McCaffrey Nikki, Agar Meera, Harlum Janeane, Karnon Jonathon, Currow David, Eckermann Simon
Flinders Clinical Effectiveness, Flinders University, Bedford Park, South Australia, Australia 5041; Palliative and Supportive Services, Flinders University, Bedford Park, South Australia, Australia.
Department of Palliative Care, Braeside Hospital, Prairiewood, New South Wales, Australia; Palliative and Supportive Services, Flinders University, Bedford Park, South Australia, Australia.
PLoS One. 2015 Mar 9;10(3):e0115544. doi: 10.1371/journal.pone.0115544. eCollection 2015.
Comparing multiple, diverse outcomes with cost-effectiveness analysis (CEA) is important, yet challenging in areas like palliative care where domains are unamenable to integration with survival. Generic multi-attribute utility values exclude important domains and non-health outcomes, while partial analyses-where outcomes are considered separately, with their joint relationship under uncertainty ignored-lead to incorrect inference regarding preferred strategies.
The objective of this paper is to consider whether such decision making can be better informed with alternative presentation and summary measures, extending methods previously shown to have advantages in multiple strategy comparison.
Multiple outcomes CEA of a home-based palliative care model (PEACH) relative to usual care is undertaken in cost disutility (CDU) space and compared with analysis on the cost-effectiveness plane. Summary measures developed for comparing strategies across potential threshold values for multiple outcomes include: expected net loss (ENL) planes quantifying differences in expected net benefit; the ENL contour identifying preferred strategies minimising ENL and their expected value of perfect information; and cost-effectiveness acceptability planes showing probability of strategies minimising ENL.
Conventional analysis suggests PEACH is cost-effective when the threshold value per additional day at home (𝕜1) exceeds $1,068 or dominated by usual care when only the proportion of home deaths is considered. In contrast, neither alternative dominate in CDU space where cost and outcomes are jointly considered, with the optimal strategy depending on threshold values. For example, PEACH minimises ENL when 𝕜1=$2,000 and 𝕜2=$2,000 (threshold value for dying at home), with a 51.6% chance of PEACH being cost-effective.
Comparison in CDU space and associated summary measures have distinct advantages to multiple domain comparisons, aiding transparent and robust joint comparison of costs and multiple effects under uncertainty across potential threshold values for effect, better informing net benefit assessment and related reimbursement and research decisions.
运用成本效益分析（CEA）比较多个不同的结果很重要，但在姑息治疗等领域具有挑战性，因为这些领域的指标难以与生存率相结合。通用的多属性效用值排除了重要指标和非健康结果，而局部分析（即分别考虑结果，忽略不确定性下的联合关系）会导致关于首选策略的错误推断。
本文的目的是探讨是否可以通过替代的呈现方式和汇总指标更好地为这类决策提供信息，扩展先前在多策略比较中显示出优势的方法。
在成本负效用（CDU）空间中对居家姑息治疗模式（PEACH）相对于常规护理进行多结果CEA，并与成本效益平面上的分析进行比较。为比较不同潜在阈值下的策略而开发的汇总指标包括：预期净损失（ENL）平面，用于量化预期净效益的差异；ENL等高线，用于确定使ENL最小化的首选策略及其完美信息的期望值；以及成本效益可接受性平面，显示使ENL最小化的策略的概率。
传统分析表明，当在家中额外天数的阈值（k1）超过1068美元时，PEACH具有成本效益；或者仅考虑在家中死亡比例时，PEACH被常规护理所主导。相比之下，在联合考虑成本和结果的CDU空间中，没有一种替代方案占主导地位，最优策略取决于阈值。例如，当k1 = 2000美元且k2 = 2000美元（在家中死亡的阈值）时，PEACH使ENL最小化，PEACH具有成本效益的概率为51.6%。
在CDU空间中的比较和相关汇总指标相对于多领域比较具有明显优势，有助于在不确定性下对成本和多种效应进行透明且稳健的联合比较，跨越效应的潜在阈值，更好地为净效益评估以及相关的报销和研究决策提供信息。
