MODEL TYPES SUBMITTED TO NICE- WHAT IS CONSIDERED APPROPRIATE BY EVIDENCE REVIEW GROUPS?
Author(s)
Hearmon NC1, Ghosh W1, Buguth B1, Kusel J2
1Costello Medical Consulting Ltd., Cambridge, UK, 2Costello Medical Consulting Ltd., London, UK
OBJECTIVES: A review of Evidence Review Group (ERG) critiques of models in recent NICE submissions was conducted to determine any associations between model type and ERG criticisms. METHODS: Full NICE single technology appraisals published over the past 12 months were reviewed and data extracted on indication, model type, whether the ERG considered the model type appropriate and details of the ERG critique. When the model structure was considered appropriate for the decision problem, this was assumed to also apply to model type. RESULTS: The 52 submissions reviewed included 47 cohort state-transition models (Markov, partitioned survival, semi-Markov and decision tree/Markov models), 1 Markov model run at both a cohort and microsimulation level, 3 Monte Carlo individual patient simulations and 1 Discretely Integrated Condition Event (DICE) model. In 34 (65.4%) cases the ERG agreed the model type was appropriate, typically due to alignment with previous models in the same or similar indications. In 9 (17.3%) cases the ERG critique of the model type was unclear or not stated. In the remaining 9 (17.3%) submissions, the ERG was unsatisfied with the model type to some extent. In 2 cases a dynamic modelling approach was considered more appropriate than a Markov (in infectious diseases), and in 2 cases a cohort model was used where patient heterogeneity was believed to be important. A total of 5 models were criticised for inflexibility in capturing key evidence, or appeared overcomplicated and lacked transparency. Additionally, the DICE model was criticised for impractical implementation and lack of clear benefit over a discrete event simulation model. CONCLUSIONS: Cohort state-transition models are generally considered appropriate by ERGs, the justification often being that they have been previously used in the disease area. Other model types are more likely in disease areas with fewer submissions, and are generally considered appropriate if their implementation is transparent and user-friendly.
Conference/Value in Health Info
2017-11, ISPOR Europe 2017, Glasgow, Scotland
Value in Health, Vol. 20, No. 9 (October 2017)
Code
PRM118
Topic
Methodological & Statistical Research
Topic Subcategory
Modeling and simulation
Disease
Multiple Diseases