Health Preference Assessment, RTI Health Solutions, Research Triangle Park, NC, USA.
UKCRC Centre of Excellence for Public Health, Queen's University Belfast, Royal Victoria Hospital, Belfast, Antrim, UK.
Med Decis Making. 2018 Jul;38(5):593-600. doi: 10.1177/0272989X18758279. Epub 2018 Apr 3.
In discrete-choice experiments (DCEs), respondents are presented with a series of scenarios and asked to select their preferred choice. In clinical decision making, DCEs allow one to calculate the maximum acceptable risk (MAR) that a respondent is willing to accept for a one-unit increase in treatment efficacy. Most published studies report the average MAR for the whole sample, without conveying any information about heterogeneity. For a sample of psychiatrists prescribing drugs for a series of hypothetical patients with schizophrenia, this article demonstrates how heterogeneity accounted for in the DCE modeling can be incorporated in the derivation of the MAR.
Psychiatrists were given information about a group of patients' responses to treatment on the Positive and Negative Syndrome Scale (PANSS) and the weight gain associated with the treatment observed in a series of 26 vignettes. We estimated a random parameters logit (RPL) model with treatment choice as the dependent variable.
Results from the RPL were used to compute the MAR for the overall sample. This was found to be equal to 4%, implying that, overall, psychiatrists were willing to accept a 4% increase in the risk of an adverse event to obtain a one-unit improvement of symptoms - measured on the PANSS. Heterogeneity was then incorporated in the MAR calculation, finding that MARs ranged between 0.5 and 9.5 across the sample of psychiatrists.
We provided psychiatrists with hypothetical scenarios, and their MAR may change when making decisions for actual patients.
This analysis aimed to show how it is possible to calculate physician-specific MARs and to discuss how MAR heterogeneity could have implications for medical practice.
在离散选择实验（DCE）中，向受访者呈现一系列情景，并要求他们选择自己偏好的选择。在临床决策中，DCE 允许人们计算出受访者愿意为治疗效果提高一个单位而接受的最大可接受风险（MAR）。大多数已发表的研究报告了整个样本的平均 MAR，而没有传达任何关于异质性的信息。对于一组精神科医生为一系列假设的精神分裂症患者开处方的样本，本文演示了如何将 DCE 建模中考虑的异质性纳入 MAR 的推导中。
向精神科医生提供了关于一组患者对治疗的反应的信息，这些信息来自阳性和阴性症状量表（PANSS），以及在 26 个情景中观察到的与治疗相关的体重增加。我们估计了一个随机参数对数（RPL）模型，以治疗选择为因变量。
从 RPL 得到的结果用于计算总体样本的 MAR。结果等于 4%，这意味着总体而言，精神科医生愿意接受不良事件风险增加 4%，以获得 PANSS 上症状改善一个单位。然后将异质性纳入 MAR 计算中，发现 MAR 在整个精神科医生样本中在 0.5 到 9.5 之间变化。
我们向精神科医生提供了假设情景，他们的 MAR 在为实际患者做出决策时可能会发生变化。
本分析旨在展示如何计算医师特定的 MAR，并讨论 MAR 异质性如何对医疗实践产生影响。
