, and other partners are implementing aggressive measures to substantially slow U.S. transmission of 2019-nCoV, including identification of U.S. cases and contacts and managing travelers arriving from mainland China to the United States. Interim guidance is available at https://www.cdc.gov/coronavirus/index.html and will be updated as more information becomes available."
The box already holds the structure of the whole report: a statement of the known, attributed to detection in another country; a statement of the new, a count and an expectation; and a statement of action that is aggressive, aims to slow rather than to stop, and rests on interim advice that "will be updated as more information becomes available". That phrase is the report's first statement of what would change its guidance. It names no particular information.
5.2The dated account of the known
The body of the report begins with a chronology. "On December 31, 2019, Chinese health officials reported a cluster of cases of acute respiratory illness in persons associated with the Hunan seafood and animal market in the city of Wuhan, Hubei Province, in central China. On January 7, 2020, Chinese health officials confirmed that a novel coronavirus (2019-nCoV) was associated with this initial cluster (1). As of February 4, 2020, a total of 20,471 confirmed cases, including 2,788 (13.6%) with severe illness,* and 425 deaths (2.1%) had been reported by the National Health Commission of China (2)."
Three kinds of time are kept apart in these sentences. There is the time of a report by Chinese officials, 31 December; the time of a confirmation of the agent, 7 January; and the time of a count, "As of February 4, 2020". The count is attributed to the body that reported it. Each figure is a figure of what "had been reported", not of what had happened.
The ratio in brackets after the deaths is one of reported deaths to reported confirmed cases on one day. Lipsitch and colleagues had set out in 2015 why such a ratio taken during an outbreak is biased: deaths lag behind cases, which pulls it down; mild cases go undetected, which pushes it up; and definitions and testing differ (Lipsitch and others 2015). The report does not discuss these biases. It gives the ratio as an arithmetic fact about reported figures, which it is; a reader who would use it as a measure of severity must bring the caution from elsewhere. The footnote to "severe illness" does give the definition used, so the report marks one of the three sources of bias: what is counted depends on a definition.
The chronology continues with three acts of declaration and restriction, each dated: "On January 30, the World Health Organization (WHO) Director-General declared that the 2019-nCoV outbreak constitutes a Public Health Emergency of International Concern.† On January 31, the U.S. Department of Health and Human Services (HHS) Secretary declared a U.S. public health emergency to respond to 2019-nCoV.§" On the same day, the report says, the president signed a proclamation that limited entry from mainland China. The declarations are listed in order and without comment, as facts that frame the response.
5.3The assumption and the admission
The paragraph that follows the chronology is the centre of the report for this article's question. It states what the measures require, on what expectation they rest, and what they are for: "These measures are being implemented in anticipation of additional 2019-nCoV cases in the United States. Although these measures might not prevent the eventual establishment of ongoing, widespread transmission of the virus in the United States, they are being implemented to 1) slow the spread of illness; 2) provide time to better prepare health care systems and the general public to be ready if widespread transmission with substantial associated illness occurs; and 3) better characterize 2019-nCoV infection to guide public health recommendations and the development of medical countermeasures including diagnostics, therapeutics, and vaccines."
Two features of this sentence matter. The first is the admission: before stating what the measures are for, the report says they "might not prevent the eventual establishment of ongoing, widespread transmission". It acts on signs and says openly that the action may fail of the outcome most wanted. The second is the threefold purpose. Only the first aim is about the disease's course; the second is about time, to prepare; the third is about knowledge, so that later recommendations rest on more. The measures are justified partly by what they will teach.
The Discussion repeats the assumption in words that make it explicit: "These measures are being implemented based on the assumption that there will be more U.S. 2019-nCoV cases occurring with potential chains of transmission, with the understanding that these measures might not prevent the eventual establishment of ongoing, widespread transmission of the virus in the United States." Here the report names a presumption as a presumption. In Neustadt and May's terms, it places the expectation of more cases in the third category, the presumed, and says so (Neustadt and May 1986).
Then it closes its first section with a promise of revision: "Public health authorities are monitoring the situation closely. As more is learned about this novel virus and this outbreak, CDC will rapidly incorporate new knowledge into guidance for action by CDC and state and local health departments." The same two sentences, with the last phrase shortened to "guidance for action", close the Discussion. The report thus says twice what would change its guidance: new knowledge. It does not say which knowledge, how much of it, or at what point. It names no threshold of cases, no finding about transmission, no date at which the measures would be looked at again.
5.4The unclear, named as unclear
The report's second paragraph sets out what was not known, and it does so in the language of uncertainty rather than of assertion: "Preliminary investigation of 2019-nCoV also suggests a zoonotic origin (6), but the exact origin has not yet been determined. Person-to-person spread is evident (7); however, how easily the virus is transmitted between persons is currently unclear." It goes on to say that the new virus resembles those that cause MERS and SARS, which spread mainly by respiratory droplets, and then gives the one figure it derives by analogy: "Based on the incubation period of illness from MERS and SARS coronaviruses, CDC believes that symptoms of 2019-nCoV infection occur within 2 to 14 days following infection." The paragraph ends with a long admission: "however, many characteristics of this novel coronavirus and how it might affect individual persons and potentially vulnerable population subgroups, such as the elderly or those with chronic health conditions, remain unclear."
The sentence on incubation repays a slow reading. The fourteen-day window that recurs later in the report, in monitoring, in the criteria for testing and in the proclamation on entry, is not by the report's own account a measured property of the new virus. It is an inference from two related viruses, introduced by the verb "believes". Reinhart Koselleck's distinction between the space of experience and the horizon of expectation (Koselleck 1985) names what the sentence does: it draws an expectation about a new disease from the remembered experience of two earlier ones, and says so. A later reader can see from the sentence itself that the window was never presented as measured.
The section on clinical guidance repeats the gap and states where the guidance came from: "Additional information about 2019-nCoV is needed to better understand transmission, disease severity, and risk to the general population. Although CDC and partners are actively learning about 2019-nCoV, initial CDC guidance is based on guidance for management and prevention of respiratory illnesses including influenza, MERS, and SARS. No vaccine or specific treatment for 2019-nCoV infection is currently available." Here the report states plainly that its guidance is borrowed. In Jasanoff's terms, this is regulatory science under time pressure (Jasanoff 1990): advice that has to be given now, on the basis of what is known about other diseases, and that is labelled interim for that reason.
The report does not say what the measures would cost, or who would bear the cost of screening, restriction of entry or quarantine. It gives one weighing of risks, in the paragraph on seasonal influenza: "Reducing the number of persons in the United States with seasonal influenza will reduce possible confusion with 2019-nCoV infection and possible additional risk to patients with seasonal influenza." It is the closest the report comes to Sunstein's point that risks lie on all sides of a decision (Sunstein 2005), and it concerns a vaccine for another disease, not the measures against the new one.
5.5The measures, dated and counted
The report gives dates for each part of the federal response. "CDC established a 2019-nCoV Incident Management Structure on January 7, 2020. On January 21, CDC activated its Emergency Operations Center to optimize coordination for domestic and international 2019-nCoV response efforts." Entry screening began on 17 January at three airports and then extended to two more; "These five airports together receive approximately 85% of all air travelers from Wuhan, China, to the United States." The report then gives the screening's yield: "As of February 1, 2020, a total of 3,099 persons on 437 flights were screened; five symptomatic travelers were referred by CDC to local health care providers for further medical evaluation, and one of these persons tested positive for 2019-nCoV."
That sentence is a count of a measure's output, dated, with its scope stated. It does not claim more than it shows. It says that one of the referred travellers tested positive; it does not say whether any infected traveller passed through screening without symptoms, and it does not claim the screening prevented anything.
The report also records an event that changed the measure's scope from outside: "On January 24, 2020, travel bans began to be instituted by the Chinese government, resulting in restricted travel in and out of Hubei Province, including the city of Wuhan, and fewer travelers undergoing entry screening in the United States." And it records the next step in the border measures: "In response to the escalating risks associated with travel from mainland China, on January 31, 2020, the Presidential Proclamation further refined the border health strategy to temporarily suspend entry, undergo additional screening, or possible quarantine for individuals that have visited China (excluding Hong Kong, Macau, and Taiwan) in the past 14 days."
Here the report names, after the fact, the reason one measure gave way to another: "the escalating risks associated with travel from mainland China". It does not say what observation showed the risk escalating, or what level had been set in advance at which screening would become suspension of entry.
The report's account of the domestic cases is more specific about how they were found: "Nine cases were in travelers from Wuhan. Eight of these nine cases were identified as a result of patients seeking clinical care for symptoms and clinicians connecting with the appropriate public health systems." The Discussion draws a conclusion from this: "The critical role that the U.S. health care system plays in halting or significantly slowing U.S. transmission of 2019-nCoV is already evident: eight of the first 11 U.S. cases were detected by clinicians collaborating with public health to test persons at risk." Set beside the screening count, these sentences say something the report does not draw out: most of the first cases were found in clinics, and the airport screening, by the report's own figures, found one. The report states both facts and does not compare them.
5.6The date of an act, corrected
The section on laboratory support dates the steps by which a test for the new virus came into use. "Chinese health officials posted the full 2019-nCoV genome sequence on January 10, 2020, to inform the development of specific diagnostic tests for this emergent coronavirus (1). Within a week, CDC developed a Clinical Laboratory Improvement Amendments–approved real-time RT-PCR test that can diagnose 2019-nCoV respiratory samples from clinical specimens." The protocol was posted on 24 January. Then: "On January 4, 2020, the Food and Drug Administration issued an Emergency Use Authorization to enable emergency use of CDC’s 2019-nCoV Real-Time RT-PCR Diagnostic Panel. To date, this test has been limited to use at CDC laboratories."
As printed, the authorisation is dated before 10 January, the day the sequence on which the test was built was posted. In the next week's issue the journal printed an erratum: "In the report “Initial Public Health Response and Interim Clinical Guidance for the 2019 Novel Coronavirus Outbreak — United States, December 31, 2019–February 4, 2020,” on page 142, the fourth sentence of the first paragraph under “Laboratory and Diagnostic Support” should have read “On February 4, 2020 the Food and Drug Administration issued an Emergency Use Authorization to enable emergency use of CDC’s 2019-nCoV Real-Time RT-PCR Diagnostic Panel.”"
The error is small and the correction prompt. It is a clean instance of the problem the parent study asks about, whether a record lets a later reader keep apart the time of an act and the time of its report. The record of the act, written within days of it, put it in January instead of February, an order the report's own chronology made impossible. Corrected, the authorisation falls on the last day of the period the report covers. A reader of the first printing alone would have placed a regulatory act on 4 January; a reader who compared the chronology inside the report would have seen that it could not be so.
5.7The record beside the report
Skeptron's record of COVID-19 concerns a different act, by a different body, on 11 March, after the report was published. Its scenario reads, in part:
Skeptron's records hold one case on COVID-19, the World Health Organization's statement of 11 March 2020; lockdowns, vaccines and national responses are not in them, and this study does not use them as evidence.
It stated that the spread of COVID-19 could be characterized as a pandemic. This designation is not an act of re-dating the first occurrence of the infection that was already underway.
The record's point is the separation of a statement's date from the date of what it describes. By 11 March, as the report of February shows, Chinese officials had reported a cluster on 31 December, a novel virus had been confirmed on 7 January, an emergency of international concern had been declared on 30 January, and the United States had counted its first eleven cases by 4 February. The record and the report agree in what they keep apart, and the report supplies the earlier dates the record does not give.
The two documents differ on one detail. The record's second question reads:
What did the WHO Secretary-General designate as a pandemic on March 11, and does this wording mean the first occurrence of the disease?
The record lists its source under the same title, a briefing by the Secretary-General on 11 March 2020; the source's web address refers to the Director-General's opening remarks. The report of February names the office as "the World Health Organization (WHO) Director-General". The record's English, in this detail, gives the head of the Organization a different title from the one the report and the source's own address use.
The record's boundary answer concerns the count given in the briefing:
The announcement point and the scope of the aggregation must be maintained. Because there are delays in diagnosis and reporting, and subsequent revisions, the record judged on the information at the time and the after-the-fact outcome are kept separately.
The question it answers is:
May the cumulative tally presented in the briefing be used as the total damage later confirmed?
The record's answer is the same caution that Lipsitch and colleagues set out for ratios made during an outbreak, applied to a total: a count announced on a day is a count of what had been diagnosed and reported by that day, and later revisions belong to a different record. The report of February had followed the same rule in its own counts, attributing each to its source and dating it "As of" a day. Neither document claims that its figures are the outbreak's size.
The record says nothing about the national response, and the report of February nothing about 11 March. Each is a partial view of a period that neither covers alone, and the platform holds only one of them.
5.8Emergency and pandemic
The report's chronology and the record's subject together mark a distinction that Wilder-Smith and Osman make explicit. The International Health Regulations of 2005 provide for a public health emergency of international concern; they do not provide for the declaration of a pandemic (Wilder-Smith and Osman 2020). The report of February records the first: on 30 January the Director-General "declared that the 2019-nCoV outbreak constitutes a Public Health Emergency of International Concern". The record of March concerns the second: by the record's account, a statement that the spread of COVID-19 could be characterised as a pandemic.
The difference in verbs is a difference in kind. A declaration under the Regulations, as Gostin sets out, is made on the advice of an emergency committee and lets the Director-General issue temporary recommendations, which do not bind states (Gostin 2014). A characterisation outside the Regulations has no such procedure. A later argument about whether the world was warned too late has to choose which date it means. The platform record, by its subject, picks the later; the federal report shows the earlier on its first page.
Fidler's account of SARS adds a layer the report touches without discussing. In his reading, China's delay in reporting showed how far the older regime depended on states' notification, while the Organization used information from non-state sources, and the episode shaped the revised Regulations (Fidler 2004). The report of February begins its chronology with a state's notification: "On December 31, 2019, Chinese health officials reported a cluster". It says nothing about the time between the cluster's start and that report; its subject is the American response. The report keeps the date of the notification, not the date of the first case.
5.9What the report knew and what would change its action
Sorted by the four questions, the report of February 2020 gives a structured picture, largely in its own words. It presents as known a set of dated reports and counts, each attributed to the body that made it, together with the evidence of person-to-person spread, the symptoms, and the absence of any vaccine or specific treatment. It presents as unclear the virus's origin, how easily it passes between persons, how it affects individuals and vulnerable groups, and how it will affect the country over time; it says its guidance is borrowed and its incubation window believed by analogy. It presumes, and says so, that more cases with chains of transmission will follow. It presumes without saying so that slowing transmission is worth the cost of the measures, since it names no cost.
And it names, as what would change its action, new knowledge in general. The guidance "will be updated as more information becomes available"; the agency "will rapidly incorporate new knowledge into guidance for action". The one change of measure it describes, from screening to suspension of entry, is attributed after the fact to "escalating risks". No threshold, observation or date is set in advance. In Neustadt and Fineberg's terms, the report states that reconsideration will happen; it does not build in the points at which it will (Neustadt and Fineberg 1978).
5.10What this case adds
To the precaution debate, the case adds a document that does several things the Late Lessons editors ask for: it acknowledges ignorance as well as uncertainty, commits to monitoring, and says openly that its guidance is borrowed from related diseases (Harremoës and others 2001). It weighs one risk against another only in its paragraph on seasonal influenza, not for the measures against the new virus (Sunstein 2005), and it names its own possible failure before its aims. To the debate over declarations, it adds a dated record of the emergency of 30 January, to set against the platform's record of the characterisation of 11 March: two acts of different legal kinds that a later argument about warning can easily run together. To the timing debate, it adds an erratum: in the next week's issue, the report's own date for one of its acts was moved from January to February.