Excess Deaths has become a popular way, globally, to assess the overall changes in reported deaths this year relative to what was expected this year. We are also watching the data with interest in South Africa.

In layman’s terms, excess deaths are the difference between reported deaths and expected deaths. For the reporting period 6 May to 7 July 2020, the total excess deaths amount to 3,159 deaths.

During the same period, the reported COVID-19 deaths in South Africa amounted to 3,354 deaths. This suggests that the reported deaths and excess deaths are in broad alignment.

We were astounded, however, when the SAMRC published a report in which it estimates recent excess natural death mortality for South Africa at almost 11,000. It suggested that this meant that actual COVID deaths were 3 times the “official” 3,000+ number.

They then used this "anomaly" to explain the gap between the SACMC death forecast and the actual numbers.

Panda first caught wind of this in the Western Cape modelling update presentation.

Business Day subsequently headlined the predictable story. https://www.businesslive.co.za/bd/national/health/2020-07-16-mrc-finds-covid-19-has-led-to-three-times-more-deaths-than-official-tally/

Our first interpretation was that lockdown deaths, rather than COVID deaths, were starting to come through.

On investigation, it seems that SAMRC calculated the excess deaths as the difference between actual deaths and the LOWER BOUND of their estimated range (rather than the best estimate itself). They present expected deaths in a range as shown below.

When comparing the actual deaths versus the lower bound, you get total excess deaths of almost 11,000 and it can be illustrated as shown.

The rationale given by SAMRC for using the lower bound rather than the best estimate is that during the hard lockdown months (April and May), the actual reported deaths were tracking the lower bound. They, therefore, assume that this lower bound is the new baseline.

The data was suspicious. Why would natural deaths drop off immediately upon implementation of lockdown? And where was the downward adjustment for “deaths with” COVID, which has been one of the most noticeable features of this epidemic? Was zero really the best they could do?

Using the lower bound as their "best estimate" is an unreasonable assumption, and the observed lower natural deaths during the hard lockdown are much better explained by reporting delays.

It was difficult to access the state bureaucracy during hard lockdown to report deaths. This doesn’t mean the deaths weren’t happening, but rather that the reporting was delayed in some cases during the hard lockdown.

This is supported by the observation that in the first quarter of 2020, actual natural deaths were running approximately 2.5% HIGHER than their best estimate. There is no convincing hypothesis which can explain why lockdown itself saw natural deaths plummet in the early months.

This overstated excess death number appears to be a deliberate attempt to inflate the excess death numbers and attribute the difference to COVID-19 and therefore imply that COVID-19 deaths are currently being significantly understated.

If this is correct and the researchers knew this, it is deeply problematic. If they didn’t know this, serious questions remain.

Either way, their estimate bears scrutiny, all the more so because the head of the modelling team whose forecasts had been blown out of the water within days of their publication is the brother of a member of the SAMRC team who produced the report of the excess death.

Is this a case of bad science being used to rationalize bad science elsewhere? We’re keen to hear opinion and for Business Day to do some investigative journalism.

It’s worth pointing out too, that PANDA does expect true excess deaths to outstrip COVID-19 reported deaths over time since emerging data already are pointing to the potential of “deaths due to lockdown”.

In the coming months, when excess deaths start to outstrip COVID-19 deaths, do not make the mistake of attributing those to COVID-19, but rather attribute them where they below - to the lockdown itself.