Division of Computer Science & Engin., College of Engineering, Louisiana State University, Baton Rouge, LA 70803, USA; Department of Medicine, Section of Hematology & Med Oncology, School of Medicine, Tulane University, New Orleans, LA 70112, USA.
Complete Decisions, LLC, Baton Rouge, LA 70810, USA.
Artif Intell Med. 2020 Jun;106:101882. doi: 10.1016/j.artmed.2020.101882. Epub 2020 May 26.
Health utilities express the perceptions patients have on the impact potential adverse events of medical treatments may have on their quality of life. Being able to accurately assess health utilities is crucial when deciding what is the best treatment when multiple and diverse treatment options exist, or when performing a cost / utility analysis. Due to the emotional and other complexities that may exist when such data are elicited, the values of the health utilities may be inaccurate and cause inconsistencies. Existing literature indicates that such inconsistencies may be very frequent. However, no method has been developed for dealing with such inconsistencies in an effective manner.
Given a set of health utilities, this paper first explores ways for determining if there are any inconsistencies in their values. It also proposes a number of quadratic optimization approaches to best estimate the actual (and hence unknown) values when a set of initial health utility values are provided by the patient and certain inconsistencies have been detected. This is achieved by readjusting the initial values in a way that is minimal and also satisfies certain consistency requirements.
The proposed methods are applied on an illustrative example related to localized prostate cancer. Data from some published studies were used to illustrate how a set of initial values can be analyzed. This analysis aims at readjusting them in a minimal manner that would also satisfy some key numerical constraints pertinent to health utility values.
The numerical results and the computational complexities of the proposed models indicate that the proposed approaches are practical as they involve quadratic optimization modeling. These approaches are novel as the problem of addressing numerical inconsistencies in the elicitation process of health utilities has not been addressed adequately. The approaches are also critical in shared decision making and also when performing cost / utility analyses because health utilities play a central role in determining the quality-adjusted life years when making decisions in these healthcare domains.
健康效用表示患者对医疗治疗潜在不良事件可能对其生活质量产生影响的感知。当存在多种不同的治疗选择，或者进行成本/效用分析时，能够准确评估健康效用至关重要。由于在获取此类数据时可能存在情感和其他复杂性，健康效用的值可能不准确，并导致不一致。现有文献表明，这种不一致可能非常频繁。然而，尚未开发出一种有效的方法来处理这种不一致。
给定一组健康效用值，本文首先探讨了确定其值是否存在不一致的方法。它还提出了一些二次优化方法，以便在患者提供一组初始健康效用值并检测到某些不一致时，最佳估计实际（即未知）值。这是通过以最小的方式重新调整初始值来实现的，并且还满足某些一致性要求。
所提出的方法应用于与局部前列腺癌相关的说明性示例。使用一些已发表的研究数据来说明如何分析一组初始值。这种分析旨在以最小的方式对其进行调整，同时满足与健康效用值相关的一些关键数值约束。
所提出模型的数值结果和计算复杂度表明，所提出的方法是实用的，因为它们涉及二次优化建模。这些方法是新颖的，因为在健康效用的启发过程中解决数值不一致的问题尚未得到充分解决。这些方法在共同决策和进行成本/效用分析时也很关键，因为健康效用在这些医疗保健领域的决策中确定质量调整生命年时起着核心作用。
