Operational Playbooks

Start with the delivery problem, not the system name.

Eight concise, searchable playbooks turn common clinical data and analysis risks into decisions, checkpoints, handoffs, and evidence expectations.

8 playbooks

Milestone readiness

Database lock readiness checklist

Protect the final production and review window by separating true lock blockers from work that can be completed after lock.

Clarify first

Authoritative data-cut and lock definition
Open analysis decisions that can change derivations
External-data and reconciliation status
Final production, QC, review, and approval windows

Failure modes

“Nearly complete” is treated as ready
Late specifications enter after review starts
Unresolved dependencies are hidden in functional trackers

Closure evidence

Lock readiness record
Open-item disposition
Final run and independent-QC evidence
Decision and approval references

Milestone readiness

Statistical programming dry-run checklist

Use the dry run as a system test of data, decisions, programming, QC, review, handoffs, and execution rather than a cosmetic preview.

Clarify first

Representative data and known limitations
Stable populations, parameters, and shells
Priority outputs and review participants
Defect classification and closure rules

Failure modes

Outputs run without representative data
Review comments do not distinguish decision gaps from coding defects
Findings are not translated into final-run controls

Closure evidence

Dry-run inventory and execution log
Review disposition
Defect trend and readiness decision
Final-run improvement actions

Change & response

Health-authority response workflow

Translate the authority’s question before programming, control data provenance, and build the smallest reproducible answer that resolves the request.

Clarify first

Exact question, intent, population, parameter, and time window
Authoritative data cut and prior submitted context
Response format, review path, and transmission deadline

Failure modes

A technically correct output answers the wrong question
Sources or programs differ from the approved cut
Compressed review weakens traceability

Closure evidence

Approved interpretation record
Provenance manifest
Independent QC package
Final approval and transmission confirmation

Change & response

Late analysis change impact assessment

Write one authoritative change statement, trace every downstream artifact, and reset affected production and review status.

Clarify first

Final approved rule and rationale
Effective point and superseded decision
Affected datasets, outputs, metadata, and narratives

Failure modes

Different teams implement different interpretations
Only the visible output is corrected
Review remains marked complete after logic changes

Closure evidence

Approved change statement
End-to-end impact inventory
Regression and independent-QC evidence
Cross-deliverable approval

Data dependency

External and vendor data readiness

Make transfer timing, content, acceptance, downstream consumers, cutoff rules, and fallback decisions explicit before the final analysis window.

Clarify first

Expected content, identifiers, format, cadence, and accountable contacts
Consumer datasets, analyses, reconciliations, and outputs
Cutoff, refresh, and late-data rules

Failure modes

A transfer arrives too late for meaningful review
Completeness is assumed rather than reconciled
Different deliverables use different transfer states

Closure evidence

Transfer commitment
Dependency map
Approved cutoff decision
Receipt and reconciliation record

Standards & validation

Define-XML readiness checklist

Treat metadata as part of the data package and reconcile it continuously with datasets, specifications, terminology, methods, and reviewer documentation.

Clarify first

Applicable Define-XML and dataset standards versions
Origins, methods, codelists, value-level metadata, and document references
Metadata ownership and validation timing

Failure modes

Define-XML is generated only at the end
Metadata reflects specifications rather than delivered datasets
Reviewer guides contradict validation dispositions

Closure evidence

Dataset-to-metadata reconciliation
Link and stylesheet checks
Validation disposition
Reviewer-guide consistency approval

Standards & validation

Submission validation finding disposition

Classify the finding, identify root cause and recurrence, and apply a consistent package-level correction or evidence-supported justification.

Clarify first

Rule, severity, standard, and acceptance basis
False-positive possibility and regional context
Related datasets, variables, metadata, and guides

Failure modes

A local fix hides a systemic inconsistency
Disposition text is not supported by evidence
Final validation, guides, and manifest are misaligned

Closure evidence

Finding classification
Root-cause and recurrence analysis
Correction or justification evidence
Final package approval

Milestone readiness

ADaM and TLF delivery risk assessment

Connect analysis decisions, data readiness, specifications, programming, independent QC, review, and interpretation across the complete delivery chain.

Clarify first

Authoritative analysis definitions and source-data state
Dataset/output dependencies and execution sequence
Production, QC, statistical review, and revision capacity

Failure modes

Output completion hides unstable upstream derivations
QC mirrors production assumptions
Review timing leaves no controlled correction window

Closure evidence

Specification and decision approval
Production/QC reconciliation
Review disposition
Final inventory and release record