Early data peek for DYMO-PD group

Outline

This is a peak at some of the early data for the Dynamics of Motivated decision making (DYMO) task in the PD group. Patients that were able were tested in counterbalanced on/off dopaminergic medication sessions.Participants completed subtasks in which they either aimed to minimise physical effort or maximise reward. They first learnt to associate 7 images with different (fixed) amounts of effort or reward (credits), then demonstrate sufficient learning of associations, then performed a main experiment in which they had to choose between either two or three options that always resulted in either a level of effort or reward.

The main experimental questions/hypotheses were:

-Do Parkinson’s patients differ to healthy controls in implicit markers of motivation - eg. sensitivity to reward or effort.

-Does dopamine medication make ppts more willing to put in cognitive effort to avoid physical effort, or make them less effort averse or cautious? Another way of asking this is does dopamine affect RT or accuracy in decisions leading to physical effort compared to leading to reward?

-Does dopamine medication alter subjective assessments of cognitive effort / choice difficulty?

-Does apathy lead to changes in decision-making. Specifically..

-Does cognitive/executive apathy lead to less willingness to put in cognitive effort in decisions that lead to effort?

-Does action initiation apathy lead to more willingness to put in cognitive effort in decisions that lead to effort, or lead to increased timeouts (choice failure, due to difficulty initiation actions particularly in difficult choices?)

As well as the behavioural task, participants complete the Dimensional apathy scale (which probes action initiation and executive apathy), apathy motivation index, MOCA, SHAPS, PHQ-8, fatigue severity scale, forward and backwards digit span, and a repurposed questions called the GCTI to look at spontaneous thought (to see if this correlates (negatively) with action initiation apathy). Patients also complete the UPDRS (Parkinson’s rating scale).

Overall task performance

Below shows main plots of the data. Main interesting findings seem to be:

-PD patients are generally slower in effortful tasks compared to reward when compared to healthy controls (particularly in easier decisions), possibly indicating a general effort aversion in keeping with their disease

-Dopamine medication does not seem to make people less effort averse (similar RT in effort task between on/off state) but it does appear to speed decisions in the reward task. This possibly indicates and is in support of literature that dopamine doesn’t affect sensitivity to effort but effects are driven by reward sensitivity

Inferences about learning

One question would be whether effects of on/off (or indeed HC vs PD) are driven by learning. E.g. if they learn associations better, they are quicker and more accurate.

Below shows the number of checkphase trials as a surrogate for learning (in this section, which follows 21 learning trials in which they encounter all possible stimulus pairings, participants must choose the best (lowest effort or highest reward) option out of two randomly-selected images - the checkphase ends after 18 correct answers, therefore the total number of checkphase trials is indicative of learning).The median number of trials look similar between HC/PD and PDon/off suggesting that learning is similar.

Another way to look at learning is to examine responses by separating trials into indivdual stimuli presented as the best option, to see their cumulative accuracy for each presentation of stimuli. For this, ordered data from the training, checkphase, and main experiment are pooled and a tally of correct answers are plotted (limited to first 25 trials, which is the least trials of the least frequent stimuli, converted onto a common scale of value (flipped values in effort task). Results from all participants, separated into groups and task (Effort vs reward) are shown below (i.e. session 1 and 2 are pooled together). It appears that learning is less effective in the effort task, likely due to the format of the task (no numeric readout of value in the effort task - here they learn by the ‘feel’ of the effort and height of a bar.

In the below plots, training, checkphase and main experiment data are pooled to demonstrate learning, however only PD patients that have done both on and off sessions are analysed and compared. Overall, learning looks similar on and off medication.Again, results are presented for each individual stimulus presented as the ‘best’ option.

Effect of distracting options - effort task

Below is one way of displaying the effect of distracting options on RT and accuracy. Only three option choices are presented. For each plot, the horizontal facet numbers represent the ‘best’ option, and vertical facet numbers represent the second best option. Therefore across the diagonal represents the most difficult (value difference of 1) choices, with decreasing difficulty fanning away from the diagonal. For each of these comparisons, choices are split based on the the value of the third distractor as low or high (e.g. if best option is 2 (MVC 30%) and second best option 3 (MVC 40%) in effort task, ‘low’ would be distractor value of 4 or 5, ‘high’ would be distractor value of 6 or 7. Any ‘mid’ level distractors are excluded). There is a suggestion that high effort distractors are more distracting (longer RT) when looking at pooled results but effects looks small.

Below shows similar plots, but looking at accuracy instead of RTs.

Effect of distracting options - reward task

Below I have made the same plots but looking at the reward task. First I have made plots of reaction times:

Below are distractor plots for accuracy for the reward task:

Subjective ratings of choice difficulty

Next I have looked at subjective ratings of choice difficulty. Following certain (prespecified) choices, participants were asked to rate (between 0 and 10) how cognitively difficult they found the choice. For each participant, this rating is z-valued in standard deviation units across both testing sessions. The first two plots are split by difficulty (value difference) and number of options. The last plot is pooled across all trials.

Force metrics in effort task - PD group only

Finally, I look at force metrics in the effort task. I have calculated maximum force, yank force, AUC.