arXiv:2504.20908v3 Announce Type: replace 
Abstract: Current subgroup identification methods typically follow a two-step approach: first estimate conditional average treatment effects and then apply thresholding or rule-based procedures to define subgroups. While intuitive, this decoupled approach fails to incorporate key constraints essential for real-world clinical decision-making, such as subgroup size and propensity overlap. These constraints operate on fundamentally different axes than CATE estimation and are not naturally accommodated within existing frameworks, thereby limiting the practical applicability of these methods. We propose a unified optimization framework that directly solves the primal constrained optimization problem to identify optimal subgroups. Our key innovation is a reformulation of the constrained primal problem as an unconstrained differentiable min-max objective, solved via a gradient descent-ascent algorithm. We theoretically establish that our solution converges to a feasible and locally optimal solution. Unlike threshold-based CATE methods that apply constraints as post-hoc filters, our approach enforces them directly during optimization. The framework is model-agnostic, compatible with a wide range of CATE estimators, and extensible to additional constraints like cost limits or fairness criteria. Extensive experiments on synthetic and real-world datasets demonstrate its effectiveness in identifying high-benefit subgroups while maintaining better satisfaction of constraints.

تم اقتراح إطار عمل جديد يسمى MOSIC لتحديد المجموعات الفرعية المثلى في اتخاذ القرارات السريرية، حيث يعالج القيود الموجودة في الأساليب الحالية التي لا تأخذ في الاعتبار القيود الأساسية مثل حجم المجموعة الفرعية وتداخل الميل. يقوم هذا الإطار الموحد بالتحسين بإعادة صياغة المشكلة الأولية المقيدة كهدف متميز غير مقيد، يمكن حله باستخدام خوارزمية النزول والصعود بالتدرج.

Se ha propuesto un nuevo marco llamado MOSIC para la identificación óptima de subgrupos en la toma de decisiones clínicas, abordando las limitaciones de los métodos existentes que no incorporan restricciones esenciales como el tamaño del subgrupo y la superposición de propensión. Este marco de optimización unificado reformula el problema primal restringido en un objetivo min-max diferenciable sin restricciones, que se puede resolver mediante un algoritmo de descenso-asenso de gradiente.

Un nouveau cadre nommé MOSIC a été proposé pour l'identification optimale de sous-groupes dans la prise de décision clinique, abordant les limitations des méthodes existantes qui ne tiennent pas compte des contraintes essentielles telles que la taille des sous-groupes et le chevauchement de propension. Ce cadre d'optimisation unifié reformule le problème primal contraint en un objectif min-max différentiable sans contrainte, qui peut être résolu à l'aide d'un algorithme de descente-ascent par gradient.

A new framework named MOSIC has been proposed for optimal subgroup identification in clinical decision-making, addressing limitations in existing methods that fail to incorporate essential constraints like subgroup size and propensity overlap. This unified optimization framework reformulates the constrained primal problem into an unconstrained differentiable min-max objective, which can be solved using a gradient descent-ascent algorithm.