Admissions Strengthening Plan (OMSAS / CanMEDS)¶

Created: 2026-02-25 Status: Active Scope: Real-world credibility, external validation, and public-facing artifacts to strengthen the project's presentation for Ontario medical school admissions.

Current State Assessment¶

Purpose: Wait Time Canada is a Health Systems Observatory that audits methodological inconsistencies in Canadian ER wait-time reporting across 4 provinces (ON, QC, AB, BC), using a strict metric ontology to tag every measurement and auto-block invalid cross-hospital comparisons via divergence warnings.

Target users: Framed as a medical school portfolio project, but has genuine utility for patients making ER decisions, ER clinicians understanding reporting context, and health policy researchers studying interprovincial heterogeneity.

Maturity: Engineering is highly mature — 33 milestones complete, 777+ tests, full CI/CD, OpenAPI spec, MkDocs deployment, French i18n, Ontario equity layer with 2021 StatsCan census tracts, divergence briefs, historical occupancy trends, structured failure taxonomy.

Evidence of use: Still minimal. There are no completed stakeholder interviews yet, no published testimonials, no user analytics, no external citations, and no GitHub community signals. The placeholder Zenodo DOI has been removed from README, but no real DOI has been registered yet. The important launch gate is now cleared: https://wait-time.ca is live with valid HTTPS and verified redirects, so Phase B can focus on public-facing follow-through rather than infrastructure rescue.

Biggest gaps: 1. No meaningful public engagement yet — the canonical URL is live, but stakeholder interaction and outward-facing launch activity still lag the engineering maturity. 2. No real-world human engagement — zero completed stakeholder conversations. 3. No published academic or semi-academic output beyond internal drafts. 4. No real Zenodo DOI has been activated yet, so the citable-software path remains parked. 5. The health equity analysis still covers only Ontario — the Advocate claim remains narrower than the rest of the platform. 6. Several outward-facing items are intentionally parked by choice (A3, A4, B3, outward C1 activation), which keeps quality high but also keeps external signal lower than the engineering maturity.

Current Posture¶

The remaining human-facing items now fall into three buckets:

1. Parked by choice
2. A3 Zenodo DOI
3. A4 stakeholder outreach
4. B3 LinkedIn launch post
5. outward C1 publication/review activation
6. Internal-only assets
7. C1 technical report draft
8. C3 equity interpretation inside that draft
9. Conditional / later
10. B2 first-party privacy-safe telemetry, but only after the VPS/logging path is mature enough to justify it
11. C2 targeted researcher outreach, but only if outward C1 activation happens later
12. C6 named reviewer acknowledgement, only if future outreach naturally produces it

This means the plan remains useful, but most remaining work is now governed by activation decisions rather than immediate execution.

Guiding Principles¶

1. Phase A items are all pre-deployment. No blockers. Start immediately.
2. Stakeholder interview (#A4) and README mission/equity/stewardship section (#A1) are the two highest-ROI actions. A1 is delivered. A4 remains high-value but should stay parked until explicitly activated.
3. The technical report (C1) is the anchor scholarly artifact. The case study, findings, and equity interpretation feed into it.
4. Don't publicize anything you can't defend. Every metric, finding, and claim must survive "how do you know that?" from a skeptical interviewer.
5. The live URL enables Phase B but does not automatically create value. The value comes from the content (case study, findings, limitations) already visible when someone visits.
6. Phase D items are genuinely optional. They improve the software but have diminishing admissions returns compared to Phases A–C. Don't sacrifice A–C quality by starting D early.

Phase A: Immediate Credibility Cleanup¶

A1: Mission, Equity, and Stewardship Section in README¶

Priority: 1 (tied with A4) | Effort: 30 min | CanMEDS: Professional, Communicator

Replace the maintainer-oriented README blurb with a short public-facing section that states the project's mission, health/data equity and EDIA posture, barrier-reduction aims, and stewardship standards. The goal is to make the site read like a serious public-interest health data observatory rather than a personal portfolio or pet project.

Artifact: ## Mission, ## Equity and Access, and ## Stewardship sections in README.

Status: Delivered 2026-03-28.

A2: Limitations Section in README¶

Priority: 1 | Effort: 1 h | CanMEDS: Scholar, Professional

Add specific, technical limitations to README. Keep limitations honest and specific rather than vague: - Scraper data now reflects an hourly GitHub Actions cadence rather than continuous ingestion. - Equity layer uses 2021 census income data, not current. - Methodology labels are inferred from provincial documentation and may not match internal hospital reporting practices. - Platform cannot account for unreported overcrowding events.

Most student projects overclaim. Honest limitations are a differentiated signal that demonstrates command of the material.

Artifact: ## Limitations section in README and docs/application-summary.md.

A3: Fix Placeholder Zenodo DOI¶

Priority: 2 | Effort: 90 min | CanMEDS: Scholar

The placeholder DOI has been removed from README, which is safer than displaying a fabricated citation signal. The remaining task is to register a real Zenodo deposit, update CITATION.cff, and restore the README badge only once the DOI is real.

Status: Repo and release preparation completed on 2026-03-28; real DOI still pending Zenodo account activation and publication. Parked until explicit user go-ahead.

A real Zenodo record is third-party-attested, timestamped evidence of the project's existence and authorship — not just a badge.

Artifact: Real Zenodo record with DOI; updated CITATION.cff and README badge.

A4: Stakeholder Interview Outreach¶

Priority: 1 (tied with A1) | Effort: 2–3 h outreach; interviews on respondents' schedules | CanMEDS: Collaborator, Health Advocate

Status: Parked pending explicit user go-ahead.

Begin outreach to ER professionals using the existing templates in docs/stakeholder-interviews/. Aim for 10–15 contacts. The conversation, not the live URL, is the point — these interviews are about methodology, so no deployment is needed.

One genuine 15-minute conversation with an ER nurse or physician, honestly documented, is the target. Don't pad it. Don't inflate a 5-minute hallway chat into "stakeholder validation."

Artifact: Filled docs/stakeholder-feedback.md with date-stamped entries; potentially a published testimonial in frontend/content/stakeholderTestimonials.ts; a logged outreach campaign describable in an ABS entry.

A5: Private Reflection Document¶

Priority: 3 | Effort: 2 h | CanMEDS: Professional, Scholar

Status: Intentionally skipped on 2026-03-28. Kept out of active scope because the value is mostly private interview prep, not a public-facing credibility artifact, and the time is better spent elsewhere unless interview preparation needs change later.

Write a private document (not committed to the repo) covering: 1. What motivated the project — the specific insight or experience. 2. One technical decision you'd revisit. 3. One clinically relevant surprise from the data. 4. One ethical tension you navigated (e.g., surfacing raw occupancy numbers that could cause patient panic). 5. What competencies you think this demonstrates.

This is interview preparation that produces articulate, reflective answers. The document is not the artifact — the interview performance is.

Artifact: None for now. Revisit only if future interview preparation clearly needs it.

A6: Ottawa–Gatineau Case Study¶

Priority: 2 | Effort: 2–3 h | CanMEDS: Scholar, Communicator, Health Advocate

Status: Delivered 2026-03-28.

Write a 2-page case study documenting the Ottawa–Gatineau methodology divergence. This is the project's clearest, most narratively compelling story: two cities across a provincial border, served by hospitals in two different provincial health systems, with patients who routinely cross the river for care — and reported wait times that measure fundamentally different things.

Before writing: Verify that Ottawa-area and Gatineau-area hospitals are active in the database (5 min query). If they're missing, identify the best alternative cross-border pair.

Structure: Context (border geography, patient behavior) → Specific metrics from each province → Why comparison is invalid (field-by-field ontology mismatch) → Clinical implication for patients → What a valid comparison would require.

Artifact: docs/case-studies/ottawa-gatineau-divergence.md; linked from README and application summary; interview-ready material.

A7: Quantified Metrics & Methodology Findings¶

Priority: 5 | Effort: 3 h | CanMEDS: Scholar, Leader

Status: Delivered 2026-03-28.

Query the database for the strongest honest metrics. Add a "By the Numbers" section to README.

Important scope constraint: Findings must be methodology-characterization only — no cross-province performance claims. The project's premise is that cross-province comparisons are invalid; writing findings that implicitly compare performance would be self-undermining.

Safe examples: - "X measurements collected since launch across 380+ hospitals." - "100% of cross-province hospital pairs have ontology mismatches on ≥2 dimensions." - "Quebec hospitals report stretcher occupancy >110% in X% of measurements."

If the raw numbers are still early or modest, lead with operational metrics (uptime, measurement count, anomaly detection events) rather than impact metrics.

Artifact: "By the Numbers" section in README; specific figures for ABS entries.

Phase A total: mostly delivered. Remaining Phase A work is intentionally parked (A3, A4) rather than active.

Phase B: Launch Follow-Through (Post-Verification)¶

B1: Restore Production Hosting¶

Priority: 0 (gate) | Effort: 2–3 h | CanMEDS: Leader

Delivered on 2026-03-12: custom-domain TLS + redirect validation completed, production smoke passes against https://wait-time.ca, and the canonical URL is now trustworthy.

Artifact: Clean HTTPS + redirect behavior at https://wait-time.ca; passing production smoke test against the canonical URL.

B2: Privacy-Safe Usage Analytics¶

Priority: 3 | Effort: 2 h | CanMEDS: Professional, Privacy

Do not add paid or third-party analytics. If this work is activated later, implement only minimal first-party aggregate usage telemetry from the direct-VPS path once that runtime is the settled long-term baseline for the project. The goal is stewardship, not marketing instrumentation.

Status: Re-scoped on 2026-03-28 away from Plausible/third-party analytics. Keep on the roadmap as a later VPS/logging stewardship item, not an active near-term deliverable.

Guardrails: No cookies, no cross-site tracking, no user profiling, no advertising identifiers, and no vendor cost. Prefer aggregated ingress/log-derived reporting over client-side scripts.

Artifact: First-party aggregate usage telemetry with an updated privacy policy once activated.

B3: Publish LinkedIn Launch Post¶

Priority: 3 | Effort: 30 min | CanMEDS: Communicator, Leader

The canonical URL is now live. Publish using https://wait-time.ca. Spend no more than 30 minutes total. The value is a public timestamp of work and asymmetric upside if it reaches health informatics professionals.

Status: Copy refreshed on 2026-03-28 to match the current mission/equity/stewardship posture and live Batch A feature set. Manual publication is parked until explicit user go-ahead.

Artifact: Public, timestamped LinkedIn post.

B4: Video Walkthrough¶

Priority: 3 | Effort: 3 h | CanMEDS: Communicator

Status: Intentionally skipped on 2026-03-28. Kept out of active scope because the time cost and presentation burden are not justified right now relative to other work, and a mediocre walkthrough would be worse than having none.

Script a 3–4 minute walkthrough. Record with decent audio (headset mic is sufficient). Cover: landing page → select two incomparable hospitals → divergence warning → methods page → equity layer. Upload as unlisted YouTube or Loom link.

This is the only format where a committee hears your voice and experiences your command of the material. Prioritize audio quality over video quality. A bad video is worse than none.

Artifact: None for now. Revisit only if future application/interview needs make the payoff clearer.

B5: GitHub Topics & Discoverability¶

Priority: 8 | Effort: 30 min | CanMEDS: Leader

Add repository topic tags (health-systems, emergency-medicine, wait-times, canada, data-observatory, open-data). Check 1–2 open data registries for activity; submit if active.

Status: Delivered 2026-03-28. The GitHub repository now points its homepage to https://wait-time.ca and uses a cleaner subject-matter topic set (canada, emergency-medicine, health-informatics, health-systems, open-data, public-health, data-observatory, wait-times, nextjs, postgresql). A quick registry screen did not justify an immediate external submission: DataPortals.org is aimed at open data portals rather than this kind of project, and DataHub is oriented toward packaged open datasets rather than a live observatory site.

Artifact: GitHub topic tags and homepage link; no external directory submission for now.

Phase B total: mostly delivered or intentionally deferred. Remaining work is a future telemetry decision (B2) plus the parked manual launch post (B3).

Phase C: Weeks 2–4 (Scholarly Artifacts & External Validation)¶

C1: Technical Report¶

Priority: 4 | Effort: 1–2 d | CanMEDS: Scholar, Health Advocate

Write a 2–4 page structured report: "Methodological Heterogeneity in Canadian Emergency Department Wait-Time Reporting: A Four-Province Audit."

Status: Draft completed on 2026-03-28 at docs/research/methodological-heterogeneity-four-province-audit-draft.md. Keep it as an internal review draft for now. Do not treat it as a finalized public scholarly artifact or push it outward until the reviewer/outreach path is explicitly reactivated.

Check existing files first: docs/'Canadian ER Wait Time Data Audit.docx' may be partially complete.

Structure: Background → The Measurement Problem → Methodology (metric ontology) → Findings (specific divergences, within-province observations) → Equity Layer Analysis (descriptive association → limitations → required study design) → Limitations → Implications.

Quality gate: Do not publish until at least one informed person has read it (from A4 interviewee or C2 researcher). Publish via SSRN or as a GitHub Pages document. The venue matters less than the quality.

The comparison class for "Scholar" is other pre-med students, not JAMA. No other pre-med applicant will have a written methodological analysis of interprovincial ER data reporting.

Artifact: Internal draft for now. Only becomes a public, citable document with a stable URL if the outward review/publication path is explicitly activated later.

C2: Targeted Researcher Outreach¶

Priority: 6 | Effort: 3 h + response time | CanMEDS: Collaborator, Scholar

Identify 3 specific researchers at ICES, INSPQ, or university health informatics labs whose published work relates to ER methodology. Reference their specific papers. Share the technical report (C1). Ask for 15–20 min methodology feedback.

Don't cold-email Ontario Health's general inbox — target individuals whose work you've read. Canadian health informatics is a small community where informed, targeted outreach can get a response.

Status: Conditional only. This item should activate only if the technical report is intentionally moved outward for external review or publication.

Artifact: Documented outreach log; potentially a named reviewer in Acknowledgements.

C3: Equity Layer Interpretive Summary¶

Priority: 6 | Effort: Merged with C1 | CanMEDS: Scholar, Health Advocate

Write as a section of the technical report (C1), not standalone. The limitations ARE the finding. Structure as: descriptive association → four specific reasons this cannot be interpreted causally (ecological fallacy, 2021/2025 temporal mismatch, confounders, census tract ≠ neighborhood) → what study design would answer the causal question.

A structured analysis of what you cannot conclude from this data is a stronger Scholar signal than a finding that ignores those limits.

Status: First draft section completed on 2026-03-28 inside docs/research/methodological-heterogeneity-four-province-audit-draft.md. Keep it internal-only with C1 for now; final status remains tied to later review/publication.

Artifact: Equity analysis section in technical report.

C4: Operational Transparency Report¶

Priority: 7 | Effort: 2 h | CanMEDS: Professional, Leader

Generate one monthly report from data_quality_snapshots and scraper_status tables. Be honest about the current hourly GitHub Actions runtime and any alert/recovery incidents. Store in docs/operations/reports/. The value is behavioral — demonstrating the practice of systematic operational review, not the specific numbers.

Status: First report completed on 2026-03-28 at docs/operations/reports/2026-03-operational-report.md.

Artifact: docs/operations/reports/YYYY-MM-operational-report.md.

C5: Incident Post-Mortem¶

Priority: 8 | Effort: 1–2 h | CanMEDS: Professional

Write the Quebec zero-value parser incident post-mortem (commit 0738054). Timeline, impact, root cause, resolution, follow-up. Store in docs/operations/incident-reports/.

Write it for the interview answer it produces: "What was the hardest problem?" needs a specific, honest answer. The document itself is secondary.

Artifact: docs/operations/incident-reports/YYYY-MM-DD-quebec-zero-value.md; interview preparation material.

Status: Delivered 2026-03-28 at docs/operations/incident-reports/2026-02-19-quebec-zero-value.md.

C6: Named Reviewer Acknowledgement¶

Priority: 7 | Effort: 30 min if applicable | CanMEDS: Collaborator

If A4 (stakeholder interview) or C2 (researcher outreach) produces someone willing to be acknowledged as a methodology reviewer, add an Acknowledgements section to README. Do not pursue independently — this is a natural extension of Items A4 and C2, not a separate task.

Don't inflate titles. A 20-minute methodology review conversation makes someone a "reviewer," not an "advisor."

Artifact: Acknowledgements section in README (conditional).

Status: Conditional only. Do not pursue independently.

Phase C total: partially delivered. C1/C3 exist as internal drafts, C4/C5 are delivered, C2 remains conditional on future outward C1 activation, and C6 remains conditional on future reviewer activity.

Phase D: If Time Permits (Weeks 5–12)¶

These items improve the project as software but have diminishing admissions returns. Only pursue after Phases A–C are complete.

D1: Comparability Matrix Upgrade¶

Status: Delivered on 2026-03-28.

The /methods comparability matrix now provides explicit pairwise field-by-field verdicts for all unique province pairs, including match/mismatch badges for each ontology dimension, match-count summaries, and short clinical-implication text. The compare API now reuses the same shared comparability utility so divergence language and methodology verdicts remain consistent across surfaces.

D2: Quebec Equity Layer Extension¶

Effort: M–L | Condition: Only after Phases A–C complete. Must apply full Ontario-equivalent rigor (ADR, suppression provenance, uncertainty bounds, non-causal framing).

A methodology that works for one province is interesting. A methodology with a proven framework that can be applied province-by-province is a system. But a rushed extension that cuts corners undermines Ontario's credibility.

D3: Per-Hospital Data Freshness Indicators¶

Status: Delivered already; confirmed during roadmap reconciliation on 2026-03-28.

The hospital API already exposes last_updated, the public map/list/details surfaces already render freshness cues, and raw export already includes timestamp_utc. No additional feature work is needed unless a future pass wants separate freshness treatment for aggregated exports.

D4: Nova Scotia Scraper (M35)¶

Effort: L | Condition: Only if NS methodology is confirmed to add a genuinely new ontology combination AND all higher-priority items are complete.

5 vs 4 provinces is not notably more impressive unless NS methodology is novel. The time cost (1–4 weeks) is disproportionate to the marginal admissions value.

D5: Grant / Competition Application¶

Effort: M | Condition: Opportunistic only. Do not actively search. Apply if a suitable, eligible competition is discovered during other activities.

Current Outcome¶

As of late March 2026, this plan has produced:

• a live, professional URL at wait-time.ca
• a clear public-facing mission, equity/EDIA, and stewardship statement
• an honest limitations section
• a concrete Ottawa–Gatineau case study
• quantified methodology findings grounded in real platform data
• a technical report draft retained internally for now
• one monthly operational report
• one incident post-mortem for interview preparation
• deliberately skipped lower-value items (A5, B4)
• deliberately parked outward-facing items (A3, A4, B3, outward C1/C2)

The remaining value in this plan is now mostly optional activation work rather than missing foundational artifacts.

Analytical Notes¶

This plan is the product of three analytical passes: 1. Original plan: 24 improvements identified, sorted by admissions value. 2. Devil's advocate pass: Each item challenged for execution risk, overstated claims, and unfavorable comparison classes. 3. Steelman pass: Each item defended with the strongest possible case, identifying overlooked value. 4. Final synthesis: Both sides weighed objectively; items re-prioritized based on defensible value and effort.

Key insights from the analysis: - The engineering foundation is already excellent. The gap is almost entirely in real-world engagement, external validation, and public-facing artifacts. - Items 2 (stakeholder interview) and 8 (README mission/equity/stewardship section) are the two highest-ROI actions. A1 is now delivered; A4 should remain parked until explicitly activated. - The technical report (C1) is the anchor scholarly artifact — the case study, findings, and equity interpretation all feed into it. - Usage analytics should remain a later, first-party VPS/logging decision rather than an immediate post-launch task. - Phase D items are genuinely optional for admissions purposes and should not compete with Phase A–C execution.