Characteristics of Study Participants
describes the baseline characteristics of the study participants. The groups did not differ in the composition of gender, ethnicity, and veteran status. However, compared to the comparison group, the group with prior history of depression had younger mean age, more years of education and a lesser proportion of married individuals. Physical health functioning (PCS), as measured by the SF-36, did not significantly differ between the two groups, but mental health functioning (MCS) was significantly worse in the prior depression group. Both groups presented lower PCS but somewhat higher MCS in comparison to the US general population, aged 18-74 years, and this appears to be a consistent finding in older adults.31,32
Chronic medical burden as measured by CDS (an objective measure of physical health) was not significantly different between the two groups, but depressive symptom severity assessed using HAM-D (a clinician rated measure of mental health) was significantly higher in the prior depression group. Although PCS and CDS, the measures of physical health, were not significantly different when compared by simple t-tests, they were significantly worse in the prior depression group when the comparison was adjusted for age and education using multivariable linear regression (see footnote of ). As to the characteristics of prior depressive episodes, the median time since last depressive episode was nearly 10 years, indicating a period of sustained full remission for most in the prior depression group. In addition, the median duration of last depressive episode was 9 months, with on average only one lifetime episode of a major depressive disorder. Nearly one-third of the prior depression group was using antidepressant medications at baseline.
Baseline Characteristics of Community-Dwelling Older Adults With Prior Depression History and Comparison Subjects With No History of Mental Illness
During the follow-up, only one comparison subject developed a depressive disorder. In contrast, among the prior depression group, the cumulative incidence of depressive disorders was 4.9% (95% confidence interval [CI] = 2.0% to 9.8%) at 6 weeks; 12.6% (95% CI=7.5% to 19.4%) at 1 year; and 16.9% (95% CI=11.0% to 24.3%) at 2 years. Notably, 54.2% of the participants who developed a depressive disorder during the follow-up had at least one episode of non-major or sub-threshold depression.
Impact of Group Status on the Change of Physical Health measured by PCS and CDS
describes the change of physical and mental health functioning over 2 years in each group. The mean PCS dropped more than 4 points in the prior depression group over the 2 years while it was essentially unchanged in those without a depression history (unadjusted B for group-by-time interaction= -0.02, SE=0.008, Z= -2.60, P=0.01). In contrast, the mean MCS remained practically unchanged in both groups throughout the study period (B= -0.004, SE=0.006, Z= -0.63, P=0.53). The individual scales followed this pattern. Scales related to the PCS declined more rapidly in the prior depression group (Physical Functioning, B= -0.02, SE=0.01, Z= -1.89, P=0.06; Role-Physical, B= -0.08, SE=0.03, Z= -2.57, P=0.01; Bodily Pain, B= -0.04, SE= 0.02, Z= -1.85, P=0.07) as compared to those without a depression history. Scales related to the MCS remained unchanged (Mental Health, B= -0.01, SE= 0.01, Z= -1.40, P=0.17; Role-Emotional, B=0.002, SE=0.03, Z=0.09, P=0.93), whereas scales related to both the PCS and MCS presented mixed results (Vitality, B= -0.02, SE=0.01, Z= -1.32, P=0.19; General Health, B= -0.0001, SE=0.01, Z= -0.01, P=0.99; Social Functioning, B= -0.05, SE=0.02, Z= -2.82, P=0.005).
Change of SF-36 summary measures (Physical Component Summary and Mental Component Summary) over time in community-dwelling older adults with prior depression history and comparison subjects with no history of mental illness
For PCS, unadjusted mixed model regression analysis revealed a group effect (B= -1.88, SE=0.92, Z= -2.05, P=0.05), time effect (B= -0.01, SE=0.005, Z= -2.60, P=0.003), and group-by-time interaction (B= -0.02, SE=0.008, Z= -2.60, P=0.01). After adjustment for sociodemographic characteristics and MCS, HAM-D and PSQI, the results remained significant (). The significant group-by-time interaction indicated a longitudinal decline of physical health functioning, which was more rapid over time in the prior depression group compared to the comparison group. When the multivariable analysis was repeated excluding those who developed depressive episodes during follow-up, significant group-by-time interaction was again found (B= -0.02, SE=0.007, Z= -2.61, P=0.01). As nearly one-third of the participants with prior depression were using antidepressants at baseline, we repeated the multivariable analysis excluding these participants. There was a significant group-by-time interaction in this case as well (B= -0.03, SE=0.008, Z= -3.36, P=0.001), indicating antidepressant use cannot explain the observed effect.
Table 2 Effect of group (control = 0 vs. prior depression = 1), time (weeks) and group-by-time interaction on physical health functioning and chronic medical burden of community-dwelling older adults with prior depression history and comparison subjects with (more ...)
describes the change in chronic medical burden (CDS) in the two groups; CDS worsened at a greater rate in those with prior depression as compared to comparison subjects. Unadjusted mixed model regression analysis for CDS showed a time effect (B= 0.004, SE=0.001, Z=4.96, P<0.001) and a group-by-time interaction (B= 0.003, SE=0.001, Z= 2.67, P=0.008), but no group effect (B= 0.38, SE= 0.25, Z=1.57, P=0.12). After adjustment for sociodemographic characteristics and MCS, HAM-D and PSQI, the results remained significant (). The significant group-by-time interaction indicated a longitudinal increase of chronic medical burden, which was more rapid in the prior depression group compared to the comparison group. As above, when the multivariable analysis was repeated excluding those who developed depressive episodes during follow-up, a significant group-by-time interaction was still found (B= 0.004, SE=0.001, Z=2.75, P=0.006). Moreover, when the multivariable analysis was repeated without those using antidepressants at baseline, a significant group-by-time interaction was also observed (B= 0.005, SE=0.002, Z=3.03, P=0.003). Finally, there was a significant correlation between PCS and CDS, as assessed in their longitudinal format, which reflects all the measurements performed during the study (r= -0.27, df=349, P<0.0001).
Change of chronic medical burden (Chronic Disease Score) over time in community-dwelling older adults with prior depression history and comparison subjects with no history of mental illness
Reverse Causality: Impact of Poor Physical Health on Depression
First, according to mixed effect linear regression analyses, there was no significant impact of baseline physical health status (poor vs. good, i.e. PCS<50 vs. PCS≥50) on the change of depressive symptoms measured by HAM-D during follow-up (unadjusted B for group-by-time interaction= 0.0005, SE=0.003, Z=0.20, P=0.85). Multivariable adjustment for sociodemographic characteristics and sleep quality further attenuated this association (adjusted B for group-by-time interaction= 0.0002, SE=0.003, Z=0.09, P=0.93). Second, although unadjusted mixed effect logistic regression analysis revealed poor physical health to be a significant risk factor for subsequent depressive episodes (unadjusted odds ratio 5.18, 95% CI 1.17 to 22.95, Z=2.16, P=0.04), this association was no longer significant after multivariable adjustment for sociodemographic characteristics and sleep quality (adjusted odds ratio 2.78, 95% CI 0.73 to 10.63, Z=1.49, P=0.14).