Why English? | Reporting TRIPOD-SRMA Prediction-Model Reviews Through Questions, Searches, Model Details, Appraisal, Synthesis and Conclusions

ORIENTATION

What this guide covers

Current fact check

TRIPOD-SRMA was published in BMJ 381:e073538 on 3 May 2023, DOI 10.1136/bmj-2022-073538, PubMed ID 37137496. The official TRIPOD site says it covers systematic reviews of prediction models, with or without quantitative synthesis, from scoping reviews to focused external-validation reviews.

What this guide does not cover

This article focuses on transparent reporting of systematic reviews and meta-analyses of prediction-model studies. It does not develop a model, perform a search, guarantee model quality, replace CHARMS or PROBAST, prescribe a pooling method, or decide whether a model is suitable for an individual or institution.

Worked example

Every chapter returns to a fictional review of models that predict whether secondary-school students may need additional attendance support, using administrative and school-context predictors without treating predictions as labels. It is fictional, low-stakes and educational; it does not report a real school, clinical decision or guaranteed outcome.

Warm up with eduKate’s guides to explaining scientific evidence and asking better questions in class.

CHAPTER 1 OF 15

1. Define the Prediction Question

Back to contents

Why this move matters

State population, setting, target outcome, time horizon, intended use and whether the review concerns diagnosis or prognosis.

Turn a review question into a traceable model question

State population, setting, target outcome, time horizon, intended use and whether the review concerns diagnosis or prognosis. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: The attendance-support review predicts a defined outcome over one school term rather than an undefined risk of disengagement. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

A broad label can join models that answer different questions. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Define the Prediction Question”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 2 OF 15

2. Declare the Review Objectives and Model Scope

Back to contents

Why this move matters

Explain whether the review maps models, compares versions, evaluates one model or synthesises performance.

Turn a review question into a traceable model question

Explain whether the review maps models, compares versions, evaluates one model or synthesises performance. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: The review maps all eligible models and separately examines external validations of one widely used tool. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

Scoping and comparative objectives need different conclusions. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Declare the Review Objectives and Model Scope”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 3 OF 15

3. Report Eligibility Criteria Precisely

Back to contents

Why this move matters

Define participants, settings, data sources, model types, development and validation studies, outcomes, reports and exclusions.

Turn a review question into a traceable model question

Define participants, settings, data sources, model types, development and validation studies, outcomes, reports and exclusions. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: Models using school records are eligible, while simple one-predictor association studies are not. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

Vague eligibility can turn an evidence review into an arbitrary collection. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Report Eligibility Criteria Precisely”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 4 OF 15

4. Make the Search Reproducible

Back to contents

Why this move matters

Report databases, registers, dates, complete strategies, supplementary methods, language handling and update searches.

Turn a review question into a traceable model question

Report databases, registers, dates, complete strategies, supplementary methods, language handling and update searches. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: Search terms combine prediction concepts, attendance outcomes and validation language, then citation tracking finds linked reports. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

Prediction-model searches can be difficult and sensitive to terminology. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Make the Search Reproducible”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 5 OF 15

5. Link Reports That Describe the Same Model

Back to contents

Why this move matters

Identify companion papers, updates, recalibrations, validations and overlapping cohorts.

Turn a review question into a traceable model question

Identify companion papers, updates, recalibrations, validations and overlapping cohorts. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: A development paper, a recalibration report and two validations are linked to one named model family. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

Counting reports as independent models inflates the evidence base. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Link Reports That Describe the Same Model”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 6 OF 15

6. Describe Participants, Settings and Data Sources

Back to contents

Why this move matters

Report recruitment, inclusion, geography, time period, sample size, missingness and data provenance.

Turn a review question into a traceable model question

Report recruitment, inclusion, geography, time period, sample size, missingness and data provenance. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: One validation uses urban school records from 2024, while another uses a national cohort from 2021. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

Similar sample sizes do not make settings interchangeable. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Describe Participants, Settings and Data Sources”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 7 OF 15

7. Report Predictors and Their Measurement

Back to contents

Why this move matters

Name candidate and final predictors, timing, coding, availability and any transformations.

Turn a review question into a traceable model question

Name candidate and final predictors, timing, coding, availability and any transformations. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: Prior absence is measured before the prediction point, and family variables are not inferred from proxies. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

A familiar predictor label can hide a different measurement. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Report Predictors and Their Measurement”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 8 OF 15

8. Report Outcomes and Prediction Horizons

Back to contents

Why this move matters

Define the target outcome, assessment method, timing, blinding and handling of competing events.

Turn a review question into a traceable model question

Define the target outcome, assessment method, timing, blinding and handling of competing events. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: The outcome is ten or more unauthorised absences within the next term, verified from records. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

Changing thresholds can create apparently different performance. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Report Outcomes and Prediction Horizons”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 9 OF 15

9. Appraise Risk of Bias and Applicability

Back to contents

Why this move matters

Report tool, signalling-question judgements, reviewers, disagreements and domain-level reasons.

Turn a review question into a traceable model question

Report tool, signalling-question judgements, reviewers, disagreements and domain-level reasons. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: Reviewers explain concerns about participant selection, predictor timing and analysis rather than publishing only a colour chart. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

A total score can hide the source of concern. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Appraise Risk of Bias and Applicability”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 10 OF 15

10. Extract Model Development and Validation Details

Back to contents

Why this move matters

Report modelling method, predictor selection, sample size, events, overfitting controls, internal validation, external validation and recalibration.

Turn a review question into a traceable model question

Report modelling method, predictor selection, sample size, events, overfitting controls, internal validation, external validation and recalibration. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: Bootstrap validation is distinguished from testing in a different school system. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

Calling both procedures validation can mislead readers. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Extract Model Development and Validation Details”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 11 OF 15

11. Report Performance Measures With Definitions

Back to contents

Why this move matters

Name discrimination, calibration, overall performance, thresholds, decision measures and uncertainty.

Turn a review question into a traceable model question

Name discrimination, calibration, overall performance, thresholds, decision measures and uncertainty. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: The review reports calibration slope and observed-to-expected ratio beside the C-statistic. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

Good ranking does not guarantee accurate probabilities. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Report Performance Measures With Definitions”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 12 OF 15

12. Explain Any Quantitative Synthesis

Back to contents

Why this move matters

Describe eligibility for pooling, transformations, hierarchical structure, heterogeneity, multiple estimates and sensitivity analyses.

Turn a review question into a traceable model question

Describe eligibility for pooling, transformations, hierarchical structure, heterogeneity, multiple estimates and sensitivity analyses. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: External validations with the same outcome horizon are synthesised while other horizons remain narrative. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

A pooled number can conceal setting and method differences. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Explain Any Quantitative Synthesis”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 13 OF 15

13. Present Results So Models Remain Traceable

Back to contents

Why this move matters

Use study flow, linked-report tables, model families, appraisal profiles and setting-specific performance.

Turn a review question into a traceable model question

Use study flow, linked-report tables, model families, appraisal profiles and setting-specific performance. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: Readers can follow one model from development through each validation and recalibration. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

A large summary table can still erase relationships. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Present Results So Models Remain Traceable”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 14 OF 15

14. Write Conclusions at the Right Strength

Back to contents

Why this move matters

Separate evidence quantity, reporting completeness, bias, performance, transportability and clinical or educational usefulness.

Turn a review question into a traceable model question

Separate evidence quantity, reporting completeness, bias, performance, transportability and clinical or educational usefulness. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: The review says evidence is promising but too biased and context-limited for routine school decisions. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

“Best model” can overstate evidence when no fair comparison exists. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Write Conclusions at the Right Strength”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

CHAPTER 15 OF 15

15. Connect Reporting to Responsible Decisions

Back to contents

Why this move matters

Explain local validation, updating, equity, data governance, human oversight and what research is needed next.

Turn a review question into a traceable model question

Explain local validation, updating, equity, data governance, human oversight and what research is needed next. Connect population, setting, outcome, time horizon, model purpose, development or validation role, eligible evidence and planned synthesis.

Follow one model across reports and settings

Worked example: A model is tested for calibration and subgroup performance before any supportive, non-punitive use. Track model identity, predictors, outcome definitions, performance measures, validation settings and duplicate cohorts before combining anything.

Keep reporting quality separate from model quality

A complete review report can reveal weak models, and an incomplete report can conceal a careful review. Transparency enables appraisal; it does not guarantee performance or usefulness.

Use TRIPOD-SRMA for the review report

Transparent review reporting does not justify automated action. The statement guides reporting of systematic reviews and meta-analyses of prediction-model research. It does not replace CHARMS extraction, PROBAST bias appraisal, statistical expertise or local validation.

Student, parent and career transfer

Make one evidence card for “Connect Reporting to Responsible Decisions”: source location, plain-language meaning, limit and next question. A parent or teacher can ask for one everyday explanation and one specialist explanation; both must preserve the same conditions. This habit supports medicine, public health, education research, statistics, data science, psychology, policy, librarianship, peer review, publishing and responsible artificial intelligence without promising a grade, admission, funding decision or job.

PRACTICE LAB

Build a TRIPOD-SRMA prediction-model review reporting decision trail

Student route

1. Choose one original report inside this guide’s scope.
2. Copy its objective and underline actor, action, context, evidence and intended decision.
3. Draw the specific reasoning chain used in this article.
4. Mark each step reported, inferred, contested or unknown.
5. Rewrite the conclusion so it preserves the narrowest important limitation.

Parent or teacher route

Ask the learner to explain how English helps us report prediction-model reviews so model questions, searches, model identity, appraisal, performance synthesis, conclusions and limits remain understandable twice: once to a classmate and once to a method specialist. Praise a precise correction and a well-stated unknown; both show disciplined attention to evidence.

Early-career route

Create a one-page TRIPOD-SRMA prediction-model review reporting audit with source link, date checked, scope, method logic, decisive wording, contrary evidence, limit and next action. A colleague should be able to continue without guessing.

Continue with eduKate’s guides to reading course and career information and giving useful feedback in group projects.

USEFUL NEXT READING

Continue with a related guide

Read the CHARMS prediction-model review guide for a related but different reader job. Return here when the question is specifically how to report prediction-model reviews so model questions, searches, model identity, appraisal, performance synthesis, conclusions and limits remain understandable. Use the linked guide for its separate reader task, and this guide for the question defined here.

For English practice across research, school and career communication, visit the Singapore English Tuition Centre learning site.

FREQUENTLY ASKED QUESTIONS

Questions students and parents often ask

What does TRIPOD-SRMA cover?

Systematic reviews and meta-analyses of multivariable prediction-model research.

Does it apply when there is no meta-analysis?

Yes. The official guidance includes systematic reviews without quantitative synthesis.

Is CHARMS the same as TRIPOD-SRMA?

No. CHARMS supports review framing and data extraction; TRIPOD-SRMA guides transparent reporting of the completed review.

What is external validation?

Evaluation of a model in data separate from those used to develop it, ideally representing the intended setting or population.

Does high discrimination make a model useful?

Not by itself. Calibration, thresholds, bias, applicability, consequences and local performance also matter.

Why does English matter?

Exact descriptions keep model identity, outcome, horizon, setting and performance from being blended into a misleading label.