Gordon B. Hazen

Sensitivity analysis and the expected value of perfect information

Measures of decision sensitivity that have been applied to medical decision problems were examined. Traditional threshold proximity methods have recently been supple mented by probabilistic sensitivity analysis, and by entropy-based measures of sen sitivity. The authors propose a fourth measure based upon the expected value of perfect information (EVPI), which they believe superior both methodologically and prag matically. Both the traditional and the newly suggested sensitivity measures focus en tirely on the likelihood of decision change without attention to corresponding changes in payoff, which are often small. Consequently, these measures can dramatically over state problem sensitivity. EVPI, on the other hand, incorporates both the probability of a decision change and the marginal benefit of such a change into a single measure, and therefore provides a superior picture of problem sensitivity. To lend support to this contention, the authors revisit three problems from the literature and compare the results of sensitivity analyses using probabilistic, entropy-based, and EVPI-based mea sures. Key words: sensitivity analysis; expected value of perfect information. (Med Decis Making 1998;18:95-109)

Partial Information, Dominance, and Potential Optimality in Multiattribute Utility Theory

When a multiattribute utility function is only partially specified by prior preference statements, what can be said about the relative desirability of actual alternatives? This question is addressed for the cases of additively separable cardinal utility with unknown scaling constants; multiplicatively separable cardinal utility with unknown scaling constants; and additively separable ordinal utility. We review previous approaches, and treat issues of consistency is the prior information consistent?, dominance does the prior preference information imply that one outcome is preferred to another? and potential optimality are there utility functions of the given form, consistent with prior preference information, under which a particular outcome is preference optimal?. In the additive cases, a key relationship between dominance and potential optimality may be derived. The paper concludes by presenting an example application to a well known nuclear siting study.

A NEW PERSPECTIVE ON MULTIPLE INTERNAL RATES OF RETURN

ABSTRACT The most commonly cited drawback to using the internal rate of return to evaluate deterministic cash flow streams is the possibility of multiple conflicting internal rates, or no internal rate at all. We claim, however, that contrary to current consensus, multiple or nonexistent internal rates are not contradictory, meaningless or invalid as rates of return. There is, moreover, no need to carefully examine a cash flow stream to rule out the possibility of multiple internal rates, or to throw out or ignore “unreasonable” rates. What we show is that when there are multiple (or even complex-valued) internal rates, each has a meaningful interpretation as a rate of return on its own underlying investment stream.It does not matter which rate is used to accept or reject the cash flow stream, as long as one identifies the underlying investment stream as a net investment or net borrowing. When we say it does not matter which rate is used, we mean that regardless of which rate is chosen, the cash-flow acceptance or rejection decision will be the same, and consistent with net present value.

A Bayesian approach to sensitivity analysis

Sensitivity analysis has traditionally been applied to decision models to quantify the stability of a preferred alternative to parametric variation. In the health literature, sensitivity measures have traditionally been based upon distance metrics, payoff variations, and probability measures. We advocate a new approach based on information value and argue that such an approach is better suited to address the decision-makers real concerns. We provide an example comparing conventional sensitivity analysis to one based on information value. This article is a US government work and is in the public domain in the United States.

Stochastic Trees A New Technique for Temporal Medical Decision Modeling

This paper introduces stochastic trees, a new modeling approach for the class of medical decision problems in which risks of mortality and morbidity may extend over time. A stochastic tree may be regarded as a continuous-time version of a Markov-cycle tree, or alternately, as a multi-state DEALE model. Optimal decisions in stochastic trees can be determined by rollback, much in the same fashion as decision trees. The author discusses how age- dependent mortality rates and declining incidence rates may be modeled using stochastic trees. Concepts are illustrated using examples from the medical literature. It is argued that stochastic trees possess important advantages over Markov-cycle trees for medical decision modeling. Key words: stochastic trees; DEALE models; decision analysis; Markov cycle trees. (Med Decis Making 1992;12:163-178)

Comparative effectiveness of donation after cardiac death versus donation after brain death liver transplantation: Recognizing who can benefit

Due to organ scarcity and wait‐list mortality, transplantation of donation after cardiac death (DCD) livers has increased. However, the group of patients benefiting from DCD liver transplantation is unknown. We studied the comparative effectiveness of DCD versus donation after brain death (DBD) liver transplantation. A Markov model was constructed to compare undergoing DCD transplantation with remaining on the wait‐list until death or DBD liver transplantation. Differences in life years, quality‐adjusted life years (QALYs), and costs according to candidate Model for End‐Stage Liver Disease (MELD) score were considered. A separate model for hepatocellular carcinoma (HCC) patients with and without MELD exception points was constructed. For patients with a MELD score <15, DCD transplantation resulted in greater costs and reduced effectiveness. Patients with a MELD score of 15 to 20 experienced an improvement in effectiveness (0.07 QALYs) with DCD liver transplantation, but the incremental cost‐effectiveness ratio (ICER) was >

Optimality Conditions in Nonconical Multiple-Objective Programming

2,000,000/QALY. Patients with MELD scores of 21 to 30 (0.25 QALYs) and >30 (0.83 QALYs) also benefited from DCD transplantation with ICERs of

Subjectively weighted linear utility

478,222/QALY and

An Extension of the Internal Rate of Return to Stochastic Cash Flows

120,144/QALY, respectively. Sensitivity analyses demonstrated stable results for MELD scores <15 and >20, but the preferred strategy for the MELD 15 to 20 category was uncertain. DCD transplantation was associated with increased costs and reduced survival for HCC patients with exception points but led to improved survival (0.26 QALYs) at a cost of

Factored Stochastic Trees A Tool for solving Complex Temporal Medical Decision Models

392,067/QALY for patients without exception points. In conclusion, DCD liver transplantation results in inferior survival for patients with a MELD score <15 and HCC patients receiving MELD exception points, but provides a survival benefit to patients with a MELD score >20 and to HCC patients without MELD exception points. Liver Transpl,18:630–640, 2012.