THE FIVE-DAY rolling average of Covid-19 cases in Ireland currently stands at 5,768 cases per day.
Although most people might be more aware of daily case numbers, the five-day average is important because throughout the pandemic, health officials have advised that averages over a specific period are the best way to track the epidemiological situation.
Another important metric both here and across the EU is the 14-day incidence rate, which tracks the average number of cases over a two-week period per 100,000 people in a given country.
The five-day average is simpler and, as it suggests, reflects the average number of daily cases recorded over the past five days. Add five days of cases together and divide by five.
Despite the emphasis on day-to-day cases, reading into them is not nearly as useful as these other two metrics.
Analysing a country’s Covid-19 situation using daily numbers can be an easy trap to fall into, because those figures have become part of everyday discussions over the past 20 months.
Much like the weather, however, looking at trends is better than focusing on a single day.
This advice was repeated by the Department of Health last weekend when 5,124 new cases were confirmed on Sunday and 7,333 new cases were confirmed on Saturday.
The huge increase on Saturday prompted questions on social media about whether the jump was down to a backlog – that is, whether the day’s cases were a fair reflection of the situation or whether it the numbers were distorted because of delays in reporting.
So is there anything to this?
‘Processing times’
On Sunday, the day after questions were raised about the country’s Covid-19 figures, the Department of Health explained how figures across multiple days can provide a better insight into analyses of Ireland’s case numbers.
In its notification to media confirming new cases, it said:
Day to day variation in the reported number of cases of Covid-19 does not impact on key trend analysis. Analysing over a longer time period such as a week or 14 days smooths out any variations introduced by processing times and provides a more stable picture of the trend.
The reference to “processing times” is worth focusing on because it refers to a delay in the system of how case numbers are notified – from the time swabs are taken, tested and ultimately reported as positive cases.
Before figures are reported to the media via the Department of Health, the Health Protection Surveillance Centre (HPSC) is notified of Covid-19 cases via an information system called Computerised Infectious Disease Reporting (CIDR).
CIDR is not just used for Covid-19 but for all notifiable diseases. Notifiable diseases are diseases which must legally be notified to the CMO when they are detected.
But because notifying these diseases involves an official process, there is a lag between the time when positive Covid tests are returned from the lab and when they are uploaded to CIDR.
And this processing lag means there is a delay between the time that a swab returns a positive result in a lab and when that same test is reported as a ‘positive case’.
Some of the process is automated, but there is also a degree of human involvement and workflow in verifying that the number of cases being processed is correct.
The HPSC says Covid-19 notifications to CIDR take place “typically one day” after positive tests are uploaded to the publicly available Covid Care Tracker, but that this time lag can “increase or decrease”.
The differing schedules for uploading results to the CIDR and the Covid Care Tracker are due to a number of reasons.
The first is to prevent de-duplication, that is, to stop two positive tests from the same person being accidentally recorded as two separate cases.
This involves a level of clinical judgment and is not an automated process. In general, about 97% of positive tests ultimately become official cases of Covid-19 and the remainder do not, cases can also be denotified following further investigation.
The second reason for the differing schedules is down to what the HPSC says is a “more manual” two-step process for uploading results to CIDR.
“Typically, the daily number of notifications to CIDR lag approximately one day behind the number of positive SARS CoV-2 results uploaded to the COVID Care Tracker,” the HPSC says in an FAQ on its website.
“However, changes in the timeliness of the processing steps described above can increase or decrease this lag.”
In effect, all of this means that swabs taken on a given day may not be reflected in daily case numbers until a number of days later.
Publicly available data
However, because positive test numbers are publicly available via the Covid Care Tracker, people may end up seeing positive tests results sooner than those tests are notified to the HPSC and then reported by the Department of Health.
This means that people looking at positive test numbers can see if officially notified cases are higher or lower over a period of days.
If notified cases are lower than positive test numbers, the reporting lag means there is a build up in the number of cases that are to be processed and ultimately notified.
This build up of cases is acknowledged by the Department of Health, but the department does not consider such build ups to constitute a “backlog” unless there is specific and irregular delays beyond the ordinary processes.
Of course, if the overall number of cases is increasing, then there will be a greater number of positive tests also waiting to be officially notified and then reported.
For example, six weeks ago when over 5,400 cases were confirmed on a Friday it was pointed out that the official cases notified in previous days had been lower than the number of positive swabs taken.
When this was pointed out, the Department of Health tweeted that there was “currently no specific delay in processing time or backlog of laboratory diagnoses”.
Similar questions were asked last Saturday when 7,333 new cases were reported and it was noted by people who focus on positive swabs that swab numbers had been higher than cases in the previous days.
But Professor Cathal Walsh of NPHET’s Irish Epidemiological Modelling Advisory Group pointed out this was not due to a specific backlog.
Speaking to The Journal, Walsh outlined that the instances recently highlighted were “part of the normal workflow”.
He adds that there is a “batch effect” when it comes to the processing of cases and that these batches may be cleared towards the end of the week.
This he compares to the “sawtooth model”, which is also seen in manufacturing processes where a demand is cleared at once. It’s a shape that can be clearly seen by looking at the swab and case data:
“It is to do sometimes with the availability of people and sometimes you see it really ramping up towards the weekend, because people want to close out the workload, so there often is additional activity towards the weekend,” Walsh said.
Sometimes people see a big spike on a Saturday and may think that it’s an effect of a lot of work on a Saturday, but they’ve actually been processed the previous day. Early on in the pandemic we saw the number of swabs the labs coped with increased on a Thursday, because people work late on a Thursday. There are various human factors at play.
Earlier in the pandemic, the “weekend effect” became common parlance due to what was seen as artificially lower cases numbers early in the week.
It is still common that a portion of tests may build up throughout the week before they are processed towards the end of the week.
It’s arguably a matter of semantics whether a normal build up of positive tests should be considered a ‘backlog’, but the Department of Health has acknowledged when an out-of-ordinary delay in cases occurs that there is a backlog.
At the turn of last year, cases were increasing so rapidly that the Department of Health reported an ongoing backlog of cases that was not cleared until 7 January. At that point the Department said the reporting of cases to CIDR had “caught up”.
Walsh says that the HSE and and the HSPC are closely related, but that the Department of Health is separate and would only be informed of a backlog if there was a delay that was out of the ordinary.
“Last December/January I think the Department knew there was a large number of cases that hadn’t been fully processed,” he said.
“In the cases that we’ve seen raised recently, they’ve been part of the normal workflow and the Department of Health wouldn’t have been aware of a delayed reporting of some cases, or a change in the average delay.”
In an announcement last night, the Department of Health said that due to what’s expected to be a large volume of cases over the coming period, daily case numbers are instead going to be based on positive swabs uploaded to the Covid Care Tracker the previous day and not officially notified cases.
This essentially removes the potential that there is a build up of positive tests over the Christmas period, with the department saying that these figures will be “provisional” before cases are officially notified to CIDR.
What does all this mean?
The disparity between the number of positive swabs and reported cases is usually small, so it does not require that they be treated differently to any significant degree.
However, it has been pointed out that a build up in positive tests over a number of days leads to concerning spikes on a single day.
While this may be the case, the build up of tests also means that cases are artificially low on the days immediately before the spike, potentially contributing to less concern than it is warranted on those days.
This is why looking at averages is considered the only accurate way of judging the level of cases, because days when cases are artificially low are balanced out when they may be artificially high.
Walsh agrees, saying this is the “point to emphasise”.
“It’s naive in the extreme to look at a single day’s cases, because you’ve got everything like the day of the week effect to different test centres and their activities, there are a huge number of factors that vary day-to-day,” he says.
And what about the unproven theory that there’s a conspiracy within the Department of Health to purposely hold off on processing tests to coincide with the announcing of Covid-19 restrictions?
Walsh laughs at the idea that labs across the country would even be able to coordinate such a plan, such is their workload.
In reality, he says the theory “demonstrates a poor understanding of underlying epidemiology and data challenges that exist” in tracking cases.
TL;DR? If you’re surprised by apparent spikes or dips in cases, just look at average case numbers over a couple of days for the bigger picture.
To embed this post, copy the code below on your site
have your say