Development and Reporting of Prediction Models: Guidance for Authors From Editors of Respiratory, Sleep, and Critical Care Journals

Abstract

Prediction models aim to use available data to predict a health state or outcome that has not yet been observed. Prediction is primarily relevant to clinical practice, but is also used in research, and administration. While prediction modeling involves estimating the relationship between patient factors and outcomes, it is distinct from casual inference. Prediction modeling thus requires unique considerations for development, validation, and updating. This document represents an effort from editors at 31 respiratory, sleep, and critical care medicine journals to consolidate contemporary best practices and recommendations related to prediction study design, conduct, and reporting. Herein, we address issues commonly encountered in submissions to our various journals. Key topics include considerations for selecting predictor variables, operationalizing variables, dealing with missing data, the importance of appropriate validation, model performance measures and their interpretation, and good reporting practices. Supplemental discussion covers emerging topics such as model fairness, competing risks, pitfalls of "modifiable risk factors", measurement error, and risk for bias. This guidance is not meant to be overly prescriptive; we acknowledge that every study is different, and no set of rules will fit all cases. Additional best practices can be found in the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines, to which we refer readers for further details.

Figure 1.

The chronology of information is critical in prediction. A prediction, Ŷ, is made at time, t_p, based on data collected up to and including that time, but no later. Ŷ is the estimate of Y, which cannot be observed until a time in the future. The times at which we can observe Y (rather than just Ŷ) fall within a prediction window (t_{e_1} to t_{e_2}), which occurs after a certain amount of lead time has elapsed. The width of observation, lead time, and prediction intervals will influence the usefulness of any prediction model.

Figure 2.

Problems with categorizing continuous variables. Consider the example of splitting respiratory rate (RR) values into “high” and “low” based on a cut-off of 30 breaths per minute. Note this makes several assumptions, namely: 1) that there is no difference between a RR of 12 and a RR of 29 (points A and B); 2) that there’s no difference between a RR of 31 and a RR of 40 (points C and D); and 3) that RRs of 29 and 30 are categorically different (points B and C).