FlatASCEND: Autoregressive Clinical Sequence Generation with Continuous Time Prediction and Association-Based Pharmacological Testing

TL;DR

FlatASCEND is a large autoregressive model for clinical sequence generation that emphasizes patient-specific conditioning to preserve pharmacological effects, but faces challenges in causal interpretation and cross-site transfer.

Contribution

The paper introduces FlatASCEND, a novel autoregressive model with flat composite tokens and a zero-inflated log-normal time head for clinical sequence generation and pharmacological testing.

Findings

Patient-specific conditioning amplifies known pharmacological effects.

Model recovers correct mechanistic directions in 40% of cases.

Reward optimization can destroy meaningful pharmacological associations.

Abstract

Autoregressive models can predict clinical events, but generating patient-conditioned multi-step trajectories that respond to intervention tokens and testing whether those responses preserve known pharmacological associations has received limited attention. We present FlatASCEND, a 14.5M-parameter autoregressive clinical sequence model using flat composite tokens and a zero-inflated log-normal time head. Standard distributional metrics (Jaccard 0.889-0.954) do not distinguish FlatASCEND from trivial baselines; the model's value lies in conditional generation from patient-specific prefixes. A prompt-shuffle ablation shows patient-specific conditioning amplifies mechanistic pharmacological effects (2.0-2.2x for steroid to glucose, diuretic to potassium) while leaving confounding-driven associations unchanged (0.9x for insulin to glucose). An incident-user framework assesses directional consistency against prior pharmacological knowledge on MIMIC-IV (N=500 per comparison): 4/10 recover correct mechanistic directions, 2 reproduce treatment-context associations, 4 are incorrect (9/10 significant, Wilcoxon p<0.05). This pattern - partial recovery under residual confounding - is consistent with learned observational associations without causal distinction. Direct preference optimisation with surrogate reward destroys all correct associations (3/3 to 0/3), illustrating reward exploitation when reward and evaluation share an outcome domain. Generative evidence is strongest for short-horizon ICU data; outpatient temporal fidelity is weaker (median 10 vs 154 days on INSPECT), and zero-shot cross-site transfer degrades without adaptation.