Synchrony of change in depressive symptoms, health status, and quality of life in persons with clinical depression

The constructs of depression, mental and physical health, and quality of life (QOL) are believed to be distinct but closely related. Depressed persons had worse health status and QOL than others in several cross-sectional comparisons [1‐6]. Cross-sectional data cannot address whether changes over time in one construct are reflected in the other areas as well, which requires experimental or longitudinal data. Only a few studies in primary care and specialty samples have examined longitudinal associations between depression and measures of health status or disability. In a sample of high utilizers of general medical care, improvement in depression over 12 months was associated with significant reductions in both days of disability due to illness and in self-rated disability [7]. In a sample of primary care patients, long-term (1 to 3 year) improvement in depressive and anxiety symptoms was associated with significant improvement in interviewer-rated social disability [8]. In a sample of specialty care patients, month-to-month changes in severity of depression were associated with similar changes in interviewer-rated psychosocial health [9]. Persons with clinical depression experienced changes in both depression and quality of life over a 6-week period [10].

Here, we expand on these earlier findings by examining a wider range of health status and quality of life measures as well as examining synchrony of change over both short (6 weeks) and intermediate (9 month) time frames. We examine synchrony between the three constructs (depression, QOL, and health status) and agreement among the instruments used to operationalize the constructs. The primary goal of this paper is to determine whether mental health, physical health, and quality of life changed over time in persons who were clinically depressed at baseline, and to determine whether the change in depressive symptoms was synchronous with change in the other measures. The second goal is to compare the performance of the various instruments in a cohort of persons who were initially clinically depressed, to determine which instruments showed the most change, and to provide some information about the psychometric properties of each instrument.

The overall aim of the Longitudinal Investigation of Depression Outcomes (LIDO) study was to characterize associations among quality of life (QOL), economic and depression outcomes, based on a multi-national observational study involving a prospective cohort of primary care patients with major depressive disorder [6]. LIDO investigators evaluated depressive symptoms, mental and physical health, and quality of life for 982 clinically depressed persons from 6 international research sites: Be'er Sheva, Israel;Barcelona, Spain; Porto Alegre, Brazil; Melbourne, Australia; St. Petersburg, Russia; and Seattle, WA, USA.

Patients making primary care visits were screened for initial eligibility, defined as a score of 16 or greater on the Center for Epidemiologic Studies- Depression scale (CESD) [11]. For patients meeting initial eligibility criteria, a baseline assessment was conducted, which included administration of a depression diagnostic instrument, the Composite International Diagnostic Interview (CIDI) [12, 13]. All subjects who were diagnosed as clinically depressed by the CIDI (i.e., satisfied DSM-IV criterion A for major depressive episode [14]) were considered eligible, but those with a known organic or major psychiatric disorder (dementia, psychosis, bipolar disorder) were excluded. Patients who currently or in the previous three months were under treatment for depression were also excluded. The reference population is thus persons with untreated clinical depression being seen in primary care settings. Those who met the inclusion criteria and were willing to enroll in the study were followed for depression, health, and QOL at baseline, 6 weeks, 3 months, and nine months.

Of the 1180 clinically depressed persons enrolled, 982 had a valid CIDI value at 9 months. Persons without a valid CIDI were similar in age and sex to those with a CIDI, but had significantly better baseline health. For the 982 persons studied here, mean age was 41 (range 17–76) and 71% were women. Fewer than one percent had missing data. Table 1 shows the grand mean and standard deviation for each variable, for all times combined. Recall that higher scores on the CESD and the QLDS domains indicate worse depression or health, respectively (denoted by an asterisk in the table), while higher scores on the other measures indicate better status. Table 1 also shows the mean value of each study variable at the four time periods. For example, the grand mean for CESD was 24.31, and the mean score decreased from 29.03 at baseline to 20.55 at 9 months, or a decrease of 8.48 CESD points. Thus, although there was improvement, the average level of depressive symptoms was still high 9 months after baseline. All of the measures showed improvement over time. The standard deviations of change scores from screener to baseline and from baseline to 6 weeks, 3 months, and 9 months are in Appendix Table 1. These may be useful for estimating the necessary sample size for future studies of clinically depressed patients.

Table 2 shows the average z-score for each variable, at each time point. The z scores represent differences from the mean in standard deviation units,(i.e., the observed value minus the grand mean divided by the overall standard deviation) using the values reported in Table 1. The signs were reversed for the CESD and QLDS, so that a higher value indicates better status for all variables. On average, all variables were below the grand mean at baseline (the z-score has a negative sign), and all but CESD and QOL:Rate were still negative at 6 weeks. All the variables were above the mean at 3 months and 9 months. Thus, all variables showed monotone improvement over time (baseline < 6 weeks < 3 months < 9 months). The largest change over time was for the CESD, which moved from .36 standard deviations below the mean at baseline to .33 standard deviations above at 9 months, or an improvement of .69 standard deviations. The mental health variables all showed large changes, the physical health variables a smaller amount of change, and the QOL variables were intermediate.

We next examined whether changes in the CESD were mirrored by changes in the other variables. The subjects were divided into four (approximately) equal groups on the basis of their 6-week change in the CESD. Table 3 shows the mean change (in the z-scores) between 0 and 6 weeks. A positive change represents improvement. The first column in Table 3 has information for those whose CESD improved by 10 or more points in 6 weeks (chosen because about 25% of the subjects improved 10 or more points). The mean improvement for the CESD for this group was of course high, because the groups were defined by CESD change; persons averaged an improvement of 1.50 standard deviations on this measure. All of the other variables in column 1 also had positive 6-week change, but the amount of change was smaller. This positive change was statistically significant (denoted by the "+" sign following the values in this column) for all but the PCS. (Change on the original scale can be calculated by multiplying the change in z-score by the appropriate standard deviation in Table 1). The fourth column of Table 3 shows the means for persons whose CESD worsened by 2 or more points in the first 6 weeks. Their CESD dropped an average of .62 standard deviations. Most of the other variables also had negative change. The only exception was PCS, which did not change. The "-" signs following numbers in this column denote variables with significant negative change. Most of the mental health and QOL measures showed a statistically significant decrease in column 4, but none of the physical health measures changed significantly.

At the end of the study all persons were assessed for clinical depression using the CIDI interview. Table 4 shows the mean 9-month change in each variable as a function of clinical depression status at 9 months. All of the measures improved significantly over time for those who were not clinically depressed at 9 months. The group with persistent depression did not change significantly on any of the measures.

The data were collected in six international sites, and there were significant differences among sites in the baseline levels of the measures [6]. In spite of this, the age-sex-education-site-adjusted changes were nearly identical to the unadjusted results. Further, graphs of changes over time were very similar across sites. For these reasons, as well as for space constraints, results are not presented separately by site or by other covariates.

We investigated whether there were changes in each measure, independent of changes in the CESD, by regressing 6-week change for each instrument on 6-week change in the CESD; for example, we regressed change in the QLDS on change in the CESD. We then tested whether the constant term in the regression was significantly different from zero. That is, did the average person with no change on the CESD have significant change in the variable of interest (QLDS in the example)? This is conceptually similar to what is shown in column 3 of Table 3. There was significant improvement in the first 6 weeks, even after controlling for change in the CESD, for MCS, MHI5, Do less:Emotional, Do less: Physical, and for WQ:Physical. There was not significant change in PCS, EVGGFP or in any QOL measure but WQ:Physical, which became non-significant if CESD squared was added to the regression. In a final analysis of the baseline data only, we regressed each QOL measure on CESD, and then used stepwise regression to select additional significant predictors from among age, sex, education, and the three physical health measures. Two or more of these candidate variables entered in every regression, and will be addressed below.

The primary goal of this paper was to determine whether mental health, physical health, and quality of life changed over time in persons who were clinically depressed at baseline, and to determine whether the change was synchronous with change in depressive symptoms. We also wanted to know whether change in depression could explain the changes in the other dimensions. The final goal was to compare the performance of the various measures of each construct.

All measures improved significantly, in synchrony with changes in clinical depression. Measures of depressive symptoms and mental health showed the greatest change, followed by quality of life, and then by physical health. The single items performed fairly well, and may be a useful alternative in some large studies. These findings may be useful for future researchers who need to choose among instruments. Additional research in these areas should further examine causality by following clinical changes in the patients. Further studies of the importance of changes in QOL in subjects with clinical depression are also needed.