Credit CDC
While these cases appear to be sporadic and likely due to bird-to-human transmission - and there are no indications of community spread - the reality is our ability to detect a community outbreak in its early stages is limited.
We've discussed this often over the years, but it requires more than a little luck to detect sporadic cases, like the ones reported this week, even in countries with well-equipped and functioning public health systems.
First, an infected person must become sick enough to seek medical care, which - depending on the flu strain - may exclude > 90% of infections. They then must have access to modern medical care, an option not available to > 40% of the world's population, and then be lucky enough to be properly tested for novel influenza.
We've seen numerous studies (see here, here, and here) suggesting that only a small fraction of novel flu (or MERS-CoV) cases are ever detected, and reported, by surveillance. Perhaps as few as 1% or 2%.
While there have been only a handful of clade 2.3.4.4b H5N1 infections reported in humans over the past 18 months, the real number of spillover events is unknown. The good news, however, is we've seen no signs of sustained of efficient community spread.
But the obvious question is, how long would it take for us to recognize community spread if it were occurring?
That obviously depends on where an outbreak occurs, its severity, and how fast it is spreading (R-Value). It would presumably be easier to detect in places London, or Tokyo, or New York City than in Outer Mongolia, or Ethiopia.
But according to yesterday's UKHSA Technical Briefing #3 on H5N1, even in a developed country like the UK, it might take weeks.
Yesterday we looked at the UK's three working `pandemic scenarios', but today we'll look at an analysis of the UK's ability to detect community transmission of a novel flu virus. As you'll see, the Expected time to detection (TTD) and the cumulative number of infections (CI) before we'd know there was community spread are sobering.
Updated 29 March 2023
(EXCERPT)
3.3 Time to detection
The above scenarios are used to derive estimates of time to detection (TTD) of avian influenza under various testing scenarios. This is based on the assumption of sustained person to person transmission, and no cross-immunity from previous influenza vaccination or infection.
Disease growth in such a situation can be simply modelled as daily growth equals R to the power of 1 over the serial interval (SI). To account for variation in the SI, it is sampled from a Weibull distribution, with mean and median of 3 days, and growth simulation run 100,000 times to yield an expected growth curve and confidence intervals.
The TTD is determined for a number of testing scenarios, estimating the point at which the outbreak will have grown large enough we expect to detect a case. This depends both on the growth rate and the severity (for hospital-based testing). Testing scenarios modelled are:
Figure 7 shows the expected TTD in each scenario with an R of 1.2, with Figure 8 showing the equivalent when R equals 2. Day 0 is the day the first case becomes infected. The TTD includes delays for: testing (2 days); admission to hospital (9 days); and admission to ICU (a further 2 days) as appropriate. For simplicity the community testing is assumed to take place in the mild severity scenario.
- sampling of asymptomatic individuals in the community with coverage of 1 in 1,000 people tested daily
- sampling of asymptomatic individuals in the community with coverage of 1 in 200 people tested daily
- testing of all hospital admissions (with influenza-like illness (ILI) symptoms)
- testing of admissions to intensive care units (ICU) only (with ILI symptoms).
With a lower R (and hence longer doubling time) it may take twice as long to detect an avian influenza case using lower coverage community testing, as it does to detect when testing hospital admissions. Therefore, there may be substantial numbers of infections in the community by the time the first case is detected.
Tables 3a and 3b show the expected number of infections by the time each testing scenario detects the first case. Being able to test in multiple settings improves the speed of detection across a range of plausible R values.
The UK now estimates it would likely take between 3 and 10 weeks before community spread would become apparent to authorities, after anywhere between a few dozen to a few thousand community infections.
And this is for the UK. The TTD in many less-developed regions would presumably take longer.
A reminder than anything we say about the current threat from H5N1, H3N8, MERS-CoV, or any other pandemic threat must carry an implied asterisk.
A disclaimer that says, ` * based on available, and likely incomplete, information. . . '.