ID: 12217
Meeting / Value in Health Info: ISPOR Twelfth Annual International Meeting
Arlington, VA, USA
May, 2007
Value in Health, Vol. 10, No.3 (May/June 2007)
Code: PMC5
Disease: Multiple Diseases/No Specific Disease
Topic: Conceptual Papers (CP)
Topic Subcategory: Cost Studies including CE/CB/CU, resources use and productivity (CS)
Title: EVALUATING AN ONLINE CALCULATOR FOR ANALYZING INCREMENTAL NET BENEFIT AND THE EXPECTED VALUE OF PERFECT INFORMATION FROM PATIENT LEVEL DATA
Author(s):

William F. McGhan, PharmD, PhD, Professor of Pharmacy & Health Policy, Namita Tundia, BS, Graduate Student, Hafsa Quadri, BS, Graduate Student, Shilpa Viswanathan, BS, Graduate Student, Andrew M. Peterson, PharmD, Chair, Department of Pharmacy Practice and Pharmacy AdministrationUniversity of the Sciences in Philadelphia, Philadelphia, PA, USA

Content:

OBJECTIVES: To evaluate an online calculator for analyzing incremental net benefit and expected value of perfect information. METHODS: An online calculator was developed that generates incremental net benefit (INB) and expected value of perfect information (EVPI) statistics and graphs from patient level cost and effectiveness data and is freely available at www.HealthStrategy.com. The calculator was compared to two other software options: Obenchain's ICEplane software which can be downloaded from www.math.iupui.edu/~indyasa/bobdown.htm and an MS Excel module developed by Nixon, Wonderling and Grieve, and downloadable through http://www.mrc-bsu.cam.ac.uk. For the comparison, three datasets were utilized from published studies dealing with fluoxetine (FLUX), a randomized test dataset (RND), and acupuncture (ACCU). RESULTS: INB values on the three datasets at various lambda threshold values (WTP) between HealthStrategy, ICEplane and Nixon were as follows: ACCU (WTP at mean ICER, 10088): 2.0 vs 2.7 vs 2.0; RND (WTP at 1.0): 2.0 vs 2.4 vs 2.0; and FLUX (WTP at 1742): 57472 vs 56570 vs 56979. ICEplane does not calculate EVPI, but the Nixon module was adapted by adding a column for the estimation of the unit normal loss integral as documented by Griffin and Chilcott. The respective HealthStrategy and Nixon module EVPI values for each dataset were as follows: ACCU (WTP at mean ICER): 67.9 vs 66.7; RND (WTP at 0.0): 0.167 vs 0.105; and FLUX (WTP at 0.0): 169.8 vs 169.8. CONCLUSION: All three software provide basic statistics and graphs including scatter plots, confidence intervals and acceptability curves. For INB and EVPI, the HealthStrategy and Nixon packages use parametric calculations requiring assumptions not always met with cost effectiveness data. For future research, more comprehensive software should be added to this comparison like Stata, the R statistical package and Winbugs, along with consideration of population EVPI and expected value of sample information (EVSI).

Back to HDS Software List