Cost-effectiveness of long-acting injectable risperidone versus flupentixol decanoate in the treatment of schizophrenia: a Markov model parameterized using administrative data

Simon Frey, Roland Linder, Georg Juckel, Tom Stargardt

We use longitudinal patient-level data from a German sickness fund with 7.26 million insured in a Markov-simulation model to assess the cost-effectiveness of long-acting injectable risperidone (LAI-RIS) compared with long-acting injectable flupentixol (LAI-FLX) in the long-term management of schizophrenia. We simulate treatment costs from the payer's perspective, hospitalization, the probability to be prescribed co-medication, and treatment discontinuation over a 2-year time horizon. Model inputs were derived from 935 patients hospitalized with schizophrenia between 2005 and 2008 who received either LAI-RIS or LAI-FLX for at least 1 month. After 2 years, 89.4% (95.8%) of patients who were initiated on LAI-RIS (LAI-FLX) discontinued the initial regimen. The number of days spent in hospital per month and patient was slightly lower with LAI-RIS (1.08 vs. 1.28 days, p<0.001). The proportion of patients receiving side-effect co-medication was lower with LAI-RIS (8.3 vs. 15.0% per month, p<0.001). Mean total costs of treatment per patient and month were 1,015 € under LAI-RIS and 395 € under LAI-FLX, resulting in an ICER of 3,088 € (95% CI [-913 €; 3,551 €]) for an avoided hospital day per patient and month in the base case scenario with a 15.1% probability of LAI-FLX being the dominant treatment strategy. Cost differences were mainly attributable to the higher drug costs of LAI-RIS. The effectiveness of LAI-RIS in preventing hospital days appears to be similar to LAI-FLX, with a slight superiority in side-effect and switching rates. This comes at the cost of substantially higher treatment expenses. From a decision-maker's point of view, the use of health insurance data as a source of input for decision models appears to be a reasonable alternative to models driven by clinical data only.