The author derived an explicit analytical link between relative excess mortality calculated from raw death counts and from age-standardised mortality rates (ASMRs). The derivation showed that P-scores based on raw deaths and on ASMRs become algebraically identical when the age distribution of the chosen standard population matches the age distribution of the population under study. The study team applied this framework to German mortality data covering the COVID-19 pandemic years 2020–2023. Expected mortality for the excess calculations was estimated by linear extrapolation of pre-pandemic trends. Data indicated that, in practice, the two approaches produced closely corresponding estimates for Germany across the period examined. Short, direct point: equivalent weighting yields equivalent P-scores. Longer, analytical point: when the standard population’s age weights differ from the study population, the two metrics diverge because ASMRs apply external weights while death counts reflect current population structure and internal age-specific mortality changes.
The investigation reported that the influence of the standard population depended on where excess mortality occurred across age groups. In 2020, when excess deaths concentrated at older ages, the effect of a standard population was particularly pronounced. The study observed that choosing a standard population with relatively younger weights tended to down-weight the excess concentrated among older age groups, altering ASMR-based P-scores compared with raw death-count P-scores. The author noted that these differences are not mysterious; they stem directly from ASMRs applying fixed external weights to age-specific mortality rates. Short sentence: weighting matters. Longer sentence: because ASMRs reweight age-specific death rates to a chosen standard distribution, the age pattern of excess mortality interacts with the selected standard population to produce systematic shifts in estimated relative excess.
The study team examined additional methodological choices and their downstream effects. Aggregating the open-ended age group (for example, grouping ages 80+) altered the observed pre-pandemic ASMR trend in the data and, by extension, changed the estimated excess mortality when using trend-based baselines. The author reported that variation in the pre-pandemic reference period had a substantial impact on calculated P-scores. Both death-count and ASMR-based P-scores changed when different baseline windows were used; however, the two approaches tended to move in a closely corresponding manner under these baseline variations. Short observation: baseline selection shifts estimates. Longer analytical note: because trend-based expected mortality depends on the chosen reference years, any change in that window modifies the extrapolated baseline and thus the computed excess, and those modifications affect both raw counts and ASMR-derived measures, albeit filtered through different weighting schemes.
The author framed practical guidance around three core methodological choices: the standard population, the age-group definitions (including how the open-ended upper ages are grouped), and the specification of the baseline for trend extrapolation. The piece argued that for trend-based excess mortality estimation these are integral choices that must be stated and justified. The study recommended using death counts as a simple, transparent reference when analyses are confined to a single population context because raw counts avoid the additional assumptions required for standardisation. In contrast, the author cautioned that ASMRs require careful selection of a standard population and age-grouping scheme, as these choices alter the effective weighting and can materially change results when excess mortality is age-structured.
Readers should note that this work is presented as a preprint and has not been certified by peer review. The author declared no competing interests and indicated that only publicly available data were used; code and data are available at the GitHub link provided in the source. The paper did not report additional methodological parameters beyond those summarized here, and no peer-review conclusions or regulatory milestones were provided in the source. Open questions documented in the manuscript include the degree to which alternative standard populations would affect cross-country comparisons and the sensitivity of different baseline specifications to non-linear pre-pandemic trends—questions the author identified as dependent on the specific analytic choices researchers make and on the age pattern of excess mortality in their study populations.