RETRACE: Resilience-Guided Trait-Conditioned Craving Estimation from Wearable Physiology in Opioid Use Disorder

Yi Xiao; Harshit Sharma; Dessa Bergen-Cico; Asif Salekin

cs.AIAug 15, 2026

RETRACE: Resilience-Guided Trait-Conditioned Craving Estimation from Wearable Physiology in Opioid Use Disorder

Authors: Yi Xiao, Harshit Sharma, Dessa Bergen-Cico, Asif Salekin

Organizations: Arizona State University, Tempe, USA · Syracuse University, Syracuse, New York, USA

Abstract

Detecting opioid craving from wearable physiological signals is critical yet difficult, with the potential to support proactive interventions for individuals with opioid use disorder (OUD). This challenge is especially pronounced under subject-independent evaluation because craving is subjective, heterogeneous, and often physiologically entangled with stress. Our empirical analysis shows that stress elicits strong and reproducible autonomic responses, while craving-related signals are weaker, sparse, and largely embedded within stress-related physiology. We further show that psychological resilience, which shapes stress regulation and craving vulnerability, is not reliably observable from short-term wearable windows, but can be captured through reusable subject-level proxies, including post-stress heart-rate recovery and autobiographical memory recall. Motivated by these findings, we introduce RETRACE, a resilience-guided trait-conditioned framework for subject-independent craving estimation from wearable physiology. RETRACE reframes craving detection as trait-conditioned physiological interpretation: rather than assuming the same physiological pattern has the same meaning across individuals, it uses resilience-related subject context to guide inference. Technically, RETRACE introduces a novel dual-encoder design that separates generalizable stress physiology from subject-specific craving interpretation. It combines a frozen stress-pretrained encoder with resilience-conditioned craving encoder, using feature-level gating and representation-level fusion to enable lightweight personalization without target-user craving labels or per-user retraining. We evaluate RETRACE on a novel multimodal OUD dataset containing wearable physiology, stress and craving annotations, and autobiographical narratives. Under LOSO setup, RETRACE achieves up to 7% absolute improvement over the strongest baseline.

Figures & tables

Group	Age	Gender	Race/Ethnicity
Control (n=50)	30.80 $\pm$ 9.60 [22–65]	Female: 35 (70.0%) Male: 15 (30.0%)	White: 30 (60.0%) Asian: 16 (32.0%) Black or African American: 3 (6.0%) Other: 1 (2.0%)
OUD (n=24)	35.67 $\pm$ 4.15 [29–47]	Female: 18 (75.0%) Male: 5 (20.8%) Non-binary: 1 (4.2%)	White: 14 (58.3%) Black or African American: 7 (29.2%) Other: 2 (8.3%) American Indian or Alaska Native: 1 (4.2%)

Table 1. Participant demographics. Values are reported as mean

\pm

SD [range] or count (%).

Figure 1. Task sequence and timeline used in the user study. Blue blocks represent non-stress baseline tasks, and red blocks represent stress-inducing tasks.

Figure 4. Feature-level physiological effects across modalities. Each column corresponds to an individual physiological feature, grouped by system and ordered within each group by the magnitude of stress-related effects. Color intensity represents the magnitude of significant effects (| $\beta$ |) estimated by GEE models, while non-significant features are masked. Rows correspond to different constructs (stress, craving, and resilience).

Physiological System	Stress	Craving	Resilience
Cardiovascular	$\beta{=}0.946$ , SE=0.284, $p{<}10^{-3}$	$\beta{=}0.380$ , SE=0.685, $p{=}0.579$	$\beta{=}{-}0.917$ , SE=1.173, $p{=}0.434$
Electrodermal	$\beta{=}3.317$ , SE=0.673, $p{<}10^{-6}$	$\beta{=}2.977$ , SE=1.816, $p{=}0.101$	$\beta{=}3.525$ , SE=2.164, $p{=}0.103$
Movement	$\beta{=}2.798$ , SE=0.471, $p{<}10^{-9}$	$\beta{=}1.478$ , SE=1.462, $p{=}0.312$	$\beta{=}1.504$ , SE=1.497, $p{=}0.315$
Thermoregulation	$\beta{=}1.314$ , SE=0.289, $p{<}10^{-5}$	$\beta{=}1.071$ , SE=0.484, $p{=}0.0269$	$\beta{=}{-}0.962$ , SE=0.650, $p{=}0.139$

Table 2. System-level physiological modulation. Reported coefficients (

\beta

) summarize population-average task effects estimated by the GEE models. Positive

\beta

values indicate higher physiological activity under the task condition, while negative values indicate reduced activity. Larger absolute

\beta

values, when observed consistently across related features or physiological modalities, reflect stronger and more coherent task-evoked physiological modulation. Standard errors (SE) quantify uncertainty in these estimates, and

p

-values indicate whether the observed effects differ reliably from zero after accounting for within-subject dependence.

Language Condition	Spearman $r$	$p$ (Spearman)	$p$ (Permutation)
Autobiographical Language	0.843	$2.9\times 10^{-17}$	$<0.001$
Context-neutral Speech	0.132	0.313	0.316

Table 3. PLS-based representational alignment between Autobiographical Language and HR recovery. Spearman rank correlations and permutation-based

p

-values are reported for associations between a one-dimensional PLS latent representation derived from Autobiographical Language embeddings and HR area-under-the-curve (AUC) following stress cessation.

Figure 7. Overview of the proposed resilience-guided architecture. A frozen stress encoder captures stable physiological structure, while a subject-conditioned encoder models adaptive variations. A resilience embedding modulates the input conditioning and representation fusion.

Model	Guidance	Acc	Macro-F1	Mean $\pm$ Std	$\boldsymbol{\Delta}$ Acc	p-value	CI
MLP	–	0.581	0.578	$0.579\pm 0.107$	0.139	0.010	[0.063, 0.215]
RBF-SVM	–	0.630	0.624	$0.623\pm 0.205$	0.094	0.105	[0.009, 0.195]
XGBoost	–	0.638	0.632	$0.629\pm 0.152$	0.088	0.160	[0.003, 0.174]
IRM	–	0.619	0.616	$0.619\pm 0.096$	0.099	0.051	[0.015, 0.188]
GroupDRO	–	0.609	0.607	$0.636\pm 0.114$	0.108	0.050	[-0.0226, 0.1903]
V-REx	–	0.623	0.622	$0.609\pm 0.115$	0.094	0.066	[0.015, 0.173]

Table 4. Baseline comparison on craving detection. Metrics are computed on pooled predictions across subjects. Mean

\pm

std is computed across subject-wise folds.

\boldsymbol{\Delta}

Acc denotes the accuracy improvement relative to the proposed model with HR dynamics guidance ( RETRACE

-

Baseline). Confidence intervals (CI) are reported for accuracy.

Variant	HR dynamics Guidance			Semantic Guidance
Variant	Acc	F1	Acc Mean $\pm$ Std	Acc	F1	Acc Mean $\pm$ Std
(1) Component Ablations
Remove fusion gate	0.619	0.617	$0.616\pm 0.129$	0.636	0.630	$0.636\pm 0.157$
Remove input gate	0.658	0.658	$0.652\pm 0.147$	0.654	0.649	$0.648\pm 0.273$
Remove frozen stress encoder	0.660	0.656	$0.659\pm 0.170$	0.623	0.620	$0.628\pm 0.141$
Remove learnable encoder	0.674	0.663	$0.670\pm 0.217$	0.654	0.645	$0.663\pm 0.211$

Table 5. Ablation study results under HR dynamics and semantic guidance.

Model	Neutral Video	Arithmetic	QA Baseline	Counting	Calm	Song Preparation	Stroop
HR dynamics Acc	0.707	0.714	0.923	0.714	0.768	0.600	0.658
Semantic Acc	0.724	0.684	1.000	0.633	0.659	0.800	0.763

Table 6. Task-level craving detection accuracy across models.

Table 12

Figure 8. Feature attribution under HR dynamics guidance for high- and low-resilience groups. AUC-based features (signed, absolute, positive, and negative) dominate attribution, while other temporal features show low importance.

Model	HR dynamics Guidance		Semantic Guidance
	BAcc	AUC	BAcc	AUC
Cross-Attn	0.635	0.635	0.500	0.599
MoE	0.635	0.635	0.583	0.583
FiLM	0.625	0.448	0.562	0.410
RETRACE	0.660	0.679	0.686	0.564

Table 9. In-the-wild evaluation results under different guidance types. Balanced accuracy (BAcc) is reported as the primary metric due to class imbalance, with AUC provided for reference.

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Component	Layers	Input $\rightarrow$ Output	Notes
Stress Encoder	4	303 $\rightarrow$ 384	Frozen, dropout = 0.2
Physiology Branch	2	303 $\rightarrow$ 384	Dropout = 0.193
Gate Projection	2	32 $\rightarrow$ 303	Sigmoid output (feature gating)
Fusion Projection	1	32 $\rightarrow$ 384	Subject-conditioned
Fusion Gate	2	3*384 $\rightarrow$ 1	Sample-wise scalar weight
Classifier	3	384 $\rightarrow$ 2	Binary logits

Appendix

Table 11. Architecture details of the best HR-AUC main model.

Figure 11. Neurobiological flow of stress–reward dysregulation in opioid use disorder (OUD). Chronic opioid dependence alters reward and stress circuits, transforming positive memory recall into aberrant stress activation, which increases craving and relapse vulnerability.

Panel of Figure 12 — Figure 12. Qualitative visualizations of autobiographical narratives based on LLM-derived, phrase-level resilience-related appraisal. Red indicates relatively positive resilience-related appraisal, blue indicates relatively negative appraisal, and gray indicates neutral language; color intensity reflects the relative strength of appraisal within each narrative.

Explore similar work

Aug 17, 2026

Representation Is Not Enough: Body-Localized Thermal Evidence for Contactless Stress and Craving Sensing in Opioid Use Disorder

Removing wearables from physiological monitoring also removes their supervision: the signal indicating where and when a stress response occurred. Contactless stress sensing therefore becomes a weakly supervised evidence-localization problem, where a clip-level label must be traced to the body regions and moments that produced it. We address this with FABLE-Therm, a weakly supervised architecture that preserves localized evidence across body regions, time, and encoder-specific representations until the final decision. FABLE-Therm fuses frozen foundation-model encoders at the embedding level, with theory explaining why localized fusion can outperform feature concatenation and prediction averaging. We study this problem in opioid use disorder (OUD), where stress is a major relapse trigger and sustained wearable use can be difficult during early recovery. Using fixed thermal video, FABLE-Therm achieves 0.938 AUROC on held-out participants, and its learned representation transfers to self-reported craving, providing, to our knowledge, the first evidence that craving can be recovered from contactless thermal video. Localized evidence also enables participant-level analysis of deployment failure. We find that improving representation alone is insufficient for equitable deployment: additional data from the underserved group would recover only about half of the cohort gap, while the remainder reflects person-to-person heterogeneity. This modality-agnostic decomposition applies to models with identifiable subpopulations. Together with the first cohort-structured contactless thermal OUD benchmark, our results show that preserving localized evidence supports both accurate sensing and principled analysis of who a model fails and why.

Sachin Deb, Harshit Sharma, Asif Salekin

Arizona State University

Weakly Supervised Learning Thermal Imaging Mental Health Algorithmic Fairness

Ask PDF

Sep 10, 2026

Patient-Reported Survey Data Improve Prediction of Opioid Use Disorder

Electronic health records (EHRs) may incompletely capture patient-reported factors associated with opioid use disorder (OUD). We evaluated whether survey data improve prediction of a first recorded OUD diagnosis among 267,747 All of Us participants with documented opioid exposure, including 15,287 OUD cases. We compared EHR-only and EHR+survey models across 6-, 12-, and 24-month look-back windows using logistic regression, random forest, XGBoost, LightGBM, multilayer perceptron, LSTM, GRU, and Transformer. Survey augmentation improved PR-AUC across all 24 model-window combinations by 0.0087-0.0505; the best 24-month LightGBM model improved from 0.6219 to 0.6603. Survey coverage increased with longer windows and differed by OUD status (24 months: 21.7% OUD-positive vs. 60.7% OUD-negative). Permutation analysis ranked survey features as the second most important information domain at 24 months in both evaluated models. Patient-reported data provide complementary predictive signals beyond structured EHRs while highlighting the importance of survey availability.

Xiyue Jiang, Zihan Ding, Grace Han +3

Department of Applied Mathematics & Statistics, Stony Brook University, Stony Brook, NY, USA · Department of Computer Science, Stony Brook University, Stony Brook, NY, USA · Department of Biomedical Informatics, Stony Brook University, Stony Brook, NY, USA +1

Clinical Outcome Prediction Electronic Health Records Clinical Risk Prediction

Ask PDF

Jun 23, 2026

Retrieval-Augmented Personalization with Foundation Models for Wearable Stress Detection

Personalization in wearable-based stress detection remains challenging due to substantial inter-individual variability in physiological and behavioral responses. While traditional approaches rely on user-specific fine-tuning or costly self-supervised pre-training on large datasets, we propose a lightweight alternative based on retrieval-augmented personalization. Our method leverages frozen, out-of-domain foundation models to retrieve similar patterns from a target user's history and encode them into a compact personalized embedding that modulates representations extracted by a lightweight transformer network. We evaluate our approach on the WESAD stress detection dataset with N=15 users, comprising wrist-worn physiological (EDA, BVP, temperature) and activity (accelerometer) signals, and report gains of +3.92% in accuracy and +4.76% in macro F1-score over a non-personalized transformer baseline, approaching supervised fine-tuning performance without requiring any labeled user data. We further show that temporal retrieval, where only prior user samples are available, achieves performance close to full intra-user retrieval, demonstrating robustness to limited user history. Finally, we explore personalization in a cross-dataset retrieval setting, leveraging embeddings from the K-Emocon dataset to personalize representations for stress detection on the WESAD dataset.

Louis Simon, Mohamed Chetouani

Institut des Systèmes Intelligents et de Robotique Sorbonne University Paris, France

Wearable Health Monitoring

Ask PDF