ISPOR Europe 2018
Barcelona, Spain
November, 2018
PRM135
Obesity
Research on Methods (RM)
Modeling Methods (MS)
DICE FOR NICE? LESSONS FROM A SINGLE TECHNOLOGY APPRAISAL
Sullivan W1, Bullement A2, Lee D3
1BresMed Health Solutions, Sheffield, UK, 2BresMed Health Solutions, Nottingham, NTT, UK, 3BresMed Health Solutions Ltd., Sheffield, UK
OBJECTIVES: To assess the suitability of a Discretely Integrated Condition Event (DICE) approach to discrete event simulation (DES) modelling in Microsoft Excel for a company submission to the National Institute for Health and Care Excellence (NICE) Single Technology Appraisal (STA) process.

METHODS: To implement the DICE methodology, variables and processes are tabulated. Using a DICE approach in Excel for a DES model that would traditionally be programmed directly in Excel’s underlying Visual Basic for Applications (VBA) code has advantages for validation and review by stakeholders more comfortable reviewing data and logic in Excel versus VBA. An application of DICE in the company submission for NICE TA494 (naltrexone-bupropion for obesity) provide evidence on its suitability for a company submission to the NICE STA process.

RESULTS: The Appraisal Consultation Document reports the Evidence Review Group (ERG) and company experiencing “extremely slow run times” and the committee concluded that “an alternative approach to implementing the DES model would be more practical for decision-making”. The ERG Report describes the ERG being able to “check formulae in the DICE sheet” but being “unable to examine the internal validity of the model according to its usual standards” and that this was mainly a “consequence of long model run times for one single deterministic analysis”. The Final Appraisal Determination stated that a “revised model [using VBA directly] ran more efficiently”. Base case model execution was approximately 200 times faster in the revised model.

CONCLUSIONS: When considering implementing DICE methodology, the speed at which the chosen software can read specifying tables and process consequences of events, and the practical implications of this, should be evaluated. Evidence from TA494 suggests that applying DICE methodology for a DES in Excel may lead to impractically long run times for a NICE STA model.