OBJECTIVE: Traditional readmission risk models relying on static discharge data have limited predictive performance and fail to capture patients' recovery trajectories after hospitalization. We sought to identify optimal modeling parameters for dynamically predicting readmission risk using post-discharge step-count data from remote monitoring devices. METHODS: We combined data for adults aged 55+ from 2 studies that collected longitudinal activity data after discharge. We constructed a patient-day dataset incorporating static demographic and clinical variables and dynamic activity features aggregated over retrospective windows of 3, 5, 7, or 10 days. Models predicted readmission or death over prospective horizons of 3, 5, 7, or 10 days, within follow-up periods of 30-180 days. Logistic regression and LightGBM models were trained using 5-fold cross-validation on an 80:20 patient-level split. RESULTS: Among 215 participants, LightGBM outperformed logistic regression across all configurations (mean AUC 0.82 vs 0.76). Performance improved with longer prospective horizons but was insensitive to retrospective window length. The LightGBM model was well-calibrated (Hosmer-Lemeshow χ2 = 2.46, P = .96), whereas logistic regression showed miscalibration (χ2 = 51.8, P < .001). In feature-importance analyses, LightGBM ranked static (length of stay, vitals, BMI) and activity (recent steps, distance) features highly, whereas logistic regression emphasized activity variables. DISCUSSION: Prediction performance was impacted by horizon length and training window, with minimal effect of retrospective window. LightGBM achieved better discrimination and calibration, supporting flexible, non-parametric methods for post-discharge risk prediction. CONCLUSION: Post-discharge step count data enhance dynamic readmission risk prediction. Optimizing temporal windows and model type improves discrimination and calibration.

In settings where ML models are used to inform the allocation of resources, agents affected by the allocation decisions might have an incentive to strategically change their features to secure better outcomes. While prior work has studied strategic responses broadly, disentangling misreporting from genuine adaptation remains a fundamental challenge. In this paper, we propose a causally-motivated approach to identify and quantify how much an agent misreports on average by distinguishing deceptive changes in their features from genuine adaptation. Our key insight is that, unlike genuine adaptation, misreported features do not causally affect downstream variables (i.e., causal descendants). We exploit this asymmetry by comparing the causal effect of misreported features on their causal descendants as derived from manipulated datasets against those from unmanipulated datasets. We formally prove identifiability of the misreporting rate and characterize the variance of our estimator. We empirically validate our theoretical results using a semi-synthetic and real Medicare dataset with misreported data, demonstrating that our approach can be employed to identify misreporting in real-world scenarios.

In many settings, machine learning models may be used to inform decisions that impact individuals or entities who interact with the model. Such entities, or *agents,* may *game* model decisions by manipulating their inputs to the model to obtain better outcomes and maximize some utility. We consider a multi-agent setting where the goal is to identify the “worst offenders:” agents that are gaming most aggressively. However, identifying such agents is difficult without knowledge of their utility function. Thus, we introduce a framework in which each agent’s tendency to game is parameterized via a scalar. We show that this gaming parameter is only partially identifiable. By recasting the problem as a causal effect estimation problem where different agents represent different “treatments,” we prove that a ranking of all agents by their gaming parameters is identifiable. We present empirical results in a synthetic data study validating the usage of causal effect estimation for gaming detection and show in a case study of diagnosis coding behavior in the U.S. that our approach highlights features associated with gaming.