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

WHILE t < T+1

% Select a drug using the ε-greedy method with increasing 1-(1/log(n+2)).

pn = rand(1);

IF (pn < (1-(1/log(n+2))))

[nil,cDrug] = max(Q{t}(cState,:));

ELSE cDrug = randi([1,size(SS{t},1)]); END
% Update the treatment history.

tHist = [tHist;cDrug];

% Simulate a next state and reward associated to R(s,a,s’).

nState = randi([1,3]);

IF nState ~= 3 % For the alive patients,

dHist = [dHist,nState]; % Update the disease history.

[~,nStateIdx] = ismember(dHist,HS{t+1},'rows');

% Evaluate the one-step reward for the transition from cState to nState where cDrug is used at t, using EvModelDC. fProb is a 1x3 matrix including the transition probabilities from cState to the next health states; and fSBP and fSBPSD are 1x2 matrices including the mean SBPs for the controlled and uncontrolled patients after treatment.

[IR,fProb,fSBP,fSBPSD] = EvModelDC...

(Scenario,t,dHist,cDrug,cProb,cSBP,cSBPSD,cMT,DrugForCVDDM);

% Generate the subsequent health states depending on the future transitions to be considered.