Early Depressed mood after stroke predicts long-term disability: the Northern Manhattan Stroke Study (NOMASS)

Abstract

Introduction

Depression is highly prevalent after stroke, and may influence recovery. We aimed to determine whether depressed mood acutely after stroke predicts subsequent disability and mortality.

Methods

As part of the Northern Manhattan Stroke Study, a population-based incident stroke case follow-up study performed in a multiethnic urban population, participants were asked about depressed mood within 7–10 days after stroke. Participants were followed every 6 months the first 2 years, and yearly thereafter for 5 years, for death and disability measured by the Barthel Index (BI). We fitted polytomous logistic regression models using canonical link to examine the association between depressed mood after stroke and disability, comparing moderate (BI 60–95) and severe (BI < 60) disability to no disability (BI ≥ 95). Cox-proportional hazards models were created to examine the association between depressed mood and mortality.

Results

A question about depressed mood within 7–10 days after stroke was asked in 340 of 655 ischemic stroke patients enrolled, and 139 reported that they felt depressed. In multivariate analyses controlling for socio-demographic factors, stroke severity, and medical conditions, depressed mood was associated with a greater odds of severe disability compared to no disability at one (OR 2.91, 95% CI 1.07–7.91) and two years (OR 3.72, 95% CI 1.29–10.71) after stroke. Depressed mood was not associated with all cause mortality or vascular death.

Conclusion

Depressed mood after stroke is associated with disability but not mortality after stroke. Early screening and intervention for mood disorders after stroke may improve outcomes and requires further research.

Introduction

Stroke is the leading cause of disability and the third leading cause of death in the United States¹. Recovery after stroke is affected by multiple factors including initial disability, the volume of the infarct or hemorrhage, the anatomic location, pre-stroke functional status, marital status and a social support network, and access to rehabilitation services^{2, 3}. Depression may be a risk factor for incident myocardial infarction (MI)⁴ and stroke⁵. It is highly prevalent after stroke, with a range from 16% to 72%⁶. Major depression has also been found to have a strong effect on recovery from stroke, as well as the likelihood of death on long-term follow up⁶. However, stroke patients are not always fully evaluated for depression. The purpose of this study was to investigate the association of depressed mood with stroke outcome in a multiethnic urban population-based cohort, and mortality after first stroke. Our hypotheses were that depressed mood after stroke would independently predict functional outcome at both 1 and 2 years after stroke, and mortality within 5 years.

Methods

Recruitment of the sample and baseline evaluation

The current sample was derived from Northern Manhattan Stroke Study (NOMASS) cases. Methods of recruitment, evaluation, and follow-up have been described previously^{7, 8}. In brief, patients were eligible if they were diagnosed with their first ever ischemic stroke between July 1^st 1993 and July 1^st 1997, were over the age of 39, had been a resident of Northern Manhattan for more than three months at the time of stroke onset, and had access to a telephone. Cases of hemorrhage or transient ischemic attack were not eligible. Stroke patients were examined by a study neurologist and a trained research physician. Participants who spoke Spanish at home were interviewed by native speakers. The interview and examination included a complete neurological examination and National Institutes of Health Stroke Scale (NIHSS), demographic information, and medical history. Participants were asked the first question on the Hamilton Depression Rating Scale (Ham-D) regarding their mood in the week after the onset of the stroke⁹. We only included participants who had their initial assessment of depressed mood within 30 days of their stroke and who could answer the question themselves to minimize the possibility that recall bias depended on long term stroke outcome.

Social support was assessed by addressing marital status and social isolation. We defined social isolation as knowing fewer than three people well enough to visit with in their homes (less than three friends). Standardized questions about vascular risk factors were adapted from the Centers for Disease Control and Prevention and the Behavioral Risk Factor Surveillance System¹⁰. Hypertension, diabetes mellitus, cardiac disease, and hypercholesterolemia were defined as described previously⁸. We did not collect data on the presence of a diagnosis, or treatment of, depression before or after stroke. Race-ethnicity was based on self-identification.

Follow-up

Participants were examined at 6 months for the first 2 years of follow up and then annually for five years to detect changes in health status, including death. Disability was measured using the Barthel Index (BI) of activities of daily living¹¹.

Statistical Analysis

Baseline characteristics were assessed in relation to depressed mood using a t test or χ² test as appropriate. We fitted polytomous logistic regression models using canonical link to calculate odds ratios (OR) and 95% confidence intervals (CI) for the association between an affirmative answer to the depressed mood question early after stroke and functional status. We defined functional status as severe disability (BI<60), moderate disability (BI 60–95), and no disability (BI≥ 95)^{12, 13}, and hypothesized that there would be a difference between the three levels. BI ≥ 95 was chosen as the reference. In the Ham-D, depressed mood is rated in four degrees of severity, but for the purposes of this analysis we categorized depressed mood as absent or present. Stroke severity was categorized based on the NIHSS as mild (NIHSS < 6), moderate (NIHSS 6–13), severe ≥ 14^{12, 13}. Measurements of disability post-stroke were censored at the time of recurrent stroke, myocardial infarction, or death. Given the declining numbers of survivors, main analyses on functional status were limited up to the first 2 years after stroke. Mortality was ascertained up to 5 years from initial stroke. Vascular death was categorized when participants died of stroke, myocardial infarction, heart failure, pulmonary embolus, or sudden cardiac death. Univariate analyses were first carried out, followed by a final model including the following sociodemographic factors which could influence stroke recovery: age, race-ethnicity, completing a high school education, having less than 3 friends, being unmarried, having Medicaid or no insurance. The final model also included possible medical contributors to stroke recovery including stroke severity (as measured by the NIHSS), diabetes¹⁴, physical inactivity¹⁵, and coronary artery disease. In secondary analyses, we further explored the association of depressed mood and functional disability at 6 months, 1.5 years and 3 years.

In order to assess whether having depressed mood ascertained or not was related to functional status, which could be a source of bias, logistic regression models were constructed for each time point using an indicator of whether each subject had ascertainment of depressed mood as the outcome and functional status as the predictors (adjusted for the same covariates in our final model).

Cox-proportional hazards models were developed to examine the association between depressed mood and all mortality, vascular death, and non-vascular death, adjusting for age, sex, race-ethnicity, education, insurance, stroke severity, diabetes, hypertension, and coronary artery disease. The study was approved by the CUMC Institutional Review Board and all participants signed written informed consent.

Results

A question on depressed mood within 7–10 days after stroke was asked in 340 of 655 (51.9%) ischemic stroke patients enrolled, and 139 of the 340 (40.9%) reported that they felt depressed. The participants who had depressed mood ascertained were more likely to be Hispanic and to have fewer than 3 friends, and less likely to have completed high school and to have a severe stroke compared to those who did not. Reasons for not including depressed mood after stroke included aphasia or death within 30 days of stroke (n = 113), impaired consciousness (n = 12), interview > 30 days after stroke (n=77), loss to follow-up before ascertainment of depressed mood (n= 37) and incomplete data (n =76). Baseline demographics of the sample used in the analysis are outlined in table 1. Participants who reported depressed mood were more likely to be physically inactive and have severe stroke than those who did not report depressed mood; non-Hispanic blacks were less likely to report depressed mood. Over 5 years of follow up there were 105 deaths, of which 45 were vascular, 49 were non-vascular, and 11 with unconfirmed cause of death.

Table 1.

Baseline characteristics of participants with stroke and depressed mood assessed in the Northern Manhattan Stroke Study (n = 340)

		N or mean	Early depressed mood present , N (proportion) or mean (standard deviation) n = 139	p-value for difference
Age		68.8	68.0 (13.3)	0.3
Race Ethnicity	Non-Hispanic White	70	34 (48.6)	reference
	Non-Hispanic Black	81	22 (27.2)	0.007
	Hispanic	182	83 (45.6)	0.7
Women		192	84 (43.8)	0.2
Men		148	55 (37.2)
Less than a high school education		239	97 (40.6)	0.8
Greater than a high school education		98	41 (41.8)
Not married/widowed		222	98 (44.1)	0.09
Married		116	40 (34.5)
Diabetes Mellitus		104	47 (45.2)	0.3
No diabetes mellitus		236	92 (39.0)
Physically inactive		191	90 (47.1)	0.006
Physically active		148	48 (32.4)
Less than 3 friends		77	37 (48.1)	0.1
Three or more friends		263	102 (38.8)
Medicaid or no insurance		176	78 (44.3)	0.2
Private insurance or medicare		164	61 (37.2)
Cardiac disease*		113	41 (36.3)	0.2
No cardiac disease		227	98 (43.2)
Stroke severity**	Mild	180	77 (42.8)	reference
	Moderate	116	36 (31.0)	0.04
	Severe	41	25 (61.0)	0.04

^*History of coronary artery disease or prior myocardial infarction.

^**Mild: NIHSS < 6; Moderate: NIHSS 6–13; Severe: NIHSS ≥ 14

The distribution of the participants’ disability by the presence of depressed mood is outlined in table 2. At one year there were 246 participants available for analysis, with 207 available at year 2 after excluding those with recurrent stroke, MI, or death. Among those with depressed mood present at baseline, there was a greater proportion of participants in the severe disability category at one and two years. In univariate analyses, depressed mood was not associated with functional disability (Model 1 in Table 3). When adjusted for socio-demographic factors (age, race-ethnicity, marital status, having fewer than 3 friends, educational achievement, and insurance status), stroke severity, and medical and behavioral comorbidities (diabetes, physical inactivity, and coronary artery disease), depressed mood was associated with a greater odds of severe disability compared to no disability at one (OR 2.91, 95% CI 1.07–7.91) and two years (OR 3.72, 95% CI 1.29–10.71) after stroke. In secondary analyses, depressed mood was associated with severe disability in fully adjusted models starting at the first point of follow up 6 months after stroke (adjusted OR 2.81, 95% CI 1.13–6.99), at 1.5 years after stroke (adjusted OR 5.64, 95% CI 1.93–16.52), and persisting until 3 years after stroke when it was no longer significantly associated (adjusted OR 3.57, 95% CI 0.89–14.23). Depressed mood was not associated with moderate disability compared to no disability in any of the analyses.

Table 2.

Distribution of Barthel Index by presence of depressed mood at year one and two of follow up.

		Depressed mood present	Depressed mood absent
One year	No disability*	51 (53.7%)	85 (56.3%)
	Moderate disability**	24 (25.3%)	47 (31.1%)
	Severe disability***	20 (21.0%)	19 (12.6%)
	Total (n = 246)	95	151
Two year	No disability*	47 (52.8%)	62 (52.5%)
	Moderate disability**	21 (23.6%)	40 (33.9%)
	Severe disability***	21 (23.6%)	16 (13.6%)
	Total (n = 207)	89	118

^*No disability: Barthel Index ≥ 95

^**Moderate disability: Barthel Index 60–95

^***Severe disability: Barthel Index < 60

Table 3.

Association between post-stroke depressed mood and disability score categories up to 2 years after stroke onset

	Model 1* OR (95% CI)		Model 2† OR (95% CI)
	< 60 compared to ≥ 95	60–95 versus ≥ 95	< 60 compared to ≥ 95	60–95 versus ≥ 95
One year	1.75 (0.86–3.60)	0.85 (0.47–1.55)	2.91 (1.07–7.91)	1.13 (0.52–2.48)
Two years	1.73 (0.82–3.68)	0.69 (0.36–1.33)	3.72 (1.29–10.71)	0.98 (0.43–2.26)

^*Unadjusted

^†Adjusted for sociodemographic factors: age, race-ethnicity, completing a high school education, having less than 3 friends, being unmarried, having Medicaid or no insurance; stroke severity physical activity, coronary artery disease, and diabetes

We retained a polytomous model since the effect size of depressed mood on severe disability compared to no disability and that on moderate disability compared to no disability were different (p = 0.05 for 1 year follow up, and p = 0.01 for 2 years follow up comparing the estimates for moderate and severe disability). In a series of logistic models examining the association between having depressed mood ascertained and functional outcomes we found no associations, indicating that having ascertainment of depressed mood did not depend on functional status.

Depressed mood was not associated with all cause mortality (adjusted HR 1.15, 95% CI 0.76–1.75), vascular death (adjusted HR 1.52, 95% CI 0.81–2.88), or non-vascular death (adjusted HR 0.78, 95% CI 0.41–1.50).

Discussion

We found that stroke patients who had depressed mood within the week following their stroke were more likely to be severely disabled up to 24 months after stroke. This disability started at the earliest time of follow up of 6 months. Our findings were independent of stroke severity, socio-demographic factors, and medical co-morbidities, but stroke severity was the strongest mediator, highlighting the importance of the initial deficit on ultimate outcome. We did not however find an effect on mortality, or on moderate disability. The reasons for not finding an effect on moderate disability could be due to inadequate statistical power to detect subtle effects, a lack of a true effect, or still unmeasured confounders. The latter point is unlikely given that we observed an effect for severe disability in univariate and adjusted models, and noted no trends in univariate analyses for moderate disability. Other studies have found that depressed mood is common after stroke, and that it correlates with poor functional outcomes⁶, as measured by activities of daily living scales¹⁶ and cognitive performance¹⁷. Depression has also been found to have a strong effect on recovery from stroke, as well as the likelihood of death on long-term follow up⁶. We find it noteworthy that a single question about mood would have such predictive value, with the understanding that stroke patients have undergone a potentially life-changing event and any depressive symptoms could be reactive in nature.

Our findings regarding depressed mood and post-stroke recovery, however, differ from other studies examining the link between post-stroke depression and outcomes in ascertaining depressed mood in the acute setting after stroke. Controversy exists regarding the appropriateness of diagnosing and therefore treating depression in the setting of an acute illness, particularly stroke. The clinician may be observing an acute “adjustment disorder”, physical symptoms (such as sleep disturbance and appetite disturbance) that could be due to stroke or depression, or the effects of brain injury itself⁶. While some of the psycho-somatic symptoms of depression after stroke do not persist (sleep disturbances, loss of libido), depressed mood and other psychological symptoms that are ascertained in the acute setting (such as anhedonia and hopelessness) remain in patients ultimately diagnosed with depression¹⁸. The distinction between “reactive” and endogenous depression may be arbitrary, as regardless of the cause, the presence of depressed mood may lead to diminished motivation to perform rehabilitation or participate in medical management for secondary stroke prevention. This is notable given that the prevalence of mood disorders increases significantly from one week to 3 months after stroke¹⁹, and may be one explanation for our observed results. Interestingly, one group demonstrated that post-stroke depression influenced functional recovery after stroke, but not motor recovery as measured by the Fugl-Meyer scale²⁰, arguing for motivation being an important component.

The finding that depressed mood after ischemic stroke affects recovery may have important clinical and treatment implications. A recent study randomized non-depressed stroke patients within three months to escitalopram versus placebo and found significant improvement in depression²¹. Another group has established the superiority of a brief psychosocial-behavioral intervention with pharmacological treatment compared to pharmacological treatment alone²². No clinical trials have been carried out to date in the more acute stroke setting (i.e. < 10 days).

The strengths of our study include evaluation of an older tri-ethnic population for whom we know less regarding stroke recovery, and a careful longitudinal design. We carried out systematic follow up during multiple time-points and were able to capture recurrent events which might influence our results. In the sample we used for this study, we ascertained the presence of depressed mood in the acute setting, which has been less frequently done.

This study has several limitations. We did not collect information on pre-stroke diagnosis of major depression, which predicts post-stroke depression in other studies. In other studies, however, the association between depression and poor functional outcomes remains when excluding those with pre-morbid depression²³. Depressed mood was assessed by a positive response to the first question on the Ham-D, but a full depression assessment was not consistently done, and thus we cannot report the effect of post-stroke major depression. However, the Ham-D has been used previously to measure incident post-stroke depression, and the prevalence of depression was 41% at one month, similar to the prevalence of depressed mood within 7–10 days of stroke in this study²⁴. In addition, depressed mood is one component of well-validated two-question screening questionnaires which have excellent sensitivity and specificity for depression^{25, 26}. There remains wide variability in the scales that are used to assess post-stroke depression, but most have yielded similar prevalence estimates⁶. We included participants with mild aphasia, and many of these scales have not been well validated in that setting. Psychomotor retardation, inattention, aprosody, and other psychiatric symptoms are commonly seen in patients with right hemisphere ischemia, and could masquerade as depressed mood, but in our analyses the side of the lesion was not associated with any of the outcomes, and it was therefore not kept in the final model. We may have been underpowered given our sample size to detect more subtle effects on moderate disability compared to severe disability, or mortality due to depressed mood. Post-stroke depressed mood was not ascertained in all participants at the beginning of this study, and those who did not have the question ascertained were more likely to have a severe stroke; nonetheless, having the question asked was not associated with functional outcome, suggesting that this was not a factor in our results. Participants were excluded if enrolled after 30 days from their event to minimize bias caused by early recovery from the index stroke affecting the answer to the mood question. We acknowledge that our assessment was carried out soon after stroke, so we may be capturing an adjustment reaction rather than major depression, though others have demonstrated an association between early psychological symptoms of depression and eventual diagnosis of depression¹⁸. We do not have information regarding post-stroke physical activity and fitness, which others have found to be strongly predictive of recovery²⁷. Use of anti-depressants after stroke was limited in our population, and also not systematically collected in all participants on follow up.

Further research into the effect of mood disorders on stroke recovery is needed. These data suggest that a single question about depressed mood can be an important predictor of post-stroke disability. It is likely that early treatment, which may include a combination of psychosocial intervention, psychotherapy, and pharmacotherapy, is beneficial for stroke recovery, but a clinical trial is needed in the acute setting.