Skills › Scientific Agent Skills › clinical-decision-support › survival_analysis_plan_template.json
survival_analysis_plan_template.json
134 lines · skills/clinical-decision-support/assets/survival_analysis_plan_template.json at 5f2bc03
1{
2 "schema_version": "1.0",
3 "metadata": {
4 "plan_id": "SYNTHETIC-SURVIVAL-PLAN-001",
5 "title": "Synthetic Aggregate Time-to-Event Analysis Plan",
6 "version": "0.1-draft",
7 "status": "draft",
8 "purpose": "Demonstrate estimand-led planning without time-to-event records.",
9 "data_cut_date": "2026-07-23",
10 "data_level": "plan_only",
11 "raw_rows_supplied": false,
12 "patient_care_use": false,
13 "human_review_required": true
14 },
15 "intended_use": "Research analysis planning only; not for patient care or live clinical use.",
16 "prohibited_uses": [
17 "No individual prognosis or classification.",
18 "No diagnosis, treatment recommendation, dosing, triage, alarm, or autonomous decision.",
19 "No claim of statistical approval, clinical validity, authorization, or compliance."
20 ],
21 "limitations": [
22 "The validator does not inspect data or fit a model.",
23 "Methods require review against the actual design and estimand.",
24 "Synthetic placeholders must be replaced before an approved research analysis."
25 ],
26 "estimand": {
27 "population": "Synthetic eligible research cohort",
28 "condition_or_comparison": "Synthetic exposure A versus synthetic exposure B",
29 "endpoint_variable": "Time from defined synthetic index to defined synthetic event",
30 "population_summary": "Hazard contrast plus restricted mean survival time difference at the prespecified horizon",
31 "time_horizon": "24 synthetic time units",
32 "rationale": "The hazard contrast addresses relative event rate; restricted mean survival time supports interpretation if proportional hazards is doubtful.",
33 "intercurrent_events": [
34 {
35 "event": "Synthetic exposure discontinuation",
36 "strategy": "Treatment-policy strategy in the primary estimand",
37 "rationale": "Preserves the prespecified research question."
38 },
39 {
40 "event": "Synthetic competing event",
41 "strategy": "Competing-event strategy with cumulative incidence",
42 "rationale": "The competing event precludes the primary event."
43 }
44 ]
45 },
46 "endpoint": {
47 "time_zero": "Synthetic index date defined before analysis",
48 "event_definition": "Prespecified synthetic event with blinded aggregate adjudication",
49 "time_scale": "Synthetic months",
50 "ascertainment": "Prespecified uniform synthetic schedule",
51 "follow_up_end": "Earliest of event, loss to follow-up, or administrative data cut",
52 "delayed_entry": "Assess and use left-truncation methods if eligibility begins after time zero",
53 "same_time_rules": "Apply prespecified deterministic ordering and document ties",
54 "censoring_rules": [
55 "Administrative censoring at the fixed data cut",
56 "Censor loss to follow-up at last verified observation with informative-censoring sensitivity analysis"
57 ],
58 "competing_events": [
59 "Synthetic competing event"
60 ]
61 },
62 "analysis": {
63 "primary_method": "cox_proportional_hazards",
64 "effect_measure": "Hazard ratio with 95% confidence interval; restricted mean survival time difference at 24 units if proportional hazards is not supported",
65 "analysis_population": "Prespecified synthetic analysis population",
66 "covariate_strategy": "Prespecified covariates based on subject-matter rationale; no univariable p-value selection",
67 "missing_data": "Describe patterns; use justified multiple-imputation or model-based approach with sensitivity analysis",
68 "multiplicity": "Prespecify primary endpoint and control or transparently label secondary and exploratory analyses",
69 "uncertainty": "Two-sided 95% confidence intervals and assumption-focused sensitivity analyses",
70 "software_and_version": "Record locked statistical software, packages, code, and versions before execution",
71 "model_diagnostics": "Functional forms, influential observations, residual checks, convergence, and calibration where predictive",
72 "proportional_hazards_assessment": "Prespecified graphical assessment and time-interaction/residual-based assessment",
73 "non_proportional_hazards_strategy": "Report time-varying effects and restricted mean survival time at the prespecified horizon",
74 "competing_risk_method": "cumulative_incidence"
75 },
76 "bias_controls": {
77 "informative_censoring": "Compare follow-up patterns and run a justified sensitivity analysis.",
78 "immortal_time": "Align eligibility, assignment, and time zero; model time-varying exposure if required.",
79 "time_dependent_confounding": "Specify whether g-methods are required for the causal estimand.",
80 "outcome_misclassification": "Document ascertainment validity and conduct plausible sensitivity analyses.",
81 "informative_visits": "Assess observation schedules and potential outcome-dependent measurement."
82 },
83 "sensitivity_analyses": [
84 "Alternative censoring assumptions tied to plausible loss-to-follow-up mechanisms",
85 "Alternative effect summary when proportional hazards is not supported",
86 "Competing-risk analysis aligned with the absolute-incidence question"
87 ],
88 "subgroups": {
89 "prespecification": "List scientifically justified subgroups before results.",
90 "interaction_testing": "Use formal interaction estimates with uncertainty.",
91 "multiplicity": "Control or label exploratory subgroup analyses.",
92 "precision_and_disclosure": "Report aggregate sizes/events and suppress unstable or disclosive outputs."
93 },
94 "validation": {
95 "internal_validation": "Use design-appropriate resampling if a prediction model is involved.",
96 "external_validation": "Plan independent temporal or geographic validation before transportability claims.",
97 "calibration": "Evaluate calibration at prespecified horizons for prediction models.",
98 "subgroup_performance": "Report performance and uncertainty across prespecified groups."
99 },
100 "human_review": {
101 "roles": [
102 "survival-analysis statistician",
103 "epidemiologist",
104 "domain expert"
105 ],
106 "completed": false,
107 "approval_boundary": "An accountable research organization must approve the final plan; approval never permits patient-care use."
108 },
109 "governance": {
110 "owner": "research analysis governance owner",
111 "change_control": "Version the estimand, endpoint, data cut, code, software, deviations, and sensitivity analyses.",
112 "auditability": "Retain the approved plan, amendments, execution logs, aggregate outputs, and reviewer decisions.",
113 "monitoring": "Review assumption violations, missingness, censoring, subgroup precision, and external validation.",
114 "retirement": "Supersede or retire the plan when the question, data source, endpoint, or method changes materially."
115 },
116 "source_ids": [
117 "ICH-E9R1",
118 "RMST"
119 ],
120 "sources": [
121 {
122 "id": "ICH-E9R1",
123 "citation": "ICH E9(R1), Addendum on Estimands and Sensitivity Analysis in Clinical Trials, 2019.",
124 "url": "https://database.ich.org/sites/default/files/E9-R1_Step4_Guideline_2019_1203.pdf",
125 "accessed": "2026-07-23"
126 },
127 {
128 "id": "RMST",
129 "citation": "Royston P, Parmar MKB. Restricted mean survival time. BMC Medical Research Methodology. 2013;13:152.",
130 "url": "https://pubmed.ncbi.nlm.nih.gov/24314264/",
131 "accessed": "2026-07-23"
132 }
133 ]
134}