TOPplus model for malignancy prediction in symptomatic breast assessment: minimised restricted analysis data

Minimised pseudonymised episode-level analysis data and preserved row-level probabilities accompanying the manuscript Presenting-symptom information for malignancy prediction in symptomatic breast assessment: derivation and internal validation of the TOPplus model. The dataset covers 7,829 eligible adult symptomatic breast-assessment episodes among 7,305 patients recorded as female, including 614 malignancies. The files comprise the analysis matrix, patient-grouped 10-fold out-of-fold probabilities for TOPplus, row-level probabilities for the explanatory five-model comparison, and probabilities for the supplementary internal temporal assessment. Names, addresses, postcodes, NHS or medical-record numbers, contact details, source identifiers, exact dates, month, free-text narratives, imaging reports and pathology reports are excluded. The data remain pseudonymised and must not be treated as anonymous public data. Version 2.3 is a documentation-only consistency correction. All four row-level CSV files are byte-for-byte unchanged from version 2.2; no outcome, predictor value, patient grouping, probability, coefficient or numerical result has changed. Access is restricted to approved research under the stated access and use conditions. Access does not authorise clinical use, autonomous decision-making, patient-level service operation, reidentification or redistribution. Aggregate outputs, executable code and current questionnaire materials are provided in the related public reproducibility-materials version 1.5: https://doi.org/10.5281/zenodo.22541062. Restricted access conditions: Access is not automatic. Requests must identify the research question and requested files; applicant and institution; lawful and ethical basis and any required approvals; information-security arrangements; intended duration of access; and publication plan. If approved, recipients must use the data only for the approved purpose; store them in an access-controlled institutional environment limited to named authorised personnel; make no attempt to identify or contact a patient or undertake linkage intended to increase identifiability; not redistribute, publish or deposit row-level data or row-level predictions; report only appropriately aggregated results and suppress small or potentially identifying cells where necessary; report suspected loss or unauthorised disclosure; delete or return the files at the end of the approved work or retention period; and cite the version-specific dataset DOI and accompanying manuscript. Record-owner approval does not override data-custodian, institutional, legal or ethics requirements. Additional data-use documentation may be required. No clinical use, autonomous decision-making or patient-level service operation is authorised.