Editors' ChoiceModeling

TRIPOD puts prediction models on a firmer footing

See allHide authors and affiliations

Science Translational Medicine  28 Jan 2015:
Vol. 7, Issue 272, pp. 272ec14
DOI: 10.1126/scitranslmed.aaa6670

Prediction models are used throughout medicine to guide decisions related to patient care: from surviving a trip to the ICU, to justifying cholesterol treatment, even to predicting the day a baby will be born. All of these models began with academic publications before they were implemented into clinical decision-making. As clinical research has expanded with the near endless discovery of biomarkers coupled with the introduction of electronic medical records and access to population-scale data, the desire to use these data to guide patient care through multivariable models has intensified. But there is a risk, as clinical decisions based on a poorly validated or incorrect prediction model may cause real harm to patients. But how can one judge whether a model has been properly constructed in the first place? Even experts have difficulty with this question. Enter Collins et al. with their description of TRIPOD, the “Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis.”

TRIPOD is a group of international statisticians and clinical researchers who came together to establish standards for reporting how prediction models are created and validated. Specifically, the TRIPOD statement recommends a thorough 22-point checklist for the reporting of multivariate prediction models and gives guidance for everything from the format of the title to the descriptions of the source and validation data sets. Although the checklist may seem to be only an incremental improvement to an otherwise well-constructed study, the hope is that through adherence to the TRIPOD standards, the general field of prediction model studies will be improved. In many ways, prediction models are the pinnacle of clinical research, not only bearing the burden of retrospective statistical rigor, but also of prospective prediction. Thus, one may ask: Will the adoption of TRIPOD standards for prediction models lead to better models with improved predictive capabilities? Given the number of statisticians involved in TRIPOD, predictably, we should know soon.

G. S. Collins et al., Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): The TRIPOD statement. Circulation 10.1161/CIRCULATIONAHA.114.014508 (2015). [Abstract]

Navigate This Article