Evaluating the timing of differences in activity related to depression symptoms across adulthood in the United States

Abstract

Background:

Relative activity deficits found in people with (verses without) depression symptoms/disorders may accumulate uniformly throughout the day, or they may tend to be expressed at specific times. Evidence for the latter would suggest times when behavioral approaches are most needed to reduce depression and its health consequences.

Methods:

We performed a secondary-data analysis of participants who contributed valid accelerometer data at the 2005–2006 National Health and Nutrition Examination Survey (n=4390). Participants were categorized according to the Patient Health Questionnaire-9 standard cut-point of ≥10 (i.e., people with versus without clinically significant depression symptoms). Average levels of accelerometer-measured activity in two-hour bins were the dependent variable in mixed models testing if the relationship between depression status and activity level differed by time of day; and if any such relations varied by age group (18–29 years, 30–44 years, 45–59 years, and 60+ years).

Results:

In adults over the age of 30, people with depression symptoms had generally lower levels of activity across the day, but these effects were most markedly pronounced in the morning hours. We found no differences in activity levels associated with prevalent depression symptoms among people 18–30 years of age.

Limitations:

Core aspects of depression pathophysiology that produce these different activity patterns and confer their effects on mood were not measured.

Conclusions:

In adults 30 years and older, efforts to ameliorate relative activity deficits associated with depression may benefit from considering the apparently outsized role of inactivity that occurs in the morning.

Introduction

Major depressive disorder is a world leading cause of disability¹ estimated to cost the United States economy $210 billion per year². Depression can have major consequences including increased rates of suicide^3–6, cardiovascular disease⁷, and dementia later in life^8,9. One prominent feature of depression that plausibly drives its incidence, and consequences, is a reduction in the level of physical activity. Such activity deficits predict depression incidence¹⁰ and major health consequences including suicide¹¹, cardiovascular disease¹², and dementia¹³. These examples of studies finding health effects of activity used measures that summarize activity levels over days or weeks (consistent with most extant studies). This consistently demonstrated relevance of overall activity, to depression and plausibly its health consequences, supports the need for additional research further characterizing activity disruption in people with depression.

It is plausible that the nature of relative activity deficits in people with depression varies by time of day. Diurnal variation in mood¹⁴ or the availability of social zeitgebers¹⁵ could make physical activity more or less challenging at certain times of the day. A majority of patients with depressive disorders endorse difficulty “getting going” in the morning¹⁶; this could reduce time overall spent being active. In agreement, a small study of older dementia caregivers used objective accelerometer measures and found that caregivers with depression symptoms were less active (than non-depressed caregivers), specifically, in the hours from 8–10 AM¹⁷. This evidence of a time period when activity levels correlate with depression symptoms suggests mornings may be a particularly high-value target for physical activity and other behavioral interventions in people with depression. However, this evidence of an objective activity deficit specific to the morning time was based on a small selected sample; and we are unaware of large-scale studies that have evaluated whether differences in activity levels associated with depression occur across the entire day or are particularly pronounced in specific time periods.

Furthermore, it is plausible that depression-related differences in the daily distribution of activity vary of life-stage. Social roles and physical health change with aging. Age itself correlates with differences in the daily distribution of activity: young adulthood appears to be associated with a delay in morning activity onset, which reverts towards activity beginning earlier in the morning with development and aging^18,19. Thus, it is plausible that depression related activity deficits are specific to particular times of day, and these differences could vary across age groups. Identifying times of day when activity deficits are particularly pronounced in people affected by depression symptoms could help guide research/intervention efforts to investigate temporally specific behavioral risks/intervention targets.

We therefore aimed to determine if the relationship between prevalent depression symptoms with activity levels varied by time of day. Given the potential of age differences in these relationships, and to potentially identify age-specific targets, we evaluated if potential time-of-day effects differed by age group.

Methods

Participants:

We conducted a secondary data analysis of a large-scale study that was designed to be representative of adults in the United States. Data were drawn from the publicly available 2005–2006 National Health and Nutrition Examination Survey. This survey included 5393 adults who were at least 18 years old, of whom 4731 completed the outcome depression measure. The analytic sample was further restricted to 4390 individuals who provided adequate activity data as defined in previously published work²⁰. Participants provided informed consent and the study was approved by the National Center for Health Statistics Ethics Review Board.

Activity patterns:

Activity data were collected using the Actigraph AM-7164 (Actigraph, Ft. Walton Beach, FL) device worn for 7 consecutive days on the right hip. Participants were asked to wear the device when awake except when swimming or bathing²¹, therefore data here reflect only waking and not resting period activity. Data, recorded in 1-minute epochs, were flagged as unreliable and excluded from the analytic sample following published processing standards for this dataset²⁰. In this analysis, we analyzed activity patterns throughout the day by summing activity counts in twelve two-hour bins similar to previous approaches^17,22.

Depression symptoms:

The nine-item Patient Health Questionnaire was used to assess the presence or absence of clinically significant depression symptoms. We dichotomized this severity measure at the standard cut-point of ≥10 that has been shown to have good predictive utility compared with a diagnosis of major depressive disorder (88% sensitivity and 88% specificity)²³. We refer to cases above and below this cut-point as cases with and without depression symptoms for simplicity.

Covariates:

We considered three covariates that might affect the relationship between PA and depression: age, gender, and ethnicity. For age, we used a 4-category polychotomous variable with the categories 18–29 years, 30–44 years, 45–59 years, and 60+ years. We defined self-reported gender by the categories ‘Male’ and ‘Female’ and self-reported ethnicity by the categories ‘Hispanic’, ‘White’, ‘Black’ and ‘Other’.

Statistical Analysis:

All analyses were conducted in R v3.6 and accounted for sampling weights. Contingency table analyses and associated Pearson chi-square tests were conducted as descriptive analyses that quantify marginal associations of variables with depression symptoms. For the primary analyses, we fit a multilevel mixed effects model for activity counts with: fixed categorical effects for time of day, depression status, age group, ethnicity and gender, all two-way interactions between time of day, depression status, and age group, the three-way interaction between time of day, depression status and age group; and a random subject effect to account for repeated measures within subject. We evaluated the need for the interactions through likelihood ratio tests and Bayesian information criterion (BIC). To illustrate our findings, we present data stratified by variables where statistically significant interactions were detected. In addition, we used post-hoc analyses via linear contrasts of estimated fixed effects parameters from the fitted model. Alll p-values from the post-hoc analyses were adjusted for multiple comparisons by controlling the false discovery rate (FDR) using the method of Benjamini and Hochberg²⁴.

Results

Descriptive Analyses:

The age of our sample (n=4390) was 44 years on average with a standard deviation of 19 years. Table 1 displays sample distributions of age group, gender, and ethnicity for each depression group. The prevalence of depression symptoms overall was 6.1% and, by age, rates were highest in the 45–59 year old age category (9.4%). Prevalence was higher in: women (7%) than men (5.1%); and in black participants (8.3%) versus the other self-reported ethnicity categories (which ranged from 4.9–6.3%).

Table 1:

Sample characteristics by depressive symptoms (n=4390).

		No Depression Symptoms, n=4122	Depressive Symptoms, n=268	p-value
Age	18 – 29	1279 (31.0)	76 (28.4)	0.001
	30 – 44	969 (23.5)	60 (22.4)
	45 – 59	796 (19.3)	83 (31.0)
	60+	1078 (26.2)	49 (18.2)
Gender	Male	2000 (48.5)	108 (40.3)	0.011
	Female	2122 (51.5)	160 (59.7)
Ethnicity	White	2027 (49.2)	105 (39.2)	0.001
	Black	961 (23.3)	87 (32.5)
	Hispanic	982 (23.8)	66 (24.6)
	Other	152 (3.7)	10 (3.7)

Number of participants (percentage) shown

Associations of depression with activity:

Table 2 displays the type III ANOVA table for the fixed effects of this fitted model. All subsequent analyses were based on this model. Both the likelihood ratio test and BIC indicated the necessity of the three-way time of day, age group, and depression symptom interaction. We therefore tested the interaction between depression status and time of day in each age group separately. There were statistically significant interactions between time of day and depression status in the three older groups (p’s all <0.00001) but not the youngest group (p=0.71).

Table 2:

Type III ANOVA table of fixed effects from the fitted mixed effects model using Satterthwaite’s method.

Factor	Sum of Squares	Numerator DF	Denominator DF	F-Value	P-value
Depression	88.4	1	4681	38.4	< 0.001
Time	30774.7	11	48556	1215.15	< 0.001
Age	82.4	3	4631	11.93	< 0.001
Ethnicity	54.9	3	4669	7.96	< 0.001
Gender	5.4	1	4400	2.37	0.124
Depression X Time	371.5	11	48556	14.67	< 0.001
Depression X Age	18.9	3	4631	2.73	0.042
Time of Day X Age	529.1	33	48556	6.96	< 0.001
Depression X Time X Age	235.9	33	48556	3.10	< 0.001

Figure 1a displays estimated differences in mean activity levels for those with depression symptoms compared to those without at each point in time separately for each age group. There were no statistically significant differences in activity between those with depressive symptoms and those without at any time of day for those age 18–29 (FDR-adjusted p-values > 0.05). For the other three age groups, we detected statistically significant differences in activity levels where people with depression symptoms (vs. not) had lower activity levels from 6am-8pm (FDR-adjusted p-values < 0.05). The estimated mean difference in daytime activity between groups, among adults aged 30+ years peaked in the morning hours in all age groups above the age of 30 (Figure 1b).

Figure 1:

(A) Estimated mean activity, and 95% point-wise confidence intervals, for participants with and those without depressive symptoms stratified by age group. (B) Estimated difference in mean activity in those with depression symptoms compared to those without at each point in time and age group. Both panels show 95% point-wise confidence intervals, and FDR-adjusted p-values for tests of differences in mean activity among depression group.

To further illustrate these findings, we plotted the cumulative percentage of the total difference in average activity levels between the groups with and without depression symptoms as a function of time bin (Figure 2). This illustrates: (1) no/slow accumulation of group differences in activity during night-time hours when people tend to sleep and were not wearing the accelerometer; and (2) that it took less time in the morning, relative to afternoon/evening hours, for similar portions of the activity deficit (e.g., 25%) to accumulate.

Figure 2:

Estimated percent of cumulative mean activity deficit in those with depressive symptoms compared to those without. Dashed lines drawn to indicate the times when each quartile of the relative activity deficits had accumulated. Note the time from reaching the first 25% of the activity deficit to 50% is considerably shorter than the time for the subsequent quartile.

Discussion

Among adults 30 years and older, we found that people with clinically significant depression symptoms (PHQ-9 scores ≥10) were less active throughout the day when compared with people who were below this threshold (PHQ-9 scores <10). Notably, effect size differences were largest during the morning hours. The relative activity deficit that was associated with prevalent depression symptoms accumulated most rapidly over a few morning hours, and continued until late in the night. This pattern, where differences in activity associated with depression start and disproportionately amass in the morning, suggests that: (1) mornings may represent a critical frame-time in which to further examine the potential causes of inactivity in depression; and (2) mornings could also provide a particularly critical, high-yield, period to target with behavioral approaches (e.g., scheduled activities).

In contrast, among adults 18–30, we failed to detect any associations between depression and activity levels. This is perplexing and we can only speculate as to why this was observed. Note that a prior analysis of NHANES data also surprisingly found that emerging adults were less active than middle-aged people²⁵. Since a high proportion of people in this younger age group likely attend school, which serves as a rigid social zeitgeber, activity patterns may have been forced to match more closely in people with/without depression symptoms. That said, analyses from other samples are required to confirm whether relative activity deficits associated with depression are indeed less prevalent in emerging adults or certain contexts.

Our main findings in people age 30+ extend existing knowledge by specifying when relative activity deficits found in depression are most pronounced, and therefore, when interventions may be needed most. An analysis of NHANES data over thirty-years ago was the first large-scale study to show that physical activity (self-reported) predicts - not only prevalent depression symptoms - but also future depression incidence²⁶. A more recent paper using accelerometer data from the same 2005–2006 NHANES sample used here estimated time spent in sedentary, light, moderate, vigorous activity states; this past study found that adults with depression spent less time in light and moderate activity states only. In light of this prior study using summary physical activity measures in the same sample used here, the relative activity deficits observed here in the morning are likely due to lower levels of light and moderate intensity activity in the morning (and not likely related to a lack of vigorous exercise).

In addition to our findings, other sources of evidence and reasoning support the notion that participation in morning activity may protect mood. Inadequate engagement with morning activity could provide immediate time and opportunity to ruminate on emotional distress. In contrast, morning activity engagement could provide distraction or positive engagement with beneficial physiological (e.g., increasing cerebral blood oxygenation) and psychological (e.g., increasing social interaction) properties. Being active in the morning (rather than inactive or asleep) could also increase light exposure, and light exposure has acute effects on mood^27,28. The composition of morning sunlight differs when compared with afternoon and evening light, especially in the winter months of northern hemispheres²⁹ (where the present study was conducted); and in people with seasonal affective disorder, morning light has greater antidepressant effects than evening light³⁰. Regularly engaging in morning activity can also stabilize the circadian clock and prolong the period when homeostatic drive is accumulated; since circadian rhythms and homeostatic drive control sleep³¹, being regularly active in the morning could improve sleep quality (thereby reducing a depression risk factor³²). It is also possible that, instead of being active, people with depression symptoms in this sample were sleeping. However, sleep in early morning hours contains the greatest proportion of Rapid Eye Movement (REM) sleep³³; therefore, “sleeping in” may increase the probability of REM fragmentation and related nocturnal mentation, which degrade the benefits of REM sleep in terms of dissolution of amygdala-mediated distress^34,35. In addition, low mood in the morning related to depression pathophysiology could also make it more difficult to behaviorally active. Thus, there are several plausible reasons why inactivity occurring in the morning is relevant to depression etiology, pathophysiology, and persistence.

Limitations:

It is important to note that, while plausible, the current report was not designed to place the observed association within the larger mechanistic context of depression or depression subtypes. A variety of potential confounders and mediators, which could explain why and when people with depression symptoms are less active in the morning, were not considered. It is plausible that people with depression symptoms were either sleeping or in bed longer (but see comments on potentially depressing effects of REM fragmentation above). Future studies are required to determine why morning activity appears to be particularly challenging for people with depression. Since the outcome was defined using a simple PHQ-9 threshold, it remains to be seen whether the associations detected were driven by a subgroup sharing a particular depression symptom profile. Potential acute and bi-directional relations over time-scales of days and weeks, linking morning inactivity with specific mood symptoms, remain to be established. Delineating the temporal relations and mechanisms linking this behavioral sign – low morning activity engagement – with specific mood symptoms/profiles could help guide interventions. In summary, to improve our ability to detect and interrupt these pathways, future studies are needed to determine why morning inactivity occurs; which mood symptoms relate to acute/chronic morning inactivity exposure; and how morning inactivity and mood are linked mechanistically.

Conclusions:

Our findings confirm prior observations that highlight the relevance of morning activity to mood. Specifically, past research indicates that being a “morning type” is protective against depression³⁶, and that people with depression often report difficulty “getting going” in the morning¹⁶. Strengths of our study include the use of objective accelerometer measures, which are not subject to affective or recall biases, and that our study sample was large and designed to be representative of people in the United States. As such, these data support the likely generalizable implications that, in adults aged 30+ in the United States, morning inactivity may represent a high-yield target for behavioral interventions to improve mood and prevent depression’s health consequences. Future work is needed to clarify the mechanisms linking morning inactivity and mood, and to evaluate the effects of increasing morning activity engagement.

Highlights.

• Depression was associated with lower activity levels in adults aged 30+.

• The largest effect size differences occurred in the morning hours.

• We found no differences in activity related to depression in adults age 18–30.