Epic: AI-surfaced clinical deltas for clinicians
F02-001
Surface lab trend anomalies
Clinician5 SP
As a clinician, I want the system to flag lab values trending outside normal ranges over the last 3 visits so that I can identify deteriorating conditions without manually comparing historical results.
Acceptance Criteria
Given a patient's HbA1c has risen from 6.2 to 7.1 to 8.4 over 3 visits, When I open their record, Then I see an amber "Trending Up" alert on HbA1c with a sparkline showing the trajectory.
Given all lab values are within normal range and stable, When I open the record, Then no alerts are shown and the labs section displays "No significant changes."
F02-002
Detect medication interaction risks
Clinician5 SP
As a clinician, I want the system to cross-reference active medications against known interaction databases so that I'm alerted to potential adverse interactions before the patient visit.
Acceptance Criteria
Given a patient is on Warfarin and a new prescription for Amiodarone is added, When I open their medication list, Then a red "Severe Interaction" alert appears with the specific risk (increased bleeding).
Given a patient has no known interactions, When I view the medication list, Then a green "No Interactions Detected" indicator is shown.
F02-003
Imaging change summary between studies
Clinician8 SP
As a clinician, I want a side-by-side comparison summary when new imaging is available versus the previous study so that I can quickly assess progression or improvement without reading full radiology reports.
Acceptance Criteria
Given a patient has a new chest CT and a prior CT from 6 months ago, When I open the imaging section, Then I see a delta summary highlighting key changes (e.g., "Nodule: 8mm → 12mm, growth detected").
Given no prior imaging exists for comparison, When I view the study, Then it displays "Baseline study — no prior available for comparison."
F02-004
Clinical delta notification preferences
Clinician3 SP
As a clinician, I want to configure which types of clinical deltas I receive alerts for so that I'm not overwhelmed by low-priority notifications and can focus on what matters to my specialty.
Acceptance Criteria
Given I am an endocrinologist, When I configure my alert preferences, Then I can enable HbA1c, glucose, and thyroid panel alerts while disabling imaging and cardiac alerts.
Given I have disabled imaging alerts, When a patient gets new imaging, Then I do not receive a notification but the data is still accessible in the full record.
Epic: Plain-language health updates for patients
F02-005
Translate lab results to plain language
Patient5 SP
As a patient, I want my lab results explained in plain English with severity indicators so that I understand what each value means for my health without needing medical training.
Acceptance Criteria
Given my HbA1c result is 8.4%, When I view it on the Health Updates tab, Then I see "Your average blood sugar over 3 months is elevated" with a red severity indicator and a recommendation to discuss with my provider.
Given all my labs are normal, When I view Health Updates, Then I see "All results within normal range" with a green indicator.
F02-006
Medication change plain-language summary
Patient3 SP
As a patient, I want to see a clear summary when my medications change so that I understand what was added, removed, or adjusted and why.
Acceptance Criteria
Given my Metformin dose was increased from 500mg to 1000mg, When I view Health Updates, Then I see "Metformin dose increased — your provider adjusted this to help manage your blood sugar levels."
Given no medication changes occurred since my last visit, When I view Health Updates, Then the medications section shows "No changes since [last visit date]."
F02-007
Health update severity triage
Patient3 SP
As a patient, I want my health updates sorted by severity (critical, attention needed, informational) so that I see the most important items first.
Acceptance Criteria
Given I have a critical lab result and a routine medication refill, When I open Health Updates, Then the critical item appears at the top with a red banner, and the routine item appears below with a blue informational indicator.
Epic: Real-world evidence from consented longitudinal data for buyers
F02-008
Longitudinal data assembly for query results
Buyer5 SP
As a data buyer, I want query results to include longitudinal patient histories (not just point-in-time snapshots) so that my models can track disease progression and treatment effectiveness over time.
Acceptance Criteria
Given I submit a Dataset query for "Type 2 Diabetes outcomes," When the results are assembled, Then each patient record includes chronological visits, lab trends, medication changes, and outcomes spanning the full consent window.
Given a patient revoked consent partway through, When their data is included, Then only the consented time period is present and a "Consent window: [start] to [end]" field is included.
F02-009
Data quality scoring per patient record
Buyer3 SP
As a data buyer, I want each patient record scored for completeness and data quality so that I can filter for high-quality records and understand the reliability of my dataset.
Acceptance Criteria
Given a patient has labs, meds, diagnoses, and imaging all populated, When their record is scored, Then it receives a "High" quality score (90%+).
Given a patient only has diagnoses and no labs or meds, When their record is scored, Then it receives a "Low" quality score (below 50%) with a breakdown showing missing categories.
F02-010
De-identification verification for dataset exports
Compliance5 SP
As a compliance officer, I want every dataset export to pass automated de-identification checks (Safe Harbor or Expert Determination) so that no PHI leaves the platform in identifiable form.
Acceptance Criteria
Given a Dataset query result is ready for export, When the de-identification pipeline runs, Then all 18 HIPAA Safe Harbor identifiers are removed or generalized and a certification report is attached.
Given a record contains a rare diagnosis that could be re-identifying in a small cohort, When the k-anonymity check runs, Then the record is suppressed if the cohort size falls below the k threshold (default k=5).
F02-011
Consent provenance metadata in query results
Buyer2 SP
As a data buyer, I want each record in my query results to include consent provenance metadata so that I can demonstrate to regulators that every data point was collected with first-party consent.
Acceptance Criteria
Given I export a dataset, When I inspect the metadata, Then each record includes consent_granted_at, consent_version, jurisdiction, and categories_consented fields.