Optimal Bayesian Sequential Sampling Rules for the Economic Evaluation of Health Technologies

Abstract

We present a Bayes sequential economic evaluation model for health technologies in which an investigator has flexibility over the timing of a decision to stop carrying out research and to conclude that one technology is preferred to another on cost-effectiveness grounds. We implement the model by using an evaluation of the treatment of bacterial sinusitis and derive approximations of the optimal stopping rule as a function of accumulated sample size. We compare the performance of the model with existing frequentist and Bayes sequential designs and investigate the sensitivity of the stopping rule to changes in the parameters of the model. Our results suggest that accounting for the dynamic nature of experimentation, together with its economic parameters, should lead to greater efficiency in resource allocation within healthcare systems.

Optimal sequential sampling with delayed observations and unknown variance

“…They allow for on-line ('earn while you learn') and off-line learning as well as the absence or presence of two types of sampling cost (marginal cost per sample, and discounting costs). Their model extends work in simulation optimization Gans 2009, Chick andFrazier 2012), biostatistics and health economics (Pertile, Forster, and La Torre 2014).…”

“…Some of these techniques are based on Bayesian principles Inoue 2001, Frazier, Powell, andDayanik 2008), and others on heuristics (Chen 1996, Branke and Schmidt 2004, Chen and Kelton 2005, Chen, Yiicesan, Dai, and Chen 2010 or frequentist techniques (Kim andNelson 2006, Xu, Nelson, andHong 2010). Significant advances are also found in others areas of operations (Gittins and Jones 1974, Bertsimas and Mersereau 2007, Caro and Gallien 2007, Ahuja and Birge 2015, biostatistics and health economics Eick 1995, Pertile, Forster, andLa Torre 2014).…”

Optimal decision rules for HTA under uncertainty: a wider, dynamic perspective

“…Pertile et al . () show that it is possible to derive optimal sequential sampling rules for technology adoption and research abandonment decisions. Although this will require decision makers to move towards a more complex decision‐making framework for HTA, we believe that the potential gains, in terms of enhanced efficiency of resource allocation, justify the effort.…”

“…As suggested by Conti and Claxton (2009), we believe that the Bayesian literature on sequential sampling (Chernoff, 1965a(Chernoff, , 1965bChick and Gans, 2009) offers potential because it allows for a multi-period perspective and is able to incorporate all of the dimensions that are discussed previously. Pertile et al (2010) show that it is possible to derive optimal sequential sampling rules for technology adoption and research abandonment decisions. Although this will require decision makers to move towards a more complex decision-making framework for HTA, we believe that the potential gains, in terms of enhanced efficiency of resource allocation, justify the effort.…”

A Bayesian Decision Theoretic Model of Sequential Experimentation with Delayed Response

“…This question is especially important given the increasing policy interest in sequential and adaptive trial designs (European Medicines Agency, ; US Food and Drug Administration, ). Concern about avoiding unnecessary recruitment to the trial, past the point at which evidence is deemed to be conclusive, means that there is a growing focus on valuing the cost of carrying out research, together with the benefits that accrue to trial participants and patients who may benefit from a new technology (Lewis et al ., ; Willan and Kowgier, ; Pertile et al ., ). Indeed, the UK National Institute for Health and Care Excellence (National Institute for Health and Care Excellence, ) examines cost and effectiveness when making trade‐offs in care, and the value‐based health movement (Porter, ) calls for increased attention to the health benefits that are obtained for a given level of expenditure.…”