Per-Month Linear 2012–19 Baselines with 95% Prediction Intervals · Data through September 2025
In-patient and/or specialised open care, patients per 100,000 per month
Data: Swedish National Board of Health and Welfare — Patient Register ·
European Standard Population 2013
Analysis by Claude Opus 4.6 (Anthropic), prompted by @dobssi on X · February 2026
The companion ICD-10 Chapter Overview identified D50–D89 (Blood & immune disorders) as breaching the 95% prediction interval in 0–19 year olds in 2021–2024, and E00–E90 (Endocrine, nutritional & metabolic) as showing age-specific divergence. No other chapter showed comparable sustained elevation in children.
This report drills into both chapters at monthly resolution by subcategory and 5-year age band. The central finding is that the D50–D89 signal is driven entirely by one subcategory — D80–D89 Immune disorders — while the remaining subcategories (D50–D79) have actually run below baseline since 2020. The parent chapter aggregate partially masks a far more dramatic subcategory signal.
A note on metrics: The overview report uses yearly unique patient counts (how many distinct children were diagnosed at least once that year), while this deep dive uses monthly patient counts (how many children were diagnosed in each calendar month). Children seen in multiple months are counted in each month, so monthly rates cannot be summed to match the yearly figure. Both approaches detect the same underlying signal; monthly resolution adds timing and seasonality information but is noisier at the aggregate chapter level. The D50–D89 parent chapter charts in this report therefore appear less dramatic than in the overview — but the D80–D89 subcategory charts below reveal the signal clearly.
Chart convention: amber bars = observed monthly rate; blue line = expected baseline (solid during 2012–2019 training, dashed when extrapolated); shaded band = 95% prediction interval (extrapolation period only).
D80–D89 Immune disorders are the dominant signal. ASMR 0–19 has been significantly above the 2012–2019 baseline for 14 consecutive quarters since Q4 2020, peaking at +80% in 2022 and remaining at +30% in 2025. Clear age gradient: strongest in 0–4 year olds (peak +127%, still +55% in 2025), progressively weaker with age.
D50–D79 (remainder of blood chapter) has run below baseline since 2020 (−15% to −25%), partially cancelling the D80–D89 spike in the parent chapter aggregate. This is why D50–D89 appears modest at the chapter level despite D80–D89 showing dramatic excess.
E10–E14 Diabetes shows a divergent age pattern: ages 0–4 rising steeply (+61% by 2025) while ages 10–19 are declining (−10% to −27%).
E65–E68 Obesity shows broad increases across ages 5–19, accelerating to +39% ASMR in 2025. Pre-existing upward trend complicates attribution, but the recent acceleration exceeds the baseline trajectory.
Catch-up effect excluded for D80–D89: No dip during 2020 restrictions; cumulative excess has climbed for five years without plateau.
D80–D89 is the subcategory driving the D50–D89 breach flagged in the overview report. At +80% ASMR excess in 2022 and still +30% in 2025, this is by far the largest sustained deviation in any paediatric diagnosis subcategory. The age gradient — strongest in the youngest, most immunologically naïve children — is consistent with an immunological mechanism.
While the percentage excess is dramatic (+127% in 0–4 year olds at peak), the absolute numbers are small relative to the population. At peak in 2022 there were approximately 1,070 excess D80–D89 patient-months in ages 0–4, equivalent to roughly 0.18% of the age group population — about 1 in 550 children. By 2024 this had fallen to ~470 excess patient-months (0.085%, or ~1 in 1,200). These are patient-months from the monthly register, not unique individuals — the number of uniquely affected children will be smaller still.
Could the D80–D89 increase simply reflect diagnoses being reclassified from D50–D79 to D80–D89, rather than a genuine increase? If so, the parent chapter D50–D89 total should remain stable. It does not — patient counts in ages 0–4 rose from 4,208 (2019) to 5,191 (2022), a 23% increase. The D50–D79 drop (~400 patients) is far smaller than the D80–D89 rise (+1,170 patients). D80–D89's share of the parent chapter jumped from 11–18% pre-pandemic to 31–37% post-pandemic. The D50–D79 decline appears to be a separate phenomenon — possibly reflecting reduced healthcare utilisation for anaemias, coagulation disorders and similar conditions — rather than reclassification into D80–D89.
Within D50–D79, the decline is spread across multiple subcategories (D50–D53 nutritional anaemias, D60–D64 aplastic anaemias, D65–D69 coagulation defects) rather than concentrated in one, further arguing against a systematic reclassification from one specific code to D80–D89.
The monthly data counts patient-months, so a child with a chronic immune disorder appearing every 3 months would contribute 4 patient-months per year. This means the D80–D89 increase could reflect: (a) genuinely more children being diagnosed, (b) existing patients being seen more frequently, or (c) both. This distinction cannot be resolved from aggregate register data — individual-level longitudinal analysis would be needed. However, the steep onset in 2021 across all age bands simultaneously, rather than a gradual increase in visit frequency, is more consistent with new diagnoses than with existing patients visiting more often.
Top: COVID-19 diagnosis waves (3-month rolling). Bottom: D80–D89 12-month rolling average. The initial D80–D89 inflection in 0–4 year olds coincides with the first major paediatric COVID wave, with the steepest acceleration following the Omicron wave (January–February 2022).
The catch-up hypothesis — that elevated rates reflect deferred diagnoses from reduced healthcare access during pandemic restrictions — can be tested against two predictions: (1) there should be a dip in D80–D89 during 2020, and (2) the subsequent spike should plateau and revert as the backlog clears. Top panel: annual excess vs baseline shows no dip in 2020 — ages 0–4 were already +11% above baseline that year. Bottom panel: cumulative excess from 2020 has climbed continuously for five years without plateau in any age group. Sweden returned mostly to normal by late 2021 and removed all recommendations in February 2022.
The parent chapter D50–D89 was the trigger for this deep dive, having breached the 95% prediction interval in the yearly overview. At monthly resolution, the parent chapter signal appears modest — this is because D80–D89 (sharply up) and D50–D79 (running below baseline since 2020) partially cancel. The charts below are included for completeness and to illustrate this masking effect.
The diabetes data reveal a divergent age pattern since 2020: rising sharply in ages 0–4 while declining in older children. The 2012–2019 baseline in ages 0–4 was mildly declining (slope −0.3/yr, R² = 0.44), driven by three consecutive low years in 2016–2018 (rates of 15.0, 14.1, 13.2 per 100k/month). The 2019 rate (15.5) was elevated relative to those preceding years but within the historical range. The decisive break came in 2021–2022 when rates reached 17.3 then 20.2 — clearly outside anything seen in the 2008–2019 period. Whether the inflection began in 2019 (pre-pandemic) or represents a noisy year followed by pandemic-era acceleration cannot be resolved from these data alone. In either case, the divergent age pattern — rising in 0–4 while declining in 10–19 — is unusual and warrants clinical investigation.
Broad-based increase across ages 5–19, accelerating markedly since 2023. Pre-existing upward trend complicates pandemic-specific attribution, but the 2024–2025 acceleration clearly exceeds the baseline trajectory.
As with D50–D89, the parent chapter aggregate masks divergent subcategory dynamics — diabetes rising in infants but falling in adolescents, obesity rising broadly. Included for completeness.
This monthly deep dive identifies three specific signals from the two ICD-10 chapters flagged in the companion overview:
1. D80–D89 Immune disorders: The most robust signal. 14 consecutive quarters of significant excess (ASMR 0–19), peaking +80% in 2022, still +30% in 2025. Clear age gradient. Not explained by catch-up. The D50–D79 deficit means the parent chapter understates the immune signal.
2. E10–E14 Diabetes in infants: Rising in 0–4 year olds (+61% by 2025) while declining in 10–19 year olds.
3. E65–E68 Obesity: Broad increase ages 5–19, +39% ASMR in 2025, accelerating beyond the pre-pandemic trend.
The companion mortality analysis found no excess mortality from natural causes in Swedish children during this period. The findings here therefore describe a morbidity increase — more children diagnosed with specific conditions — without a corresponding mortality signal.
Per-month linear regression on 2012–2019 (8 training years, 12 separate models per calendar month for each diagnosis × age combination). 95% prediction intervals shown on extrapolation period (2020+) only. Both sexes combined. ASMR 0–19 uses European Standard Population 2013 weights truncated to 0–19: 0–4 = 5,000; 5–9, 10–14, 15–19 = 5,500 each. Population denominators from Socialstyrelsen monthly estimates by 5-year age band. Quarterly charts average monthly rates within each quarter.