Sequential drug decision problems in long-term medical conditions: a case Study of Primary Hypertension Eunju Kim ba, ma, msc

ELSE t=5; END % For the patients died. END % Reach a terminal state. END % The end of learning in time t. % Determine the best drug for each health state in t based on the Q-values. FOR t = 1:T FOR h = 1:size(Q{t},1) [v,idx] = sort(Q{t}(h,:),'descend'); FOR a = 1:size(v,2) IF The feasibility assumptions is satisfied, Break END [OptV(:,:),OptSol(:,:)] = max(Q{t},[],2); END END END

Figure ‎6.. Pseudo-code of the Q-learning used for the hypertension SDDP model

Chapter 7.Modelling sequential drug decision problem for hypertension: Results

7.1Chapter overview

This chapter starts with the sub-section of model validity, which describes how the hypertension SDDP model was checked for technical errors and validated internally and externally. The rest of this chapter consists of the outcomes of the hypertension SDDP model depending on the optimisation method applied, which are enumeration, SA, GA and RL. The main outcomes include the optimal solution, the total net benefit where the optimal solution is used and the time or computational expense. Sensitivity analyses present whether the optimal or near optimal solutions are robust to changing the value of key parameters in each method.