Observable Neural ODEs for Identifiable Causal Forecasting in Continuous Time

TL;DR

This paper introduces Observable Neural ODEs (ObsNODEs), a novel continuous-time model that ensures latent states are reconstructible from observations, enabling causal forecasting in the presence of hidden confounders.

Contribution

The work links control-theoretic observability to causal identifiability and proposes ObsNODEs for continuous-time causal outcome prediction with strong empirical results.

Findings

ObsNODEs outperform recent sequence models on synthetic and real-world data.

The model enables causal forecasting under hidden confounders in continuous time.

A continuous-time adjustment formula links potential outcomes to observed data.

Abstract

Causal inference in continuous-time sequential decision problems is challenged by hidden confounders. We show that, in latent state-space models with time-varying interventions, observability of the latent dynamics from observed data is necessary for identifying dynamic treatment effects, linking control-theoretic observability to causal identifiability, even when hidden confounders affect both treatments and outcomes. We derive a continuous-time adjustment formula expressing potential outcome distributions under treatment trajectories via the measurement model, latent dynamics, and the filtering distribution over latent states given observed histories. We propose Observable Neural ODEs (ObsNODEs), Neural ODE models in observable normal form for causal forecasting. ObsNODEs learn continuous-time dynamics with states reconstructible from observations, enabling outcome prediction under alternative treatment paths. Experiments on synthetic cancer data, semi-synthetic data based on MIMIC-IV, and real-world sepsis data show strong performance over recent sequence models.