Covid-19 conundrums and considerations.Insight
By Insight Actuaries & Consultants
Adapted from an article written for the SA Actuary magazine
Written in June, published in July and updated for this blog dated 10 August 2020
The novel Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) which is the causative agent of Coronavirus Disease 2019 (Covid-19) has been a pandemic of several kinds. First the epidemiological kind, as declared by the WHO on 11 March 2020, but also a pandemic of data visualisation dashboards, projection models, sensational news and strong opinions. It has been a polarising wedge in politics, society, and even the actuarial profession where we are not shy to offer differing opinions.
The world watched in morbid fascination as China locked down the Wuhan district. Being classified as a novel virus is what drove a lot of the concern. Novelty in virology terms means no inherent immunity, no treatment, and no vaccine. Its characteristics placed it in the sweet spot of danger – moderate infection reproduction rate, short serial interval, a long enough incubation period to spread widely before the onset of symptoms, and a low to moderate fatality rate so that it did not kill all its hosts. The ongoing uncertainty caused unprecedented policy reactions with mixed success. Modellers did not cover themselves in glory with few early models coming near reality. Countries experienced vastly different disease progression, some having no or extremely low fatalities (Vietnam for instance) while others have experienced sharp excess mortality (the UK for instance); some were hit early and hard such as Italy and Spain. Some countries got a late start but could not avoid the morbid mortality climb nonetheless (the USA). Other countries show a slow and steady rise and even slower decline, such as India, or Brazil which has had an average of 1,000 Covid-19 fatalities a day for over two months after a long slow start.
Much remains unknown about the nature of the virus although so many new academic papers appear pre-peer review every day it is tough to keep up. We know the disease disproportionately affects the elderly and largely leaves children alone, which is unusual for a respiratory disease. People with certain comorbidities also experience more severe outcomes and increased fatality risk, especially diabetes and hypertension. The overall infection fatality rate remains unknown as the disease presents many conundrums such as the relatively high proportion of individuals who exhibit no or only very mild symptoms and so go untested and untreated. Low side estimates are in the 0.2% range, while high side estimates are 1% depending in large part on the risk profile of those infected. Comparing across countries, case fatality rates vary widely and are difficult to compare as they depend on each country’s testing capacity and protocols which have varied widely.
The Actuarial Society’s Covid-19 modelling group was pulled together in March 2020 from a small group of volunteers who wished to contribute to the profession’s dialogue on the emerging pandemic. The groups first modelling effort was released for public comment on 28 April 2020. Unfortunately, the media picked the worst scenario shown as their headline although the in the report it was clear that the high scenario figure was to show sensitivity to key a key assumption rather than describe a likely outcome. Much feedback was received, some related to form and some on substantive issues. At the time of writing this article, the second model release was being finalised after protracted discussion. The views within the modelling group are wide, and it has not been possible to reach consensus on some fundamental disease dynamics that affect spread. Individual actuaries will continue to take their own views.
One contentious point is the question of where the disease’s attack rate (the proportion of the population infected) is likely to end, and what drives that rate. Is the epidemic over for countries who have ‘ turned the corner’, despite serology studies in many of those countries being below antibodies in 20% of the sampled population, or even below 10% – far below the level implied by the (often misunderstood) herd immunity threshold. Are these serology test plateaus because of social and governmental actions, the so called non pharmaceutical intervention (NPI) effects of social distancing, mask wearing, and general increase in hand and surface sanitising? Or are the effects due to some innate immunity that varies widely across the world and within communities? One clear implicit assumption in the compartmental model used by the ASSA working group that required attention was the implicit assumption of heterogeneity. While we had allowed for age-based disease severity and mortality once infected, there was not allowance for heterogeneity in the infection process. This heterogeneity may arise from different sources – variations in contact rates, variations in susceptibility, and variations in infectivity. Variation in contact rates is a known and widely accepted phenomenon. Medical experts seem to disagree on whether natural immunity responses, such as those at the cellular level, would affect disease severity, or whether they could shield certain individuals from becoming infected at all. It is not a binary question as viral loads vary and affect immune response and testing accuracy. Allowing for greater heterogeneity changes the progression of infection, lowers the herd immunity threshold and the final population attack rate. Conceptually, those at higher risk of becoming infected are captured earlier and once removed from the susceptible reservoir leave a less susceptible group.
This results in higher rates of transmission early on, followed by slower disease spread trailing off. Could this effect explain what has been observed in other countries with varying levels of NPI intervention? Are country reactions and stringent policy steps taken to curb spread responsible for saving hundreds of thousands of lives worldwide? Given the observed variations in country mortality rates for countries that have adopted similar interventions, it is a difficult case to make. It seems more likely that the observed variation in country experience arises from some combination of factors. There remains much confounding information at a macro and micro level. Socio-economically poorer communities within countries do comparatively worse, but low- and middle-income countries have generally performed better in international comparisons.
In South Africa’s case, early lockdown held back the tide of Covid-19 like a sandcastle wall on the beach, noting that this was the stated intention – to buy time. Eventually the viral spread broke through and spread through the Western Cape and then the other provinces despite one of the most stringent lockdown policies globally. It then proceeded to peak and turn in the Western Cape despite continued easing of lockdown thereafter.
Scientists and modellers alike have made calls for re-opening the economy as the requisite time had been bought but the WHO has urged an abundance of caution which has influenced decision makers. The National Epidemiological Consortium Modelling team’s central estimate of around 40,000 deaths was well established in April (the report released publicly in May). Whenever I have had the opportunity to present on the issue of Covid-19 modelling I take time to stress that the effect of the disease should be seen in context. South Africa has high premature mortality for our age profile due to our high burden of HIV/AIDS, TB, NCDs, violent deaths and other causes. Close to 500,000 South Africans will die in 2020, circa 1,300 a day. The Thembisa model outputs suggest over 60,000 deaths from HIV/AIDS and the disease places a heavy and ongoing burden on our health system. However, we do not report these daily deaths to the country on websites and social media.
The economics of healthcare are unavoidably cruel as apportionment of limited funds mean rationing is inevitable. Saving lives costs money – some easily and cheaply such as investing in improving the social determinants of health, clean water, sanitation, food security; and in other cases more expensive such as an innovative cancer drug that may cost hundreds of thousands of Rands and extend life 6 to 12 months. Health economists use a term called ‘willingness to spend’ when considering healthcare spending priorities (amongst other factors). At face value the willingness to spend on Covid-19 has been very high compared to other healthcare needs driven in part by fear and uncertainty and in part by the ever-present information deluge. In South Africa’s case while direct spending on Covid-19 has been significant, the cost in overall economic terms is astronomically higher. The interventions taken to curb a tide of Covid-19 deaths make these likely the most expensive lives we have ever saved.