New (updated) analysis confirms early insights into Omicron variant's lowered tendency towards serious illness

Authors
Shirley Collie (Chief Healthcare Analytics Actuary, Discovery Health), Lebohang Radebe (Data Scientist, Discovery Health), Tommy Chen (Actuary, Discovery Health) and Michael Cohen (Actuarial Analyst, Discovery Health)

New findings, in summary

This analysis (spanning data from 15 December 2021 to 14 February 2022) compares the odds of developing serious illness when infected with the Omicron variant versus infection with the Delta variant (which drove South Africa's third wave of infection in July 2021).

Compared to the Delta-led wave, the Omicron-led wave resulted in:

• 44% lower odds of COVID-19 hospital admission
• 62% lower odds of COVID-19 ICU admission

Four dominant variants have driven South Africa's four waves of infection from early 2020 to date

The following table shows the start and end dates selected for the dominant period of each variant of concern for the purposes of this comparison. These periods were established through research by experts involved in the Network for Genomic Surveillance in South Africa .

Periods over which four distinct SARS-CoV-2 variants were dominant in South Africa

Dominant variant	Start date	End date
D614G	1 March 2020	19 November 2020
Beta	20 November 2020	16 May 2021
Delta	17 May 2021	14 November 2021
Omicron	15 November 2021*	Analysis conducted up to 15 February 2022

*This variant was first reported to the World Health Organization on 24 November 2021; however, detection of first infections of the variant date back to 9 November 2021.

Recap of our initial analysis - first three weeks of the Omicron-led, fourth wave of COVID-19 infection in South Africa

Over the course of December 2021 and into early 2022, the highly transmissible Omicron variant replaced the formerly dominant Delta variant in South Africa.

Omicron subsequently drove an exponential rise in COVID-19 cases in South Africa's fourth wave of infection.

At the time, given the increased transmissibility of the Omicron variant compared to prior COVID-19 viral variants, it quickly became important to predict the severity of illness this variant might cause, and the consequent COVID-19 hospital admission rate that could result over the fourth wave.

Discovery Health's actuarial team investigated the early data available on disease severity linked to Omicron infection in a group of 380,430 members of Discovery Health administered medical schemes.

• With more than 3.7 million medical scheme members under administration, Discovery Health has extensive claims data and private health system utilisation data and is uniquely positioned to generate at-scale, real-world insights into the impact of COVID-19.

This initial research was carried out as part of a wider analysis - the first at-scale, real-world analysis of Omicron's impact, reported on here . This analysis included early, preliminary data and clarified that the insights and the outlook shared could change over the course of the fourth wave.

Let's take a look at picture that emerged in the weeks that followed. Our new data extends the initial three-week analysis to a three-month period. We can now provide a reflection on the severity of disease caused by the Omicron variant up to 15 February 2022, so covering South Africa's fourth wave peak of infection and subsequent decline.

Increased transmissibility of Omicron variant evident among medical scheme members

When the Omicron variant first emerged, initial reports indicated greater transmissibility than past dominant variants.

We saw that compared to prior months, from 15 November 2021 to 15 February 2022, there was a significant increase in the number of confirmed positive COVID-19 infections among members of medical schemes administered by Discovery Health.

• The seven-day rolling test positivity rate was 39% on 9 July 2021, during the peak of the Delta wave, relative to a peak of 46% on 26 December 2021, during the peak of the Omicron wave.
• The Omicron wave peak case rate surpassed the peak seven-day rolling average of daily new cases seen in the Delta wave amongst scheme members. And the seven-day rolling average went from a peak of 3,679 daily new cases (and 23,284 nationally) on 13 December 2021 down to 600 daily new cases (and 2,531 nationally) on 15 February 2022.

Comparing severity of clinical outcomes - Omicron variant versus Delta variant

We set out to investigate how the severity of clinical outcomes recorded during the Omicron period of COVID-19 infection compared to outcomes recorded in the previous wave of COVID-19 infection, driven by the Delta variant.

Our investigation considered only adult medical scheme members (aged 18 years or older).

Our methodology

• We use 'admissions as a percentage of documented infections' as a proxy for clinical severity.
• How did we define a COVID-19 admission? As any admission likely to be for COVID-19 treatment, such as pneumonia, within a 14-day window of testing positive for COVID-19.
• Hospital admissions unlikely to be related to COVID-19 treatment were excluded, for example, admission for removal of cataracts, fractures, or hernia repair.

In our December 2021 data analysis, we assessed the risk of general admission following confirmed COVID-19 infection, using a Cox proportional hazard model.

Given that there were later enough cases in which the outcome of admissions was known (due to sufficient post-observation, follow-up time), we used logistic regression models to assess the odds of COVID-19 admissions to any hospital ward, and COVID-19 admissions to ICU wards, specifically.

We adjusted for the following confounders:

• Vaccination type and status (such as Johnson & Johnson or Pfizer COVID-19 vaccine, number of doses, and time since last dose)
• Age (<40, then in 10-year bands up to age >=80)
• Sex
• Number of comorbidities considered to be risk factors for serious COVID-19 illness (0,1,2,3+)¹ (see Appendix 1)
• Dominant variant at the time of the member's COVID-19 infection
• Last documented infection from (likely) previous dominant variant
• Rolling seven-day positivity rate

We assessed and interpreted the ' dominant variant at infection date' coefficient from the logistic regressions to assess odds of admission in the Omicron-led wave, relative to the Delta-led wave.

Establishing severity of disease: Removing the confounder of underlying seropositivity (immunity from prior infection or vaccination) to establish true variant virulence

Recent work (published in the New England Journal of Medicine on 17 February 2022) points out challenges in measuring the severity of the Omicron variant - or this variant's intrinsic virulence.

• Authors note that, "Although the reduced CFR [case fatality rate] seen in the early weeks of South Africa's omicron-variant wave is better than the alternative, much of the observed difference relates to increased immunity among the people being infected. More time and careful comparisons controlling for age, pre-existing immunity, detection bias, lag time, hospital capacity, and other factors will be required to determine omicron's intrinsic virulence."

Our analysis included 497,509 medical scheme members.

As time passed, the percentage of medical scheme members who had developed underlying immunity to COVID-19 varied - specifically between waves of infection.

• During the Delta-driven third wave, 76.7% (n = 160,979) of members included in our study had no prior documented infection or vaccination (the two sources of COVID-19 immunity).
• By the time the Omicron-driven wave hit, only 30.5% (n = 30,443) of individuals included in our study had no prior documented infection or vaccination.

We performed a sensitivity analysis by running models on two sets of people.

We did this to tease apart the effect of changing seropositivity (the level of immunity to COVID-19 a person has when they are infected) on the severity of illness people experienced in the face of each variant.

1. First, we modelled odds using all our data - in other words for every member in our sample.
2. Second, we looked at the impact of prior infection alone on the odds of experiencing severe COVID-19 illness.

We restricted the analysis to people who were infected with COVID-19 during the Delta- and Omicron-driven waves and who had had a documented prior infection in the first wave (D614G-driven) or second (Beta-driven) wave.

In this way we could look at the severity of disease people would experience if they already had immunity to COVID-19 at the time of infection versus the severity of disease for those with no prior, documented infection.

Given the general population includes people who have had both confirmed and unknown COVID-19 infection (both of which would result in immunity post infection), by restricting our analysis to those with a documented prior infection, we could create a like-for-like comparison (removing the confounder of underlying seropositivity or existing immunity).

Our results

Overall, we analysed:

• 497,509 documented infections
• 68,972 admissions
• 59,853 COVID-19 admissions to hospital
• 14,204 COVID-19 admissions to ICU.

Proportion of documented COVID-19-positive cases in which the member was admitted

Using hospital admission rates as a proxy for the severity of the SARS-CoV-2 virus dominant at periods (as a proportion of confirmed documented infections) it is apparent that the severity of disease caused by Omicron variant was lower than that documented in South Africa's third wave, when Delta was the dominant circulating variant.

In the Omicron-driven wave, we found (Figure 1):

• An overall admission rate of 14%
• A COVID-19 admission rate at 12%
• A COVID-19 ICU admission at 3%

When restricting the analysis to include only those people who (a) were infected with COVID-19 in the Delta or Omicron-driven waves and (b) who had a documented prior positive infection dating back to the first wave (D614G-driven) or second (Beta-driven) wave, we are left with 13,378 documented infections (n = 7,705, 58% from Delta) and we see (Figure 1, right side) the following during the Omicron wave:

• 808 admissions overall (6%)
• 491 COVID-19 admissions (4%)
• 79 admissions to ICU (1%)

Figure 1 . Proportion of admissions by severity for all people versus those who got COVID-19 in the first and second waves of infection, and were reinfected in the third or fourth waves

• Comparative likelihood of COVID-19 admission in the Omicron-driven wave relative to the Delta wave, after adjusting for confounders

After adjusting for confounders (as defined in the methodology section above - vaccination type, age, sex, comorbidities etc) the data shows that, for the Omicron-led wave relative to the Delta-led wave (Figure 2) for all individuals in our sample, there was:

• An 18% drop in the odds of overall admission to hospital
• A 44% drop in the odds of COVID-19 admission to hospital
• A 62% drop in the odds of COVID-19 ICU admission.

After further limiting the analysis to only those members who had a past infection in either the (first) D614G-driven or (second) Beta-driven waves of infections, and adjusting for the confounders listed in our methodology above, the data shows (Figure 2) that for the Omicron-led wave relative to the Delta-led wave:

• There was a 6% increase in the odds of overall admission
• There was a 6% drop in the odds of COVID-19 admission
• There was a 60% drop in the odds of COVID-19 ICU admission.

These results are only statistically significant in the COVID-19 ICU admissions model at a 95% significance level.

This means that individuals with documented prior infection experienced a reduction in risk of ICU admission in the Omicron-driven wave relative to the Delta-driven wave, consistent with the full population (that is, those with and without documented prior infection).

So, people's underlying seropositivity (immunity) was not the only reason for the reduction in disease severity risk observed in the Omicron wave.

Figure 2 . Odds of admissions during Omicron-led fourth wave compared to the Delta-led third wave

Final thoughts: Changing severity or changing immunity?

• In summary, after adjusting for vaccination status (keeping in mind that fully vaccinated clients had a 70% lower odds of COVID-19 admission in the Omicron-led wave ), age, sex, number of comorbidities that increase the risk of experiencing serious COVID-19 illness, the seven-day rolling average test positivity rate and documented infection, we observed a 44% decrease in the odds of COVID-19 admission and a 64% decrease in the odds of COVID-19 ICU admission in the Omicron-driven wave relative to Delta-driven wave of infection.

A known limitation in studies related to severity of variants, is that that many global studies have shown higher seropositivity rates exist in populations than those indicated through national surveillance statistics.

• A South African study published in December 2021 estimated that up to 95% of COVID-19 infections were not reported to national surveillance.

In this case, the full extent to which underlying seroprevalence levels of SARS-CoV-2 antibodies within the Discovery Health Medical Scheme member base contribute to our observations is unclear and complex to investigate.

However, our robust analysis and results indicate:

1. A decrease in all forms of admission to hospital during the Omicron-led fourth wave compared to the Delta-led third wave
2. A greater decrease in the odds of serious illness, as the level of care required increases.

Interestingly, for members who had a prior COVID-19 infection in the D614G- or Beta-led waves we do indeed observe a decrease in the odds of COVID-19 admission during the Omicron-driven fourth wave, compared to the Delta-led wave.

This is the case even though we expect that the immunity they gained from prior infection in waves one and two would have waned by wave four, which took place over a year after the first two waves.

This indicates that the observed decrease in odds of COVID-19 admission in the Omicron-led wave may not be purely associated with increasing seropositivity but is likely also linked to the decrease pathogenicity of the Omicron variant. However, given our small sample sizes, this was only statistically significant in our ICU model.

Appendix 1

The following list of comorbidities (presented in alphabetical order) has been adapted from the Prescribed Minimum Benefits Chronic Disease List (CDL) and the Centers for Disease Control and Prevention's list of conditions associated with increased risk of severe COVID-19 .

Number	Category	Conditions
1	Cancer	Cancer
2	Cardiovascular disease	Cardiac failure Cardiomyopathy Coronary artery disease Dysrhythmias Peripheral arterial disease Cerebrovascular disease (including stroke)
3	Chronic renal disease	Chronic renal disease
4	Chronic respiratory disease	Asthma Chronic obstructive pulmonary disease (COPD) Bronchiectasis
5	Diabetes mellitus	Diabetes mellitus 1 Diabetes mellitus 2
6	HIV	HIV
7	Hypertension	Hypertension
8	Liver disease	Alcoholic liver disease Fatty liver disease Cirrhosis
9	Neurological disorders	Epilepsy Parkinson's disease Dementia (any cause, including Alzheimer's disease)
10	Overweight / obesity	BMI >25
11	Severe mental disorders	Bipolar mood disorder Schizophrenia
12	Solid organ transplant	History of kidney, liver, heart, or lung transplant

Interested in knowing more or reporting on these findings?

Please email MEDIA_RELATIONS_TEAM@discovery.co.za to request any updated data available since publication and to obtain any further context required.

Did you find this post interesting?

You may also be interested in reading our related post - our 14 December 2021 press release summarising our broader analysis of the Omicron variant with, predominantly, coverage of the way in which the two-dose Pfizer-BioNTech vaccination provides 70% protection against severe complications of COVID-19 requiring hospitalisation, and 33% protection against COVID-19 infection, during the current Omicron wave.

Disclaimer