NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Meltzer DO, Hoomans T, Chung JW, et al. Minimal Modeling Approaches to Value of Information Analysis for Health Research [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2011 Jun. (Methods Future Research Needs Reports, No. 6.)

Cover of Minimal Modeling Approaches to Value of Information Analysis for Health Research

Minimal Modeling Approaches to Value of Information Analysis for Health Research [Internet].

Show details

Appendix ADetails of Application 1 and Application 2

Table A1Details of application 1 (azithromycin vs. amoxicillin/clavulanate in acute bacterial sinusitis) and application 2 (erlotinib and gemcitabine vs. gemcitabine alone in pancreatic cancer)

ApplicationApplication 1: Azithromycin vs. Amoxicillin/Clavulanate in Acute Bacterial SinusitisApplication 2: Erlotinib and Gemcitabine vs. Gemcitabine Alone in Pancreatic Cancer
Modeling of health outcomes (i.e. limited modeling component)No modeling, manual replication of curve for symptom reliefNo modeling, manual replication of survival curve
Approach to value of information calculationsBootstrapping/simulation, nonparametricBootstrapping/simulation, nonparametric
ApplicationAzithromycin vs. amoxicillin/clavulanate in patients with acute bacterial sinusitisErlotinib and gemcitabine vs. gemcitabine in patients with pancreatic cancer
PerspectiveThird-party payerThird-party payer

Symptom resolution: Comparative effectiveness study [41)


Drug price: Drug Topics Red Book (42)


Survival curves: Phase III trial (46)


Costs: assumptions

Incidence [prevalence]10M25k [35k]
Time horizon10 years5 year
(Cost-)effectiveness results

Effectiveness = −0.57 days to symptom resolution


NB: $18.2


Effectiveness = +0.08 life-years



Value of information resultsEffects only: pEVPI = $40M
Both cost and effects: pEVPI =$250M
Effects only: pEVPI = $0
Both cost and effects: pEVPI = $400M
Willingness-to-payWTP = $73.20/day to symptom resolutionWTP = $180k/life-year

NB = net benefit; pEVPI = population expected value of information; WTP = willingness-to-pay

Brief description of the methods use to develop the survival curves in the two examples: After enlarging a hard copy of the published survival figure, we drew a rectangular grid of lines on the figure in order to read the treatment-specific survival fraction corresponding to each unit increment of time (a day for the sinusitis example [Appendix B] and a half-month for the pancreatic cancer example [Appendix C]). We then took an empty dataset with n observations, n corresponding to the treatment-specific sample sizes of the respective study, and assigned a time of event to each observation (representing one patient) based on the survival fraction and the corresponding time, thereby generating an individual level outcomes dataset. We validated this approach by replicating the point estimate of the hazard ratio reported in the original analysis using our generated individual-level data. We also performed a nonparametric bootstrap of the individual level data to replicate the 95 percent confidence interval of the hazard ratio reported in the original analyses. These results are presented in Appendix B and Appendix C. These bootstrap results also produced the distribution of comparative effects that we use for the value of information analyses.

PubReader format: click here to try


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (1.1M)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...