TRAVEL ALERT ✈️: U.S. Department of State raised its travel advisory for China from Level 3 to *Level 4*, which means "do not travel to China." Previously alert level 3 on Monday, which means "reconsider travel to China due to the novel coronavirus." travel.state.gov/content/travel… 
2) New airlines suspending China ✈️ amid the outbreak: Air France (Until Feb. 9), British Airways (until end February), Air Seoul, Egyptair, Lion Air, Lufthansa (till end February), Swiss Airlines (till end February) and Austrian Airlines (till end February).
3) AIRPORT 🦠 SCREENING: “for every 100 infected travellers planning to take a 12 hour flight, only 9 will be detected at entry screening upon arrival” lshtm.ac.uk/newsevents/new…
4) However, phrased another way, “thermal scanning at airports detects less than 1 in 5 passengers arriving from a 12 hour flight who are infected with the new coronavirus” (link above). Difference between 20% and 9% is that some are discovered before landing.
5) “Our work reinforces that thermal scanning cannot detect every traveller infected with this new coronavirus. Other policies that can decrease the risk of transmission are crucial.” says @LSHTM researcher (link above).
6) Testing issues - although the below is an anecdote (I don’t usually like anecdotes cuz one can’t conclude anything), this unreliability is technically and issue of “sensitivity” and “specificity” in epidemiology. No test is perfect, all have errors.
7) Regarding 🦠lab test sensitivity (% of true cases the test IDs) and specificity (%negs of true neg), we don’t know it perfectly yet for new #2019Cov yet, but for SARS RT-PCR test it was only 80% sens, 90% spec in early onset. This is actually quite poor ncbi.nlm.nih.gov/pmc/articles/P…
8) In epidemiology of screening, not just sensitivity% and specificity% important, but the *PPV* (Positive Predictive Value - % of Tested+ who are actually True+) is even more key. But what happens if you screen everyone randomly, such as busy cold/flu season, something weird...
9) ...weird happens when you mass-screen in population w/ few #coronavirus cases (ie random testing in gen pop with common cold/flu) is that PPV % value can be *super low* despite semi-good sens/spec if low prev. For math why sensitivity=80% & specificity=90% can yield bad PPV...
10) THE 🦠 SCREENING DILEMMA: so despite 80% PCR virus test sensitivity + 90% specificity (defined above), if virus has 2% prevalence, then PPV=13.8%, meaning only 1 in 7 positive test results will be true positives😮! Even if virus generously has 10% prev, PPV still only 44%!😥
11) 🦠 SCREENING PROBLEM (Part 2)—So what does the above math mean??? Even if good lab🧪 test for #coronavirus, the low PPV of test tells us we could have a LOT of +Test results that are not True+ (unless 🦠 prevalence super high in population tested). It’s a v challenging issue.
12) 🦠 SCREENING PROBLEM (Part3)—To be clear, this does not mean don’t get screened or don’t get tested! It just adds challenges for rapid field tests on mass populations (Eg airports or random street) that have low 🦠 prevalence. In a good CDC lab, a sample can be tested again.
13) 🦠SCREENING PROBLEM (Part4)—Now all the “confirmed” cases we see are all true cases because CDC labs use more advanced tests + repeat 2 stage tests. But this is also why it takes a long time to confirm #coronaviruses cases, because not all clinics can do the advanced tests.
14) 🦠SCREENING🧵(bottomline)—because 🦠 must be stopped even before symptoms, we need to do more rapid field testing. But math shows the challenge of any mass screening because of *bad PPV* when 🦠carriers are rare. Such why this virus will be here infecting us for a long time🤧
15) Maybe the “Screening 🧵” above was challenging for some to follow (but I believe you can all flex some #math🧢 to rise to occasion), here are additional reading on basics of screening parameters in epidemiology, because epidemiology is for everyone!💪🏼 online.stat.psu.edu/stat507/node/7…
