Analyzing weekly all-cause mortality data from 31 European countries (2015 onward, COVID-19 period excluded from baseline), NeMMo — an evolution of the widely-used EuroMOMO model — produced higher expected mortality baselines, tighter prediction intervals, and larger maximum Z-scores than its predecessor. Crucially, Z-scores and P-scores during non-pandemic weeks were closer to zero while being more elevated during pandemic weeks, improving the signal-to-noise ratio for detecting genuine mortality crises. Incorporating population offsets revealed negative linear mortality trends across all countries, consistent with genuine longevity gains once demographics are properly controlled.
Accurate excess mortality estimation matters far beyond pandemic accounting. It underpins detection of heatwave events, drug overdose epidemics, and the delayed health consequences of chronic disease — making algorithmic improvements here consequential for public health response times globally. NeMMo's use of periodic B-splines rather than sinusoidal functions captures heterogeneous seasonal mortality patterns that a single sine curve systematically misses, a meaningful methodological advance. The data-driven baseline selection — minimizing residual skewness rather than relying on fixed calendar windows — reduces the arbitrary assumptions that have historically made excess mortality estimates contentious across research groups.
Limitations are real: this is a methods paper, not an etiological study, and the proof is in routine adoption rather than a single validation exercise. The open-source R package lowers the barrier to replication. As a preprint not yet peer-reviewed, these performance claims require independent validation before NeMMo supplants EuroMOMO in official surveillance systems.