A Hidden Markov Framework for Time-to-Flare Prediction in Crohn’s Disease Under Endogenous Laboratory Sampling: A Simulation Study

This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible. I present a hidden Markov model framework for longitudinal Crohn's disease monitoring with state-dependent laboratory sampling. The model combines daily Gaussian wearable summaries, intermittent log-normal laboratory biomarkers, and the laboratory-draw indicator in one likelihood; laboratory values contribute only on draw days. At each landmark day, I distinguish an unconditional first-passage distribution with current-flare mass at day zero from a transient-conditional distribution whose mean is bounded by the Remission- and Mild-state hitting-time means. Parameters are estimated without simulator state labels by Baum–Welch expectation maximization. A 17-member responsibility-conditioned empirical-Bayes mixture propagates parameter variation; the event-history comparator uses a 17-member patient-bootstrap ensemble. The censored log score retains censored landmarks and the probability tail beyond the observation window; fixed-horizon Brier scores, calibration, coverage, and sharpness complete the evaluation. Across ten simulations of 120 patients (40 held-out test patients each), the endogenous-sampling model obtained a censored log score (lower is better) of 2.261 (95% confidence interval 2.162–2.370), a four-horizon mean Brier score of 0.1199 (0.1118–0.1278), and 94.6% coverage for nominal 90% intervals. Comparator-minus-proposed differences for the otherwise identical model without the draw indicator were +0.0010 in log score (−0.0001 to 0.0021) and +0.00010 in mean Brier score (−0.00011 to 0.00031). The time-to-flare interface can be estimated without oracle states and evaluated, but the analysis does not establish a practically important gain from the draw indicator in this well-specified regime.

Authors

Institutions

Publication Details

Published
2026-11-15
DOI
https://doi.org/10.33774/coe-2026-3kwkg
Primary Topic
Inflammatory Bowel Disease
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

A Hidden Markov Framework for Time-to-Flare Prediction in Crohn’s Disease Under Endogenous Laboratory Sampling: A Simulation Study

Rishi Jasti
Inflammatory Bowel Disease
preprint

A Hidden Markov Framework for Time-to-Flare Prediction in Crohn’s Disease Under Endogenous Laboratory Sampling: A Simulation Study

Rishi Jasti
preprint en

Abstract

This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible. I present a hidden Markov model framework for longitudinal Crohn's disease monitoring with state-dependent laboratory sampling. The model combines daily Gaussian wearable summaries, intermittent log-normal laboratory biomarkers, and the laboratory-draw indicator in one likelihood; laboratory values contribute only on draw days. At each landmark day, I distinguish an unconditional first-passage distribution with current-flare mass at day zero from a transient-conditional distribution whose mean is bounded by the Remission- and Mild-state hitting-time means. Parameters are estimated without simulator state labels by Baum–Welch expectation maximization. A 17-member responsibility-conditioned empirical-Bayes mixture propagates parameter variation; the event-history comparator uses a 17-member patient-bootstrap ensemble. The censored log score retains censored landmarks and the probability tail beyond the observation window; fixed-horizon Brier scores, calibration, coverage, and sharpness complete the evaluation. Across ten simulations of 120 patients (40 held-out test patients each), the endogenous-sampling model obtained a censored log score (lower is better) of 2.261 (95% confidence interval 2.162–2.370), a four-horizon mean Brier score of 0.1199 (0.1118–0.1278), and 94.6% coverage for nominal 90% intervals. Comparator-minus-proposed differences for the otherwise identical model without the draw indicator were +0.0010 in log score (−0.0001 to 0.0021) and +0.00010 in mean Brier score (−0.00011 to 0.00031). The time-to-flare interface can be estimated without oracle states and evaluated, but the analysis does not establish a practically important gain from the draw indicator in this well-specified regime.

North Carolina School of Science and Mathematics (US)
Inflammatory Bowel Disease
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon