Project 001 generates fictional NACRS and DAD records to study hospitalization and inpatient length of stay. The same NACRS cohort supports adult-asthma propensity-score matching, outcome regression, and hospital-level regional comparisons. Ten memos document the methods, findings, and limits.
The research
The original questions use CTAS, the Canadian Triage and Acuity Scale (levels 1–5, where level 1 is most urgent). A07–A09 use adults from the same NACRS file to study asthma, admission, and hospital-region differences.
Among emergency visits, do CTAS and the time before the disposition decision affect whether the patient is hospitalized?
Among hospitalized visits that link to a DAD record, do CTAS and total emergency stay affect inpatient length of stay in days?
Among comparable adult ED visits, how does admission risk differ when asthma is the main diagnosis?
Adults aged ≥18 are selected from the shared 30,000-visit NACRS file. Matching is exact within hospitals, with baseline propensity scores. Outcome models reuse the fixed pairs, and A09 fits hospital random-intercept logistic models for Calgary, Edmonton, and other regions, including surrounding towns.
CTAS is downstream of the episode in this design. The causal interpretation applies to the stated simulation; the effect is not a clinical finding.
Open asthma analysis (A07) Outcome regression (A08)The data
Every value is randomly generated. The records imitate the structure — the record grain, field families, and linkage — of two real Alberta administrative sources, without using any real patient information.
National Ambulatory Care Reporting System. Here, one row is an emergency department visit — arrival, triage, disposition timing, and whether the patient was admitted.
Discharge Abstract Database. Here, one row is an inpatient stay, including the derived length of stay. Question 2 uses the subset linked from a hospitalized NACRS visit.
The method
Each memo hands off to the next — from constructing the derived variables, through describing the files, to the two models and the summary.
The deliverables
Start with A06 for the original study summary, A07 for the adult-asthma matching simulation, A08 for post-matching outcome regression, and A09 for regional GLMM comparisons. Each card opens a rendered report.
Builds emergency stay, hospitalization, and inpatient length of stay from the source fields, and records each definition.
Open memo A01Descriptive summary of the analytic NACRS file — 30,000 records, 50 variables.
Open memo A02Descriptive summary of the analytic DAD file — 60,000 records, 64 variables.
Open memo A03 · Q1Logistic regression of hospitalization on pre-disposition stay and CTAS. Total stay and other covariates sit in the appendix.
Open memo A04Links hospitalized NACRS visits to DAD abstracts by PHN and a six-hour window, then compares linked with unlinked visits.
Open memo A05 · Q2Linear regression of inpatient days on emergency stay and CTAS in the linked cohort, with an acute-stay sensitivity analysis.
Open memo A06 · SummaryStudy design, data, statistical methods, key findings, and limitations across Memos A00 through A05.
Open memo A07 · AsthmaAdults from the shared NACRS cohort, within-hospital propensity-score matching, hospital-clustered admission risks, and known-effect validation.
Open memo A08 · Outcome modelsLogistic outcome models on A07's fixed pairs, standardized risks, hospital-clustered uncertainty, and repeated-simulation checks against known truth.
Open memo A09 · Regional GLMMThirty real ED hospital reference names; entirely synthetic patients and results. Hospital random intercepts, case-mix standardized risks, and an asthma-by-region interaction.
Open memo RepoPublic repository, variable dictionary, and the GitHub-safe reporting examples in code_examples/.
Open on GitHubSupporting reference
Project 001 is the project. The Alberta Health Data Atlas is only the reference used to keep the synthetic records realistic — it holds no patient-level values and is not a second project.
A Chinese–English guide to nine Alberta administrative datasets and clinical information systems. Project 001 uses it to decide record grain, field families, and linkage before generating the synthetic NACRS and DAD examples — it holds no patient-level values.
For developers
The working R Markdown and analytic data stay local, so the memos cannot be rebuilt from the GitHub copy alone. What is public are the sanitized, project-relative reporting examples in code_examples/ and the shared variable dictionary.
# install example dependencies Rscript code_examples/setup_dependencies.R # render the public example reports Rscript code_examples/render_all_rmd.R # place synthetic workbooks first: # code_examples/example_data/synthetic_nacrs.xlsx # code_examples/example_data/synthetic_dad.xlsx
Project & client
All names and organizations below are fictional. They do not identify real people, clients, employers, or institutions.
All data are randomly generated and entirely synthetic. They contain no real patient records, personal health information, client data, or operational information. Hospital names are real reference labels; patient records, activity levels, and results are simulated and do not describe those hospitals.
The data are for software demonstration, education, and portfolio presentation only. They must not be used for clinical, policy, financial, or operational decision-making.
The Alberta Health Data Atlas is an independent educational resource and is not an official Alberta Health Services, University of Calgary, Government of Alberta, or CIHI publication. It contains no patient-level data and does not copy the official data-dictionary workbooks.