Daily Affective Dynamics in Major Depressive Disorder: The Role of Daily Stressors and Positive Events

Affective Variability and MDD

Daily affective variability (i.e., indexed as a within-person variability of affect within a person) captures the overall range of affect levels a person may experience across a particular time period (Ram & Gerstorf, 2009). Greater variability in affect may be associated with worse health outcomes. Studies support a cost linked to moving in and out of an affective “homeostasis” with higher fluctuations predicting poorer mental and physical health (Piazza et al., 2013). Temporal fluctuations in PA are associated with elevated mortality risk in older adults (Ong & Steptoe, 2020).

Affective variability has generally been found to be significantly higher in depressed individuals than in healthy controls (Bos et al., 2019). However, there are discrepancies in the literature on whether NA and PA are significantly variable in depressed individuals. One study showed individuals with more severe forms of depression demonstrated lower variability of PA. Another study, however, observed greater variability of PA in MDD (Heininga et al., 2017). Other studies on affective fluctuations showed that depressed patients had higher variability in NA (Peeters et al., 2006) and PA in comparison to controls (Gilbert, 2012; Gruber et al., 2013; Peeters et al., 2006). Patients with a current disorder of depression and/or anxiety disorders were found to have higher fluctuations of NA (Lamers et al., 2018) and PA than remitted patients and controls (Schoevers et al., 2021). Most studies have not included large samples affected by mood disorders, nor have they systematically investigated the comorbid role of anxiety disorders (Pfaltz et al., 2010). Such comparisons are essential for understanding how affective regulation differs between depression and comorbid anxiety disorder (O’Leary et al., 2017). This study focuses on affective variability in MDD by assessing day-to-day within-person fluctuations controlling for comorbid anxiety disorders. We also measure affective variability in a novel way by implementing multilevel heterogeneous variance models that simultaneously model level and variability in the same analyses.

Affective Reactivity and MDD

Our second focus in the study is the role of context (i.e., how affect changes in response to daily stressful or positive events). The change in affect due to daily context (i.e., a stressful event or a positive event) is termed affective reactivity (Charles et al., 2009). The role of context in the depression literature has focused primarily on negative events such as stress or major past adverse events. Previous studies examining affective reactivity following daily life events in MDD have yielded mixed results. One study reported increased NA in reaction to daily negative events among individuals with prior depression as compared with controls (Husky et al., 2009), while another study found smaller increases in NA in response to negative events in individuals with MDD (Peeters et al., 2003). The latter study also reported that decreases in PA following stressors were of significantly smaller magnitude in MDD compared with controls, while increases in PA following positive events were larger. A third study found no significant differences between participants with and without depression regarding changes in PA following either positive or negative events, or relative to increases in NA following negative events (Bylsma et al., 2011). The one relatively consistent yet surprising finding in examining affect reactivity in depression is the “mood brightening” effect. The mood brightening (MB) effect observed in ecological investigations of affective reactivity in depression is an unusual theoretical and empirical phenomenon in the depression literature. The MB effect refers to findings that individuals with MDD report larger decreases in negative affect (NA) in response to positive events in daily life compared to individuals with no history of MDD (Bylsma et al., 2011; Peeters et al., 2003; Thompson et al., 2012). Interestingly, these studies also observed that the MB effect was observed even after controlling for baseline mood, suggesting that this effect could not be accounted for by high levels of current NA among individuals with MDD. Another study observed a similar MB effect for depressed participants, but these individuals did not differ from controls concerning PA following positive events, or positive or negative affect following negative events (Khazanov et al., 2019). In summary, with the exception of the mood brightening effect that was observed across three investigations (Bylsma et al., 2011; Peeters et al., 2003; Thompson et al., 2012), the findings of previous EMA studies are inconsistent concerning differences in affective reactivity between individuals with and without depression. Such inconsistency may be explained by numerous factors including the presence of comorbidity, the nature of the event itself, and the clinical state of the participants at the time of the study (aan het Rot et al., 2012; Bylsma et al., 2011; Tan et al., 2012; Vrshek-Schallhorn et al., 2015).

In summary, the study of affective dynamics in MDD is warranted given the mixed results in the literature and the importance of affective dynamics in well-being. Implementing ecological sampling paradigms to advance the causes and maintenance of MDD and improve treatment interventions is supported (Telford et al., 2012). A meta-analysis has shown that overall, low psychological well-being co-occurs with more variable, unstable, and inert affect, with more pronounced results for negative compared to positive affect (Houben et al., 2015). The current study examines the distinctive nature of affective variability and affective reactivity, among MDD vs. non-MDD groups, particularly focusing on positive affective reactivity given the mood brightening effect observed in the literature.

Participants and Procedure

The current study used data drawn from the second wave of the National Study of Daily Experiences (NSDE), the daily diary project of the Midlife in the United States (MIDUS) study (Ryff & Almeida, 2017; Ryff et al., 2018, 2021). The MIDUS study is a national survey examining health and well-being across middle and older adulthood. The original sample at the first wave of the study (MIDUS I) was recruited from 1995 to 1996 via a random-digit-dialing (RDD) procedure, resulting in 7,108 non-institutionalized and English-speaking adults aged between 25 and 74. In 2004 to 2006, a longitudinal follow-up was conducted (MIDUS II) which included 4,963 participants who completed the baseline phone interview (aged 35–86 years; 75.4% retention rate). Of those, 81% completed self-administered questionnaires (SAQs). Due to the lack of participants as racial minorities, especially those who identified as Black (90.1% white and 4.6% Black in MIDUS II vs. 75.1% white and 12.3% Black in the U.S. population), this sample was not fully representative of the U.S. population. To improve the diversity representation of the MIDUS II sample, a supplemental sample of 592 participants who mostly identified as Black (93.4%) were recruited from Milwaukee County, WI. The Milwaukee oversample completed the initial baseline interview (70.7%) and 67.2% of them also completed the SAQs. A random subsample of MIDUS II participants were invited to enroll in the daily diary study, and 2,022 participants completed the telephone interviews about their daily experiences on eight consecutive days (Almeida et al., 2009). Approximately 96% of participants completed at least six interview days, resulting in 92% of retention rate (i.e., 14,912 days out of a possible 16,176 days). Further data documentation is available at http://doi.org/10.3886/ICPSR26841.v2.

Of the 2,022 participants with 14,895 interview days, we included 1,970 participants with 12,225 interview days in the analyses who completed at least two consecutive interview days. The median time gap between the main survey and daily diary assessment for the current sample was 1.6 years (19 months), where about 50% had less than a 1-year gap and 70% had less than a 2-year gap. We used demographic variables drawn from the MIDUS II main survey, and participants who did not provide information on demographic characteristics (n = 52) were excluded in the analyses. The final sample of this study were aged 57 years on average and 57% were female. In addition, 85% of the sample identified their racial group as white, 11% Black, and 4% others (i.e., Asian, native American, native Hawaiian, or others), and 31% of the sample reported their highest level of education completed as high school or less.

Measures

Data Analysis

Multilevel models were estimated to examine our hypotheses on differences in daily affective variability and affective reactivity to daily events between MDD and non-MDD groups. Affective variability was operationalized as a within-person variability of daily affect across eight study days, and affective reactivity was operationalized as the degree of change in affect in response to daily events compared to days without any events. For the analyses, level-1 continuous predictors were person-mean centered and level-2 continuous predictors were grand-mean centered to interpret estimated parameters as deviations from person-mean and grand-mean, respectively. First, to examine whether MDD has a greater daily affective variability, we tested multilevel models with heterogeneous within-person variance. This approach is powerful and parsimonious in that it allows modeling for within-person fluctuation in daily affect while controlling for all within- and between-person covariates (Liu et al., 2018). Level 1 residuals of daily affect estimated from these models represent within-person variability of daily affect. For example, equations for a model with within-person heterogeneous variance estimating daily negative affect are as follows:

• Level 1 (day-level):

  $$\begin{array}{c}{DailyNegativeAffect}_{\mathit{di}}={\beta }_{0i}+e_{\mathit{di}},e_{\mathit{di}}\sim N\left(0,,,R\right)\\ R=\left[\begin{array}{cc}{\sigma }_{eg=0}^2& 0\\ 0& {\sigma }_{eg=1}^2\end{array}\right]\end{array}$$

• Level 2 (person-level):

  $${\beta }_{0i}={\gamma }_{00}+{\gamma }_{01}{(MajorDepressionDisorder)}_i+u_{0i},u_{0i}\sim N\left(0,,,G\right)$$

At level 1, ${\beta }_{0i}$ represents the intercept which indicates the person i’s average amount of negative affect and $e_{\mathit{di}}$ is a residual error representing leftover level-1 variance in negative affect for person i. At level 2, ${\gamma }_{00}$ represents the sample mean of negative affect, ${\gamma }_{01}$ represents the effect of MDD on the person-specific intercept ${\beta }_{0i}$, and $u_{0i}$ represents a deviation of person i’s mean from the sample mean of negative affect. R matrix indicates the variance–covariance matrix of level-1 (i.e., within-person) residuals, where ${\sigma }_{\mathit{eg}}^2$ differs across two groups (i.e., MDD and non-MDD groups). The models with heterogeneous within-variance were compared with common variance model based on Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), of which lower values indicate a better-fitting-model.

To examine differences in affective reactivity to daily events among MDD and non-MDD groups, we estimated multilevel models with cross-level interaction terms between daily events and MDD. For example, equations for a model examining interaction between daily events and MDD on daily negative affect are as follows:

• Level 1 (day-level):

  $$\begin{array}{c}{DailyNegativeAffect}_{\mathit{di}}=\\ {\beta }_{0i}+{\beta }_{1i}{(DailyStressor)}_{\mathit{di}}+{\beta }_{2i}{(DailyPositiveEvent)}_{\mathit{di}}+e_{\mathit{di}}\end{array}$$

• Level 2 (person-level):

  $$\begin{array}{c}{\beta }_{0i}={\gamma }_{00}+{\gamma }_{01}{(MDD)}_i+u_{0i}\\ {\beta }_{1i}={\gamma }_{10}+{\gamma }_{11}{\left(M,D,D\right)}_i+u_{1i}\\ {\beta }_{2i}={\gamma }_{20}+{\gamma }_{21}{(MDD)}_i+u_{2i}\end{array}$$

At level 1, ${\beta }_{1i}$ and ${\beta }_{2i}$ are the slopes representing a change in person i’s daily negative affect on stressor days and positive event days, respectively. At level 2, ${\gamma }_{11}$ and ${\gamma }_{21}$ capture the interaction effects between daily events (i.e., daily stressors and daily positive events) and MDD. $u_{1i}$ and $u_{2i}$ represent leftover level-2 variance in the regression coefficients, ${\beta }_{1i}$ and ${\beta }_{2i}$, respectively, that is not explained by MDD. Random effects (i.e., random slopes) were included in the estimation for daily stressors and daily positive events, which allows changes in daily affect in response to daily events to differ across individuals. The aforementioned covariates were also included in the analyses.

Descriptive Statistics and Correlations

Participants’ sociodemographic characteristics and descriptive statistics for daily experiences are presented in Table 1. Among the sample of 1,970 participants, 202 participants (10.3%) were diagnosed with MDD in the past 12 months and 1,768 participants (89.7%) were not diagnosed with MDD. MDD participants were more likely to be younger (M = 52.2 vs. 57.2) and female (72.3% vs. 55.5%) than non-MDD participants, and 35% of MDD participants had an anxiety disorder compared to 5% among non-MDD participants. On 76.9% of interview days, both MDD and non-MDD participants reported at least one stressor or positive event. Stressors occurred on 44% of interview days among MDD participants and 35% of days among non-MDD participants. Positive events occurred on 63% of interview days among MDD participants and 69% of days among non-MDD participants.

Table 2 shows patterns of associations among the daily diary variables (i.e., within-person and between-person correlation coefficients). The intraclass coefficients indicate that between-person differences account for 77% of variance in daily positive affect, 53% in daily negative affect, 25% in daily stressors, and 41% in daily positive events. At the between-person level, the correlation coefficients between MDD status and daily events were moderate to weak, indicating that MDD participants were more likely to report higher NA, lower daily PA, higher stressor exposure, and lower positive event exposure. At both between-person and within-person levels, the correlation matrix indicates that higher daily PA was correlated with lower daily NA, having fewer stressors, and more positive events. In addition, lower daily NA and reports of positive events were correlated with having stressors, respectively, at both between-person and within-person levels.

Daily Affective Variability

Table 3 shows the results from multilevel models with heterogeneous within-person variance. The fixed-effect estimates for association between MDD and daily affect indicate that MDD group reported higher average levels of NA (est. = 0.182, SE = 0.019, p < .001) and lower average levels of PA (est. = -0.423, SE = 0.055, p < .001). For within-person variance components, MDD group reported greater variability in both NA (0.131 vs. 0.037) and daily PA (0.244 vs. 0.133; See Fig. 1).

Affective Reactivity to Daily Events

Table 4 displays the results from multilevel models for affective reactivity to daily stressors and positive events. For positive affective reactivity, the interaction between daily positive events and MDD was significant (Est. = 0.181, SE = 0.038, p < .001). Figure 2a shows that there was a greater increase in PA on days with any positive events in MDD participants (simple slope: est. = 0.253, SE = 0.037, p < .001) than non-MDD participants (simple slope: est. = 0.072, SE = 0.010, p < .001). The interaction between daily stressors and MDD on PA was marginally significant (est. = -0.068, SE = 0.035, p = .054). Although the result did not reach statistical significance level of 0.05, we conducted simple slope analyses to examine the difference between positive affective reactivity to daily stressors in MDD vs. non-MDD groups (Fig. 2b). Simple slope analyses revealed that there was a decrease in PA on stressor days in both MDD participants (Est. = -0.199, SE = 0.034, p < .001) and non-MDD participants (Est. = -0.131, SE = 0.009, p < .001).

For negative affective reactivity, the interaction between MDD and daily positive events (est. = -0.103, SE = 0.024, p < .001) and between MDD and daily stressors (est. = 0.114, SE = 0.026, p < .001) were significantly associated with NA. Figure 2c indicates that there was a greater decrease in NA on a positive-event day among MDD participants (simple slope: est. = -0.087, SE = 0.015, p < .001) than non-MDD participants (simple slope: est. = 0.002, SE = 0.006, p = .70). In addition, as depicted in Fig. 2d, there was a greater increase in NA on stressor-days among MDD participants (simple slope: est. = 0.263, SE = 0.025, p < .001) compared to non-MDD participants (simple slope: est. = 0.149, SE = 0.006, p < .001).

Affective Variability

The higher levels of NA and lower levels of PA in depressed individuals is unsurprising and expected. However, despite the belief of a stable depressed mood and affect state, depressed individuals still experience higher levels of affective variability for both negative and positive affect. We found that, overall, the MDD group experienced 3.5 times higher levels of negative affective variability and close to twice as much higher levels of positive affective variability compared to the non-MDD group. The higher NA variability is supported by the literature (Koval et al., 2013; Panaite et al., 2020), although similar studies did not find significant variability in PA (Panaite et al., 2020). Positive affective dynamic indices have increasingly gained importance as a marker for well-being and positive mental health. Given that the MDD group was experiencing such higher variability levels for both negative and positive affect, interventions targeting the decrease in affective variability may be warranted, focusing on positive affective variability.

Affective Reactivity to Daily Events

Earlier work by Almeida and colleagues has emphasized the importance of affective dynamics indices as a predictor of both physical and mental health outcomes (Almeida, 2005; Almeida et al., 2020; Charles et al., 2013; Piazza et al., 2013; Sin et al., 2020). In our study, MDD participants experienced not only higher negative affect and lower daily positive affect but also higher stressor exposure (9% more reported stressors) and lower positive event exposure (6% less reported positive events) compared to non-MDD participants. This implies that, within this sample, the MDD group experiences at least one extra stressor event per week and at least one less positive event per week. On a yearly basis, that translates to 52 extra stressors and 52 fewer positive events for the MDD group compared to the non-MDD group. This finding is in line with the stress generation hypothesis (Hammen, 1991, 2006), where depressed individuals are more likely to experience stressful life events than non-depressed ones because they behave in ways contributing to stressful environments influenced by their beliefs, expectations, and personal characteristics (Alloy et al., 2010). With the caveat that the 8-day sample might not be representative of the participants’ year, this reflects how the MDD group navigates and perceives their daily lives, highlighting the importance of context and external events that happen on a daily basis.

The investigation of how affect is impacted by the occurrence of both stressors and positive events in depressed individuals was a novel aspect of this study. Here too, we found that the MDD group’s NA increased and PA decreased significantly in response to a stressful event compared to the non-MDD group. Thus, the already elevated levels of NA are further exacerbated in the event of a daily stressor in depressed individuals. Similarly, the PA levels which were lower than those of non-depressed individuals prior to the stressor also decreased significantly when the daily stressor occurred. The impact of event-related affective changes on daily functioning is a promising research area that warrants further investigation. Research has demonstrated that reduced PA in the event of a daily stressor is likely to maintain and even predict future depression relapse (Rackoff & Newman, 2020).

In accordance with the mood brightening theory, depressed individuals experienced an increase in PA and decrease in NA significantly more than the non-depressed group in response to positive events. Thus, the MDD group appears to experience a greater advantage or gain from positive events than the non-MDD group, at least in terms of the magnitude of change in NA and PA following daily positive events (Panaite et al., 2018). This challenges the theoretical view of depression as involving intractable and high levels of NA (Koval et al., 2013). The findings also contradict most lab-based studies which generally detect an association of blunted affect irrespective of context in MDD (Rottenberg, 2005). From a therapeutic standpoint, the findings on affective reactivity to positive events were both informative and promising. The caveat, however, is that non-depressed individuals have significantly higher levels of PA and lower levels of NA to begin with and thus there is limited room for change. On the other hand, depressed individuals have room for change, and introducing coping mechanisms to not only heighten these affect changes but also to maintain them is a vital research area that requires further investigation. Recent innovative interventions are now targeting increases in positive affect, not just decreases in negative affect, to improve transdiagnostic symptoms of depression, anhedonia, and anxiety in mental health patients (Craske et al., 2019; Dunn et al., 2019).

Limitations

Several limitations of this study should be noted. First, our study is based on a community population; therefore, generalization to mental health populations needs to be cautious warranting replication in a clinical population of MDD patients. Second, other comorbid psychiatric symptoms, including bipolar disorder, were not controlled other than generalized anxiety disorder and panic disorder. Importantly, bipolar disorder shares overlapping diagnostic criteria with MDD and is well known to be sensitive to positive affect. Due to the lack of information on bipolar disorder, we could not rule out whether some individuals within the MDD group actually had bipolar disorder, which in turn could account for some of the positive affect findings (Painter et al., 2019). Future studies could address whether other mental disorders have similar impacts on daily affective dynamics and reactivity. Third, the timeline for diagnosis of MDD was not fully known due to the time gaps between the diagnostic interview and daily diary assessments. Therefore, it is unclear how many of the MDD group were symptomatic and fulfilled diagnostic criteria at the time of completing the daily diary interviews, limiting generalization of the results to clinical samples. Further, the cross-sectional observational design of the current study does not inform about causal relationships (i.e., is the higher affective variability and reactivity in MDD caused by the development of MDD symptoms or vice versa?). Future studies using longitudinal or experimental design and multiple assessments could help to examine the directionality of the associations. Additionally, our measures on daily affect and daily events are based on subjective self-reported questionnaires which might produce reporting biases. Studies that utilize objective indicators of daily events and well-being could inform how these biases created by self-reported measures may impact the results. Next, DSM III R criteria was the basis for MDD diagnosis at the time of the study. The major difference between DSM III R and DSM 5 criteria would be the absence of the following criterion for MDD diagnosis in the DSM 5: “The symptoms are not better accounted for by bereavement (i.e., after the loss of a loved one, the symptoms persist for longer than 2 months or are characterized by marked functional impairment, morbid preoccupation with worthlessness, suicidal ideation, psychotic symptoms, or psychomotor retardation).” Thus, the DSM III R excludes individuals from MDD who are experiencing symptoms of grief/sadness related to bereavement while the DSM 5 criteria do not. Finally, marginally significant interaction effects of daily stressor and MDD status on positive affect might indicate that the result did not produce statistically meaningful results. However, we found that there was a difference in simple slopes, which indicates that there might be meaningful difference in positive affective reactivity to daily stressors between two groups. Nevertheless, these results need to be interpreted with caution and replicated in different samples in future investigations. Future studies should measure cumulative stressors rather than relying on a binary measure of stressors.

Future Directions

Despite these limitations, our study has important implications for future investigation. Our study could warrant future studies in two areas: intervention-focused and symptom-trajectory-focused. Intervention-focused studies would benefit from investigating the impact of increasing affective stability with a particular focus on increasing PA. This may be achieved through interventions directed at increasing positive experiences in the lives of people with MDD. The second area of investigation focuses on the trajectory of changes in depressive symptoms based on longitudinal studies examining the link between the emergence of depressive symptoms, their maintenance, and the remission of symptoms based on changes in affective dynamics. This would provide clues on the stability of the mental health diagnosis over time and the factors associated with it (Keyes et al., 2010). The robustness, scope, and relative specificity of the brightening effect highlights the need to resolve conflicting findings across laboratory and ecological studies to advance the understanding of affective dynamic indices in MDD (Khazanov et al., 2019).

Funding

Support for this project is through The Pennsylvania State University Social Science Research Institute to DM. MIDUS has been funded by the John D. and Catherine T. Foundation Research Network and National Institute on Aging (P01-AG020166 & U19-AG051426) DA.

Conflict of Interest

The authors have no competing interests to declare that are relevant to the content of this article.

Data Availability

The datasets generated during and analyzed during the current study are publicly available in the National Archive of Computerized Data on Aging (NACDA) repository https://www.icpsr.umich.edu/web/NACDA/studies/26841/versions/V2 (Ryff & Almeida, 2017; Ryff et al., 2018, 2021).

Code Availability

Not applicable.

Author Contribution

Not applicable.

Ethical Approval

The study was approved by the IRB of the institution responsible for data collection and all respondents consented to their participation.

Consent to Participate

Not applicable.

Consent for Publication

Not applicable.