# AIRS Longitudinal — Completed Exemplar BRD

> **What this is**: A fully authored research-grade BRD. Every section is filled in to the level of detail a real principal investigator would sign off on. It is the *finished state* a BRD reaches when the discipline has been applied end to end.
>
> **How to use it**: Read it as the worked example for what a complete BRD looks like before attempting the four starter exercises ([CloudRevenue](cloudrevenue-starter.md), [M365Marketing](m365marketing-starter.md), [SupportInsights](supportinsights-starter.md), [EnterpriseGovernance](enterprisegovernance-starter.md)). The point is not to match this BRD field for field; it is to see what *done* looks like before you try to recognize *not yet done* in your own draft.

---

## Business Requirements Document

### AIRS Longitudinal Study — Cohort Outcomes Dashboard

- **Project**: Longitudinal outcomes dashboard for the AIRS (AI in Research Studies) cohort, supporting hypothesis testing across the Year 1 through Year 5 follow-up windows
- **Stakeholder**: Research Principal Investigator
- **Date**: 2026-04-12
- **Priority**: High

### 1. Executive Summary

The AIRS Longitudinal Study tracks a cohort of 1,847 participants enrolled between 2021 and 2023, with planned follow-up at Years 1, 3, and 5 post-enrollment. This dashboard supports the principal investigator and the study's senior research team in monitoring participant retention, evaluating outcome distributions across the four pre-registered hypotheses, and identifying subgroups that warrant deeper analysis before the Year 3 publication window closes. The dashboard is the operational artifact behind the planned 2026 publication submission and the routine quarterly investigator reviews that drive cohort-management decisions.

### 2. Business Context

The Year 3 follow-up window completed enrollment-side data collection in Q4 2025. The investigator team is preparing the primary publication for submission in Q3 2026 and requires a consolidated dashboard to support hypothesis testing, subgroup exploration, and the senior-author review process. Current tooling consists of disconnected R notebooks and spreadsheet pivots maintained by individual research assistants; the dashboard consolidates these into a single source of truth, removes the cross-RA reconciliation overhead, and standardizes the visualizations used in investigator meetings.

### 3. Key Business Questions

***Retention and cohort integrity***

1. What percentage of the Year 0 cohort completed the Year 3 follow-up assessment, broken down by enrollment site and by demographic stratum?
2. Among participants lost to follow-up between Year 1 and Year 3, what is the distribution of last-known status (withdrew, unreachable, deceased, moved out of region)?

***Pre-registered hypothesis testing***

1. For each of the four pre-registered hypotheses (H1–H4), what is the effect size in the Year 3 cohort, with 95% confidence intervals?
2. Are the H1 and H2 effects robust to the planned sensitivity analyses (complete-case versus multiple-imputation; per-protocol versus intention-to-treat)?

***Subgroup exploration (post-hoc, flagged as exploratory)***

1. Does the H1 effect vary across the three pre-specified subgroups (enrollment cohort year; baseline severity tertile; site type)?
2. Are there interaction effects between H2 and H3 that warrant a separate exploratory analysis in the discussion section?

### 4. Success Criteria

| Criterion | Measurement |
| --- | --- |
| **Publication-readiness** | All Table 1 (cohort characteristics), Table 2 (primary outcomes), and Table 3 (sensitivity analyses) values in the planned publication can be reproduced from the dashboard with one click each. PI confirms reproducibility against the analysis statistician's R output. |
| **Investigator-meeting cadence** | The dashboard is the primary visual artifact in the quarterly investigator review meetings, replacing the prior practice of slide compilation. Confirmed by zero slide-prep hours in the two meetings following deployment. |
| **Subgroup-exploration speed** | A senior investigator can filter to a pre-specified subgroup and see all primary-outcome panels updated within three seconds. Tested against deuteranopia-simulated views to confirm color-encoded subgroup distinctions remain accessible. |
| **Audit trail** | Every value displayed has a one-click drill-through to the underlying participant-level rows (with PHI redacted at the display layer). Confirmed by the data-governance officer's spot-check of ten randomly selected display values. |

### 5. Stakeholder Requirements

#### Primary Users

- **Principal Investigator**: Daily use during the publication-prep window (Q2–Q3 2026); weekly otherwise. Uses the dashboard for hypothesis-test result review, subgroup exploration, and publication-table reproduction.
- **Senior Investigators (n=4)**: Weekly use during the publication-prep window. Use the dashboard during the quarterly investigator review meetings and for independent verification of hypothesis-test results before approving the manuscript draft.
- **Analysis Statistician**: Bi-weekly use. Uses the dashboard to verify that the published values match the R analysis output; identifies discrepancies for reconciliation.

#### Secondary Users

- **Research Coordinators (n=3)**: Monthly use. Reference the retention dashboards for cohort-management discussions; do not use the hypothesis-testing surfaces.
- **Data Governance Officer**: Quarterly use. References the audit-trail surface during the standing data-governance review.

### 6. Data Requirements

| Requirement | Detail |
| --- | --- |
| **Data Sources** | (1) AIRS REDCap study database (participant-level outcomes, Years 0–3); (2) AIRS enrollment and consent ledger (REDCap); (3) site-level enrollment registry (Excel, maintained by the lead research coordinator); (4) externally-linked vital-status registry (quarterly extract, restricted access) |
| **Refresh Frequency** | Daily refresh during the publication-prep window (Q2–Q3 2026); weekly otherwise. The vital-status extract refreshes quarterly with the registry's release cadence. |
| **Historical Horizon** | Full study horizon: enrollment date forward through current Year 3 follow-up window. Year 5 follow-up data will extend the horizon as it arrives (planned 2027–2028). |
| **Data Sensitivity** | Restricted. Contains PHI (date of birth, enrollment site, outcome assessments). PHI is masked at the display layer; only the underlying participant-level drill-through (gated to PI + Analysis Statistician roles) reveals identifiable fields. Display-layer aggregations are de-identified per the study's HIPAA Safe Harbor implementation. |
| **Known Quality Issues** | Site 4 enrolled later (2023 vs the cohort baseline 2021–2022) and has thinner Year 3 follow-up density; flag in cohort-characteristics tables. The vital-status registry has a 60-day lag for newly-deceased participants; flag the lag in the retention panels. |

### 7. Deliverables

| Item | Detail |
| --- | --- |
| **Format** | Power BI workspace with three reports: (1) Cohort Integrity, (2) Primary Hypothesis Testing, (3) Subgroup Exploration. Each report has an associated dataset with row-level security gating PHI access to PI + Analysis Statistician roles. |
| **Access Method** | Power BI Service, deployed to the AIRS Research workspace. Investigators access via direct workspace membership; research coordinators access via a published app with the PHI surfaces hidden. |
| **Mobile Required** | No. Investigator review uses laptop or desktop only. Mobile view explicitly deferred to a future iteration. |
| **Target Go-Live** | 2026-05-15 (six weeks before the planned manuscript-draft circulation date) |
| **Review Cadence** | Quarterly investigator review (replaces prior slide-based meeting format). Dashboard reviewed against the manuscript Table 1–3 values at each pre-submission checkpoint. |

### 8. Constraints & Assumptions

- **Constraint**: All PHI display rules must be reviewed and signed off by the Data Governance Officer before deployment to the AIRS Research workspace.
- **Constraint**: The dashboard cannot expose the externally-linked vital-status registry to any user outside the PI + Analysis Statistician roles; the registry's data-use agreement requires named-individual access logging.
- **Constraint**: The dashboard must support reproducibility of all values that appear in the planned publication; any disagreement between the dashboard and the analysis statistician's R output is a release blocker.
- **Assumption**: Site 4's thinner Year 3 follow-up density will not invalidate the primary hypothesis tests; sensitivity analyses (with and without Site 4) will be reported in the manuscript supplement.
- **Assumption**: The Year 5 follow-up data will extend the existing dashboard schema rather than requiring a separate report; the schema is designed to accept additional time-window columns.

### 9. Research Ethics (study-specific extension)

| Item | Detail |
| --- | --- |
| **IRB protocol** | AIRS Protocol v4.2, approved 2026-01-08, valid through 2027-01-08 |
| **Data Use Agreement** | DUA-2023-AIRS-VS for vital-status linkage (named-individual access log required; quarterly compliance review) |
| **PHI handling** | HIPAA Safe Harbor de-identification at display layer; participant-level drill-through restricted to PI + Analysis Statistician via row-level security; access logged via Power BI audit trail; logs reviewed quarterly by Data Governance Officer |
| **Publication ethics** | All exploratory subgroup analyses are flagged as post-hoc in both the dashboard surface and the manuscript; pre-registered hypotheses (H1–H4) are reported regardless of result direction; null results receive identical visual treatment to positive results |

### Approval

| Role | Name | Date | Signature |
| --- | --- | --- | --- |
| Principal Investigator | [PI Name] | | |
| Senior Investigator (analytic lead) | [Sr. Investigator Name] | | |
| Analysis Statistician | [Statistician Name] | | |
| Data Governance Officer | [DGO Name] | | |
| Lead Analyst (BRD author) | [Analyst Name] | | |

---

**Source**: Adapted from Appendix E § "The reference BRDs" of *The Defensible Decision*. The AIRS dataset is the completed exemplar; the four starter BRDs ([CloudRevenue](cloudrevenue-starter.md), [M365Marketing](m365marketing-starter.md), [SupportInsights](supportinsights-starter.md), [EnterpriseGovernance](enterprisegovernance-starter.md)) ship with Sections 1–3 in place and Sections 4–8 left for the reader to complete using the transcript-driven elicitation loop.