Purpose
Compare your case exposures to typical population exposures from Foodbook to prioritise hypotheses during outbreak investigations.
How references are computed
-
References use Foodbook microdata with
Survey weights account for sampling design and ensure population-representative estimates from the Foodbook study
survey weights
(as in OMD's Stata workflow).
-
If multiple PTs are selected, a single
Weighted average across selected provinces/territories, maintaining population representativeness
combined reference
is computed across them.
- You can optionally limit the reference by Age Group (0-9, 10-19, 20-64, 65+) and Month.
- Defaults like "Canada" and "All" auto-deselect once you add other selections.
Analysis outputs
-
Proportion of cases exposed: (Yes + Probably) / (Yes + Probably + No)
Observed %
: (Yes + Probably) / (Yes + Probably + No) in your cases.
-
Weighted population exposure percentage from Foodbook survey data, filtered by your selected parameters
Reference %
: Weighted population exposure % from Foodbook for your selected filters (rounded to 1 decimal).
-
Statistical significance: probability of observing this exposure level if the outbreak cases were similar to the general population. Lower values indicate stronger association.
P-Value
:
Statistical test comparing case exposure proportion to population baseline (upper tail test)
Binomial test
of observed vs reference % (upper tail).
-
Alert (p≤0.05): Strong evidence of association. Borderline (p≤0.10): Moderate evidence. Not Significant (p>0.10): Weak evidence. Insufficient Data: Missing data or zero cases.
Classification
: Alert (≤0.05), Borderline (≤0.10), Not Significant, or Insufficient Data.
Advanced (CEDARS upload)
- Upload the CEDARS Excel export. Expected sheets: ‘case exposure answer’ and ‘Salmonella Case’.
- Columns are auto-detected even if the wording changes (we normalise names).
- We analyse confirmed cases if available in the linelist sheet.
Good practice
- Interpret alongside lab, trace-back, and epi linkage evidence.
- Small totals can be unstable; look for patterns across multiple signals.
- Foodbook is de-identified open data; no direct identifiers are used.
Roadmap
- Optional Bayesian results to complement binomial tests (per OMD collaboration).