THE COMPUTATIONAL METHODS REVEALING TRUE COVID-19 DEADLY BURDENS

The choosen/random notes provide additional context for the three core methods without disclosing proprietary calculation details:

 ...Even only 1/2 of the main mathematical 'Method A' is a method inside the method; it does not let get very precise results, but is enough to test and debunk any age structure -if it is not composed of truly premature deaths, showing a degree of inconsistencies. Sensitivity tests incorporating potential disease-specific risk factors (prevalence and severity) for every* CCW-listed condition (*with the possible exception for Alzheimer's Disease and Non-Alzheimer’s Dementia), show that such adjustments cannot reduce the observed inconsistencies in the official COVID-19 death structure (concerns the U.S.); on the contrary, they would strengthen the required impossibly large suppression of mortality in healthier subgroups within an age band, bolstering that the official structure diverges strongly from biological and actuarial patterns, and making necessary to decrease the total average age of victims yet more to counteract. An additional layer of refinement (sensitivity tests) arises from synergy among conditions; even when matching only severe CCW conditions between a younger highly multimorbid individual and an importantly older one (with the same RLE, as less multimorbid), the younger case carries a substantially higher competing-risk burden mainly from the larger volume of concurrent mild conditions (thus his severe conditions must be more severe/risky), but also from other severe ones; this implies that the severe conditions in younger highly multimorbid fatal victims must be even more lethal - further increasing the importance of the frail younger tail and strengthening the inconsistencies in observed age structures. This additional layer is the reason that only for Alzheimer’s Disease and Non-Alzheimer’s Dementia, which have the steepest prevalence gradients, strong synergy effects have chances not to offset or exceed the older ones' advantage in raw severity prevalence (in 'Method A'). /To make the construction yet better we finally took into accout 'a share of men seeking care (IfR x severe fraction) to a share of women seeking care' (M/F), which 'M/F' was ~1.15 for aged 85+, ~1.30 for aged 75-84, but huge ~1.55 for aged 65-74, ~ 1.25 for aged 55-64; the effect of a smaller share of older men having high numbers of CCW-listed conditions is weaker than the contrary effect of men having a smaller RLE for the same number of conditions./. However, in this case (Covid-19 in the U.S.) the structure gets essentially disproved even before any RMT (and adjusted illness rates for frail subgroups) are applied, and the required suppression is already much too strong; this demonstrates that the implausibility is structural and not dependent on the precise value of RMT.

...The methods are new ideas. Even something as obvious as the relationship between the increase in average age and the number of conditions in premature virus-driven deaths has never been properly considered (a starting point for a bit easier 'Method C'). Consensus explanations must reconcile the observed age shift with realistic chronic-condition gradients - an area where pure medical intuition often falls short of rigorous demographic/actuarial chaining. For high selectivity, shown as a high average age, there must be seen a much increased average number of conditions too (interdependence). We also distinguish a still high RLE due to being still relatively young from a still high RLE in a very healthy importantly older one. RLE at any age depends on the number of chronic conditions, and the differences are still big even among old people - e.g., for Americans aged 75 their RLE was, on average, only ~4 years if they had >15 of the 30 CCW-listed chronic conditions, but 16.5 years if they had 0 - 3 conditions; first conditions little reduce their RLE. Interestingly, with increasing total condition count, the difference in age-related average RLE gradually decreases, e.g., to become virtually none for those with 15 or more conditions, regardless of whether someone is 75 or 'only' 67 years old. RLE as low as under 2 years (after excluding those already in a clear terminal state) is very rare and most often requires extreme multimorbidity in aged 90+. 'Method C' proves that the required average number of chronic conditions for aged 75 or 67 needed to support the observed very high average death age of epidemic decedents (76.58 y. in 2020 in the U.S., 34 - 35 years more than the average age of the infected) remain very strongly elevated (even with conservative RMs and basing only on total count), compared to the population norm at that age (...the construction of this method does not take into account other types of limitations resulting from other methods, otherwise any big increase in condition count might prove insufficient). The very powerful sensitivity test, assuming no Risk Multipliers (or where needed, RM = 1.0) and a conservative adjustment of illness rates for frail subgroups, was also performed; although visibly lower, the required average number of chronic conditions at age 75 remains strongly elevated (by x1.70 - x1.75) compared to the population norm. This further supports the conclusion that the official COVID-19 death structure diverged strongly from expected biological and actuarial patterns. There are two different ways to perform 'Method C'; the second way is more difficult, but it is only needed if only more rough baseline data is available. /Even if 'Method A' is constructed for RLE vs. MD, and fully works without any knowledge about disease burden among actual epidemic victims, it can be used in the context of disease burden too./

...Attempts to explain the observed official structure of 'Covid-19 victims' in the U.S. through specific a high-risk disease or combinations of high-risk conditions (theoretically e.g., diabetes + heart disease + obesity) does not help rescue it. The wider combination the yet bigger another constraint (on top of: total condition count - severity - synergy effects) that pushes the required average total condition count yet higher, not lower; only at much higher total condition counts does the mathematical likelihood of having a potential high-risk specific combination rise substantially. When the total number of conditions is low, the probability of a particular high-risk condition being present is lower than its average prevalence at a given age.

...Death certificate data provides an additional consistency check. In 2020, official COVID-19 deaths listed an average of clearly under 4 conditions (in the U.S.). Systems overload during major events, and its impact on the average number of conditions per death certificate, are included by us. Even allowing for known under-documentation (also outside overwhelmed periods) this number remains much lower than would be expected if the majority of deaths were truly premature virus-driven cases; in the number some mostly secondary conditions (Pneumonia, Respiratory failure, Adult respiratory distress syndrome, Sepsis) were also included. This supports the view that the official death structure is dominated by natural and harvested (of those who were in a terminal state before infection) deaths; secondary conditions should not be counted as pre-existing chronic conditions when assessing true frailty; without them the number should fall to about 3.0. ...Further enrichment of Method C using detailed clinical records (beyond death certificates) would be valuable, but such data faces significant access barriers for non-governmental entities in the United States. Access to comprehensive hospital, claims, or electronic health records would allow for yet more precise quantification of the true share of premature virus-driven deaths; large entities in other countries may have better access to such detailed data in their jurisdictions, then enabling it, after accounting for age structure and selection effects.

...Two separate layers should be applied. First, the relative volume of a frail subgroup is adjusted upward to reflect their higher (compared to the age-normal level) realized infection probability (due to greater healthcare contact and institutional exposure). Second, the Risk Multiplier is applied as the higher risk of death given infection, capturing a denser frailty typical of younger individuals reaching the same low remaining life expectancy.

...As the average years of life lost (YLL) increases, the dispersion in remaining life expectancy among victims must disproportionately strong widen. A pathogen capable of causing higher average prematurity cannot become less able to kill the frailest individuals. Higher average YLL necessarily implies a greater contribution (share) from younger and less multimorbid groups; this relationship is a natural mathematical consequence of the bounded distribution of remaining life expectancy within very old age groups. An average YLL could not be only 4 - 5 years for the smallest plausible average prematurity, as the mortality rate among the rest of the elderly cannot be >100 times smaller than among the frailest ones (with RLE ⩽3 years).

...In a stable population with constant birth rates and unchanging mortality conditions, life expectancy at birth (LE) would equal the average age at death observed in any given year. In the real United States, however, a gap exists between adjusted LE at birth for natural deaths and the average age of natural deaths in a given year. This difference is primarily driven by ongoing population growth (contributing directly and indirectly), which creates a younger overall age structure. It contributes directly by increasing the share of younger people, who, although they die less often, still lower the average age at death in a given year. Also, even with the same expected remaining life expectancy (mostly a very short one) at any age, some individuals die earlier than their average (entering a terminal state sooner), while those who live longer are not yet observed in the current year’s death statistics. Another contributing factor is the gradual historical increase in LE.

...BASIC RISK MULTIPLIER: BRM represents the additional mortality risk from infection when frailty is atypical for a person’s age. When two individuals have the same short remaining life expectancy (RLE), the younger one usually has more “dense” and pathological frailty - caused by aggressive underlying disease processes and stronger synergies between conditions. This makes younger frail people disproportionately vulnerable to acute infections compared to older individuals with equivalent RLE but more gradual, age-related decline. As a result, the effective risk multiplier is over 1 and tends to increase with decreasing age. This effect helps explain why true virus-driven deaths must show an additional younger skew relative to the natural-death age distribution, under observed infection patterns (infection rates not rising, but declining at a very advanced age).

...True virus-driven deaths must show an importantly lower average age than natural deaths (injuries and infant mortality do not belong), due to prematurity itself, Risk Multipliers, the general decline in infection rates at advanced ages, and small impacts of the remaining three already minor factors - two of which increase, but one decreases that average age. At any stage of increasing average RLE-measured prematurity (starting from only the theoretical 'close to zero'), the decline in the average age of victims cannot be stopped (the method A); the average age of victims falls considerably faster than TLE rises (the method B). But for deaths wrongly attributed to the virus (natural and harvested deaths, many of which occurred in individuals already in a terminal state) the difference in realized infection rates between younger frail and older individuals shrinks strongly. In a terminal state, social activity and mobility collapse for everyone, reducing the relative advantage (intensified by healthcare contact and institutional exposure) frail but non-terminal younger individuals would otherwise have; the usual age-related differences in infection rates among the elderly are largely eliminated. This, together with Risk Multipliers falling toward 1.0, almost entirely eliminates the additional younger skew observed for true virus-driven deaths, however the most important factor for an always higher average age of natural/harvested deaths stays the lack of real prematurity. Deaths wrongly attributed to the virus can have an average age that is similar to, or even a bit higher than, the average age of natural deaths. For example, when the average age of natural deaths is around 78–79 years (higher than in the U.S.), the average age of wrongly attributed deaths may reach even 80-81 years. This can occur if there is more frequent testing of the very oldest individuals, what can further raise the observed average age; but this factor does not apply in the same way to true virus-driven deaths, because for such deaths it can only partly reduce the speed of decline in official infection/illness rates at advanced ages, and it has very little impact on final results of the analyses (the methods A/B/C).

...A major difference also exists in the expected number of chronic conditions. For true virus-driven premature deaths, mortality at any given age is very selective for individuals with higher condition counts (and correspondingly lower remaining life expectancy). Shares of younger and older victims, within the overall age structure, only inform about the strength of selection and do not reduce this requirement for a specific age, as they are external data for any given age. In contrast, for natural deaths and harvested deaths (of those in a terminal state) wrongly attributed to the virus this additional selectivity disappears; these deaths occur among people who were already very close to the end of their natural lifespan, regardless of whether they had few or many conditions, and these deaths were sure to happen soon, not only among a very small part of them. As a result, the average number of chronic conditions among such deaths is only modestly higher than the age-specific population norm (the higher age the smaller average relative increase, while simultaneously numbers of deaths rise). For genuine virus-driven deaths, the average number of conditions must be substantially elevated, and cannot match the the near-normal levels seen in natural or harvested deaths of old ones. When we know the age structure of potential virus-driven deaths and necessary background data, 'Method C' allows for accurate calculation of the lowest plausible number of conditions necessary for a given age to make this structure possible (accurate if an analysed age is not very close to the top of the age structure).

...Younger frail individuals aged 'X' should have higher RLE-distribution than older individuals aged 'Y' (e.g., 85 years) at equivalent average remaining life expectancy, resulting in a bigger contribution to 'the low-RLE tail' compared with the older individuals. It increases any Risk Multiplier yet a bit more.

...For individuals under age 60, a very small upward adjustment to any average age at death is appropriate (in Method A) to account for conditions such as early-stage cancer and certain liver diseases that strongly shorten remaining life expectancy with relatively limited impact on physiological reserves at the time of infection. In the remaining age-range, even if a share of aged 60+ in deaths is strongly bigger, the effect of early-stage cancer is very small too, due to differing shares (falling slowly at first, but quickly at ages 75+), age distributions of cancer mortality there, and because at ages 60+ aggressive cancers, that drastically shorten RLE, tend to concentrate among people who are already highly multimorbid and so already have short RLE (much shorter than what is normal for age).

...Any real structure of victims of viruses such as influenza or Covid-19 must give an average age at death significantly lower than the average age of natural deaths. Even for the weakest variant of a virus the average age at death should be several years lower; this is undeniable, and the best way to see it is by executing the methods A/B. Unfortunately, for 'official victims' of Covid-19 in the U.S., there was very small average prematurity. Additionally, their structure in 2020 quite strongly resembled the adjusted pattern of natural deaths in 2019 (fatal injuries and infant mortality eliminated); if you divided both into 10-year age groups (but the lowest age groups in our analysis are: <1 and 1-14; next 15-24, 25-34, etc.; the highest one: 85+), no share-difference reached full 2%, 1% was exceeded in all age groups of the 55++ age range, but only once 1.5% was exceeded; in all age groups younger than 55 years old the difference was <0.2%. It is also worth noting that the smaller the expected number of truly premature virus-driven deaths relative to the total excess mortality, the larger the proportion of harvested deaths that are likely caused by secondary effects rather than the virus itself. Those secondary effects could include stress and panic, disrupted medical care (closed clinics or restricted access), suboptimal treatment protocols, and delayed diagnoses. The latter factor would mostly act with a longer delay and therefore result mainly after 2020, while the main analysis focuses on 2020. 

...A common conceptual error is to treat age groups as static categories when interpreting death age structures. Comparisons such as “people aged >80 versus those aged 70–79 or <70” can create the misleading impression that an extremely high average age of official virus-attributed deaths is biologically plausible. In reality, age structures must be understood dynamically. The large majority of people who reach age 80 or older (not being in a terminal state) were not very frail at younger ages; they represent the survivors who had earlier much better-than-average health trajectories. Conversely, individuals who are already highly frail at younger ages (for example importantly under 70) and soon die naturally typically have much worse-than-average health trajectories, and therefore will not appear in the future among the oldest age groups, because they do not live long enough to reach them. 

...Comparative data concerning the pre-pandemic normal prevalences of conditions in U.S. society, at different ages, are based on CMS and ICD-coded, for younger ones extrapolated with e.g. NHANES; the average %-prevalence of later added CCW conditions is higher, 9 is less than half of the older 2008-CCW conditions, but still the new 9 weight just over 0.6 of the older ones, for the detailed age structure of 'official Covid-19 victims'. 

...When there is the same ratio of RLE between infected ones, the mortality difference (MD) must be restrainedly rising (not decreasing nor the same) with both RLE increasing, because the difference in lost physiological reserves increases, among aged >60. Of course, if a healthier group is importantly younger than 55/60 years old (with not much worse than age-normal RLE), MD should rise considerably more.

...On Consensus and Rigorous Analysis: Consensus is a valuable practical tool for coordination when rigorous evidence is unavailable. However, when a clear mathematical or actuarial derivation yields specific, bounded, and testable results that conflict with prevailing consensus, the rational approach is to prioritize the stronger method and treat the consensus as provisional. Broad consensus dilutes precision by incorporating opinions of widely varying competence. The methods A/B/C are a mathematically rigorous way to separate attributable mortality from natural baseline and harvesting effects. The methods were in 2026 repeated and validated (including calculations), in detail, with most advanced AI systems.