Suboptimally DecompartmentalizedUnderstanding a global health crisis from the confines of home July 2020
Photo: istock.com/Morsa Images
As actuaries like to tout, we assess future risk, which is wildly different than predicting the future. Last year, the pandemic risk was an intellectually interesting sideline to pricing, a beach read about the Spanish flu1 and theoretical solvency scenario—not an all-consuming professional and personal task as in 2020.
Being a life and health actuary, the personal and professional have never been entirely discrete. Just as Miss Daisy never stopped teaching (I know, this reference is dated), I assume I am not the only actuary dulling a lively dinner party (I know, dated on two counts) by warning a friend of overtreatment and the business interests of hospitals and dentists. Some of us struggle to be an easygoing patient or casual consumer of public health statistics, and others find it difficult to leave a family member’s struggle with a disease out of their decision-making at the office. Let’s optimistically call this cross-pollination. But never, for many of us, has the line felt more blurred.
As I navigate through emerging statistics that ultimately affect the insurance industry’s performance, I intertwine my thoughts at home—because that’s where most of us across the globe are or have recently been at the time of writing—suboptimally decompartmentalized.
What Kind of Disease Are We Talking About?
Here in Cologne, Germany, actuaries and nonactuaries alike played the odds and attended Carnival in late February. We hoped this illness was seasonal; we selfishly hoped our age, relative lack of comorbidities and sauerkraut would protect us. Then, invincibility gave way to paranoia, which cleared supermarket shelves, and slowly matured into handwashing, social distancing and even the occasional properly handled mask.
While many think back to SARS, MERS and our on-schedule, 1-in-100-year event of 1918, my mother recently recalled over FaceTime the paranoia of the poliovirus. U.S. President Franklin D. Roosevelt’s paralysis stopped the nation, perhaps like Boris Johnson’s hospitalization in the United Kingdom. Yet my mother remembers mainly being rueful. Since children were highly susceptible, there was no leaving the house in summer, no friends, no air conditioning and definitely no swimming pools. Rue ended when her cousin was put in an iron lung at the age of 6 and died at age 12.2 Panic in the small town and in the nation evolved from there. Were cats or mosquitoes spreading the virus? So, 72,000 cats were killed, and DDT insecticide was sprayed.3 The worst outbreak came in 1952—though the 3,100 deaths and 21,000 paralyzed in the United States make us wistful today. A year later, the Salk vaccine brought overwhelming relief to my mother’s town, and by 1957, the annual polio cases in the United States were reduced to 5,600.4
Seventy years later, we don’t talk about polio except when ticking the box for standard immunizations. As late as March 2020, the media, medical professionals and many epidemiologists were still comparing COVID-19 with the seasonal flu, though a few “alarmist” modelers and peers were starting to think about 1918.
Here we are. Perhaps the pessimists will still be wrong. We are not in a war per se, and our men are not in the trenches—though some hospital workers (men and women) surely feel that way in their hallways. Today we have ventilators, personal protective equipment, evolving cocktails on the shelf and biotechnicians hard at work, though many patients in the surges of April never made it to the hospital. Today we hope for either a prophylactic or a Salk’s miracle to return us to normality, yet the timeline for scaling a vaccine is historically years. Above all, perhaps, today we have real-time data.
As of early June, most of the northern hemisphere appeared to be coming down from the first wave, and the death rates—even in the hardest hit areas of the United States and Europe—thankfully pale in comparison to those of the 1918–1920 flu pandemic (by the lowest estimates, by more than 9 deaths per mille ultimately).5 However, it is difficult to find comfort in numbers at this point. COVID-19 is young, and its nature and ultimate toll is of immediate concern to us—but this is all based on volatile assumptions, even with epidemic SEIR (Susceptible, Exposed, Infectious, Recovered) models recalibrated to the first wave.
Whatever its closest comparison, it is the kind of virus that has caused downtown streets to ring with ambulances, unemployment to soar, retail and beyond to panic at their top line, and insurers to panic at their bottom line. It is also the kind that tempts authors to publish life hacks and well-intended tips for working parents to just do a “partner swap,” divide your day into three or learn a new hobby.6 I am thus providing no advice in this article, but rather some observations and hopefully a bit of humanity as we grapple with difficult actuarial questions and beyond.
How Transmissible Is This? How Many Infections Are Currently in the Population?
This is a serious public health question. The statistic on which epidemiologists operate but actuaries generally find difficult to transform is an R0 (number of people infected by one case). This is a measure of transmissibility. For the 1918–1920 flu pandemic, this was estimated between 1.4–2.8, with a mean of 2.0.7 Without major government intervention, the R0 for COVID-19 appears to be much higher, with initial estimates hovering above 3. This virus is highly contagious, but in recent months, countries and regions have manipulated their Rt and experienced a range between nearly 0 and 3, depending on the timing and stringency of social distancing or other restrictions. For example, Germany’s balancing strategy seems to be centered on maintaining an Rt below 1.
Weighing the Risks
While I cannot disclose Gen Re’s internal modeling and pricing deliberations related to COVID-19, I can disclose a blurred moment in the life of an exhausted actuary.
May 3, 2020, 9:37 p.m.: We will be interpreting regional differences in the years to come. For a couple of minutes in the international news cycle, eyes were on Germany. (Why is the death rate low? Jumbled theories on Karneval and age mix, only the healthy went skiing in Lombardy and Ischgl, wonderful testing availability despite the fact that my friend and anesthesiologist working the ventilators had trouble getting tested initially … Strong social contract and adherence to government interventions, robust ICU capacity, efficient federal health care system, are there oddities in coding practices on deaths, could there be low viral load across the country … or is the worst yet to come?)
Putting together a puzzle, at the same time as needing a grocery odyssey to get coffee beans to think the puzzle through: Is it Russian roulette to go to the store tomorrow? Am I already immune because my children had low-grade fevers and maybe a cough last month? That’s another puzzle. No, it couldn’t have been—they never lost their ability to taste hidden spinach immediately. It can’t be COVID-19. But we are still out of spinach or anything remotely colored, so grocery tomorrow.
And when I go, I’d like to know whether more or less than 0.22 percent of unwelcome aisle-sharers are SARS-CoV-2-positive? Higher, because despite high testing in Germany, there are surely many asymptomatic people and also many symptomatic unreporteds. Lower, because many of the known cases are in isolation either at home (or are they?) or in the hospital. Higher, because we have a higher case density in our region. Much lower, because many of the known cases are since recovered, and besides, the high transmitters are the ones rebelliously picnicking in the parks, not doing essentials. Net net, my actuarial estimate is 0.22 percent (95% CI 0%–15%).
In terms of the number of active infections in the population, many of us track the Johns Hopkins compilations or our own dashboards daily. In May, Germany had more than 180,000 cases, or just over a 0.22 percent infection rate.8 So, in a supermarket with daily foot traffic of 5,000 surface-touchers, does this mean that nearly 11 people are contagious? For further mulling on grocery store risk, please see the “Weighing the Risks” sidebar.
What Is the Ultimate Attack Rate?
The more actuarially relevant question is what the ultimate infection rate is—this is the driver of projections. German Chancellor Angela Merkel put this at 67 percent early on.9 The Imperial College of London has an unmitigated scenario with an ultimate attack rate near 80 percent.10 Parts of Asia have exhibited a controlled curve far below this, at least for a first wave.
Depending on the geography, we collectively have had whiplash at varying weeks over the outlook on the ultimate attack rate. Should last Tuesday’s announced suppression measure affect my long-term outlook and pricing for a given country? Should I be tracking the infection case trend or the unemployment rate? Other reasonable questions that are difficult for an actuary to digest, with political risks well beyond the biometric sphere of comfort, include:
- The Imperial College Report 9 has an 18–24-month outlook, and its Report 12 looks at a shorter period (250 days, for a first wave). Can we safely assume that after that period, a vaccine is both developed and widely accessible?
- What is the true infection rate? How many false negatives and positives are in the infection numbers, as well as in recent antibody trials? And are the false negatives attributable to purported reinfections? And how long does immunity last after infection? Is immunity on asympomatics the same? What are mutation risks with higher potential virulence?
- Which cultures and sectors will adapt to life at home, on video? Or will economic and public resistance force lockdowns to be permanently abandoned, no matter further illness peaks? Is there another way forward, besides the “unmitigated” scenario? Taiwan and even dense South Korea both seem to have shown sustained control in infections—despite clusters, despite mostly falling lower on Oxford’s Stringency Index11 and despite the economy being closer to business-as-usual. Have they just been lucky so far, in being able to get ahead of the test-and-trace game? Or, perhaps, is the best way forward the mass use of professional masks and well-worn tracing apps or well-timed SMS alerts?
- Will travel restrictions over 24 months or longer be internationally coordinated? If at least some governments are publicly or privately targeting herd immunity or doing dark pension liability or election math, then perhaps a theory of international convergence is valid, and the best individual countries can do is to maintain the Rt to a level where critical care beds are not overwhelmed.
It is no comfort that recalibrating SEIR models to different (all plausible) Rt vectors given recent experience shows bottom-line direct mortality impact in the short term—anywhere from <1 percent to more than three- or four-fold. All in all, the timing and degree of restrictions, vigilance and political leadership, as well as habits of the public, will dramatically influence the trajectory that our markets are on—and these are risks that even the western frontiers of pricing actuaries are uncomfortable predicting. Yet, it is our responsibility.
What Is the True Mortality Rate?
Though this, at first glance, appears to be a simple question, with published statistics12,13 hovering between 0.6 percent and 2 percent or higher, the basis of statistics causes complications. Many countries only have a national death rate, which is a unit rate, so age adjustment is the first of many problems. In certain places where infections continue to increase yet deaths fall, this may also be an age (and health status) mix shift—increasing awareness and adapted behaviors according to who is at risk.
Some countries have age-dependent numbers, but both the numerator and denominator are all on different bases. Recently we saw upward adjustments in New York, the United Kingdom and China to account for previously unreported deaths, with deaths outside of the hospital and sometimes care homes not being counted. In Italy, postmortem testing seems to be practiced, whereas classification of a death as due to COVID-19 varies elsewhere. Countries and regions with poor testing may have a high case fatality rate (CFR, which are deaths over total confirmed cases), but a much lower infection fatality rate (deaths over total infections).
In cases where the CFR has run high—above 10 percent in some hard-hit areas—this may be a combination of severity mix on the denominator due to underreporting, lack of care due to surging or general avoidance of hospital chaos. In addition, regions that are earlier in an outbreak may have active cases that may result in death, but the mortality rate still appears benign.
Regardless of reported statistics and coding differences, the proof is in the pudding: How will total deaths look? The glimpses of peak weeks from the U.K. Office for National Statistics and U.S. Centers for Disease Control and Prevention data look concerning, with +60 percent and +100 percent in the United Kingdom and the New York area, respectively, and with other places such as Ecuador looking much worse.14 If we spread out these worst moments over a year, these additions to expected mortality may be but a blip—but only if the first wave stays as an isolated, single wave, and we have already seen signs of further outbreaks in multiple regions, and perhaps a prolonged double-humped wave emerging in the United States at the time of writing.
Who Is Vulnerable? Difference on Underwritten Lives, in Life?
So far in this article, I’ve been talking about COVID-19 as it relates to the general population. For valuation purposes and, in particular, pricing new fully underwritten business, the effect of underwriting against comorbidities will grant some protection, though viewing projections on an actual-to-expected (A/E) basis dampens this. Findings were rocky at first, with no age adjustments and limited or skewed cases. But now, findings are starting to point toward respiratory conditions faring surprisingly well, while cardiovascular conditions, diabetes and obesity show high prevalence among hospital admissions and with poorer prognosis.15 Socioeconomics is another factor, with sum-weighted portfolios leaning toward higher socioeconomic status. So far, statistics in the United Kingdom point toward lower infection rates, though not necessarily lower CFR. But in markets without the National Health Service (NHS), the CFR may also diverge.16
What Is the Effect on Other Claims, Health Costs and Morbidity?
How COVID-19 will affect health care costs, health status and other morbidities—beyond the immediate hospitalization of those infected—is a broad topic just beginning to emerge. In terms of reimbursement of out-of-pocket health costs, this appears mitigated by some governments picking up the tab for health costs related to COVID-19, but this is not the case everywhere and is not guaranteed as sustained policy. For longer-lasting effects on health status, the disease is too young for longitudinal studies to begin to play out. But there are possibly lasting neuro, cardiac, respiratory and other consequences that make a clean break after 12 to 24 months all the more questionable.
As witnessed with the 2003 SARS-CoV-1 outbreak in Asia, we observe dampened health care service utilization of up to 30 percent, as noted in the Society of Actuaries’ continuing COVID-19 research.17 While this may sound good for a quarterly loss ratio, it only means that needed treatment is being pushed down the line—and with potentially devastating consequences.
COVID-19 Effects on Activity Levels
Working out is a great way to manage stress and feel a sense of accomplishment, with regular physical activity having positive effects on your overall well-being by boosting your mood through increased serotonin levels in your brain.
Before the alarming levels of the spread and severity of COVID-19 were known or pandemic declarations were made, in mid-February 2020 we launched a companywide international pilot at Gen Re that studied the behavioral impact of and adherence to a new heart rate-based activity metric—personal activity intelligence (PAI)—with the intention of improving the health and well-being of those who participated.
Then, there are the effects of isolation on mental health, addiction, domestic violence and potentially less physical activity. On one level, I personally feel grateful for the comforts and joy of home and lack of isolation, but I admit the lack of childcare has affected my sleep and certainly physical activity (see sidebar “COVID-19 Effects on Activity Levels” to see if others have been affected as well). Some people have more serious impacts and shock due to economic stress, and this is one key reason that most governments have proven eager to return from lockdown.
There is room for measured optimism, too. As COVID-19 becomes an economic question well beyond the direct pandemic risk, the associated risks—starting with lower traffic fatalities, cleaner air and other countercyclical effects—should also be considered. South Africa, weeks into lockdown, reported no increased cases of COVID-19 or natural deaths, but a historic drop in non-natural causes of death (which would reflect external causes both violent and nonviolent).18 Southeast Asia followed suit during its lockdown periods. Japan, which saw dramatic economic downturn in the late 1990s and stagnation beyond, had seen large increases in suicides. However, correlations with other causes appeared to be favorable in recession, particularly with ages 45 to 64 for heart disease and accident.19
But right now, we are still in the thick of tracking infections, unreported infections, deaths and unreported deaths—by geography and, if we are lucky, by age, health status and behaviors such as smoking or activity levels. So, although we are surrounded by the consequences less stark than life-and-death on a personal scale, actuarially we are not yet past the ends of our noses in terms of downstream consequences.
- 1. Spinney, Laura. 2017. Pale Rider: The Spanish Flu of 1918 and How It Changed the World. London: Jonathan Cape. ↩
- 2. RetroIndy: Riley Hospital for Children. Indianapolis Star, November 6, 2013 (accessed April 3, 2020). ↩
- 3. Shors, Teri. 2017. Understanding Viruses, 39–40. 3rd ed. Burlington, Massachusetts: Jones & Bartlett. ↩
- 4. History of Polio. Wikipedia (accessed April 22, 2020). ↩
- 5. Spreeuwenberg, Peter et al. 2018. Reassessing the Global Mortality Burden of the 1918 Influenza Pandemic. American Journal of Epidemiology 187, no. 12:2,561–2,567. ↩
- 6. Thompson, Avni Patel. A Guide for Working (From Home) Parents. Harvard Business Review, March 19, 2020 (accessed April 3, 2020). ↩
- 7. Pyrek, Kelly M. 100 Years After the Spanish Flu: Lessons Learns and Challenges for the Future. Infection Control Today, October 11, 2018 (accessed April 3, 2020). ↩
- 8. Johns Hopkins University. COVID-19 Dashboard by the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University (accessed May 27, 2020). ↩
- 9. Merkel zur Corona-Krise: “Wir müssen Zeit gewinnen.“ Tagesschau, March 11, 2020 (accessed April 3, 2020). ↩
- 10. Ferguson, Neil, Daniel Laydon, Gemma Nedjati-Gilani, et al. Report 9: Impact of Non-pharmaceutical Interventions (NPIs) to Reduce COVID-19 Mortality and Healthcare Demand. Imperial College of London, March 16, 2020 (accessed May 27, 2020). ↩
- 11. Coronavirus Government Response Tracker. Oxford University (accessed May 27, 2020). ↩
- 12. Rajgor, Dimple D., Meng Har Lee, Sophia Archuleta, Natasha Bagdasarian, and Swee Chye Quek. The Many Estimates of the COVID-19 Case Fatality Rate. The Lancet, March 27, 2020 (accessed April 3, 2020). ↩
- 13. Verity, Robert, et al. Estimates of the Severity of COVID-19 Disease. medRxiv, March 13, 2020 (accessed April 3, 2020). ↩
- 14. Cabrera, José, and Anatoly Kurmanaev. Ecuador’s Death Toll During Outbreak Is Among the Worst in the World. The New York Times, April 23, 2020 (accessed May 27, 2020). ↩
- 15. Grasselli, Giacomo. 2020. Baseline Characteristics and Outcomes of 1591 Patients Infected With SARS-CoV-2 Admitted to ICUs of the Lombardy Region, Italy. Journal of the American Medical Association 323, no. 16:1,574–1,581. ↩
- 16. Williamson, Elizabeth, et al. OpenSAFELY: Factors Associated With COVID-19-related Hospital Death in the Linked Electronic Health Records of 17 Million Adult NHS Patients. medRxiv, May 7, 2020 (accessed June 26, 2020). ↩
- 17. Hall, R. Dale, et al. Society of Actuaries Research Brief: Impact of COVID-19. Society of Actuaries, May 15, 2020 (accessed May 27, 2020). ↩
- 18. Report on Weekly Deaths in South Africa. South African Medical Research Council, July 8, 2020 (accessed May 2, 2020). ↩
- 19. Granados, José A. Tapia. 2008. Macroeconomic Fluctuations and Mortality in Postwar Japan. Demography 45, no. 2:323–343. ↩
Copyright © 2020 by the Society of Actuaries, Schaumburg, Illinois.