OBJECTIVES: Partitioned survival models (PSM) and state transition models (STM) are the most common modelling approaches in health technology appraisals (HTAs) for treatment of chronic, progressive, and non-communicable diseases including cancers. Due to limitations of PSMs, the NICE DSU TSD19 recommended STMs using multi-state modelling (MSM) techniques (June 2017). This study aimed at reviewing the use of MSM in NICE HTAs.

METHODS: A review of NICE TAs since June 2017 and all HSTs was performed to identify economic models which used MSM. Information was extracted on: justification for MSM; data sources to inform transition probabilities (TP); MSM for TP estimation and extrapolation; software used; feedback from Evidence Review Groups (ERGs).

RESULTS: Economic models in 14 of 308 NICE TAs used MSM (12 TAs used MSM within STMs; two TAs (in cancers) used a hybrid model structure consisting of both PSM and MSM). Two of 16 HSTs used MSM within STMs. MSM was used in STMs to estimate TP for each transition reflecting explicit disease process and treatment as well as covariate effects therefore to improve transparency around process underlying TP estimation and extrapolation and to facilitate meaningful sensitivity analysis. In the majority of HTAs, ERGs accepted MSM as appropriate to represent disease progression, with the main concerns being inappropriate implementation, including data used and survival model selection. For HSTs, MSM was criticised as being overly complex and may ‘over fit’ data.

CONCLUSIONS: There is a growing but limited number of models for NICE HTAs using MSM. As a modelling approach for consideration in cancer and other chronic, progressive, non-communicable diseases, MSM is feasible to implement, acceptable to HTA bodies and should be used more to inform decision making on cost effectiveness. HTA guidelines on the appropriate use of MSM and research on MSM using published survival data are urgently needed.