Promoting Gratitude as a Resource for Sustainable Mental Health: Results of a 3-Armed Randomized Controlled Trial up to 6 Months Follow-up

The current paper describes results from three of the four conditions of a large parallel, randomized controlled trial in the Netherlands about the efficacy of different happiness exercises.¹ Practicing various gratitude exercises during 6 weeks was compared with performing self-kindness acts during 6 weeks (active control group) and waiting for performing a well-being intervention-exercise (waitlist control group). Participants completed online assessments at baseline (T0), post-test (T1), 6 weeks follow-up (T2) and 6 months follow-up (T3). A power analysis showed that 78 participants per condition were needed to obtain an effect size of at least d = .45 on well-being (Davis et al. 2016; Kerr et al. 2015) assuming a two-tailed test with alpha of .05 and a power of .80 (1 – β). This study was approved by the Ethics Committee of the University of Twente (BCE17240) and registered in the Dutch Trial Register (NTR6786).

In September 2017, advertisements were placed via different media, recruiting participants via Facebook/LinkedIn (n = 425), a popular psychology magazine (n = 156) and national and regional newspapers (n = 72). The recruitment message was: ‘Can your well-being use a boost? Join this study on the effects of happiness exercises from the University of Twente for free’. Participants needed to be at least 18 years old, have a sufficient Internet connection and a valid email address, and they had to master the Dutch language to complete questionnaires and follow the intervention instructions. Interested participants applied online for the study. Participants who completed the online informed consent procedure were automatically redirected to an online screening questionnaire. Eligible participants had no severe depressive or anxiety symptoms, indicated by a score < 34 on the Center for Epidemiological Studies Depression (CES-D) questionnaire (Bouma et al. 1995; Santor et al. 1995; Radloff 1977) and a score < 15 on the Generalized Anxiety Disorder-7 (GAD-7) questionnaire (Donker et al. 2009; Spitzer et al. 2006). For the current study, people who were already flourishing (n = 37), as measured with the Mental Health Continuum-Short Form (Keyes et al. 2008; Lamers et al. 2011), were excluded to create a more homogeneous group. Analyses including the people who were flourishing at baseline showed similar results as presented here. After completing the baseline survey, participants were randomly assigned to each condition using random numbers from randomizer.org. Randomization was stratified by gender, education level (low, intermediate, high) and flourishing (possess at least one of the three aspects of emotional well-being and at least six of the eleven aspects of social and psychological well-being) to ensure equal distribution of these characteristics among the conditions.

The final sample consisted of 217 participants with an allocation ratio of 1:1:1. The mean age of the participants was 48.6 years (SD = 9.6), with a range between 18 and 67 years. Participants were mainly female (89.9%), higher educated (78.3%), of Dutch nationality (88.5%), married (48.8%), living with others (74.2%) and in paid employment (73.7%). Socio-demographic characteristics did not significantly differ between groups, except for Dutch nationality which was more pronounced in the waitlist condition compared to the other conditions, χ²(2) = 6.59, p = .037. Also, a marginal difference was found for living with children showing that fewer participants in the waitlist condition were living with children compared to the other conditions, χ²(2) = 5.19, p = .075 (Table 1). Intervention instructions were sent to the participants on each Sunday, during 6 weeks. If necessary, reminders were sent to increase adherence. At the T3 assessment, participants were informed that they could win one of 20 books about gratitude if they completed that last survey.

The 6 weeks gratitude intervention consisted of evidence-based gratitude exercises which were mainly writing exercises (Emmons and McCullough 2003; Emmons and Stern 2013), psycho-education (e.g. There are two types of gratitude; (1) more general appreciation such as for your living standards, talents or possessions, and (2) appreciation for others, such as about how friendly or supporting others have been for you.) and questions that trigger reflection about cultivating gratitude (e.g. What did you feel when practicing the gratitude exercise of this week? What did it mean for your daily activities?). The weekly gratitude exercises and general instructions are displayed in Table 2. On average, we expected that participants would invest around 15 min a day on 5 days per week (75 min per week).

Participants in the active control condition received the same instructions each week for 6 weeks in total. They were instructed to perform five self-kindness activities on one day per week: ‘In our daily lives, we all perform acts of kindness for others, but we often neglect to do nice things for ourselves—small and simple things that require relatively little extra money or effort. Examples include treating yourself on a special coffee or on your favorite pastry, magazine or meal, taking a 5-min break from work or study, give yourself a compliment or go for a walk. Now and then, you can treat yourself on something bigger, such as a massage or spending time on your favorite hobby. Within the next five days, you pick one day as your kindness day. On this day, you are to perform five nice things for yourself, all five in one day. These acts of kindness should be something out of the ordinary that you do for yourself with a little extra effort and may or may not be similar to the acts listed above.’ On average, we expected that participants would invest around 10 min per behavioral activity and sometimes one activity of 1 h, all on one day per week (75 min per week on average).

Participants in the waitlist control condition were told that they were allocated to the flexible group: ‘You can choose the activity that fits best to your needs to improve your happiness and well-being. However, before you can choose this activity, we would like to monitor your normal fluctuations in your level of well-being.’ At 6 weeks follow-up, the participants in this group chose their activity for which they received the instructions during the following 6 weeks.

Descriptive statistics of demographic variables and outcome measures at baseline were calculated and compared between conditions using χ²-tests and univariate analysis of variance (ANOVA) in SPSS version 25.0, using two-tailed tests and p < .05. Χ²-tests and ANOVAs were also used to determine whether the number of drop-outs differed per condition and whether drop-outs (i.e. participants who completed only the baseline assessment) differed from participants who completed at least two assessments. Adherence was defined as reporting at least four grateful things or activities for themselves during at least 4 of the 6 weeks. The difference in adherence between the gratitude and active control condition was examined using χ²-tests.

Changes in outcome measures over time were examined using multilevel growth curve modeling in R (version 0.99.902, NLME package) to account for repeated measures nested within individuals (Singer and Willett 2003). This method also complies with the intention-to-treat principal because all randomized individuals are included in the analyses. First, we tested the efficacy of the 6-week gratitude intervention compared to active control and waitlist control up to 6 weeks follow-up using an unconditional growth curve model, specifying linear and quadratic changes over time. The unconditional quadratic growth model best fitted the data on all outcomes (> Δχ² (10) = 18.19, p = .001). Therefore, the baseline unconditional quadratic growth model was compared with the hypothesis-testing models. Time was centered on the second time point (post-test) and each condition was dummy-coded.

Finally, post-hoc Cohen’s d effect sizes were calculated for significant and marginally significant findings from multilevel growth curve analyses and based on completers only data. For this, means at post-test and follow-up assessments were adjusted for baseline differences using analysis of covariance, in which the post-test or follow-up mean scores were entered as dependent variable and the corresponding baseline mean score as covariates. Between-group effect sizes were then calculated by subtracting the adjusted mean scores of the gratitude condition from the adjusted mean score of one of the control conditions at each time point, and dividing the mean difference by the pooled standard deviation. The 95% confidence intervals (CI’s) of Cohen’s d were computed in R.

At baseline, participants in the gratitude and active control condition were strongly convinced about the utility of doing a well-being exercise (M = 7.0, SD = 1.52), which decreased somewhat after the first week of practicing their exercise (M = 6.7, SD = 1.65). Participants were also highly motivated to practice in the upcoming weeks (M = 8.0, SD = 1.10), which was still satisfactorily after the first week intervention (M = 7.6, SD = 1.78). More importantly, expectations and motivation did not significantly differ between both conditions at any assessment (p > .652), except for the expectations after one week practice. The participants in the gratitude intervention were significantly more convinced about the utility of this exercise for their well-being compared to active control (M_gratitude = 7.16, SD = 1.33; M_{activecontrol} = 6.31, SD = 1.81, F(130) = 9.26, p = .003).

A total of 176 participants (81.1%) completed at least two of the four questionnaires. Drop-out was highest at 6 months follow-up (36.9%). The total number of drop-outs, as well as the number of drop-outs at post-test and at 6 weeks follow-up, were significantly lower in the waitlist condition compared to the other two conditions, χ²(2) = 12.51, p = .002; χ²(2) = 13.97, p = .001; χ²(2) = 11.69, p = .003 respectively. Drop-outs were significantly younger, F(215) = 10.44, p = .001 and they scored significantly lower on generic dispositional gratitude, F(215) = 6.31, p = .013 and marginally lower on appreciation of others F(215) = 3.70, p = .056. No other differences on socio-demographics or baseline measures were found between drop-outs and completers.

Almost half of the participants of the intervention and active control condition adhered to the program (47.9%). However, adherence was significantly greater for the 6-week gratitude intervention compared to performing self-kindness, 57.5% vs. 38.4%; χ²(1) = 5.38, p = .020. This difference in favor of the gratitude intervention was also visible during most weeks (ps < .007), except for week 3 and 4 where adherence did not significantly differ between groups (ps > .185). Overall adherence was highest in week 1 (78.1%) and week 2 (63%), and lowest in week 4 (41.8%) and week 5 (42.5%).

The raw means and standard deviations for all outcomes on each assessment are presented in Table 3. The gratitude intervention was more effective in enhancing well-being, γ₂₁ = − .21, S.E. = .08, t(315) = − 2.63, p = .009, appreciation of simple pleasures, γ₂₁ = − .30, S.E. = .12, t(312) = − 2.57, p = .011, appreciation of the contribution of others, γ₂₁ = − .34, S.E. = .16, t(312) = − 2.08, p = .038, and grateful mood, γ₂₁ = − .43, S.E. = .20, t(310) = − 2.22, p = .027, compared to the active control condition (Table 4). Mainly significant quadratic increases over time were found, indicating that the gratitude intervention led to stronger effects up to post-test, followed by a slight decline when participants were no longer directly stimulated to practice gratitude. However, well-being also showed a marginal linear increase over time, γ₁₁ = .10, S.E. = .05, t(315) = 1.81, p = .072. No significant changes over time were found for generic dispositional gratitude as measured by the GQ-6, sense of abundance, depression and anxiety compared to active control. Effect sizes at post-test were d = .63 (95% CI = .23–1.02) for mental well-being, d = .61 for appreciation of simple pleasure (95% CI = .21–1.01), d = .48 for appreciation of contribution of others (95% CI = .09–.87) and d = .53 for grateful mood (95% CI = .13–.92). At 6 weeks follow-up, effect sizes were d = .40 (95% CI = − .01–.81) for mental well-being, d = .28 for appreciation of simple pleasure (95% CI = − .13–.69), d = .09 for appreciation of contribution of others (95% CI = − .32–.50) and d = .15 for grateful mood (95% CI = − .26–.56).

Practicing gratitude led to significant linear and quadratic increases over time relative to waitlist control for well-being, γ₁₁ = .18, S.E. = .05, t(315) = 3.55, p < .001; γ₂₁ = − .28, S.E. = .07, t(315) = − 3.80, p < .001 and appreciation of the contribution of others, γ₁₁ = .18, S.E. = .09, t(312) = 1.97, p = .050; γ₂₁ = − .33, S.E. = .16, t(312) = − 2.13, p = .034. In addition, greater quadratic changes and marginal linear changes in favor of practicing gratitude were found for grateful mood, γ₂₁ = − .50, S.E. = .18, t(310) = − 2.69, p = .008; γ₁₁ = .20, S.E. = .11, t(310) = 1.73, p = .084 and for sense of abundance, γ₁₁ = .15, S.E. = .07, t(312) = 2.12, p = .035; γ₂₁ = − .20, S.E. = .11, t(312) = − 1.78, p = .076. Furthermore, practicing gratitude led to significant quadratic, but not linear, changes over time in generic dispositional gratitude, γ₂₁ = − .26, S.E. = .011 t(313) = − 2.35, p = .020 and in greater linear, but not quadratic, changes in appreciation of simple pleasures, γ₁₁ = .23, S.E. = .08, t(312) = 2.76, p = .006. Finally, marginal linear decreases were found for anxiety symptoms, γ₁₁ = − .69, S.E. = .38, t(310) = − 1.82, p = .070, while changes in depressive symptoms were not significant compared to waitlist control. At post-test, effect sizes were d = .93 (95% CI = .54–1.31) for mental well-being, d = .61 for appreciation of simple pleasure (95% CI = .23–.98), d = .65 for appreciation of contribution of others (95% CI = .27–1.02), d = .80 for grateful mood (95% CI = .42–1.18), d = .53 for sense of abundance (95% CI = .16–.90), d = .33 for anxiety symptoms (95% CI = − .04–.70) and d = .51 for dispositional gratitude (95% CI = .14–.88). Effect sizes at 6 weeks follow-up were d = .66 (95% CI = .27–1.05) for mental well-being, d = .61 for appreciation of simple pleasure (95% CI = .22–1.00), d = .49 for appreciation of contribution of others (95% CI = .10–.87), d = .47 for grateful mood (95% CI = .08–.85), d = .41 for sense of abundance (95% CI = .03–.79), d = .27 for anxiety symptoms (95% CI = − .11–.65) and d = .31 for dispositional gratitude (95% CI = − .07–.69).

Unexpectedly, practicing self-kindness (the active control condition) was marginally more effective in decreasing anxiety and depressive symptoms compared to waitlist control. Marginal linear shifts in depressive symptoms were found (γ₁₂ = 1.50, S.E. = .89, t(310) = 1.68, p = .094) and marginal quadratic shifts in anxiety symptoms (γ₂₂ = 1.04, S.E. = .63, t(310) = 1.67, p = .096). Effect sizes were d = .29 (95% CI = − .08–.66) for depressive symptoms and d = .45 for anxiety symptoms (95% CI = .08–.82) at post-test. At 6 weeks follow-up, effect sizes were d = .37 (95% CI = − .03–.77) for depressive symptoms and d = .32 for anxiety symptoms (95% CI = − .08–.71). These findings show that reducing anxiety and depressive symptoms was most successful when participants practiced self-kindness during 6 weeks compared to doing nothing, while practicing gratitude did not lead to a significant reduction in these mental health indicators.

To test the overall efficacy of practicing gratitude up to 6-months, we repeated the analyses using a quadratic growth model wherein we compared the gratitude condition solely with the active control condition as the waitlist had started their intervention after 6-weeks follow-up. Practicing gratitude led to significantly greater linear and quadratic changes over time in mental well-being, γ₁₁ = .13, S.E. = .05, t(278) = 2.34, p = .020; γ₂₁ = − .08, S.E. = .04, t(278) = − 1.98, p = .048, and to greater quadratic changes and marginal linear changes over time in simple pleasure appreciation, γ₂₁ = − .12, S.E. = .06, t(275) = − 2.15, p = .033; γ₁₁ = .14, S.E. = .08, t(275) = 1.75, p = .081. No other significant changes were found, suggesting that practicing gratitude was not more successful than practicing self-kindness in improving generic dispositional gratitude, sense of abundance, appreciation of the contribution of others and grateful mood, and in reducing depressive symptoms and anxiety symptoms over the long-run. Effect sizes at 6-months follow-up were d = .31 for mental well-being (95% CI = − .12–.73) and d = .17 for appreciation of simple pleasure (95% CI = − .26–.60).

Participants in the gratitude intervention were significantly more satisfied with their gratitude exercises (M = 23.67, SD = 5.14) compared to active control (M = 19.75, SD = 4.80) as reported at post-test, F(101) = 15.85, p < .001. When satisfaction was expressed in a grade from 1 to 10, a similar result was found with a 7.2 (SD = 1.62) for practicing gratitude and a 5.9 (SD (1.82) for practicing self-kindness, F(101) = 15.32, p < .001. However, both groups spent a similar amount of time to practice their well-being exercise (p = .505) of which most participants practiced 30 to 60 min per week (34.3%; gratitude = 35.3%, self-kindness = 33.3%) or 15–30 min per week (31.4%; gratitude = 27.5%, self-kindness = 35.3%). In addition, 20.6% of the participants practiced between 1 and 2 h per week (gratitude = 25.5%, self-kindness = 15.7%) and 6.9% of the participants spent more than 2 h per week on the intervention (gratitude = 2.0%, self-kindness = 11.8%).

The findings demonstrated that practicing gratitude was more effective in improving mental well-being in comparison to both active and waitlist control. The moderate to large effects on mental well-being relative to active and waitlist control at post-intervention were larger than has generally been found in studies on the efficacy of gratitude interventions (Davis et al. 2016). One explanation may be that the participants experienced suboptimal levels of well-being, allowing more room for improvement. Many studies have been evaluating the impact of gratitude interventions in healthy populations, thereby minimalizing the potential benefits due to a ceiling effect (Davis et al. 2016). A second explanation for the stronger effects found in the current study in comparison to earlier studies may be that, to date, most gratitude interventions were of short duration, i.e. one or two weeks focusing on a single activity such as writing a gratitude letter or making a gratitude list. Finding the optimal dosage of interventions is an important challenge in the field of positive psychology (Lyubomirsky and Layous 2013) and the current study indicates that extending gratitude interventions to 6 weeks, including various gratitude activities may enhance their impact. Directly comparing gratitude interventions of 2, 4 and 6 weeks would give more direct evidence of their optimal duration. A third explanation for the larger effects is that the exercises in the gratitude intervention were much more varied than in other studies, making the intervention less vulnerable to hedonic adaption (Sheldon and Lyubomirsky 2012; Lyubomirsky et al. 2005). Participants in the gratitude condition were instructed to practice different exercises at a daily basis, which might have kept the participants more motivated. In fact, adherence was greater in the gratitude intervention compared to active control and the participants who received the gratitude intervention were significantly more satisfied with this intervention compared to those who were instructed to perform five nice things for themselves on one day each week during 6 weeks. However, these findings also indicate that the found effects in comparison to active control might have been due to the frequency of practice and not to the type of intervention. Although the needed time to perform weekly gratitude or self-kindness activities did not differ in the current study, future research should use an active control activity that also controls for the variability of the gratitude intervention to ensure its efficacy also in comparison to another active intervention.

Interestingly, the findings of this study showed that the effects of practicing gratitude on mental well-being were maintained up to 6 months follow-up. This is one of the first studies to assess the longer term impact of gratitude interventions (Davis et al. 2016). This sustainable impact on mental well-being is relevant as lower levels of mental well-being have been found to increase the risk for first and recurrent incidence of psychopathology and diagnosed mental disorders (e.g. Schotanus-Dijkstra et al. 2017; Lamers et al. 2015; Keyes et al. 2010; Wood and Joseph 2010). Given the large numbers of people with suboptimal well-being (Schotanus-Dijkstra et al. 2016) and with distress and common mental disorders in the general population (de Graaf et al. 2012; Whiteford et al. 2013), there is a need for easy-to-administer and low-intensity interventions in the context of public mental health (e.g. Huppert 2004; Schotanus-Dijkstra et al. 2019a; Fledderus et al. 2012; Kobau et al. 2011). The participants in the present study practiced gratitude without any guidance, suggesting the intervention can be administered at low costs.

The results also showed that levels of symptomatology of depression and anxiety diminished in the gratitude intervention group, but that these changes were not significantly different in comparison with active or waitlist control. The non-significant findings of gratitude versus waitlist control were unexpected as earlier studies have found that gratitude promotes positive emotions and coping-styles such positive reframing and acceptance which are associated with reduced distress (e.g. Wood et al. 2007; Lambert et al. 2012). One explanation is that the participant in the waitlist condition also improved, resulting in comparable levels of distress in both groups.

Also, reductions in distress were found to be marginally larger for the participants practicing self-kindness in comparison to waitlist on the short-term. This finding may be explained by the fact that the participants in the active control condition were invited to do five nice things for themselves on one day a week for six weeks. These activities may have supported participants in developing a more compassionate stance towards themselves (Nelson et al. 2016), which has been shown to be positively associated with reduced distress in earlier studies (Kirby et al. 2017; MacBeth and Gumley 2012). Another explanation is that performing acts of kindness for oneself will enhance the experience of positive emotions. Amplifying positive emotions has been shown to be an effective pathway in promoting well-being, increasing resilience and diminishing distress (Fredrickson 2001, 1998; Fredrickson et al. 2003). In sum, these findings indicate that practicing self-kindness might also be a low-benefit intervention, but then on targeting depressive and anxiety symptoms in people with reduced levels of well-being and low to moderate levels of distress (Kirby et al. 2017; Shin and Lyubomirsky 2017). However, more research is needed as the effects of self-kindness might differ from self-indulgence of which the latter was intended in the current study.

To our knowledge this is one of the first studies to directly compare the impact of a gratitude intervention on various measures of gratitude. The outcome instruments included generic dispositional gratitude (GQ, McCullough et al. 2002), three specific dispositional characteristics of grateful persons: sense of abundance, appreciating simple pleasures and relational gratitude (GRAT, Watkins et al. 2003) and gratitude as a mood (McCullough et al. 2004). The findings demonstrated that the gratitude intervention led to greater increases in all of these outcomes compared to waitlist control, of which the effects at post-test and at 6-weeks follow-up were mainly of moderate to large magnitude. The gratitude intervention was also more effective compared to active control on appreciating simple pleasures, relational gratitude and grateful mood. However, only the overall efficacy of appreciating simple pleasures maintained up to 6-months follow-up, although the difference at this latter time-point was also (marginally) in favor of practicing gratitude compared to active control for relational gratitude and grateful mood.

The current findings are in contrast with a recent meta-analysis which demonstrated the absence of effects on gratitude in comparison with measurement-only control (Davis et al. 2016). A possible explanation for this discrepancy is the longer duration of the intervention in the current study in comparison to most interventions included in the meta-analysis. In addition, most studies evaluated the impact of gratitude interventions on (generic) dispositional gratitude (Davis et al. 2016), while the current study is one of the first to also report the effects of gratitude interventions on gratitude as mood. Dispositional gratitude represents a more stable trait that may be harder to change over time. This may explain the limited impact of gratitude interventions on dispositional gratitude (Davis et al. 2016), although the findings of the current study suggest that a variable 6-week gratitude intervention can effectively improve dispositional gratitude as well. In contrast to dispositional gratitude, moods could be more easily influenced by intentions to be grateful (McCullough et al. 2004). For example, in the current gratitude intervention, the participants were invited to start each day, after waking-up, with focusing on their intention to be appreciative of small good things in life and of positive contributions of others. This may have contributed to a lasting appreciative perspective on one’s life and a resulting grateful mood. Also, though of shorter duration and less stable than traits, moods certainly last longer than emotional states and thus represent an intermediate level between gratitude as an affective trait and gratitude as an emotion (Rosenberg 1998). The presence of a grateful mood will promote the occurrence of positive emotions and contribute to upward spirals in daily life (McCullough et al. 2004; Catalino and Fredrickson 2011; Watkins et al. 2015). Our findings suggest that grateful mood is a good proximal outcome of gratitude interventions in addition to dispositional gratitude (Davis et al. 2016; Wood et al. 2010).

Next to grateful mood, strongest effects were found for the appreciation of the contribution of others. This finding suggests that the effects on relational gratitude is a core aspect of gratitude, for example through the recognition and appreciation of receiving specific benefits from other persons (e.g. McCullough et al. 2002; Wood et al. 2009; Watkins et al. 2003). Participants practicing gratitude have become more appreciative of the contribution of others to their well-being in comparison to the participants practicing self-kindness. This is relevant because many recent studies have demonstrated the positive impact of experiencing and expressing gratitude on maintaining relationships (e.g. Grant and Gino 2010; Algoe and Zhaoyang 2016; Lambert and Fincham 2011) and improving relationships (e.g. Algoe 2012; Algoe et al. 2010; Bartlett et al. 2012).

Lastly, the effects on appreciation of simple pleasures were sustained up to 6 months follow-up in comparison with active control. A longer term comparison with waitlist control was not possible but might have yielded even more and stronger results as was also found up to 6-week follow-up. These findings are relevant because recent research suggests that gratitude is an adaptive resource promoting and maintaining mental health, also in the presence of negative life-events (e.g. Wood et al. 2010; Disabato et al. 2016; Millstein et al. 2016; Sirois and Wood 2017; Kashdan et al. 2006). However, longer term comparison to a more equal active control group and to waitlist control is needed to draw firm conclusion on the sustainability of the gratitude effects of practicing gratitude.