COVID-19 testing

Polymerase chain reaction (PCR) is a process that amplifies (replicates) a small, well-defined segment of DNA many hundreds of thousands of times, creating enough of it for analysis. Test samples are treated with certain chemicals[13][14] that allow DNA to be extracted. Reverse transcription converts RNA into DNA.

Reverse transcription polymerase chain reaction (RT-PCR) first uses reverse transcription to obtain DNA, followed by PCR to amplify that DNA, creating enough to be analyzed.[14] RT-PCR can thereby detect SARS-CoV-2, which contains only RNA. The RT-PCR process generally requires a few hours.[15]

Real-time PCR (qPCR)[16] provides advantages including automation, higher-throughput and more reliable instrumentation. It has become the preferred method.[17][18]

The combined technique has been described as real-time RT-PCR[19] or quantitative RT-PCR[20] and is sometimes abbreviated qRT-PCR[21] rRT-PCR[22] or RT-qPCR,[23] although sometimes RT-PCR or PCR are used. The Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines propose the term RT-qPCR,[24] but not all authors adhere to this.

Samples can obtained by various methods, including a nasopharyngeal swab, sputum (coughed up material),[25] throat swabs,[26] deep airway material collected via suction catheter[26] or saliva.[27][28] Drosten et al. remarked that for 2003 SARS, "from a diagnostic point of view, it is important to note that nasal and throat swabs seem less suitable for diagnosis, since these materials contain considerably less viral RNA than sputum, and the virus may escape detection if only these materials are tested."[29]

The likelihood of detecting the virus depends on collection method and how much time has passed since infection. According to Drosten tests performed with throat swabs are reliable only in the first week. Thereafter the virus may abandon the throat and multiply in the lungs. In the second week, sputum or deep airways collection is preferred.[26]

Collecting saliva may be as effective as nasal and throat swabs,[27] although this is not certain.[30][28] Sampling saliva may reduce the risk for health care professionals by eliminating close physical interaction.[31] It is also more comfortable for the patient.[32] Quarantined people can collect their own samples.[31] A saliva test's diagnostic value depends on sample site (deep throat, oral cavity, or salivary glands).[28] One study found that saliva yielded greater sensitivity and consistency when compared with swab samples.[33]

Viral burden measured in upper respiratory specimens declines after symptom onset.[34]

• 
  Demonstration of a nasopharyngeal swab for COVID-19 testing
• 
  Demonstration of a throat swab for COVID-19 testing
• 
  A PCR machine

Isothermal nucleic acid amplification tests also amplify the virus's genome. They are faster than PCR because they don't involve repeated heating and cooling cycles. These tests typically detect DNA using fluorescent tags, which are read out with specialized machines. CRISPR gene editing technology was modified to perform the detection: if the CRISPR enzyme attaches to the sequence, it colors a paper strip. The researchers expect the resulting test to be cheap and easy to use in point-of-care settings.[35][36] The test amplifies RNA directly, without the RNA-to-DNA conversion step of RT-PCR.[37]

An antigen is the part of a pathogen that elicits an immune response. Antigen tests look for antigen proteins from the viral surface. In the case of a coronavirus, these are usually proteins from the surface spikes.[38] One of the challenges is to find a target unique to SARS-CoV-2.[39]

Antigen tests may be one way to scale up testing to much greater levels.[38] Isothermal nucleic acid amplification tests can process only one sample at a time per machine. RT-PCR tests are accurate but require too much time, energy and trained personnel to run the tests.[38] "There will never be the ability on a [PCR] test to do 300 million tests a day or to test everybody before they go to work or to school," Deborah Birx, head of the White House Coronavirus Task Force, said on 17 April 2020. "But there might be with the antigen test."[40]

A nasopharyngeal swab is exposed to paper strips containing artificial antibodies designed to bind to coronavirus antigens. Antigens bind to the strips and give a visual readout. The process takes less than 30 minutes, can deliver results at point of care, and does not require expensive equipment or extensive training.[38]

Swabs of respiratory viruses often lack enough antigen material to be detectable.[39] This is especially true for asymptomatic patients who have little if any nasal discharge. Viral proteins are not amplified in an antigen test.[38][41] According to the WHO the sensitivity of similar antigen tests for respiratory diseases like the flu ranges between 34% and 80%. "Based on this information, half or more of COVID-19 infected patients might be missed by such tests, depending on the group of patients tested," the WHO said. Many scientists doubt whether an antigen test can be made reliable enough in time to be useful against COVID-19.[41] According to the FDA, positive results from antigen tests are highly accurate, but there is a higher chance of false negatives, so negative results do not rule out infection. Therefore, negative results from an antigen test may need to be confirmed with a PCR test.[42]

Typical visible features on CT initially include bilateral multilobar ground-glass opacities with a peripheral or posterior distribution.[43] Subpleural dominance, crazy paving, and consolidation may develop as the disease evolves.[43][44] Chest CT scans and chest x-rays are not recommended for diagnosing COVID-19. Radiologic findings in COVID-19 lack specificity.[45][43]

Left: Automated analyzer for immunoassays, used, for example, to find SARS-CoV-2 antibodies. Right: Example of quantitative results for SARS-CoV-2 antibody test.

Serology tests test blood and can identify once-infected people who have recovered.[46]

Part of the immune system's response to infection is the production of antibodies, including IgM and IgG. According to the FDA, IgM antibodies to SARS-CoV-2 are generally detectable several days after initial infection, although levels over the course of infection and beyond are not well characterized.[47] SARS-CoV-2 IgG antibodies generally become detectable 10–14 days after infection, sometimes earlier, and normally peak around 28 days after infection onset.[48][49] Antibodies emerge too late to serve as acute infection markers. Antibodies for some diseases persist in the bloodstream for many years, while others quickly fade away.[38]

Antibody tests can be used to estimate the once-infected percentage of a population. SARS-CoV-2 antibodies' potency and protective period are unknown.[4][50]

RDTs typically use a small, portable, positive/negative lateral flow assay that can be executed at point of care. RDTs may process blood samples, saliva samples, or nasal swab fluids. RDTs produce colored lines to indicate positive or negative results.[51]

ELISAs can be qualitative or quantitative and generally require a lab. These tests usually use whole blood, plasma, or serum samples. A plate is coated with a viral protein, such as a SARS-CoV-2 spike protein. Samples are incubated with the protein, allowing any antibodies to bind to it. The antibody-protein complex can then be detected with another wash of antibodies that produce a color/fluorescent readout.[51]

Neutralization assays assess whether sample antibodies prevent viral infection in test cells. These tests sample blood, plasma or serum. The test cultures cells that allow viral reproduction (e.g., VeroE6 cells). By varying antibody concentrations, researchers can visualize and quantify how many test antibodies block virus replication.[51]

Chemiluminescent immunoassays are quantitative lab tests. They sample blood, plasma, or serum. Samples are mixed with a known viral protein, buffer reagents and specific, enzyme-labeled antibodies. The result is luminescent. A chemiluminescent microparticle immunoassay uses magnetic, protein-coated microparticles. Antibodies react to the viral protein, forming a complex. Secondary enzyme-labeled antibodies are added and bind to these complexes. The resulting chemical reaction produces light. The radiance is used to calculate the number of antibodies. This test can identify multiple types of antibodies, including IgG, IgM, and IgA.[51]

Most if not all large scale COVID-19 antibody testing looks for binding antibodies only and does not measure the more important neutralizing antibodies (NAb).[52][53][54] A NAb is an antibody that defends a cell from an infectious particle by neutralizing its biological effects. Neutralization renders the particle no longer infectious or pathogenic.[55] A binding antibody binds to the pathogen but the pathogen remains infective; the purpose can be to flag the pathogen for destruction by the immune system.[56] It may even enhance infectivity by interacting with receptors on macrophages.[57] Since most COVID-19 antibody tests return a positive result if they find only binding antibodies, these tests cannot indicate that the subject has generated protective NAbs that protect against re-infection.[53][54]

It is expected that binding antibodies imply the presence of NAbs[54] and for many viral diseases total antibody responses correlate somewhat with NAb responses[58] but this is not established for COVID-19. A study of 175 recovered patients in China who experienced mild symptoms reported that 10 individuals had no detectable NAbs at discharge, or thereafter. How these patients recovered without the help of NAbs and whether they were at risk of re-infection was not addressed.[53] An additional source of uncertainty is that even if NAbs are present, viruses such as HIV can evade NAb responses.[52]

Studies have indicated that NAbs to the original SARS virus (the predecessor to the current SARS-CoV-2) can remain active for two years[59] and are gone after six years.[60] Nevertheless, memory cells including Memory B cells and Memory T cells[61] can last much longer and may have the ability to reduce reinfection severity.[60]

• 
  A Point of Care Test in Peru. A blood droplet is collected by a pipette.
• 
  Blood from pipette is then placed onto a COVID-19 rapid diagnostic test device.
• 
  The rapid diagnostic test shows reactions of IgG and IgM antibodies.

Following recovery, many patients no longer have detectable viral RNA in upper respiratory specimens. Among those who do, RNA concentrations three days following recovery are generally below the range in which replication-competent virus has been reliably isolated.[62]

No clear correlation has been described between length of illness and duration of post-recovery shedding of viral RNA in upper respiratory specimens.[63]

Scientists from China first released information on the viral genome on 11 January 2020. That day the Malaysian Institute for Medical Research (IMR) produced "primers and probes" specific to a SARS-CoV-2 RT-PCR test.[77] The IMR's materials were used to diagnose Malaysia's first patient on 24 January.[78] BGI Group was one of the first companies to receive emergency use approval from China's National Medical Products Administration for a nucleic acid test.[79]

Public Health England announced a test on the 10th,[80] using a real-time RT-PCR (RdRp gene) assay based on oral swabs.[81] The test detected the presence of any type of coronavirus, including specifically identifying SARS-CoV-2. It was rolled out to twelve laboratories across the United Kingdom on 10 February.[82]

The first case in South Korea was confirmed on 19 January.[83]

The German nucleic acid testing protocol was published on the 17th. Another early PCR test was developed by Charité University hospital in Berlin, working with academic collaborators in Europe and Hong Kong, and published on the 23rd. It used rtRT-PCR, and formed the basis of 250,000 kits distributed by the World Health Organization (WHO).[84]

In the US, the CDC developed its SARS-CoV-2 Real Time PCR Diagnostic Panel.[85] The protocol became available on the 28th.[86] One of three tests in early kits failed due to faulty reagents.

In Russia, the first COVID‑19 test was developed by the State Research Center of Virology and Biotechnology VECTOR. Production began on 24 January.[87]

CDC refused to let other labs process tests that month; allowing an average of fewer than 100 samples/day to be processed. Tests using two components were not determined to be reliable until the 28th, and only then were state and local laboratories permitted to begin testing.[88] The test was approved by the FDA under an EUA.South Korean company Kogenebiotech's clinical grade, nucleic acid test (PowerChek Coronavirus) was approved by Korea Centers for Disease Control and Prevention (KCDC) on 4 February.[89]

In Wuhan BGI opened a makeshift 2000-sq-meter emergency detection laboratory named "Huo-Yan" (Chinese: 火眼, "Fire Eye") on the 5th.[90][91] It processed more than 10,000 samples/day.[92][91] Construction required 5 days.[93] The Wuhan Laboratory was followed by Huo-Yan labs in Shenzhen, Tianjin, Beijing, and Shanghai, in a total of 12 cities across China.

On 11 February the test was approved by the Federal Service for Surveillance in Healthcare.[94]

By 4 March China reached 50,000 tests per day.[95]

On the 16th WHO called for ramping up testing programmes as the best way to slow the spread.[96][97]

Shortages of reagent and other supplies became a bottleneck for mass testing in the EU and UK[98] and the US.[99][100]

On 12 March, Mayo Clinic announced a nucleic acid test.[101]

Early in March, China reported accuracy problems with its PCR tests.[102] A study examined 1070 samples from 205 Wuhan patients and reported varied sensitivity according to the methods and location of sample collection. Samples from bronchoalveolar lavage fluid specimens returned the highest sensitivity.[103] The authors argued that CT scans showed even higher sensitivity.[104]

US commercial labs began testing in early March. As of the 5th LabCorp announced nationwide availability of COVID‑19 testing based on RT-PCR.[105] Quest Diagnostics made nationwide testing available as of 9 March.[106] US testing demand grew rapidly, causing backlogs of hundreds of thousands of tests at private US labs. Supplies of swabs and chemical reagents continued strained.[107] On 25 May, the US required each state to take responsibility for meeting its testing needs.[108] In March, the FDA issued EUAs for nucleic acid tests to Hologic (3/16),[109] Abbott Laboratories (3/18),[110] Thermo Fisher Scientific (3/19)[111] Cepheid (3/21)[112][113] and LabCorp (4/30).[110]

On 27 March the FDA issued an EUA for a test using Abbott Laboratories' ID Now assay.[114]

By 31 March United Arab Emirates was testing more of its population per head than any other country.[115] UAE implemented a combination of drive-through sample collection, and a mass-throughput laboratory from Group 42 and BGI. The lab conduced tens of thousands RT-PCR tests per day and was the first to be operational at that scale other than China.[116]

Several European countries initially conducted more tests than the US.[117][118] By 19 March drive-in tests were offered in several large cities.[119] As of 26 March, German Health Minister Jens Spahn estimated that Germany was conducting 200,000 tests per week.[120] As of the end of March at least 483,295 samples were tested and 33,491 (6.9%) had tested positive.[121]

80% of test kits that Czechia purchased from China gave wrong results.[122][123] Slovakia purchased 1.2 million antibody-based test kits from China that were found to be inaccurate.[124] China accused Czechia and Slovakia of incorrect use of those tests.[125] Ateş Kara of the Turkish Health Ministry said the test kits Turkey purchased from China had a "high error rate".[126][127]

Germany has a large medical diagnostics industry, with more than a hundred testing labs that provided the technology and infrastructure to enable rapid increases in testing. Costs are borne by insurance when the test is ordered by a physician.[128] According to the president of the Robert Koch Institute, as of 22 March, Germany had capacity for 160,000 tests per week.[129]

Spain purchased test kits from Chinese firm Shenzhen Bioeasy Biotechnology Co Ltd, but found that results were unacceptable. The maker explained that the incorrect results may stem from failure to collect samples or use the kits correctly. The Spanish ministry switched to another vendor, Shenzhen Bioeasy.[130]

By month end, testing had surpassed 200k/week.[131]

The FDA gave an EUA for the US' first antibody test on the 2nd.[132][50]

As of 7 April, WHO had accepted two diagnostic tests for procurement under the Emergency Use Listing procedure (EUL).[133]

On 13 April, Health Canada approved a nucleic acid test from Spartan Bioscience. Institutions may "test patients" with a handheld DNA analyzer "and receive results without having to send samples away to a [central] lab".[134][135]

By the start of April, the United Kingdom was delivering around 10,000 swab tests per day.[136] The British NHS announced that it was piloting a scheme to test suspected cases at home, to remove the risk of one patient infecting others at a hospital or disinfecting an ambulance used to transport a patient.[137]

The UK purchased 3.5 million antibody test kits from China, but in early April 2020 announced these were not usable.[138][139] On 21 April 2020, the Indian Council of Medical Research (ICMR) advised Indian states to stop using test kits purchased from China after receiving complaints from one state. Rajasthan health minister Raghu Sharma on 21 April said the kits gave only 5.4 percent accurate results.[140]

Antibody survey results found from 2% to 30% positive.[141] On preliminary data, WHO concluded that 2% to 3% of the world population had developed antibodies.[142]

By month end, testing had surpassed 750k/week.[131]

In May antibody tests were conducted on 5,603 major league baseball employees and 0.7% tested positive, showing that they had been infected. 70% of those who tested positive had had no symptoms.[143][144][145] The US was conducting an average of 2.5 million tests per week for the week ending 17 May. This grew to 3.2 million by 14 June.[146][147]

Attempts to culture virus from upper respiratory specimens were largely unsuccessful when viral burden is low but detectable (i.e., Ct values higher than 33-35).[72]

On 3 May, Roche received an EUA for a selective ELISA serology test.[148][149]

On 8 May, the FDA granted its first EUA for antigen test: "Sofia 2 SARS Antigen FIA" by Quidel Corp.[150][42]

The FDA announced on 14 May a review of 15 adverse event reports about the Abbott ID Now device for low sensitivity.[151]

On 21 May, researchers at Ben-Gurion University in Israel reported a one-minute coronavirus test with 90% accuracy, based on the "change in the resonance in the THz spectral range" shown by the coronavirus through THz spectroscopy.[152]

Nearly two million antibody tests imported into Australia and costing $20 million were declared unusable.[153][154][155]

In early May Harvard's Global Health Institute estimated that the US needed to test more than 900k per day.[156][157] Other recommendations ranged up to 23m per day.[158][159][160][161]

On 29 May Siemens received an EUA for its anti-spike RBD-targeting serology test that it believes detects neutralizing antibodies.[162]

By month end, testing had surpassed 1.4m/week.[131]

In early June, researchers announced a nucleic acid diagnostic test using reverse transcription-loop-mediated isothermal amplification (RT-LAMP), an existing technology used in pathogenic microorganism identification, genetically modified ingredients, tumor detection, and embryo sex identification. The test identified virus in samples of serum, urine, saliva, oropharyngeal swabs and nasopharyngeal swabs. Once commercialized the test has the potential to provide rapid (30-45 minute) diagnosis at point of care. The test was 100% selective and highly sensitive, detecting virus at a concentration of .06 fg/ml.[163]

As of 14 June 2020, the percentage testing positive in the US as a whole had fallen below 5%.[164] As of late June, test numbers crossed 600k/day.[146]

In drive-through testing, a healthcare professional takes a sample using appropriate precautions, such as wearing PPE.[165][166] Drive-through centers helped South Korea accelerate its testing program.[167]

In Hong Kong test subjects can stay home and receive a specimen tube. They spit into it, return it and later get the result.[168]

In Israel, researchers at Technion and Rambam Hospital developed a method for testing samples from 64 patients simultaneously, by pooling the samples and only testing further if the combined sample was positive.[169][170][171] Pool testing was then adopted in Israel, Germany, South Korea,[172] Nebraska,[173] China[174] and the Indian states of Uttar Pradesh,[175] West Bengal,[176] Punjab,[177] Chhattisgarh[178] and Maharashtra.[179]

Open source, multiplexed designs released by Origami Assays can test as many as 1122 patient samples using only 93 assays.[180] These balanced designs can be run in small laboratories without robotic liquid handlers.

One study proposed a rapid immune response assay as a screening test, with a confirmatory nucleic acid test for diagnosis, followed by a rapid antibody test to determine course of action and assess population exposure/herd immunity.[181]

Required testing levels are a function of disease spread. The more the cases, the more tests are needed to manage the outbreak. COVID-19 tends to grow exponentially at the beginning of an outbreak, meaning that the number of required tests initially also grows exponentially. If properly targeted testing grows more rapidly than cases, it can be contained.

WHO recommends increasing testing until fewer than 10% are positive in any given jurisdiction.[182]

Economist Paul Romer reported that the US has the technical capacity to scale up to 20 million tests per day, which is his estimate of the scale needed to fully remobilize the economy.[159] The Edmond J. Safra Center for Ethics estimated on 4 April that this capacity could be available by late July.[183] Romer pointed to single-molecule real-time sequencing equipment from Pacific Biosciences[184][159] and to the Ion Torrent Next-Generation Sequencing equipment from ThermoFisher Scientific.[185][159] According to Romer, "Recent research papers suggest that any one of these has the potential to scale up to millions of tests per day." This plan requires removing regulatory hurdles. Romer estimated that $100 billion would cover the costs.[159]

Romer also claimed that high test accuracy is not required if tests are administered frequently enough. He ran model simulations in which 7% of the population is tested every day using a test with a 20% false negative rate and a 1% false positive rate. The average person would be tested roughly every two weeks. Those who tested positive would go into quarantine. Romer's simulation indicated that the fraction of the population that is infected at any given time (known as the attack rate) peaks reaches roughly 8% in about thirty days before gradually declining, in most runs reaching zero at 500 days, with cumulative prevalence remaining below 20%.[186]

Number of tests done per day in the US.
Blue: CDC lab
Orange: Public health lab
Gray: Data incomplete due to reporting lag
Not shown: Testing at private labs; total exceeded 100,000 per day by 27 March.[187]

Tests developed in China, France, Germany, Hong Kong, Japan, the United Kingdom, and the US targeted different parts of the viral genome. WHO adopted the German system for manufacturing kits sent to low-income countries without the resources to develop their own.

PowerChek Coronavirus looks for the "E" gene shared by all beta coronaviruses, and the RdRp gene specific to SARS-CoV-2.[188]

US President Donald Trump displays a COVID-19 testing kit from Abbott Laboratories in March 2020.

Abbott Laboratories' ID Now nucleic acid test uses isothermal amplification technology.[114] The assay amplifies a unique region of the virus's RdRp gene; the resulting copies are then detected with "fluorescently-labeled molecular beacons".[189] The test kit uses the company's "toaster-size" ID Now device, which is widely deployed in the US.[190] The device can be used in laboratories or in point of care settings, and provides results in 13 minutes or less.[189]

Primerdesign offers its Genesig Real-Time PCR Coronavirus (COVID‑19). Cobas SARS-CoV-2 Qualitative assay runs on the Cobas® 6800/8800 Systems by Roche Molecular Systems. They are offered by the United Nations and other procurement agencies.

Quidel's "Sofia 2 SARS Antigen FIA"[150][42] is a lateral flow test that uses monoclonal antibodies to detect the virus's nucleocapsid (N) protein.[191] The result is read out by the company's Sofia 2 device using immunofluorescence.[191] The test is simpler and cheaper but less accurate than nucleic acid tests. It can be deployed in laboratories or at point of care and gives results in 15 minutes.[150] A false negative result occurs if the sample's antigen level is positive but below the test's detection limit, requiring confirmation with a nucleic acid test.[191]

Antibodies are usually detectable 14 days after the onset of the infection. Multiple jurisdictions survey their populations using these tests.[192][193] The test requires a blood draw.

Private US labs including Quest Diagnostics and LabCorp offer antibody testing upon request.[194]

Antibody tests are available in various European countries.[195]

Roche offers a selective ELISA serology test.[196]

Sensitivity indicates whether the test accurately identifies whether the virus is present. Each test requires a minimum level of viral load in order to produce a positive result. A 90% sensitive test will correctly identify 90% of infections, missing the other 10% (a false negative). Even relatively high sensitivity rates can produce high rates of false negatives in populations with low incidence rates.[197]

Specificity indicates how well-targeted the test is to the virus in question. Highly specific tests pick up only the virus in question. Non-selective tests pick up other viruses as well. A 90% specific test will correctly identify 90% of those who are uninfected, leaving 10% with a false positive result.[197]

Low-specificity tests have a low positive predictive value (PPV) when prevalence is low. For example, suppose incidence is 5%. Testing 100 people at random using a test that has a specificity of 95% would yield on average 5 people who are actually negative who would incorrectly test positive. Since 5% of the subjects actually are positive, another five would also test positive correctly, totaling 10 positive results. Thus, the PPV is 50%,[198] an outcome no different from a coin toss. In this situation retesting those with a positive result increases the PPV to 94.5%, meaning that only 4.5% of the second tests would return the incorrect result, on average less than 1 incorrect result.[199]

Improper sample collection, exemplified by failure to acquire enough sample and failure to insert a swab deep into the nose. This results in insufficient viral load, one cause of low clinical sensitivity.

The time course of infection also affects accuracy. Samples may be collected before the virus has had a chance to establish itself or after the body has stopped its progress and begun to eliminate it.

Improper storage for too long a time can cause RNA breakdown and lead to wrong results as viral particles disintegrate.[200]

Improper design and manufacture can yield inaccurate results. Millions of tests made in China were rejected by various countries throughout the period of March 2020 through May 2020.

Test makers typically report the accuracy levels of their tests when seeking approval from authorities. In some jurisdictions, these results are cross-validated by additional assessments. Reported results may not be achieved in clinical settings due to such operational inconsistencies.

Detection of SARS-CoV-2 by nasal swab over six weeks in patients who experienced mild to moderate illness

RT-PCR is the most accurate diagnostic test.[102] It typically has high sensitivity and specificity in a laboratory setting: however, in one study sensitivity dropped to 66-88% clinically.[201]

In one study sensitivity was highest at week one (100%), followed by 89.3%, 66.1%, 32.1%, 5.4% and zero by week six.[202][203]

A Dutch CDC-led laboratory investigation compared 7 PCR kits.[204] Test kits made by BGI, R-Biopharm AG, BGI, KH Medical and Seegene showed high sensitivity.[205]

High sensitivity kits are recommended to assess people without symptoms, while lower sensitivity tests are adequate when diagnosing symptomatic patients.[204]

One study reported that the ID Now COVID-19 test showed sensitivity of 85.2%. Abbott responded that the issue could have been caused by analysis delays.[206] Another study rejected the test in their clinical setting because of this low sensitivity.[207]

Iceland[211] managed the pandemic with aggressive contact tracing, inbound travel restrictions, testing, and quarantining, but with less aggressive lock-downs.

Researchers tested the entire population of Vò, the site of Italy's first COVID‑19 death. They tested about 3,400 people twice, at an interval of ten days. About half the people testing positive had no symptoms. All discovered cases were quarantined. Along with restricting travel to the commune, new infections were completely eliminated.[212]

Unlike other Asian countries, Japan did not experience a pandemic of SARS or MERS, so the country's PCR testing system was not well developed.[213][214] Japan preferentially tested patients with severe illness and their close contacts at the beginning. Japan's Novel Coronavirus Expert Meeting chose cluster measures to identify infections clusters.[213][214] The Expert Meeting analyzed the outbreak from Wuhan and identified conditions leading to clusters (closed spaces, crowded spaces and close-contact), and asked people to avoid them.[214][215]

In January, contact tracers took action shortly after the first infection was found. Only administrative tests were carried out at first, until insurance began covering PCR tests on 6 March. Private companies began to test, and the test system gradually expanded.[213][216]

On 3 April, those with positive tests were legally permitted to recuperate at home or in a hotel if they had asymptomatic or mild illness, ending the hospital bed shortage.[217] The first wave (from China) was contained,[218] but a second wave (caused by returnees from Europe and the US) in mid-March led to spreading infection in April.[214] On 7 April, Japan declared a state of emergency, (less strict than a lockdown, because it did not block cities or restrict outings).[214][217][219] On 13 May, antigen test kits became covered by insurance, and were combined with a PCR test for diagnosis.[220][221]

Japan's PCR test count per capita remained far smaller than in some other countries even though its positive test rate was lower. Excess mortality was observed in March.[215][219][222] The Expert Meeting stated, "The Japanese health care system originally carries out pneumonia surveillance, allowing it to detect most of the severely ill patients who develop pneumonia. There are a large number of CT scanners in Japan and they have spread to small hospitals all over the country, so pneumonia patients are rarely missed. In that sense, it meets the same standards as other countries that mainly carry out PCR tests."[215][222] The group recommended using CT scans data and doctor's findings for diagnosis.[223][224] On the Diamond Princess cruise ship, many people who initially tested negative later tested positive. Half of coronavirus-positives there who remained mild or asymptomatic had pneumonia findings on CT scans and their CT image showed a frosted glass shadow that is characteristic of infection.[225][226]

On 27 April, Russia tested 3 million people and had 183,000 positive results.[227] On 28 April Anna Popova, head of Federal Service for Surveillance in Healthcare (Roszdravnadzor) stated that 506 laboratories were testing; that 45% of those who tested positive had no symptoms; that 5% of patients had a severe form; and 40% of infections were from family members. Illness improved from six days to one day after symptoms appeared. Antibody testing was carried out on 3,200 Moscow doctors, finding 20% immunity.[228]

With contact tracing, inbound travel restrictions, testing, and quarantining, Singapore arrested the initial spread without complete lockdown.[229]

South Korea's broad testing approach helped reduce spread. Testing capacity, largely in private sector labs, was built up over several years by the South Korean government in the early 2000s.[230]

The government exploited the resident registration number (RRN) system. Authorities mobilized young men who were eligible for military service as social service agents, security and public health doctors. Public health doctors were mainly dispatched to public health centers and life treatment centers where mildly ill patients were accommodated. They performed PCR tests and managed mild patients. Social service agents worked in pharmacies to fill staff shortages. Korea's 10k PCR tests per million residents was the world's highest as of 13 April rising to 20k by mid-June. Twenty-seven Korean companies exported test kits worth $48.6 million in March, and were asked to provide test kits or humanitarian assistance by more than 120 countries. Korean authorities set up a treatment center to isolate and manage patients with asymptomatic and minor illnesses in one facility in order to vacate hospital beds for the more severely ill.

Centers were sited mainly at national facilities and corporate training centers. The failure of Korea's MERS quarantine in May 2015 left Korea more prepared for COVID-19 than countries that did not face that pandemic. Then President Park Geun-hye allowed Korean CDC-approved private sector testing for infectious diseases in 2016. Korea already had a system for isolating, testing and treating infectious disease patients separately from others. Patients with respiratory illness but no epidemiological relevance were treated at the National Hospital, and those with epidemiological relevance were treated at selected clinics.[83][231][232][233][234][235][236][237][238]

Korea established a large scale drive-through/walk-through" test testing program. However, the most common method was “mobile examination”. In Daegu City, 54% of samples were collected by 23 March in home or hospital. Collecting samples door-to-door of avoided the risk of travel by possibly infected patients, but required additional staff. Korea solved the problem by drafting more than 2,700 public insurance doctors.[83][234][233]

The government disclosed personal information to the public via KCDC without patient consent. The authorities used digital surveillance to trace possible spread.[231][234][235][237][238][239][240][241][242] [243]

Health insurance IDs and national identification card numbers were used to trace contacts.[244][245][246][247]

New York State's control measures consisted of PCR tests, stay-at-home measures and strengthening the healthcare system. On 29 February before its first case, the state allowed testing at the Wordsworth Center. They managed to convince the CDC to approve tests at state laboratories and the FDA to approve a test kit. As of 13 March the state was conducting more than 1,000 daily tests, growing to 10,000/day on 19 March. In April, the number exceeded 20,000. Many people queued at hospitals to get tested. On 21 March New York City health officials directed medical providers to test only those entering the hospital, for lack of PPE.[237][248][249][250][251]

Following an outbreak, 94% of the 4,800 aircraft carrier crew were tested. Roughly 60 percent of the 600-plus sailors who tested positive were asymptomatic.[252] Five infected sailors who completed quarantine subsequently developed flu-like symptoms and again tested positive.[253]

A shortage of trained medical laboratory scientists, assay reagents, analyzers, transport medium, and PPE coupled with high demand had limited initially limited the availability of testing and led to significantly increased turnaround times.

Testing strategies vary by country and over time,[254] with some countries testing very widely,[7] while others have at times focused narrowly on only testing the seriously ill.[255] The country that tests only people showing symptoms will have a higher figure for "% (Confirmed cases as percentage of tested samples or tested cases)" than the country that also tests others.[256] If two countries are alike in every respect, including which people they test, the one that tests more people will have a higher "Confirmed / million people". Studies have also found that countries that test more, relative to the number of deaths, have lower estimated case fatality rates[8] and younger age distributions of cases.[10]

COVID-19 testing statistics by country

Country	Date[lower-alpha 1]	Tested	Units[lower-alpha 2]	Confirmed (cases)	%	Tested / million people	Confirmed / million people	Ref.
Afghanistan	31 May	38,460	samples	15,205	39.5	988	391	[257]
Albania	9 June	16,716	samples	1,299	7.8	5,839	454	[258]
Algeria	20 June	22,467	samples	11,631	51.8	515	267	[259]
Argentina	28 June	329,036	samples	55,343	17.1	7,425	1,272	[260]
Armenia	24 May	50,397		6,661	13.2	17,074	2,257	[261]
Australia	28 June	2,379,175	samples	7,686	0.32	93,668	303	[262]
Austria	29 June	606,375		17,631	2.9	68,112	1,980	[263]
Azerbaijan	31 May	298,648		5,494	1.8	30,173	555	[264]
Bahrain	21 June	477,788		21,764	4.6	304,431	13,867	[265]
Bangladesh	26 June	699,941	samples	130,474	18.6	4,250	792	[266]
Barbados	23 May	4,664		92	2.0	16,249	321	[267]
Belarus	24 June	901,904	samples	59,945	6.6	95,019	6,315	[268]
Belgium	23 May	768,706	samples	56,810	7.4	66,752	4,933	[269]
Bhutan	5 June	18,821	samples	48	0.13	25,375	65	[270]
Bolivia	20 June	56,964	cases	23,512	41.3	4,984	2,057	[271]
Bosnia and Herzegovina	24 June	86,273	samples	3,676	4.3	25,215	1,074	[272]
Brazil	26 June	11,340,524	samples	1,280,054	11.3	53,965	6,091	[273][274]
Bulgaria	24 May	74,096	samples	2,427	3.3	10,661	349	[275]
Burkina Faso	21 May	6,642	samples	814	12.3	318	39	[276][277]
Cameroon	10 May	10,268		2,579	25.1	387	97	[278]
Canada	26 June	2,598,243	cases	102,794	4.0	68,565	2,713	[279]
Chile	25 June	1,025,081	samples	259,064	25.3	53,750	13,584	[280]
China	22 June	90,410,000	samples	90,083	0.09	62,814	63	[281][282][283][284]
Colombia	21 June	604,273	samples	68,652	11.4	12,522	1,423	[285]
Costa Rica	21 June	35,369	samples	2,213	6.3	7,075	443	[286]
Croatia	9 June	69,115	cases	2,247	3.3	16,956	551	[287]
Cuba	21 June	151,164	samples	2,312	1.5	13,346	204	[288]
Cyprus[lower-alpha 3]	23 May	99,733	samples	927	0.93	115,530	1,074	[289]
Czechia	26 June	541,212	samples	11,038	2.0	50,609	1,032	[290]
Denmark	22 June	753,812	cases	12,727	1.7	127,137	2,147	[291]
DR Congo	2 June	9,508		3,195	33.6	106	36	[292]
Ecuador	20 June	138,880	samples	50,183	36.1	8,129	2,937	[293]
Egypt	2 June	141,002	samples	26,384	18.7	1,409	264	[292]
El Salvador	21 June	140,429	samples	4,626	3.3	21,650	713	[294]
Estonia	8 June	90,621	samples	1,940	2.1	68,220	1,460	[295]
Ethiopia	23 June	223,341	samples	4,848	2.17	1,943	42	[296]
Finland	19 June	229,843	samples	7,142	3.1	41,464	1,288	[297]
France[lower-alpha 4]	2 June	1,493,333		142,843	9.6	22,281	2,131	[298][299][300]
Germany	9 June	4,694,147	samples	184,861	3.9	56,454	2,223	[301]
Ghana	7 May	137,924		3,091	2.2	4,439	99	[302]
Greece	10 May	98,877	samples	4,015	4.1	9,182	373	[303]
Grenada	18 June	5,465	samples	23	0.42	49,034	206	[304]
Guinea	6 May	7,369	cases	1,856	25.2	561	141	[278]
Hungary	8 June	210,749	samples	4,014	1.9	21,816	416	[305]
Iceland	21 June	69,190	samples	1,823	2.6	189,947	5,005	[306]
India	30 June	8,608,654	samples	566,840	6.6	6,239	411	[307][308]
Indonesia	28 June	456,636	cases	54,010	11.8	1,706	202	[309]
Iran	5 June	1,040,289	samples	167,156	16.1	12,506	2,009	[310]
Iraq	7 May	120,604	samples	2,543	2.1	2,998	63	[311]
Ireland	27 June	419,021	samples	25,414	6.1	85,141	5,164	[312]
Israel	24 June	886,068		21,666	2.4	96,574	2,361	[313]
Italy	24 June	5,107,093	samples	239,410	4.7	84,611	3,966	[314]
Ivory Coast	25 May	18,303		2,423	13.2	694	92	[278]
Jamaica	12 May	7,465	samples	507	6.8	2,739	186	[315]
Japan	17 June	599,339	samples	17,382	2.9	4,751	138	[316]
Kazakhstan	31 May	812,881	samples	10,858	1.3	43,578	582	[317]
Kenya	10 May	32,097	samples	672	2.1	675	14	[318]
Kosovo	9 June	17,199	samples	1,269	7.4	9,500	701	[319]
Kyrgyzstan	5 May	58,030	samples	843	1.5	9,082	132	[320]
Latvia	8 June	118,322	samples	1,088	0.92	61,627	567	[321]
Lebanon	12 May	53,269		870	1.6	7,804	127	[322]
Libya	2 June	6,871	cases	156	2.3	1,001	23	[292]
Lithuania	24 June	403,841	samples	1,804	0.45	144,522	646	[323]
Luxembourg	29 June	178,422	samples	4,256	2.4	284,970	6,798	[324]
Madagascar	2 June	11,311	cases	826	7.3	431	31	[292]
Malawi	21 June	10,880	samples	749	6.9	569	39	[325]
Malaysia	16 June	659,632	cases	8,505	1.3	20,128	260	[326]
Maldives	22 May	17,174	samples	1,274	7.4	43,758	3,246	[327]
Malta	27 May	64,334	samples	612	0.95	130,347	1,240	[328]
Mauritius	2 June	120,358	samples	335	0.28	95,071	265	[329]
Mexico	28 June	483,953	cases	212,802	44.0	3,762	1,654	[330]
Montenegro	5 May	8,534	samples	324	3.8	13,520	513	[331]
Morocco	2 June	221,154	cases	7,859	3.6	5,992	213	[332]
Mozambique	10 May	3,923		91	2.3	126	2.9	[333]
Myanmar	31 May	24,710	samples	224	0.91	454	4.1	[334]
Nepal	24 June	457,950		10,099	2.2	16,300	359	[335]
Netherlands	3 June	378,912	cases	46,942	12.4	21,745	2,694	[336]
New Zealand	19 June	335,167	samples	1,159	0.35	67,254	233	[337]
Nigeria	7 June	76,802	samples	12,486	16.3	375	61	[338]
North Korea	17 April	740	cases	0	0	29	0	[339]
North Macedonia	1 June	30,302		2,315	7.6	14,588	1,115	[340]
Northern Cyprus[lower-alpha 5]	30 May	31,346		108	0.34	96,153	331	[341]
Norway	22 May	310,635	cases	8,745	2.8	57,872	1,629	[342]
Oman	25 June	175,278	samples	34,902	19.9	37,715	7,510	[343]
Pakistan	20 June	1,239,153	samples	202,955	16.4	5,611	919	[344]
Palestine	30 June	92,206	samples	2443	2.6	18,252	484	[345]
Panama	9 June	80,720		17,233	21.3	19,325	4,126	[346]
Paraguay	19 June	54,278	samples	1,336	2.5	7,610	187	[347]
Peru	29 June	1,661,324	samples	282,365	17.0	50,613	8,602	[348]
Philippines	28 June	708,534	samples	35,455	5.0	7,016	351	[349]
Poland	1 June	931,520	samples	23,987	2.6	24,267	625	[350]
Portugal	23 May	689,705	samples	30,471	4.4	67,114	2,965	[351]
Qatar	26 June	341,914	cases	92,784	27.1	118,677	32,205	[352]
Romania	28 May	410,000	samples	18,791	4.6	21,132	969	[353]
Russia	28 June	19,334,442	samples	641,156	3.3	131,755	4,369	[354][355]
Rwanda	21 June	111,257		728	0.65	8,590	56	[356]
Saint Lucia	25 May	898		18	2.0	4,937	99	[357]
Saudi Arabia	9 June	976,815		105,283	10.8	28,058	3,024	[358]
Senegal	11 May	22,481		1,886	8.4	1,418	119	[359]
Serbia	28 May	233,479	cases	11,300	4.8	33,528	1,623	[360]
Singapore	22 June	684,359	samples	42,313	6.2	119,987	7,419	[361][362]
Slovakia	29 June	208,904	samples	1,664	0.80	38,276	305	[363]
Slovenia	7 June	84,684	samples	1,495	1.8	40,440	714	[364]
South Africa	6 June	850,871	cases	43,434	5.1	14,477	739	[365]
South Korea	28 June	1,232,690	cases	12,715	1.0	23,839	246	[366]
Spain	4 June	4,465,338	samples	287,740	6.4	95,550	6,157	[367][368]
Sri Lanka	16 June	90,010	cases	1,915	2.1	4,154	88	[369]
Sudan	2 June	7,825	samples	5,173	66.1	178	118	[292]
Sweden	17 June	382,804	cases	52,189	13.6	37,066	5,053	[370]
Switzerland[lower-alpha 6]	16 June	465,722	samples	31,146	6.7	54,099	3,618	[371]
Taiwan[lower-alpha 7]	29 June	76,491	cases	447	0.58	3,241	19	[372]
Tanzania	6 June	2,680		509	19.0	45	8.5	[278]
Thailand	7 May	89,791	cases	2,992	3.3	1,293	43	[373]
Trinidad and Tobago	29 June	4,969	samples	126	2.3	3,643	92	[374]
Tunisia	2 June	52,874	cases	1,086	2.1	4,474	92	[375]
Turkey	29 June	3,331,158	cases	198,613	6.0	40,060	2,388	[376]
Uganda	21 June	166,917	samples	770	0.46	3,649	17	[377]
Ukraine	23 May	456,509	samples	29,070	6.4	10,861	692	[378]
United Arab Emirates	23 June	3,055,566		45,683	1.5	318,310	4,759	[379]
United Kingdom	29 June	9,290,215	samples	311,965	3.4	137,540	4,619	[380]
United States	28 June	32,000,928	samples	2,596,771	8.1	96,684	7,846	[381][382]
Uruguay	19 June	56,234	samples	853	1.5	16,204	246	[383]
Uzbekistan	7 May	371,000	samples	2,298	0.62	10,900	68	[384]
Venezuela	20 June	1,129,176	samples	3,790	0.34	39,089	131	[385]
Vietnam	15 June	261,004	samples	334	0.13	2,644	3.4	[386]
Zimbabwe	21 June	55,709		463	0.83	3,748	31	[387]
Local time. For some countries it is unclear whether they report samples or cases. One person tested twice is recorded as one case and two samples. Excluding Northern Cyprus. Testing data from 4 May to 12 May is missing because of the transition to the new reporting system SI-DEP. Northern Cyprus is not recognized as a sovereign state by any country except Turkey. Includes data for Liechtenstein. The UN does not recognize Taiwan as a sovereign state.