arXiv subjects

Shasha Han

Publications and source records attributed to Shasha Han.

5 recordsLinked to original sources

arXiv2026

DySRec: Dynamic Context-Aware Psychometric Scale Recommendation via Multi-Agent Collaboration

Choosing suitable psychometric scales is an essential and difficult step in psychological consultation, which requires clinicians to integrate patient information, behaviors, and dynamic contextual information. Existing systems mainly use static pipelines to choose scale, or directly predict symptoms according to user inputs, limiting their ability to support dynamic assessment, risk management, and transparent decision-making. To address these limitations, we propose DySRec, a multi-agent conversational system for dynamic psychometric scale recommendation. DySRec operates as an interactive chatbot that engages users in multi-turn dialogue, models scale selection as a continuous conversational decision process, and coordinates specialized agents to maintain user context, recommend assessment scales, monitor psychological risk, and log decision trajectories. In this way, DySRec can integrate and capture heterogeneous signals, including semantic, interaction behaviors, assessment history, and content state, to dynamically update user representations and calculate scale-context compatibility score for recommending most matched scales. Moreover, DySRec incorporates a closed-loop refinement mechanism. Recommendation agent will feedback the missing or uncertain attributes and guide the conversation to elicit the targeted information. In this paper, we showcase the prototype design and architecture of DySRec, and this system has been verified in a real-world application.

cs.HC↗

arXiv2025

Identification and Estimation of Long-Term Treatment Effects with Monotone Missing

Estimating long-term treatment effects has a wide range of applications in various domains. A key feature in this context is that collecting long-term outcomes typically involves a multi-stage process and is subject to monotone missing, where individuals missing at an earlier stage remain missing at subsequent stages. Despite its prevalence, monotone missing has been rarely explored in previous studies on estimating long-term treatment effects. In this paper, we address this gap by introducing the sequential missingness assumption for identification. We propose three novel estimation methods, including inverse probability weighting, sequential regression imputation, and sequential marginal structural model (SeqMSM). Considering that the SeqMSM method may suffer from high variance due to severe data sparsity caused by monotone missing, we further propose a novel balancing-enhanced approach, BalanceNet, to improve the stability and accuracy of the estimation methods. Extensive experiments on two widely used benchmark datasets demonstrate the effectiveness of our proposed methods.

cs.LG↗

arXiv2024

Inference for Cumulative Incidences and Treatment Effects in Randomized Controlled Trials with Time-to-Event Outcomes under ICH E9 (R1)

In randomized controlled trials (RCTs) that focus on time-to-event outcomes, intercurrent events can arise in two ways: as semi-competing events, which modify the hazard of the primary outcome events, or as competing events, which make the definition of the primary outcome events unclear. Although five strategies have been proposed in the ICH E9 (R1) addendum to address intercurrent events in RCTs, these strategies are not easily applicable to time-to-event outcomes when aiming for causal interpretations. In this study, we show how to define, estimate, and make inferences concerning objectives that have causal interpretations within these contexts. Specifically, we derive the mathematical formulations of the causal estimands corresponding to the five strategies and clarify the data structure needed to identify these causal estimands. Furthermore, we introduce nonparametric methods for estimating and making inferences about these causal estimands, including the asymptotic variance of estimators and hypothesis tests. Finally, we illustrate our methods using data from the LEADER Trial, which aims to investigate the effect of liraglutide on cardiovascular outcomes.

stat.ME↗

arXiv2021

Data-Driven Inpatient Bed Assignment Using the P Model

Problem definition: Emergency department (ED) boarding refers to the practice of holding patients in the ED after they have been admitted to hospital wards, usually resulting from insufficient inpatient resources. Boarded patients may compete with new patients for medical resources in the ED, compromising the quality of emergency care. A common expedient for mitigating boarding is patient overflowing, i.e., sending patients to beds in other specialties or accommodation classes, which may compromise the quality of inpatient care and bring on operational challenges. We study inpatient bed assignment to shorten boarding times without excessive patient overflowing. Methodology: We use a queue with multiple customer classes and multiple server pools to model hospital wards. Exploiting patient flow data from a hospital, we propose a computationally tractable approach to formulating the bed assignment problem, where the joint probability of all waiting patients meeting their respective delay targets is maximized. Results: By dynamically adjusting the overflow rate, the proposed approach is capable not only of reducing patients' waiting times, but also of mitigating the time-of-day effect on boarding times. In numerical experiments, our approach greatly outperforms both early discharge policies and threshold-based overflowing policies, which are commonly used in practice. Managerial implications: We provide a practicable approach to solving the bed assignment problem. This data-driven approach captures critical features of patient flow management, while the resulting optimization problem is practically solvable. The proposed approach is a useful tool for the control of queueing systems with time-sensitive service requirements.

math.OC↗

arXiv2020

Contrast Specific Propensity Scores

Basic propensity score methodology is designed to balance multivariate pre-treatment covariates when comparing one active treatment with one control treatment. Practical settings often involve comparing more than two treatments, where more complicated contrasts than the basic treatment-control one,(1,-1), are relevant. Here, we propose the use of contrast-specific propensity scores (CSPS). CSPS allow the creation of treatment groups of units that are balanced with respect to bifurcations of the specified contrasts and the multivariate space spanned by them.

stat.ME↗