Reward/Punishment reversal learning in older suicide attempters

Abstract

Objective: Suicide rates are high in old age, and the contribution of cognitive risk factors remains poorly understood. Suicide may be viewed as an outcome of an altered decision process. The authors hypothesized that impairment in reward/punishment-based learning, a component of affective decision making, is associated with attempted suicide in late-life depression. They expected that suicide attempters would discount past reward/punishment history, focusing excessively on the most recent rewards and punishments. The authors further hypothesized that this impairment could be dissociated from executive abilities, such as forward planning.

Method: The authors assessed reward/punishment-based learning using the probabilistic reversal learning task in 65 individuals age 60 and older: suicide attempters, suicide ideators, nonsuicidal depressed elderly, and nondepressed comparison subjects. The authors used a reinforcement learning computational model to decompose reward/punishment processing over time. The Stockings of Cambridge test served as a control measure of executive function.

Results: Suicide attempters but not suicide ideators showed impaired probabilistic reversal learning compared to both nonsuicidal depressed elderly and nondepressed comparison subjects, after controlling for effects of education, global cognitive function, and substance use. Model-based analyses revealed that suicide attempters discounted previous history to a higher degree relative to comparison subjects, basing their choice largely on reward/punishment received on the last trial. Groups did not differ in their performance on the Stockings of Cambridge test.

Conclusions: Older suicide attempters display impaired reward/punishment-based learning. The authors propose a hypothesis that older suicide attempters make overly present-focused decisions, ignoring past experiences. Modification of this "myopia for the past" may have therapeutic potential.

Conflict of interest statement

All authors report no competing interests.

Figures

Figure 1

Probabilistic reversal learning task

Figure 2. Overall probabilistic reversal learning performance: proportion of participants reaching the learning criterion in each stage

The bars display the proportion of participants in each group who reached a pre-defined 8-trial learning criterion. All groups demonstrated similar probabilistic learning ability in the acquisition stage (chi-square test: χ2(3)=0.57, p=0.90), but performance in the reversal stage differed considerably between groups (χ2(3)=13.1, p=0.004)

Figure 3. Model-based analyses of probabilistic reversal learning: reliance on past reinforcement history (decay) and learning from punishments

3a, 3d: model-derived meta-learning parameters by group. 3b, 3c, 3e, 3f: correlations with behavioral indices illustrate the meaning of meta-learning parameters. Memory: attention to prior reinforcement history vs. last trial. Learning rate from punishments: impact of punishment on the last trial on subsequent choice. 3a. Analyses based on the fit-parameters of the computational model revealed a lower memory among suicide attempters, compared to non-depressed controls (omnibus ANOVA F(3,61)=2.77, p=0.049, Tukey HSD post-hoc: suicide attempters<controls, p=0.039). That is, attempters relied less on their previous reinforcement history in making their decisions and more on feedback on the last trial, compared to non-depressed controls. As expected, memory was negatively correlated with the total number of switches in subjects’ choice (3b) and with the number of probabilistic switches (switches following non-contingent negative feedback, 3c). 3d. While the three depressed groups, particularly suicide ideators, tended to have a lower learning rate from punishments, group differences were not significant: F(3,61)=2.52, p=0.066, suicide ideators<controls, p=0.087. This seemingly surprising trend was due to perseverative errors in the three depressed groups (means shown in Table 2): learning rate from punishments was negatively correlated with the number of perseverative errors (3e), i.e. subjects who did not switch their choice in response to punishment made multiple perseverative errors. Learning rate from punishments was positively correlated with the proportion of switches in response to non-contingent punishment (probabilistic switches) among all switches (3f).