# 3384 - Canadian rotavirus designs on an outcome with no known vaccine pathway: asthma in the CCDSS

**Documented verdict (CAN_noQC_RV_newly_identified_cnics): CONSISTENT - asthma incidence r 0.71 (0.48-1.06) includes 1 (autism 0.99 (0.32-1.89) on the same cells).** Asthma is CONSISTENT: no non-specific same-system bias detected at this precision; this does not validate the autism estimate (a prior_dominated flag means the control had little power).

**QC_RV_ASD_cnics: NOT OBTAINED on identical cells.** Its 125 cells are INSPQ SISMACQ age-band prevalence (1-4, 5-9, 10-14, 15-17, 18-24), not CCDSS cells, and the CCDSS publishes Quebec only as the pooled 1-19 band. The report-only Quebec CCDSS 1-19 analogue (QC_RV_ASD_cnics model on 24 cells) gives asthma prevalence 0.92 (0.25-3.21) against autism 1.37 (0.33-5.34) on the same cells (class CONSISTENT, analogue; never the QC reading).

**CAN_CIHI_INJURY_HOSP_FY2016_2024: NOT FITTED.** no province/territory x age cross-classification in any of the nine workbooks (province tables are all ages; age tables are Canada-only or Ontario and Alberta combined), so the identical CAN_noQC cells (province x ages 1-19 x FY) cannot be built; only FY2016-17..2023-24 would overlap the block in any case.

Generated by `07_analysis_code/3384_negative_controls_canada.py` (every number below is computed by that script and written to `3384_negative_controls_canada.csv` / `3384_negative_controls_canada.json`). Figure `08_figures/3384_negative_controls_canada.png` (hollow light markers = UNINFORMATIVE). Fits cached under `project_results/negative_controls_3330/3384/`. Tests: `07_analysis_code/test_3384_negative_controls_canada.py`. Downloads: `07_analysis_code/3384_acquire_canada_negative_controls.py`. Controls are outcome_class 'other' and never enter the autism hierarchy; asthma is not an injury, so no row here can join the 3330 C_INJ set (3330 sensitivity S3 is the only home of a non-injury control).

## Downloads (approved in DOWNLOAD_REQUEST.md)

| file | bytes | sha256 | url | retrieved (UTC) |
|---|---:|---|---|---|
| `ccdss_asthma_i6_s4_bd_eng.csv` | 876,020 | `77422a6590c88298...` | https://health-infobase.canada.ca/ccdss/Download.aspx?i=6&s=4&b=d&l=eng | 2026-09-30T03:11:33Z |
| `ccdss_asthma_i6_s5_bd_eng.csv` | 920,313 | `28c1cff2ace1b7e9...` | https://health-infobase.canada.ca/ccdss/Download.aspx?i=6&s=5&b=d&l=eng | 2026-09-30T03:12:07Z |
| `cihi_injury_trauma_2016-2017_2016-2017-injury-and-trauma-quick-stats-en-web.xlsx` | 150,311 | `8ec704a09e15df1b...` | https://www.cihi.ca/sites/default/files/document/2016-2017-injury-and-trauma-quick-stats-en-web.xlsx | 2026-09-30T03:12:10Z |
| `cihi_injury_trauma_2017-2018_2017-2018-injury-and-trauma-quick-stats-en-web.xlsx` | 145,459 | `a6f2d68d0e65f7b3...` | https://www.cihi.ca/sites/default/files/document/2017-2018-injury-and-trauma-quick-stats-en-web.xlsx | 2026-09-30T03:12:13Z |
| `cihi_injury_trauma_2018-2019_injury-trauma-emergency-dept-hospitalization-stats-2018-2019-en.xlsx` | 172,592 | `cf169a040578afd8...` | https://www.cihi.ca/sites/default/files/document/injury-trauma-emergency-dept-hospitalization-stats-2018-2019-en.xlsx | 2026-09-30T03:12:15Z |
| `cihi_injury_trauma_2019-2020_injury-trauma-emergency-dept-hospitalizations-stats-2019-2020-en.xlsx` | 209,216 | `48b3df0f9fcf61b5...` | https://www.cihi.ca/sites/default/files/document/injury-trauma-emergency-dept-hospitalizations-stats-2019-2020-en.xlsx | 2026-09-30T03:12:17Z |
| `cihi_injury_trauma_2020-2021_injury-trauma-emergency-dept-hospitalizations-2020-2021-data-tables-en.xlsx` | 304,561 | `8a77330b6d9984a6...` | https://www.cihi.ca/sites/default/files/document/injury-trauma-emergency-dept-hospitalizations-2020-2021-data-tables-en.xlsx | 2026-09-30T03:12:19Z |
| `cihi_injury_trauma_2021-2022_injury-trauma-emergency-dept-hospitalizations-2021-2022-data-tables-en.xlsx` | 246,005 | `982079e17e2f2df1...` | https://www.cihi.ca/sites/default/files/document/injury-trauma-emergency-dept-hospitalizations-2021-2022-data-tables-en.xlsx | 2026-09-30T03:12:21Z |
| `cihi_injury_trauma_2022-2023_injury-trauma-emergency-dept-hospitalizations-2022-2023-data-tables-en.xlsx` | 1,228,174 | `eab85fd02d134aa6...` | https://www.cihi.ca/sites/default/files/document/injury-trauma-emergency-dept-hospitalizations-2022-2023-data-tables-en.xlsx | 2026-09-30T03:12:24Z |
| `cihi_injury_trauma_2023-2024_injury-trauma-emergency-dept-hospitalizations-2023-2024-data-tables-en.xlsx` | 1,227,021 | `a4920374e2d1c19d...` | https://www.cihi.ca/sites/default/files/document/injury-trauma-emergency-dept-hospitalizations-2023-2024-data-tables-en.xlsx | 2026-09-30T03:12:26Z |
| `cihi_injury_trauma_2024-2025_injury-trauma-emergency-dept-hospitalizations-2024-2025-data-tables-en.xlsx` | 154,228 | `b91a0dfa03314869...` | https://www.cihi.ca/sites/default/files/document/injury-trauma-emergency-dept-hospitalizations-2024-2025-data-tables-en.xlsx | 2026-09-30T03:12:28Z |

Each file has a `.meta.json` sidecar (url, final URL after redirect, retrieval time, sha256, bytes, HTTP headers, terms, approval). The CCDSS endpoint `Download.aspx` now 301-redirects to `/ccdss/Download`; both are recorded. CIHI sizes equal the discovery HEAD sizes. Terms: CIHI Terms of Use (free use for non-commercial research); CCDSS licence text could not be re-read on 2026-09-29 (Health Infobase terms page 404, Canada.ca terms page did not respond) - recorded in the sidecar.

## Interpretation rule (documented)

Rule file `qa/negative_controls_20260925/3384_rule.json`, written 2026-09-29 22:18 CDT (2026-09-30T03:18Z) (file mtime 2026-09-29T22:18:45-05:00), before any fit (first fit run `fit_20260929_222454.log`, started 2026-09-29T22:24:54-05:00; the kept fits ran 2026-09-29T22:27:29-05:00 to 2026-09-29T22:28:58-05:00). SHA-256 `9aefddbba6bd4fae45e240ce2dcb2c9efd94e0f7ec77a0fc86ef85e77028cd13` (recorded at writing in `3384_rule.sha256`: `9aefddbba6bd4fae...`); all 25 fit receipts carry it (25 fitted by the current code); 2 fit log(s), 2 printing the same hash.

Classes, in order: UNINFORMATIVE if the control's 95% CrI log width > 2.5 x the matching autism CrI's on the same cells; else CONSISTENT if the CrI includes 1; else BIAS SIGNAL (direction reported). Placebo (future+3), pretrend, harm injection, two-sided family and DDD are reported beside and never change the class.

Plausibility. DOWNLOAD_REQUEST.md / the discovery record do not state that rotavirus vaccination cannot cause asthma. They say: 'Fallback only, because no public Canadian injury series fits the cells. Epilepsy and type 1 diabetes are not clean rotavirus negative controls, so asthma is the least-bad CCDSS choice.' The entry therefore ranks asthma as the cleanest available CCDSS condition; it does not argue implausibility. No established biological pathway from rotavirus vaccination to asthma is known to this task; expected r = 1. Specified caveats: (i) asthma shares the CCDSS case-finding machinery (physician billing and hospital discharge codes, provincial billing changes, COVID-period utilisation), which is what a same-system control is meant to detect, so a BIAS SIGNAL indicates non-specific system drift aligned with the launch timing, not a vaccine effect on asthma; (ii) any indirect pathway (e.g. fewer gastroenteritis-driven physician contacts at which asthma could first be coded) would be small and in the r < 1 direction; (iii) asthma is not an injury, so this row can never enter the 3330 C_INJ set (3330 sensitivity S3 'with non-injury' is its only home).

## Cells and cell losses

- Asthma incidence (CCDSS, 1-19): 166 of 166 CAN_noQC cells kept; lost 0.
- Asthma prevalence (CCDSS, 1-19): 166 of 166 CAN_noQC cells kept; lost 0.
- Quebec analogue: 24 of 24 Quebec CCDSS 1-19 cells (FY2000-01..2023-24) kept; lost 0.
- Asthma 1-19 is published for every CAN_noQC province-year (24 of 24 per province). The only suppressed asthma 1-19 cells in the export are CCDSS jurisdiction exclusions outside the block: Newfoundland and Labrador 8 of 24, New Brunswick 24 of 24, Yukon 10 of 24, Nunavut 5 of 24.
- Relative to QC_RV_ASD_cnics all 125 SISMACQ cells are lost (no CCDSS age bands for Quebec); the analogue is a different cell set.

## Results (reference setting, prior N(0, ln 2))

| context / cells | fit | r (95% CrI) | P(r > 1) | width ratio vs autism | class | 3330 harmonized (descr.) | future+3 placebo | pretrend lead |
|---|---|---|---:|---:|---|---|---|---|
| CAN_noQC, all 166 | autism (published, reproduced) | 0.99 (0.32-1.89) | 0.492 |  |  |  | PASS (1.60 (0.71-2.96), 145 paired) | PASS (1.02 (0.38-2.18), 104 cells) |
| CAN_noQC, 166 | **asthma incidence (system reading)** | 0.71 (0.48-1.06) | 0.036 | 0.45 | CONSISTENT | CONSISTENT | PASS (1.28 (0.72-1.93), 145 paired) | PASS (0.97 (0.45-1.65), 104 cells) |
| CAN_noQC, 166 | DDD autism / asthma incidence | 1.34 (0.47-2.20) | 0.830 |  |  |  | PASS (1.15 (0.46-2.42), 145 paired) | PASS (0.97 (0.35-2.09), 104 cells) |
| CAN_noQC, 166 | asthma prevalence (sensitivity) | 1.08 (0.86-1.39) | 0.779 | 0.27 | CONSISTENT | CONSISTENT | FAIL (1.26 (1.03-1.60), 145 paired) | PASS (1.21 (0.92-1.61), 104 cells) |
| Quebec CCDSS 1-19, 24 | autism prevalence (analogue refit) | 1.37 (0.33-5.34) | 0.668 |  |  |  | PASS (1.29 (0.31-5.03), 21 paired) | NOT TESTABLE (fewer than 20 untreated paired rows; 13 untreated cells) |
| Quebec CCDSS 1-19, 24 | asthma prevalence (analogue) | 0.92 (0.25-3.21) | 0.452 | 0.91 | CONSISTENT | UNINFORMATIVE | PASS (0.96 (0.26-3.43), 21 paired) | NOT TESTABLE (fewer than 20 untreated paired rows; 13 untreated cells) |
| Quebec CCDSS 1-19, 24 | DDD autism / asthma prevalence | 1.51 (0.35-6.05) | 0.711 |  |  |  | PASS (1.37 (0.32-5.36), 21 paired) | NOT TESTABLE (fewer than 20 untreated paired rows; 13 untreated cells) |

Published pooled posteriors (3320, not refitted): Canada Rotavirus (Excl. Quebec) 1.02 (0.46-1.77); Quebec Rotavirus 1.06 (0.75-1.45).

- Asthma incidence (CCDSS, 1-19): Kish grid points 40, boundary mass 1.94e-21, prior-dominated False, median on the same side of 1 as autism (0.99 (0.32-1.89)). Class under the 12 kernel-setting x prior rows: 10 of 12 CONSISTENT; rho_0_85|reference BIAS SIGNAL (0.65 (0.44-0.96) vs autism 0.97 (0.31-1.85)); rho_0_85|wide BIAS SIGNAL (0.64 (0.39-0.94) vs autism 0.88 (0.14-1.90)).
- Asthma prevalence (CCDSS, 1-19): Kish grid points 27, boundary mass 7.09e-32, prior-dominated False, median on the opposite side of 1 as autism (0.99 (0.32-1.89)). Class under the 12 kernel-setting x prior rows: 12 of 12 CONSISTENT.
- Quebec CCDSS 1-19 analogue (prevalence): Kish grid points 155, boundary mass 1.61e-18, prior-dominated True, median on the opposite side of 1 as autism (1.37 (0.33-5.34)). Class under the 12 kernel-setting x prior rows: 12 of 12 CONSISTENT.
- Asthma incidence harm injection (design sensitivity on this outcome): Asimov PASS (medians 1.03 / 1.54 / 2.05 / 5.16 for R = 1/2/3/5), plasmode PASS (R = 2 median 1.99), bootstrap pass rate 1.00. Two-sided family: strict PASS, Bonferroni PASS.
- Autism (published, reproduced) two-sided family: strict FAIL, Bonferroni PASS (strict failures: future4:reference 1.83 [1.04, 3.14]; future4:linear_only 1.83 [1.04, 3.14]; future4:spline_only 1.83 [1.04, 3.14]).

## Reproduction of the published CAN_noQC autism block

| check | result | passes |
|---|---|---|
| block input sha256 | `a4877115b377629b...` vs gate record `a4877115b377629b...` | True |
| curves (4 settings + variants) | max abs diff 4.06e-10 | True |
| unpooled rows | 0.992 (0.320-1.894); max rel diff 2.02e-12 | True |
| future+3 placebo (24 rows) | max rel diff 9.35e-12 | True |
| harm injection | max rel diff 0.00e+00; verdict PASS (published PASS) | True |
| two-sided family (5 members) | max rel diff 0.00e+00; membership identical True | True |
| data | autism y = log(2157 rate / 1e5) on every cell; asthma matched once per cell [166, 166]; independent csv parse equal [True, True]; sidecar hashes [True, True] | True |

Validated: **True**. Quebec analogue construction check: the analogue's exposure / nuisance code reproduces QC_RV_ASD_cnics's own exposure (max abs diff 0.0e+00) and nuisance columns (0.0e+00) on its 125 SISMACQ cells; the Quebec autism prevalence of the 2157 panel equals the raw CCDSS export: True. The autism pretrend lead is new here (the published record is NOT_TESTED).

## CIHI injury tables (CAN_CIHI_INJURY_HOSP_FY2016_2024)

9 workbooks, 87 data sheets; 30 with province/territory columns, 45 with age groups, 0 cross-classifying province x age. Age tables: Table 10  Sport-related brain injury emergency department visits by age and sex, Ontario and Alberta; Table 2  Injury hospitalizations by age and sex, Canada; Table 3  Injury hospitalizations by age and sex, Canada; Table 3  Unintentional fall emergency department visits by age, sex and place of occurrence, submitting provinces/territory,*; Table 4  Injury hospitalizations by age and sex, Canada; Table 5  Cause of sport and winter injury hospitalizations by age and sex, Canada; Table 6  Cause of sport and winter injury hospitalizations by age and sex, Canada; Table 6  Sport-related brain injury hospitalizations by age and sex, Canada; Table 7  Cause of sport and winter injury hospitalizations by age and sex, Canada; Table 7  Sport-related brain injury hospitalizations by age and sex, Canada; Table 8  Unintentional fall emergency department visits by age, sex and place of occurrence, submitting provinces/territory,*; Table 8a  Sport-related brain injury emergency department visits by age and sex, Ontario and Alberta; Table 8b  Sport-related brain injury hospitalizations by age and sex, Canada; Table 9  Sport-related brain injury emergency department visits by age and sex, Ontario and Alberta; Table 9  Unintentional fall emergency department visits by age, sex and place of occurrence, submitting provinces/territory,*. Fiscal years 2016-2017..2024-2025; 8 overlap the block's FY2000-2023. Inventory: `project_results/negative_controls_3330/3384/cihi/cihi_table_inventory.csv`. NOT FITTED: no province/territory x age cross-classification in any of the nine workbooks (province tables are all ages; age tables are Canada-only or Ontario and Alberta combined), so the identical CAN_noQC cells (province x ages 1-19 x FY) cannot be built; only FY2016-17..2023-24 would overlap the block in any case. Full coverage would need a fee-based CIHI custom data request (DAD/HMDB by province x age x FY), a user step.

## Caveats

- Asthma is a same-system non-injury control: it shares CCDSS case-finding (physician billing and hospital codes), so it detects ascertainment drift that lines up with the launch timing; it cannot detect autism-specific confounding (e.g. autism-service expansion).
- CCDSS counts are randomly rounded to multiples of 5 and rates are published as integers per 100,000 (incidence) or to 0.01% (prevalence); for asthma (415-1,893 per 100,000) rounding is negligible.
- The exposure scenarios include an age map built from national pre-2010 AUTISM case-age weights; it is kept unchanged for the control (identical exposure path), although asthma's own age profile differs.
- The Quebec analogue is one series of 24 cells; its autism fit is new (not the published QC estimate) and it shares the RAMQ 2016 billing change (a nuisance term) with QC_RV_ASD.
- Not obtained (user steps, not attempted): INSPQ SISMACQ asthma by the QC_RV_ASD age bands (not in the approved list) would give the identical QC cells; a CIHI custom request would give province x age injury hospitalizations.