Executive Summary

This report examines age-standardised natural causes mortality rates (ASMR) in Sweden for the period 2020–2024. Natural causes exclude external causes of death (ICD-10 V01–Y98), isolating disease-related mortality. Age-standardised rates use the 2013 European Standard Population.

The primary baseline is a quadratic (2nd-degree polynomial) fitted to 2000–2019. This captures the well-documented pattern of decelerating mortality improvement — rapid gains in the early 2000s giving way to slower improvement through the 2010s. A linear 2010–19 baseline is presented as a sensitivity check; however, as discussed in Section 3, linear extrapolation of recent steep declines can produce biologically implausible projections that generate spurious breaches.

95% Prediction Interval Breaches

Sensitivity: Linear Baseline

The linear 2010–19 baseline produces one additional breach not seen with the quadratic:

Section 1: Children (Ages 0–14)

Child mortality from natural causes is examined at three 5-year age bands (0–4, 5–9, 10–14) using crude rates, plus a combined 0–14 ASMR. The small numbers involved (typically 25–350 deaths per year per band) create wide prediction intervals, making statistically significant deviations very unlikely.

Summary Table — Children

Values: Actual (excess vs quadratic baseline). All rates per 100,000.

Charts — Children

Blue: actual. Red dashed: quadratic 2000–19. Grey dotted: linear 2010–19. Orange band: 95% PI (quadratic). Red dots: above-PI breaches.

Age 0–4

The 0–4 group shows continued strong mortality improvement. By 2024 the rate (42.3) sits 3.7 below the quadratic baseline (46.0). The quadratic captures the decelerating decline well (R² = 0.84), and 2024 falls comfortably within the PI.

Age 5–9

The 5–9 band shows the largest percentage excess in 2024 (+2.7 above the quadratic's 4.0), but this reflects just 41 deaths versus a predicted ~25. The R² of 0.23 confirms neither model captures much signal in this extremely noisy series. The 95% PI extends from roughly 0 to 8 per 100,000 — almost any value is "expected."

Age 10–14

The quadratic captures a decelerating decline (R² = 0.31) and predicts a flattening around 6.2 per 100,000. The 2024 actual of 7.1 sits closer to this quadratic prediction than to the linear's 5.7, and comfortably within the PI under either model.

Age 0–14 Combined (ASMR)

The combined 0–14 ASMR smooths single-band noise. The quadratic fits well (R² = 0.87) and all 2020–2024 values sit close to baseline, with the largest excess being just +1.2 in 2024. The dominant influence is the 0–4 band, which carries the largest ESP weight and highest absolute rates.

Section 2: Broad Age Groups (ASMR)

Larger populations produce tighter prediction intervals, making genuine deviations more detectable. The quadratic baseline is better suited here because it models the decelerating improvement seen across all adult age groups over 2000–2019.

Summary Table — ASMR by Age Group

Values: Actual (excess vs quadratic baseline). Red cells = 95% PI breach above.

Charts — ASMR by Age Group

Age 0–19

No breaches under either baseline. The 2020–2024 values track slightly above the quadratic throughout (+0.3 to +1.0), but well within the PI. The R² of 0.88 reflects a good fit to the decelerating decline pattern.

Age 20–64 — Two Breaches (Quadratic)

Under the quadratic baseline (R² = 0.99), the 20–64 group breaches in 2020 and 2021 only — the acute pandemic period. By 2022 the rate has returned within bounds, and by 2024 it sits 0.8 below the central baseline estimate at 121.4 — the all-time lowest value in the series.

The linear 2010–19 baseline additionally flags 2023 as a breach. However, the 2023 value of 130.0 was itself the second-lowest value ever recorded for this age group. The linear model demanded a rate of 123.9, which would have required a 9.6% decline from 2019 in just four years, extrapolating to zero mortality by 2056. The quadratic correctly identifies this as normal variation around a decelerating trend.

Age 65+

Breaches only in 2020 (4081 vs upper PI of 3934), driven by the pandemic's disproportionate impact on the elderly. Recovery was rapid: by 2021 the rate was already within bounds, and by 2024 it sits 148 below the central quadratic estimate at 3469 — indicating mortality improvement has fully resumed and arguably accelerated.

All Ages

Mirrors the 65+ pattern: one breach in 2020, rapid return within bounds, and by 2024 a rate of 751 that is 29 below the central baseline. The dominant driver is the 65+ group.

Section 3: Baseline Choice and the Limits of Linear Extrapolation

A critical finding of this analysis is that baseline choice can create or eliminate apparent breaches — not because the data changes, but because of differing assumptions about the trajectory of mortality improvement.

The Problem with Linear Baselines

Linear baselines fitted to periods of rapid improvement (such as 2010–2019 for the 20–64 group) assume that the pace of improvement continues indefinitely. For several sub-bands within the 20–64 group, this produces absurd extrapolations:

• 50–54 natural causes: linear baseline reaches zero in 2045
• 55–59 natural causes: linear baseline reaches zero in 2049
• 25–29 natural causes: linear baseline reaches zero in 2044
• 20–64 composite: linear baseline reaches zero in 2056

When the baseline is heading for biological impossibility, even record-low mortality rates can appear as "excess." The 50–54 and 55–59 bands breached the linear PI in every year from 2020 to 2023, yet their 2023 values were the second-lowest ever recorded. Their 2024 values are all-time lows.

Why Quadratic is More Appropriate

Mortality improvement typically follows a pattern of diminishing returns: large initial gains from public health measures, medical advances, and behavioural changes give way to slower improvement as the "easy wins" are exhausted. The quadratic 2000–2019 baseline captures this decelerating curve, using 20 years of data rather than 10.

For the 20–64 group, the quadratic (R² = 0.995) fits the training data better than the linear (R² = 0.984) and produces sensible extrapolations. Crucially, it does not flag the second-lowest-ever mortality rate as a statistically significant excess.

For child age groups, both baselines produce identical conclusions (zero breaches), so the choice is less consequential. However, the quadratic is preferred for consistency and because it more realistically models the flattening of improvement visible in bands like 10–14.

Conclusions

Using a quadratic 2000–2019 baseline with 95% prediction intervals:

1. The pandemic impact is visible in 2020 across the 20–64, 65+, and all-ages groups, with an additional breach in 2021 for 20–64. All breaches are above baseline.
2. Child mortality (0–14) shows no statistically significant deviation from baseline at any point in 2020–2024, under either the quadratic or linear baseline.
3. By 2024, every age group has returned within or below the central baseline estimate. The 65+ and all-ages rates are below baseline, indicating resumed mortality improvement.
4. No below-baseline breach occurs in any series at any point, meaning there is no evidence of a statistically significant mortality deficit in this period.
5. The one additional breach flagged by the linear sensitivity check (20–64 in 2023) occurs at a value that was the second-lowest ever recorded, driven by the linear model's biologically implausible extrapolation of the steep 2010s decline.