Mormad said:
Daviewolf83 said:
I put this quick chart together to show the daily hospitalizations, plotted on a logarithmic scale. The purpose, as I have explained in past posts, is so you can better tell the rate of change in a graph. I wish I had more recent data, but NCDHHS has only provided the daily hospitalization numbers through August 9.
As you can see from this chart, the rate of daily admissions is slowing down from July. I need to see more recent data to draw any conclusions, but I am optimistic it is showing that the peak for hospitalizations is not too far from now. I will provide an update to this chart on Tuesday, when NCDHHS should provide their next update.
I so appreciate your time and your graphs and interpretations, Davie. I do have some questions about this graph.
1. The y-axis jumps from 1 to 10 to 100 to 1000. It's exponential. So, is the graph flattening once admissions are over 100/day giving me a false sense of security, since a small rise corresponds to a much larger n? Or am i looking at it wrong?
2. How accurate is the data? We've questioned the data the entire pandemic, and depending on personal views, people here and elsewhere feel the data is anywhere from completely fabricated, to inaccurate due to delays in reporting, to inaccurate due to over or under reporting, to completely accurate and trustworthy. You and i know it's somewhere in between. When this came out the 9th, how accurately did it reflect data in real time since we know numbers change very rapidly at times, and at those times hospitals are much less accurate or timely in their reporting?
3. Locally, we've seen a huge uptick in admissions. Others, like packPA report similar issues where they are. You follow the state much more closely than i do. My local graph will likely see a big upturn if the numbers reported are accurate and timely when the data is released Tuesday. We've been outliers before here. Do you expect this across the state, and if so, how long are you predicting to actually peak and will that prediction change if the numbers have rapidly risen? I've seen your predictions and I've read Monica's, but not sure what analytics you're using to determine your prediction. But i have learned to trust your insight here and I'm trying to better understand so i can help lead decision making here.
All good questions. I will try to give you some answers and insights, plus I will provide another graph I find provides an interesting cut of the data.
1. The flattening of the graph means the rate of change in the daily increase in admissions to the hospital is stabilizing. It is a base 10 logarithmic plot of the y-axis. I do this to examine if the initial surge is stabilizing. What I have seen in the past that this flattening precedes a peak in the data. It is just a way to try and predict if we are coming close to a more stable increases and a potential peak in hospitalizations. Think of it more as a reflection of the change in the slope of the daily admissions curve. This hospitalization curve has been much more steep than the Winter Wave curve which I believe reflects the increase the Delta's ability to infect people.
2. The accuracy with anything coming from NCDHHS has always been suspect which is why I prefer this data. I am plotting are the daily hospital admission numbers that show up in the Hospitalization Demographics portion of the Covid dashboard. This data does have a built in lag and it typically lags by about 6 days. For example, the last data reported are admissions on 8/9. I do like this data, since it is more accurate than the daily hospitalization numbers reported. The daily hospitalization numbers have a few issues and the biggest of these is the daily fluctuation in the percentage of hospitals reporting. For example, over the last 14 days, the average percentage of hospitals reporting has been 96%, but during this period it has ranged from 91% to 97% of hospitals reporting. This is why I look at daily admissions more than I do the current population.
The other issue with the total number currently hospitalized is this - it is sticky data. In other words, there are factors involved that cause these numbers to not entirely reflect what is happening. How long is the average stay in the hospital? What is the standard deviation for this length of stay? How stable is the standard deviation? Can admits outpace the rate of discharge? When they do, what is the impact on the total hospitalized. For example, the number of those hospitalized my be increasing because the rate of admissions over a few days may be outpacing discharges. It only takes a few days of this to really bump up the total population. Due to all of these unknown variables (unknown to me, but I am sure others in the hospital system have access to them), I am using daily admissions.
The other reason I use daily admissions of hospitals is due to the inaccuracies in the daily case numbers reported. The daily case numbers are dependent on the number of tests being administered and people's access to testing. We also know the reported cases do not capture all active cases. They only capture the cases of people who decide to get tested. The actual number of cases is definitely higher than what gets reported. Given these variables, I find daily hospitalizations to be a much more meaningful number to measure and report. It does not suffer the same issues as case reporting and it does not have the stickiness of the daily hospitalized number.
3. Both you and PackPA are in unique locations in the state, as you can see from the first graph below. As you an see, your hospital group is in an area seeing much higher rates (meaning slope of curve) than many of the other hospital groups in the state. I can see how this would impact your thinking about what is happening and it is somewhat atypical for the rest of the state. There are some areas of the state where it appears the hospitalizations are starting to flatten out, but this is not true for our area. Your curve (THPC) is the second highest in the state, trailing only the MHPC group. I am not entirely certain, but I believe PackPA may cover some of the hospitals in the MHCP group. These two groups are most definitely seeing more hospitalization than any of the other groups. The third highest hit group (MCRHC) represents an area of the state with some of the lowest vaccination rates, so it should not be surprising it is third. The area of the state where I live is the most heavily vaccinated area and this is reflected in the curve. Despite having the largest populated city (Raleigh), the hospitals in this group (CapRAC and DHCP) have much lower hospitalizations (even when you combine them) and also appear to be the ones that are flattening out.
As far as my predictions are concerned and as you can see from the chart below, I do believe large portions of the state are starting to peak. I am using last year's Summer Wave curve to try and approximate when we will hit the peak, but the Delta variant and the fact it is more infectious, has cause me to slide my predictions more to the right than I originally anticipated. The mask mandates going into effect are also giving me something to consider, since it is possible they could prolong the Summer Wave, beyond where it would end up if we were to let it behave naturally. What we saw when Delta hit the UK is a very sharp increase (something we have seen here as well), one with a much higher slope than the Winter Wave, but also a much more rapid decrease. In other words, the faster it climbs, the faster it falls. The mask mandates my slow the rise, but they may slow the spread, but the spread will still occur. As you have said before, the virus is going to virus.
Last Summer, the wave peaked in early August and by the first part of September, it was over. This year, the wave started later (roughly mid-July) and this is why I have thought it would peak by the end of August - roughly delayed by two weeks from last Summer. If you look at the hospitalizations curves for many areas of the state, it appears this prediction could hold. For your area and and those in the southwestern, western and southern areas of the state, it could happen a little later, but the speed by which it ends could be faster.
4. As a bonus chart, I have updated the chart that shows the estimated (I assume cases are a valid case for 14 days) percentage of Covid infected people who are in the hospital, mapped against total hospitalizations. As you can see from the chart, the estimated percentage of those hospitalized is almost flat, with a value of 4.2%. This percentage continues to fall, even though the hospitalizations are increasing. This is another indicator I use to reflect how much more severe the Delta variant is than the other variants. Based on this declining percentage, my opinion is it is not more severe and it does not lead to greater hospitalizations from severity. The increase in hospitalizations are actually lower, from a percentage basis, than you would expect to see, given its increased ability to infect people.
