Reverse-Causation Risk

Scientific meaning

What it measures

Assesses reverse-causation risk using evidence appropriate to causal inference, distinguishing random uncertainty from systematic error.

Why it matters

Material weakness in reverse-causation risk can change the direction, magnitude, certainty, or biological interpretation of the research conclusion.

Measurement criteria

Prespecified signaling questions; direction and likely magnitude of distortion; domain-level and overall judgment; sensitivity to plausible bias.

Typical methods of assessment

Structured critical appraisal; quantitative estimation with uncertainty; prespecified thresholds; independent replication; sensitivity and subgroup analyses.

Computation profile

Computation mode

RiskAssessmentComputation

Readiness

RubricOrModelRequired

Formula status

ContextSpecificProtocolRequired

Method family

risk assessment or calibrated probability estimation

Output datatype

xsd:string_or_decimal

Unit or scale

Ordinal risk rating or quantitative percentage/probability; lower is generally better

Numeric range

Context dependent

Directionality

LowerIsBetter

Required inputs

• risk_factors
• assessment_context
• rubric_or_model_version
• evidence_records

Optional inputs

• weights
• thresholds
• calibration_dataset

Machine-readable expression

{
  "language": "BEMO-Expression-JSON",
  "operator": "external_protocol",
  "protocolRef": "REQUIRED",
  "allowed_outputs": [
    "ordinal_risk_category",
    "probability",
    "percentage"
  ]
}

Computation policies

Aggregation

Not specified in source; must be defined and versioned before composite use.

Normalization

None by default; any normalization must be justified, versioned, and validated.

Missing data

Must be declared before computation; report missingness; no silent imputation; perform sensitivity analysis when material.

Uncertainty required

Yes

Uncertainty method

Confidence interval, credible interval, bootstrap distribution, inter-rater reliability, or sensitivity analysis as scientifically applicable.

Confidence interval required

Yes

Threshold policy

Prespecify, justify, version, and sensitivity-test thresholds; do not derive and evaluate on the same data without correction.

Decision thresholds

No universal threshold declared

Quality control

Source provenance; input validation; duplicate control; uncertainty reporting; independent or orthogonal validation where applicable.

Calibration

Required for probability outputs; rubric reliability required for ordinal outputs.

Applicability and validation

Evidence object

Result / experiment / study / body of evidence, as applicable

Evidence level

Result / experiment / study / body of evidence, as applicable

Study types

Randomized and observational etiologic studies, natural experiments, target-trial emulations

Domain

Causal Inference

Required data sources

Context-specific; use source references, registered studies, primary data, and authoritative biomedical resources as applicable.

Validation status

Source maturity: Established; BEMO computation profile requires independent validation.

Benchmark

A representative, versioned benchmark set is required before production use.

Reference standard

Use an independent reference standard when one exists; document expert-adjudicated alternatives.

External validation required

Yes

Interpretation safeguards

Common misinterpretations

Treating reverse-causation risk as interchangeable with overall study quality, or interpreting a favorable value as proof that all other bias and validity domains are satisfactory.

Limitations

Depends on context, operational definition, thresholds, data quality, and evaluator judgment; it should not be used as a stand-alone summary score without domain-level evidence.

Dependency status

Not yet curated. Closely related metrics are represented; causal or computational dependencies require expert curation.

Source provenance

Source workbook

Pure_Biomedical_Evidence_Research_Metrics_Part_2(3).xlsx

Source worksheet

Biomedical Evidence Metrics

Source row

112

Source record SHA-256

17fc4b74f3434a4ae22f908b09530b9a1be33f82af499c89239235d78d7e1375

Definition source

Pure_Biomedical_Evidence_Research_Metrics_Part_2(3).xlsx row 112; see source_references

Ontology IRI

https://w3id.org/bemo/BEMO_2000105

Profile IRI

https://w3id.org/bemo/profile/BEMO_2000105

SHACL shape

https://w3id.org/bemo/shapes/BEMO_2000105_AssessmentShape

Scientific references (5)

1. https://www.riskofbias.info/welcome/home/current-version-of-robins-iSource framework, standard, guideline, or originSource-preserved; URL reachability not asserted by this package
2. https://www.riskofbias.info/welcome/robins-e-toolSource framework, standard, guideline, or originSource-preserved; URL reachability not asserted by this package
3. https://www.gradeworkinggroup.org/Source framework, standard, guideline, or originSource-preserved; URL reachability not asserted by this package
4. https://www.equator-network.org/reporting-guidelines/strobe/Source framework, standard, guideline, or originSource-preserved; URL reachability not asserted by this package
5. https://database.ich.org/sites/default/files/E9_Guideline.pdfSource framework, standard, guideline, or originSource-preserved; URL reachability not asserted by this package

Governance and FAIR status

Lifecycle

Candidate

Approval

Draft

Owner

BEMO Project

Curator

Unassigned

Reviewer

Unassigned

License

Pending owner approval; CC BY 4.0 recommended for OBO compatibility.

Findable

Provisional: stable local ID assigned; public namespace registration pending.

Accessible

Provisional: package is distributable; permanent public release location pending.

Interoperable

Yes: OWL 2, RDF, SKOS, SHACL, JSON-LD, CSV, and OBO-style identifiers.

Reusable

Provisional: rich metadata supplied; open-license owner approval pending.

Governance note

Candidate term pending scientific, ontology-engineering, and computation-method review.

Editor note

Source content preserved verbatim; computation metadata is a generated default and must be curated before normative use.