MétaCan
Menu
Back to cohort

Sequential sample size calculations and learning curves safeguard the robust development of a clinical prediction model for individuals

2025· article· en· W4417505513 on OpenAlexaff
Amardeep Legha, Joie Ensor, Rebecca Whittle, Lucinda Archer, Ben Van Calster, Evangelia Christodoulou, Kym I E Snell, Mohsen Sadatsafavi, Gary S. Collins, Richard D Riley

Bibliographic record

VenueJournal of Clinical Epidemiology · 2025
Typearticle
Languageen
FieldComputer Science
TopicMachine Learning in Healthcare
Canadian institutionsUniversity of British Columbia
FundersNational Institute for Health and Care ResearchEngineering and Physical Sciences Research CouncilDepartment of Health and Social CareMedical Research CouncilBirmingham Biomedical Research CentreUniversity Hospitals Birmingham NHS Foundation Trust
KeywordsSample size determinationSafeguardSample (material)Large samplePredictive modellingEarly stopping

Abstract

fetched live from OpenAlex

BACKGROUND AND OBJECTIVES: When recruiting participants to a new study developing a clinical prediction model (CPM), sample size calculations are typically conducted before data collection based on sensible assumptions. This leads to a fixed sample size, but if the assumptions are inaccurate, the actual sample size required to develop a reliable model may be higher or even lower. To safeguard against this, adaptive sample size approaches have been proposed, based on sequential evaluation of (changes in) a model's predictive performance. The objective of the study was to illustrate and extend sequential sample size calculations for CPM development by (i) proposing stopping rules for prospective data collection based on minimizing uncertainty (instability) and misclassification of individual-level predictions and (ii) showcasing how it safeguards against inaccurate fixed sample size calculations. METHODS: Using the sequential approach repeats the predefined model development strategy every time a chosen number (eg, 100) of participants are recruited and adequately followed up. At each stage, CPM performance is evaluated using bootstrapping, leading to prediction and classification stability statistics and plots, alongside optimism-adjusted measures of calibration and discrimination. Learning curves display the trend of results against sample size and recruitment is stopped when a chosen stopping rule is met. RESULTS: Our approach is illustrated for model development of acute kidney injury using (penalized) logistic regression CPMs. Before recruitment based on perceived sensible assumptions, the fixed sample size calculation suggests recruiting 342 patients to minimize overfitting; however, during data collection, the sequential approach reveals that a much larger sample size of 1100 is required to minimize overfitting (targeting a bootstrap-corrected calibration slope ≥0.9). If the stopping rule criteria also target small uncertainty and misclassification probability of individual predictions, the sequential approach suggests an even larger sample size of about 1800. CONCLUSION: For CPM development studies involving prospective data collection, a sequential sample size approach allows users to dynamically monitor individual-level prediction and classification instability. This helps determine when enough participants have been recruited and safeguards against using inaccurate assumptions in a sample size calculation before data collection. Engagement with patients and other stakeholders is crucial to identify sensible context-specific stopping rules for robust individual predictions.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame distilled prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.

metaresearch head score (Codex)0.064
metaresearch head score (Gemma)0.311
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesMetaresearch
Consensus categoriesMetaresearch
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Observational · Consensus signal: none
GenreCandidate signal: Methods · Consensus signal: Methods
Teacher disagreement score0.485
Threshold uncertainty score0.963

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0640.311
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0020.000
Bibliometrics0.0000.000
Science and technology studies0.0000.000
Scholarly communication0.0000.000
Open science0.0010.000
Research integrity0.0000.002
Insufficient payload (model declined to judge)0.0000.000

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.358
GPT teacher head0.534
Teacher spread0.176 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; both teacher heads agree on what is shown here.

Study designObservational
Domainnot available
GenreMethods

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations0
Published2025
Admission routes1
Has abstractyes

Explore more

Same venueJournal of Clinical EpidemiologySame topicMachine Learning in HealthcareFrench-language works237,207