INTERNAL, EXTERNAL, AND CROSS-MODEL VALIDATION OF A MULTI-OUTCOME DECISION MODEL FOR PARKINSON'S DISEASE

Author(s)

Siebert U1, Bornschein B2, Dodel R3, 1 Harvard Medical School, Boston, MA, USA; 2 University of Munich, Munich, Germany; 3 Friedrich-Wilhelms-University, Bonn, Germany

OBJECTIVES: We have recently reported on a generic, multi-outcome disease model for Parkinson's disease (PD). Now we present first results of internal, external and cross-model validation. METHODS: Our lifetime PD Markov model simulates a hypothetical cohort of patients moving through health states reflecting patient characteristics that would be observed in the absence of treatment (Hoehn&Yahr "off" states [HYoff]). We used HYoff I-V and death as Markov states. The model is designed to simultaneously predict multiple outcomes, e.g. time in Hoehn&Yahr “on” states (HYon) observed under treatment, quality-adjusted life expectancy (QALE), or complication rates. As internal validation, we compared time in HYoff stages predicted by our model to results reported in the progression study used to derive our input parameters. As external validation, we compared model results of mean times in HYoff and HYon states with extern literature data not used in our model. Finally, we cross-validated our model comparing QALE under levodopa treatment with QALE of other published models reporting this outcome. RESULTS: Internal validation of HYoff input data showed a 97.4-99.9% accuracy. Although external validation of average HYoff progression rates overestimated external population data from Hoehn & Yahr (1967) by 19%, the mean HYon progression rate predicted by our model (0.42 HY stages/y) matched well with estimates reported in the literature (0.40 HY stages/y). After restricting our model to a 5-year time horizon, discounted QALYs exceeded those from 2 other published models by 24% and 35%. This differences were mostly attributable to different Markov state-specific utilities. As other Markov models for drug treatment did not evaluate QALE, we could not cross-validate for this outcome. CONCLUSIONS: Our PD model is internal valid and closely reproduces external data for progression under standard treatment. Variability in QALE are due to a combination of different model design, state-specific utilities, and underlying study populations.

Conference/Value in Health Info

2004-10, ISPOR Europe 2004, Hamburg, Germany

Value in Health, Vol. 7, No. 6 (November/December 2004)

Code

PNL25

Topic

Methodological & Statistical Research

Topic Subcategory

Modeling and simulation

Disease

Neurological Disorders

Explore Related HEOR by Topic


Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on ispor.org. Update my browser now

×