|
evilweasel posted:The amount it decreased hospital stays and deaths was both about the same percentage wise but the deaths number was not statistically significant, for reasons I don’t quite understand. I thought remembered reading that too and went looking for the answer. Was that this study from China? In this one no outcome was statistically significant, though numerically people did seem to improve. I know it isn't what you're asking, but they mention that they gamed out their test and found they needed 325 people based on some assumption about uncertainty. They only got 237 so the effect would have to be much more dramatic than expected to result in a statistically significant difference in outcomes. Since I can't find the study that mentions the outcomes you mention I'll take a stab at an answer: the (point estimate) percent decline in a number can be misleading. My guess is the similarity in percentage improvement is just a coincidence and the denominator matters. The ability to distinguish between two group depends on error estimates. Error estimates are not necessarily dependent on percentage of change, so the absolute number of items in a testing group matter. Also, measuring small changes in initially small percentages is hard. A death rate of 1.75% and 2% in 200 patients is 3.5 and 4 people, respectively. Obviously 3.5 people can't die, so that number could be (but not limited to) 3 or 4. In one case the point estimate would be a 25% difference in the number of deaths, but not statistically significant because it could just as easily been 0% difference. However, if there were several thousand people in the group you could probably make a statistically significant claim. The trick is to pay attention to the error bars. In the above linked paper the "confidence interval". One of the criticisms of that Stanford paper from a month ago was that they provided a range, but it wasn't a confidence interval, it was two point estimates. The 3.5 and 4 in the above example, without mention that including the error bars would have resulted in 3.5+/- 4 and 4 +/- 4, or whatever it was, so they didn't tell people the significance of the estimates.
|
# ? May 8, 2020 15:53 |
|
|
# ? May 30, 2024 11:52 |
https://twitter.com/JenniferJJacobs/status/1258770798564855817 with one of trump's valets testing positive yesterday, i think this now increases the likelihood to nontrivial that there's an actual outbreak among white house staff
|
|
# ? May 8, 2020 15:55 |
Stickman posted:New cases per day going down under lockdowns and up when restrictions are lifted isn’t exactly some kind of controversial “belief”, nor is uncontrolled outbreaks leading to a lot of death. That’s exactly what we’ve seen in all the data so far. while looking more at this i found their breakdowns by state https://twitter.com/ColumbiaMSPH/status/1258483035168743432 even when you don't have the numbers going down from NY/NJ, it still predicts a relative lull for a couple weeks before the new cases from reopening takes off. this isn't really controversial since we should all know by now that there's a reasonably large presymptomatic period even when you're already infected, but it gives us a good baseline as to what to expect from Georgia and Texas and soon Florida/etc: roughly 2-3 more weeks of a plateau or slight rise, followed by things getting Bad if the reopening isn't undone
|
|
# ? May 8, 2020 16:00 |
|
eke out posted:https://twitter.com/JenniferJJacobs/status/1258770798564855817 There’s no way Trump isn’t immune considering all of the exposure he’s had to it since March.
|
# ? May 8, 2020 16:04 |
|
eke out posted:this isn't really controversial since we should all know by now that there's a reasonably large presymptomatic period even when you're already infected, but it gives us a good baseline as to what to expect from Georgia and Texas and soon Florida/etc: roughly 2-3 more weeks of a plateau or slight rise, followed by things getting Bad if the reopening isn't undone A buddy/ex-colleague of mine working 911 in SW FL came down with Covid 5 or 6 weeks ago has been asymptomatic for over 3 weeks and has still been testing positive from nasal swabs. He’s still out on paid admin leave because he is still probably contagious from viral shedding. THREE WEEKS ASYMPTOMATIC on the tail end of the illness. That’s crazy.
|
# ? May 8, 2020 16:06 |
|
Boot and Rally posted:I thought remembered reading that too and went looking for the answer. Was that this study from China? In this one no outcome was statistically significant, though numerically people did seem to improve. I know it isn't what you're asking, but they mention that they gamed out their test and found they needed 325 people based on some assumption about uncertainty. They only got 237 so the effect would have to be much more dramatic than expected to result in a statistically significant difference in outcomes. It was this more recent study with something over a thousand patients (I’m not sure if the exact number has been publish yet. Their sample had an ~30% decrease in median hospitalization time (15 to 11 days) and also an ~30% drop in deaths (11% to 8%). As you say, the equivalent sample effect sizes are probably a coincidence. The difference in p-values comes down to statistical tests for differences in median stay having more power than tests to detect differences in percentages, especially when those percentages are close to 0 or 100%. And that’s just because a patients length of stay carries a lot more information about the population’s median length of stay than knowing whether an individual patient died or not carries about the population death rate. Stickman fucked around with this message at 16:14 on May 8, 2020 |
# ? May 8, 2020 16:12 |
|
edit: better answer above
evilweasel fucked around with this message at 16:16 on May 8, 2020 |
# ? May 8, 2020 16:13 |
|
Stickman posted:It was [url= https://www.healio.com/infectious-d...tandard-of-care]this more recent study[/url] with something over a thousand patients (I’m not sure if the exact number has been publish yet. Their sample had an ~30% decrease in median hospitalization time (15 to 11 days) and also an ~30% drop in deaths (11% to 8%). Is there a way to know what decrease in death rates could have been measured by a study of this size? It feels like, knowing going in death rates are about 11% in the placebo population, you can figure out what the minimum effect you can detect is. So if the drug cut deaths to a quarter (i.e. 11% -> ~3%), I assume it could detect that, so we effectively rule out a lot of the "home run!" death rate cuts. But if the drug genuinely cut deaths by somewhere around 30% the study is just plain underpowered to detect that, but we can calculate what amount of people we'd need to effectively determine a smaller effect like that.
|
# ? May 8, 2020 16:16 |
|
Assuming the placebo population was the same size as the population receiving treatment ir would have had to roughly cut the mortality rate in half for the results to be significant.
|
# ? May 8, 2020 16:29 |
|
evilweasel posted:Is there a way to know what decrease in death rates could have been measured by a study of this size? It feels like, knowing going in death rates are about 11% in the placebo population, you can figure out what the minimum effect you can detect is. So if the drug cut deaths to a quarter (i.e. 11% -> ~3%), I assume it could detect that, so we effectively rule out a lot of the "home run!" death rate cuts. But if the drug genuinely cut deaths by somewhere around 30% the study is just plain underpowered to detect that, but we can calculate what amount of people we'd need to effectively determine a smaller effect like that. For mortality rate, the stats 101 statistical test would just be difference of the empirical mortality rates in the treated and untreated populations. Looks like they had 1000 patients, if it's split 500/500 treated/untreated, I find the standard deviation is 4%, so something around 3% = 11 - 2*4 would have been declared significant. For a statistical power calculation to detect an effect size of 3%, 80% of the time they need ~13000 total patients to detect an effect size of 3% (at p=0.05). Note: i am not a statistician so smarter people may say different I think Simplex above assumed 2000 total patients prokaryote fucked around with this message at 17:57 on May 19, 2020 |
# ? May 8, 2020 16:45 |
|
https://twitter.com/carlquintanilla/status/1258772141195100161
|
# ? May 8, 2020 16:57 |
|
No loving way he sticks to this.
|
# ? May 8, 2020 17:20 |
|
Solkanar512 posted:No loving way he sticks to this. Ideally this will result in Congress presenting him with a must-sign bill that was developed entirely without White House input
|
# ? May 8, 2020 17:22 |
|
Solkanar512 posted:No loving way he sticks to this. But if he does? Protests. Riots.
|
# ? May 8, 2020 17:25 |
|
MightyBigMinus posted:the thing is, there's no way that random american regional ISPs, especially more rural ones, can handle tens of thousands of concurrent video streams. they already choke if too many people try to watch game of thrones on a sunday night at once. Most video companies follow Netflix and install local servers downstream of network bottlenecks. I am most likely watching netflix from a server located at my ISP hub or in some cases the last mile cabinet. These boxes often work on "Request and Rebroadcast" mode so that only 1 copy of game of thrones needs to go through the network's pipes and after that its a few small data packets to control content access and transmission. Can't say HBO did this, but Netflix learned this a long time ago and partnered with most ISPs to reduce the expensive traffic. We mostly learned about the issue due to a few ISPs refusing to work with Netflix to create artificial congestion on the interconnects.
|
# ? May 8, 2020 17:46 |
|
evilweasel posted:Is there a way to know what decrease in death rates could have been measured by a study of this size? It feels like, knowing going in death rates are about 11% in the placebo population, you can figure out what the minimum effect you can detect is. So if the drug cut deaths to a quarter (i.e. 11% -> ~3%), I assume it could detect that, so we effectively rule out a lot of the "home run!" death rate cuts. But if the drug genuinely cut deaths by somewhere around 30% the study is just plain underpowered to detect that, but we can calculate what amount of people we'd need to effectively determine a smaller effect like that. Yes. Sometimes the math shows "we need 10000 events to see the difference between 1% and 2%. Since only 200 are sick we can't tell the difference." That doesn't mean a test with 200 people is useless, after 20 similar studies all over the world it might be possible to aggregate the data and see something significant.
|
# ? May 8, 2020 17:55 |
|
tk posted:Giving them 30 days notice* is just formalizing the “asking them to leave” portion. To actually evict them (lock them out / have the police come remove them) you need a court order. That’s likely to be more difficult than normal right now, and in the meantime you’re living with the rear end in a top hat that you’re trying to evict. This would be the kind of thing where it would be extremely useful to hire a specifically well connected lawyer if at all financially feasible since it's actually a matter of life and death
|
# ? May 8, 2020 18:09 |
I don't see why continuing to let citizens suffer and not support them is good for the economy when it's already crippled and then you risk endangering those who can work even more. Too bad there isn't international police who can arrest everyone who is just working to save their bottom lines.
|
|
# ? May 8, 2020 18:10 |
|
I'm surprised this article hasn't been getting more attention. Published in the NEJM on May 1. https://www.ncbi.nlm.nih.gov/pubmed/32356626 Cardiovascular Disease, Drug Therapy, and Mortality in Covid-19. quote:RESULTS: I guess ACE Inhibitors just aren't sexy enough to grab headlines? They're already incredibly widely used and have clear mortality benefit for various disease states (post-MI, heart failure, CVD, CKD). The way it's worded is a little strange, because the point of the study was to investigate if ACE-Inhibitors increased the risk of death (as some suspected due to its mechanism of action and how covid enters cells), but what they found was that it significantly decreased the risk of death from 6.1% down to 2.1%. Statins also significantly reduced the death risk to 4.2%, no other common drug class investigated showed significance in effect on death rates. Sure it's a retrospective study without good matching, but a reduction in death rate from 6% down to 2% is a lot more impressive than a lot of other magic bullets everyone is grasping at. In fact, with matching its benefit might be even more pronounced as patients on ACE Inhibitors are more likely to have existing cardiac conditions than an average patient. It should be noted that the suggested protective benefit here is the patient already being on an ACE-Inhibitor when hospitalized resulting in fewer deaths, not necessarily implying that starting already hospitalized patients would be beneficial. It also says nothing about the the effect of being on the drug on likelihood of contacting covid or it progressing to the point that hospitalization is necessary. It also matches well with an increasingly popular theory on how covid causes some of it's worst damage in severe cases. The covid virus enters cells via the ACE2 receptor, which itself is an enzyme responsible for converting the extremely potent vasoconstrictor Angiotensin2 into other forms of angiotensin which tend to be vasodilators. It seems likely that when covid infects the target cells via the ACE2 receptor it either cleaves away the ACE2 enzyme from the cells and/or causes a downregulation of expression of the enzyme locally in an effort to block viral entry. The net result is a local increase of Angiotensin2, causing local vasodilation, and progressive damage to the pulmonary and circulatory systems. The drug class of ACE-Inhibitors work by blocking the conversion of Angiotensin1 into the potent Angiotensin2. It's easy to grasp how this would help ameliorate some of the most damaging effects of local decrease of ACE2 enzymes. There are also studies underway on whether a recombinant form of the ACE2 enzyme could be administered locally to beneficial effect.
|
# ? May 8, 2020 18:11 |
|
cage-free egghead posted:I don't see why continuing to let citizens suffer and not support them is good for the economy when it's already crippled and then you risk endangering those who can work even more. It's a bit like the fire department rolling up on your burning kitchen, hosing it down for 30 seconds, then driving around to the other side of the house to wait and see if the fire spreads in that direction.
|
# ? May 8, 2020 18:18 |
|
Trying to starve the states to force them to reopen.
|
# ? May 8, 2020 18:20 |
|
https://twitter.com/mattyglesias/status/1258793791512809473
|
# ? May 8, 2020 18:23 |
|
RandomBlue posted:Trying to starve the states to force them to reopen. That's what it seems like, yes. It's going to horrifically backfire. On who I don't know though. Ice Phisherman fucked around with this message at 18:27 on May 8, 2020 |
# ? May 8, 2020 18:24 |
|
Yes and this is why we are all in serious trouble
|
# ? May 8, 2020 18:28 |
|
evilweasel posted:Is there a way to know what decrease in death rates could have been measured by a study of this size? Well, in life insurance the rule of thumb is around 1000 occurrences for full credibility in experience studies. I would guess epidemiologists have a more subtle model though.
|
# ? May 8, 2020 18:39 |
|
eke out posted:this isn't really controversial since we should all know by now that there's a reasonably large presymptomatic period even when you're already infected, but it gives us a good baseline as to what to expect from Georgia and Texas and soon Florida/etc: roughly 2-3 more weeks of a plateau or slight rise, followed by things getting Bad if the reopening isn't undone Here in Georgia my job just called me back in to re-open our private corporate dining facility in a week, and my mom is still working at a grocery store and living in my attic, so I am just absolutely thrilled at these projections. If unemployment in Georgia actually worked and could extend maybe six months I'd just loving quit.
|
# ? May 8, 2020 18:46 |
|
Captain Invictus posted:You were discussing this lovely roommate earlier in the thread, right? Did you confront him like was suggested initially, or boot him out since he's not actually on the lease iirc? If not, why the gently caress haven't you? You'd rather continue to be nonconfrontational and die from his hubris? Yeah, that was me! I did speak with him after the initial discussion and he had agreed to not bring his gf over. That was about a month ago. So fast forward to now and I guess he is beginning to get antsy. I plan to speak with him again because as you can see from the last discussion, he seems to be at his personal quarantine limit, and usually to each their own, but in this case I’m once again scared since I’m so high risk and it’s really quite stressful to have to fight for my life like this.
|
# ? May 8, 2020 18:48 |
|
It appears that the meatpacking industry is garbage worldwide. https://twitter.com/danieldrepper/status/1258480850829746176 EDIT: Over 10% of the workers at the plant are infected and the company wants to stay open. Similar long-term issues as the U.S. with eroding health and safety and dwindling inspection coverage. EDIT EDIT: Sorry about the German! Dick Trauma fucked around with this message at 18:54 on May 8, 2020 |
# ? May 8, 2020 18:50 |
|
mark cuban... good? no, he bad. BAD!
|
# ? May 8, 2020 18:51 |
|
Dick Trauma posted:It appears that the meatpacking industry is garbage worldwide. For those who don't speak German, Google translate. quote:Despite the strong outbreak of Corona among employees, the Westfleisch company does not want to close its plant in Coesfeld. By Thursday afternoon, 129 employees had tested positive:
|
# ? May 8, 2020 18:52 |
|
enraged_camel posted:mark cuban... good? I like that he did the thing because I think the outcome is ultimately helpful - but I kind of want to understand his motivations behind doing the thing in the first place.
|
# ? May 8, 2020 19:00 |
|
PyRosflam posted:Most video companies follow Netflix and install local servers downstream of network bottlenecks. I am most likely watching netflix from a server located at my ISP hub or in some cases the last mile cabinet. this works great when you're only streaming content one direction, not so well in Zoom's usecases. The upstream will always be the issue, unless you're some sick lunatic like me with symmetrical gigabit fiber on an unsaturated mountaintop network. Residential internet has for years been designed to serve a much bigger downstream demand than upstream; streaming and the demands of modern gaming and FaceTime has changed this, but COVID is definitely putting more stress on it. Cabbages and VHS fucked around with this message at 19:05 on May 8, 2020 |
# ? May 8, 2020 19:02 |
|
FreelanceSocialist posted:I like that he did the thing because I think the outcome is ultimately helpful - but I kind of want to understand his motivations behind doing the thing in the first place. If people are dying and they go into lockdown again, rich people lose money.
|
# ? May 8, 2020 19:04 |
|
Jaxyon posted:If people are dying and they go into lockdown again, rich people lose money. yes but what if they just don't go into lockdown again? Not everywhere is NYC that's dense enough to need port-a-morgues; I can well imagine that wide swaths of the country just accept a bunch of excess mortality for a while, because at worst it's probably going to ~double-triple the usual annual mortality (that would be killing 1-2% of the population of the country) and it will skew heavily towards nursing home populations and similar. This isn't Captain Trips. I hate to say it but I do not think the handful of dyed in the wool base people I still have any contact with, would necessarily be swayed by their own grand/parent dying. "Well, he had a good life, and he had heart disease, and he was 67, so....." Of course, this could lead to insanity like state quarantines in blue states on red state visitors, and the like. I'm not even sure if that's legal but I'm already seeing people in my feeds in places like Washington calling for it. Sweden is now saying "wow we're sure surprised how many more people have died here than Denmark", but they're still not changing their stance or response from what I can tell.
|
# ? May 8, 2020 19:09 |
|
Dick Trauma posted:It appears that the meatpacking industry is garbage worldwide. Difference to the US is: A: The plant is now shut down; B: After the first couple of cases they tested everybody in the plant, so they caught most of those cases early, C: They are now testing all employees of all meatpacking plants statewide, and D: The county that plant is located in is not allowed to relax lockdown restrictions that are being relaxed elsewhere.
|
# ? May 8, 2020 19:10 |
|
Cabbages and Kings posted:
Guam, Alaska, Hawaii currently have two week quarantines for incoming visitors. I'm not sure if anyone in the contiguous US has something though.
|
# ? May 8, 2020 19:22 |
|
3rdEyeDeuteranopia posted:Guam, Alaska, Hawaii currently have two week quarantines for incoming visitors. I'm not sure if anyone in the contiguous US has something though. They do but it’s impossible to enforce.
|
# ? May 8, 2020 19:23 |
|
Pittsburgh is going to start easing the stay at home order on the 15th We've been one of the few cities that took this seriously and got ahead of it and it's good to know we're now at the spot where we can slowly start reopening responsibly. Unlike say, Atlanta which is going to be a horror show
|
# ? May 8, 2020 19:26 |
|
PyRosflam posted:Most video companies follow Netflix and install local servers downstream of network bottlenecks. I am most likely watching netflix from a server located at my ISP hub or in some cases the last mile cabinet. These boxes often work on "Request and Rebroadcast" mode so that only 1 copy of game of thrones needs to go through the network's pipes and after that its a few small data packets to control content access and transmission. yea cool story bro now invent an appliance that can cache realtime multi-node video chat
|
# ? May 8, 2020 19:26 |
|
|
# ? May 30, 2024 11:52 |
|
Shimrra Jamaane posted:They do but it’s impossible to enforce. Exactly. See: Maine.
|
# ? May 8, 2020 19:31 |