Can undervaluing delayed rewards predict risk of relapse during recovery?

Relapse is a relatively common (40% – 60%) part of the recovery process. Characterizing the predictors of relapse can help us better identify individuals at increased risk and inform new approaches for preventing it. This study suggests that one measure of impulsivity (i.e., delay discounting: the overvaluing of smaller immediate rewards over larger later rewards) may be a useful marker for identifying individuals who lack confidence in their ability to remain abstinent and are, therefore, at increased risk for relapse. 


Delay discountingthe tendency to overvalue smaller immediate rewards over larger later rewards, is a measure of impulsivity that is shown to predict treatment outcomesDelay discounting is often thought of as a behavioral marker of addictionsuch that individuals may engage in substance use for its immediate rewarding effects with less consideration of its long-term consequences and the future benefits of abstaining. Still, it is unclear if this measure of impulsivity can be used as a behavioral marker of relapse during recovery. Assessing the confidence in one’s ability to abstain from substance use (i.e.abstinence self-efficacy) is one way to measure risk of relapse. At present, it is unclear whether abstinence self-efficacy and delay discounting are related in the context of addiction recovery. If delay discounting predicts abstinence self-efficacy, it could ultimately help to identify recovering individuals at increased risk of relapse and provide them with continuing care approaches to improve delay discounting and support continued recovery. The current study aimed to address this gap in the literature by assessing the relationship between delay discountingrecovery-relevant factors, and abstinence self-efficacy. 


Participants were recruited from the International Quit & Recovery Registry, an online research participation registry for individuals in recovery. The study was cross-sectional and individuals who are members of the registry and were included in this study self-report recovery from one or more substances. The final study sample consisted of 216 adults in recovery from substance use disorders. Participants completed questionnaires to assess their current recovery status, with respect to their primary substance (preferred substance of choice)Participants were asked if they had used their primary substance in the past 30 days (i.e., relapse), if they considered their use of the substance to be ongoing (i.e., current ongoing use), and when they last used the substance (i.e., days in recovery). Abstinence self-efficacy was measured with the Relapse Situation Efficacy Questionnaire, which assesses one’s perceived ability to remain abstinent in various situations and emotional states. An adjusting-delay task was used to measure delay discountingwhereby participants hypothetically chose between an immediate $500.00 reward and a delayed $1,000.00 reward. The delay started at 3 weeks and either increased or decreased in subsequent trials depending on the participant’s initial choice. Responses were used to calculate a discounting rate (i.e., 1 divided by the delay time expected to reduce the larger reward’s value by 50%) for each participant. A higher discounting rate indicated that a participant devalued the larger future rewards to a greater degree. 

Figure 1. The image above is an example of the series of choices a participant might be presented with across 5 trials and how the delay might change over the course of an adjusting-delay task depending on the participant’s choice (depicted as boxes with a green outline).

Individual relationships between each of these variables were first assessedThereafter, the authors conducted a single analysis looking at potential predictors (recovery duration, delay discounting, abstinence, gender, income, education, and primary substance) of abstinence self-efficacy. Participants were primarily White (80%) and resided in the United States (82%). On average, participants were 48 years oldhad 5 years of recovery, and 1,780 days since their last relapse. Ongoing substance use was reported by 10% of participants. The most commonly reported substance of choice included alcohol (63%), stimulants (15%), non-prescription opioids (10%), and cannabis (6%).   


Delay discounting and abstinence self-efficacy were related to recovery duration.

Lower rates of discounting (i.e., less impulsivity) and higher abstinence self-efficacy (i.e., confidence to remain abstinent) were associated with longer recovery durations (i.e., more days of continuous abstinence).

Delay discounting, recovery status, and demographics independently predicted abstinence self-efficacy.

Lower rates of discounting (i.e., valuing larger future rewards to a greater degree) predicted higher abstinence self-efficacyRegarding recovery status, current abstinence and longer recovery durations were associated with higher abstinence self-efficacy. Higher income and a higher level of education also predicted higher abstinence self-efficacy. Lower rates of discounting (i.e., valuing larger future rewards to a greater degree) independently predicted higher abstinence self-efficacy.


This study suggests that recovering individuals who are less impulsive (i.e., lower discounting), more educated, and currently abstinent, with longer recovery durations and higher income, may be more confident in their ability to remain abstinent (i.e., greater abstinence self-efficacy). Thinking less about the future in the context of rewards and feeling less confident in one’s recovery could ultimately be detrimental, as these individuals may be at greater risk of choosing substance use for its immediate rewarding effects as opposed to choosing abstinence for its long-term future benefits. Continued care interventions emphasizing future-focused thoughts and behaviors may be important for enhancing abstinence self-efficacy and facilitating continued recovery in these individuals. For example, studies have demonstrated the benefits of neurocognitive training tasks that help rebuild cognitive functions to reduce impulsivity. 12-step participation may also help to reduce impulsivity, and reductions in impulsivity may in turn play a role in the effectiveness of recovery engagement (e.g., 12-step program participation) and outcomes

Given that participants in this study had an average of 5 years in recovery (i.e., sustained long-term remission), supporting individuals with impulsive tendencies may be that much more important during early recovery periods. Education and income are easily measured characteristics that may also help to identify those with lower abstinence self-efficacy and increased risk for relapse. Furthermore, this study suggests that as time goes on, individuals might become more confident in their ability to continue along a successful recovery trajectory. Continuing care and mutual-help programs may help increase abstinence self-efficacy at earlier recovery durations. Greater self-efficacy might also, in turn, support ongoing abstinence, as the authors found that greater confidence in one’s ability to remain abstinent (i.e., abstinence self-efficacy) was associated with more days in recovery. Consistent with previous research, lower impulsivity (i.e., discounting) was also related to longer recovery durations. This suggests that individuals with higher impulsivity may be less likely to achieve longer recovery durations and need more continuing care to support and promote abstinence. Given the relationship between delay discounting and abstinence self-efficacy, and the importance of high abstinence self-efficacy for supporting recovery, addressing impulsive decision making might be one way to enhance confidence in one’s ability to remain abstinent and subsequently achieve longer recovery durations.   

  1. Older participants reported greater abstinence self-efficacy, and authors were unable to include age in their models for statistical reasons. Given that older participants have the potential for longer recovery durations, it is unclear whether the relationship between abstinence self-efficacy and recovery durations is influenced by age. Research also suggests the interplay between self-efficacy and motivation in predicting treatment outcomes. Therefore, understanding the contributions of multiple factors, both their influence and interactions with each other and the primary measures of interest, is needed to better characterize recovery processes and outcomes. 
  2. This study did not assess actual abstinence, as opposed to abstinence self-efficacy, or the use of substances that were not the primary substance of choice. Therefore, it is unclear whether these outcomes would differ if actual abstinence were assessed longitudinally or if abstinence from all substances was examined. Furthermore, measures of recovery were self-reported and so investigations that biologically confirm substance use or abstinence are needed. 
  3. The study was cross-sectional and participants were primarily White. Additional research is needed to determine if these relationships are observed in longitudinal samples and apply to other ethnicities/racial backgrounds. Though primary substance was assessed as a predictor of abstinence self-efficacy, primary substances other than alcohol were not widely represented in this study and additional research is needed to determine whether these outcomes are observed in other substance use populations. 


  • For individuals and families seeking recovery: This study suggests that recovering individuals who are more impulsive (i.e.,have higher rates of delay discounting) may be less confident in their ability to remain abstinent (i.e., lower abstinence self-efficacy). A tendency to choose the smaller immediate reward over the larger later reward could therefore place these individuals at increased risk for relapse, as they may be more likely to choose substance use for its immediate rewarding effects. Continued care interventions that teach these individuals to implement future-focused thoughts and behaviors, such as cognitive training exercises, 12-step involvement, and distraction- or mindfulness– based exercises might ultimately improve delay discounting, and in turn enhance abstinence self-efficacy and facilitate continued recovery. Additional research is needed to replicate these findings and determine if delay discounting predicts actual abstinence as opposed to abstinence self-efficacy, and whether it is a better predictor itself of future relapse than abstinence self-efficacy, or whether self-efficacy amediator of delay discounting is helping to explain its predictive effects that it has shown on future substance use. Still, this study provides an important foundation for better understanding the predictors of recovery and this area of research will ultimately help guide new approaches to enhance recovery for individuals and families seeking it.
  • For treatment professionals and treatment systems: This study suggests that recovering individuals who are more impulsive (i.e., higher delay discounting) may be less confident in their ability to remain abstinent (i.e., lower abstinence self-efficacy). Although not a direct measure of relapse risk, abstinence self-efficacy is suggestedto be related to it. Identifying more impulsive patients in recovery and implementing therapeutic approaches that encourage future-focused thoughts and behaviors might help to increase abstinence self-efficacy and, in turn, protect against relapse. Recovering individuals who are less educated, have lower incomes, continue to use substances, and have less time in recovery may also be at increased risk for low abstinence self-efficacy and additional recovery support may be needed for these individuals to avoid substance use and achieve longer recovery durations.
  • For scientists: The authors of this study found that recovering individuals with higher rates of delay discounting, less education, lower incomes, and shorter recovery durations who are not currently abstinent may have lower abstinence self-efficacy. Additional research is needed to replicate and extend these findings. Though abstinence self-efficacy is a predictor of relapse risk, direct and longitudinal measurement of relapse in the context of this work is needed. Studies of more complex interactions can also help to best understand the factors that moderate treatment and recovery mechanisms. Moreover, investigation is needed to determine whether cognitive remediation or clinician guided therapies that teach future-focused thoughts and behaviors (e.g., mindfulness-based exercises) can improve delay discounting, and in turn enhance abstinence self-efficacy and reduce risk for relapse. 
  • For policy makers:Studies like this help us to identify predictors of successful recovery and guide new potential continuing care approaches. With approximately 40% to 60% of individuals experiencing a lapse in recovery (return to substance use), it is essential to identify the factors contributing to it. The current study suggests that recovering individuals who are more impulsive (i.e., higher rates of delay discounting), less educated, and not currently abstinent, with shorter recovery durations and lower income may be less confident in their ability to remain abstinent (i.e., lower abstinence self-efficacy). These individuals may need additional support to enhance successful recovery outcomes. Funding and policies to support continued care interventions that teach future-focused thoughts and behaviors (e.g., cognitive training exercises, long-term 12-step involvement, distraction- or mindfulness-based practices) might improve delay discounting, and in turn enhance abstinence self-efficacy and facilitate long-term recovery. Additional research is needed to replicate these outcomes and broaden our understanding of recovery and its predictors.


Athamneh, L. N., DeHart, W. B., Pope, D., Mellis, A. M., Snider, S. E., Kaplan, B. A., & Bickel, W. K. (2019). The phenotype of recovery III: Delay discounting predicts abstinence self-efficacy among individuals in recovery from substance use disorders.Psychology of Addictive Behaviors33(3), 310-317. doi: 10.1037/adb0000460