8.3.1Individual-based model
The structural limitations of the Markov state-transition model may be solved by IBMs such as DES. While Markov state-transition models conceptualise the problem as a series of states that a cohort of objects can move from one state to another at each fixed cycle, DES does not need to define a fixed cycle length. DES moves forward by the occurrence of an event, which is governed by the parameters describing time-to-event[75, 389]. This may reduce complexity in SDDPs, where a series of actions take place according to the patient’s condition over time. For example, the hypertension SDDP model used the successive decision tree to depict the sequential decision making process. The number of possible health states in a Markov state-transition model was exponentially increased for each additional cycle to embed previous history in each state. Because of the computational complexity increased by the number of possible health states in each period, the hypertension SDDP model had to restrict the number of drug switching period. However, DES may reduce the complexity where there are no events happening, for example, a patient in DES may have nothing happening for several cycles defined in the Markov state-transition model.
Furthermore, DES does not have the memory limitations of the Markov model. Entities in DES can carry their history as attributes, whereas all history information needs to be explicitly embedded in the states for cohort models[66]. By more easily embedding memory and historical information into the entities, DES provides flexibility to address the complex relationship between individual patients’ various current and historical risk factors and the final outcome. These values may be updated while the patients go through the entire simulation. The updated information is stored as entity attributes and can be used to determine the transition to the subsequent disease pathway, as well as to estimate costs and effectiveness.
The disadvantage of DES is that it normally takes more time to obtain results than cohort models because DES usually requires sampling a large number of patients and/or multiple replications to increase the precision of model outputs[75, 389]. With a given total time to spend, increasing computational time in the evaluation model means spending less time to search the decision space. DES also requires a much greater number of calculations and detailed data to apply the various risks, decisions, resource use, and other elements that are variable[66, 75, 389]. Therefore, there will be a trade-off between the flexibility of model and the effort required to build a model.
Dostları ilə paylaş: |