Prediction meets causal inference: the role of treatment in clinical prediction models

Abstract

In this paper we study approaches for dealing with treatment when developing a clinical prediction model. Analogous to the estimand framework recently proposed by the European Medicines Agency for clinical trials, we propose a 'predictimand' framework of different questions that may be of interest when predicting risk in relation to treatment started after baseline. We provide a formal definition of the estimands matching these questions, give examples of settings in which each is useful and discuss appropriate estimators including their assumptions. We illustrate the impact of the predictimand choice in a dataset of patients with end-stage kidney disease. We argue that clearly defining the estimand is equally important in prediction research as in causal inference.

Predicting Prostate Cancer Death with Different Pretreatment Risk Stratification Tools: A Head-to-head Comparison in a Nationwide Cohort Study

“…Van Geloven et al proposed a somewhat different approach (307), which was inspired by the European Medicines Agency framework for dealing with additional treatments started after baseline, and other post-baseline but pre-outcome events, in clinical trials (308). In this approach, the choice of strategy for accounting for time-dependent treatment should be based on the question the researcher wants to address using the prediction model (307). An overview of the four proposed strategies is presented in Table 6.1.…”

Survival and analysis of predictors of mortality in patients undergoing replacement renal therapy: a 20-year cohort

“…In both scientific and clinical practice, prediction aims at accurately predicting the risk of an outcome using multiple predictors collectively, where the final prediction model is usually based on statistically significant, but not necessarily causal, associations in the data at hand [ 21 ]. Choice in a dataset have impact of the predictimand of patients with ESRD, so, defining the estimand is equally important in prediction research as in causal inference [ 22 ].…”

Improving clinical management of COVID-19: the role of prediction models

“…Although it would be ideal, it is not straightforward to predict the risk of deterioration with and without intervention. 8 Readers and users should keep in mind that the risks predicted by the 4C Deterioration model reflect the probability of deterioration of a patient receiving similar care to that of patients in the development cohort, who were hospitalised in the UK between February and August, 2020. It follows that the risk model is likely to need updates as changes in the infected population and care occur over time.…”