Longitudinal Association of Midlife Vision Impairment and Depressive Symptoms: The Study of Women’s Health Across the Nation, Michigan Site

Abstract

Objective:

Poor vision affects physical health but the relationship with depressive symptoms among midlife adults (40–65 years), who often present with early stage vision impairment (VI), is not well understood. The goal of this study was to assess the impact of vision on depressive symptoms during midlife.

Methods:

The Michigan site of the Study of Women’s Health Across the Nation conducted assessments of distance visual acuity at six consecutive, near-annual follow-up visits. At each visit, depressive symptoms (Center for Epidemiological Studies-Depression Scale) were assessed. VI was defined as mild (20/30 to 20/60) or moderate-severe (20/70 or worse). Multivariable logistic regression models using generalized estimating equations were used to assess the association of VI and reporting of depressive symptoms at the subsequent visit.

Results:

At analytic baseline, the mean age of participants (N=226) was 50.0 years (standard deviation=2.6). More than half (53.5%) of women had mild VI and 8.0% had moderate-severe VI. Adjusting for age, pre-existing depressive symptoms, race, education, economic strain, body mass index, and smoking, participants with mild and moderate-severe VI had 68% (95% C (0.97–2.90)) and 2.55-fold (95% CI 1.13–5.75) higher odds of reporting depressive symptoms at their subsequent study visit as compared to women without VI. Further adjustment for diabetes, hypertension and osteoarthritis attenuated the estimates and the associations were no longer statistically significant.

Conclusion:

VI was associated with increased odds of future depressive symptoms among mid-life women. Timely detection and appropriate correction of VI may be important to consider in maintaining the mental health status of midlife women.

INTRODUCTION

Midlife (40–64 years) is a life stage characterized by changes in health, both physical and mental. Midlife mental health status is often challenged by high stress levels and surfacing of age-related health conditions that restrict one’s ability to function effectively. This age group has the highest prevalence of depression compared to all other age groups, and women have higher depression rates than men.¹ For women in particular, this is a vulnerable life stage for depression due to onset of menopausal transition.^2,3

Initiation of prevention and treatment measures for midlife depression are possible if the correlates of midlife depression are well established. Evidence from the literature suggests that poor vision or vision impairment (VI) is associated with higher risk of depression among elderly populations.^4–8 However, many vision-threatening ocular conditions emerge prior to old age, as evidenced by the tripling of vision impairment prevalence through the mid-life years.^9,10 These conditions include common and correctable ones such as refractive errors and cataracts and other more serious, chronic eye diseases such as glaucoma, diabetic retinopathy, hypertensive retinopathy and macular degeneration.^11–14

Despite this demonstrable increase in the prevalence of both depression and common eye disorders that can compromise vision during midlife, knowledge about the impact of vision on depressive symptoms among midlife adults is limited. Several studies have reported that VI has a negative impact on psychological well-being¹⁵⁻²¹yet precise estimates among midlife populations are lacking due to the limited number of studies that assessed midlife vision and depression. Furthermore, the nature of the longitudinal association and directionality of the relationship between VI and depression among midlife adults is largely unknown. Thus, the goal of this study was to assess the longitudinal association between VI and depressive symptoms among a population-based cohort of midlife women.

METHODS

Study sample.

The sample for this study was derived from the Study of Women’s Health Across the Nation (SWAN), Michigan site. SWAN is a longitudinal, multi-ethnic study of women’s health at midlife. Michigan is one of the seven clinical sites for SWAN and includes a population-based sample from two Detroit area communities. A total of 543 women were recruited into the Michigan SWAN (MI-SWAN) cohort in 1996. Eligibility criteria at baseline included 42–52 years of age, having an intact uterus and at least one menstrual period in the previous 3 months, and no use of hormone therapy in the previous 3 months.²² Starting in 2001 (follow-up visit 4), the Michigan site initiated a site-specific protocol to assess vision. Of the 349 Michigan SWAN women who participated in the 2001 study visit in-person, the analytic sample here includes 226 women who completed the distance vision testing with their glasses or contact lenses or did not need glasses or contact lens correction for distance. Women were excluded from the analytic sample if they did not participate in the distance vision assessment (n=4), completed the vision assessment without their glasses or contact lenses (if needed) (n=110) or had missing or discordant data about use of glasses or contact lenses for the vision assessment (n=9). Women who were excluded because of testing without correction (if needed) were more likely to be African American and were slightly older than women in the analytic sample or the SWAN cohort overall but there were no other differences in key covariates (Supplemental Table 1).

Vision assessment.

Distance visual acuity was assessed using the occupational model of the Titmus 2a vision screener (Titmus Optical Inc., Chester, Virginia). The Titmus vision screener is a stereoscopic instrument that is widely used in occupational and school health settings to measure vision. It has excellent sensitivity for detection of VI.²³ Participants were tested while seated and wearing their glasses or contact lenses during the test if they normally wore visual correction. Visual acuity was measured binocularly with the instrument set to the distance vision setting. For the test, a self-lit slide with a group of diamond shaped figures was shown to the participant while looking through the eyepiece of the instrument. On a given slide, each diamond shaped figure contained four rings, three of which had a break in the oblique axis and one was unbroken. Participants were asked to name the position of the unbroken ring. The test began by the participant identifying the most easily identifiable ring and ended when two consecutive answers were incorrect. Distance visual acuity was recorded as a Snellen equivalent, with the numerator representing the distance in feet at which participant identified the smallest shape and denominator representing the distance at which a normal sighted person is expected to identify the same shape. Using definitions developed by the World Health Organization, no VI was defined as distance visual acuity better than 20/30, mild VI was defined as distance visual acuity 20/30 to 20/60, and moderate to severe VI was defined as distance visual acuity 20/70 or worse.²⁴

Depressive Symptoms.

Depressive symptoms were assessed at analytic baseline (2001) and at near-annual follow-up visits using the Center for Epidemiologic Studies Depression (CESD) scale, a self-reported 20-item questionnaire. The CESD questionnaire was designed to measure the frequency of depressive symptoms across four domains (depressed affect, positive affect, somatic activity, and interpersonal relations) during the previous week on a 4-point scale of 0 (rarely) to 3 (most or all of the time).²⁵ Responses to the 20 items were summed for a total score ranging from 0–60. A score of 16 or higher was considered to be indicative of clinically relevant depressive symptoms.²⁶ The CESD has been shown to be valid and reliable among diverse populations.^26–28

Covariates.

Age at each visit was calculated based on date of birth and visit date. Race was self-reported at SWAN baseline as African American or white. Economic strain at SWAN baseline was categorized as none, some, or high based upon self-reported difficulty paying for basic needs. Level of education was categorized based upon self-report at SWAN baseline as high school or less, more than high school, and college or post-college. At the follow-up visits, height and weight were measured using a stadiometer and calibrated balance beam scale, respectively, and used to calculate body mass index (BMI) as weight (kilograms) divided by the square of height (meters). Current smoking status at each study visit (yes/no) was based on self-report. Diabetes, hypertension and osteoarthritis were determined at each study visit based on a self-report of doctor-diagnosed disease.

Statistical analysis.

Baseline population characteristics were assessed by tabulating frequencies and percentages for VI, depressive symptoms and categorical covariates. Mean and standard deviations were calculated for continuous covariates. Bivariate analyses were conducted to identify covariates that were statistically significantly associated with VI and depressive symptoms. To assess the longitudinal association of depressive symptoms and VI, logistic regression models using generalized estimating equation with an autoregressive correlation structure were employed. The generalized estimating equations account for the within subject correlation of repeated measures and allows for multiple incidents of depressive symptoms. Unadjusted analyses were performed to assess the association of VI at time t (follow-up visits from 2001–2007) with depressive symptoms at the subsequent visit (time t +1) (follow-up visits from 2002–2009). Adjusted models were built by adding covariates identified as potential confounders from literature review and bivariate analyses. Race, economic strain and education were characterized at SWAN baseline whereas the remaining covariates were time-varying and modeled concurrent with vision assessment. Covariates were retained in the final model if there was a statistically significant association with depressive symptoms, if they changed the magnitude of the estimate for VI by 10% or more, or if they were deemed important a priori based upon a literature review. To assess whether change in VI was associated with depressive symptoms, models assessing the interaction of VI and time were examined. To assess the short- and long-term relationship between VI and depressive symptoms, stratified models including data from baseline through 2, 4, and 6 years of follow-up were considered separately. Statistical significance was defined as a p-value < 0.05. Statistical analyses were performed using SAS Version 9.3 (SAS institute, Cary, North Carolina).

RESULTS

Characteristics of the study population are presented in Table 1. At the analytic baseline (2001), more than half of women had mild VI (53.5%, n=121) and 8.0% had moderate to severe VI (n=18). Women with no VI were statistically-significantly more likely to have attended college as compared to women with mild or moderate/severe VI (41.0%, 20.5%, 22.2%, respectively; p=0.008). Diabetes prevalence was also statistically significantly positively associated with VI status (p=0.009). There were no differences in age, race, economic strain, smoking status, menopausal status, hypertension or osteoarthritis by VI status (Table 1).

Table 1:

Characteristics of study population at baseline and associations with distance vision impairment and depressive symptoms at baseline, Michigan Study of Women’s Health Across the Nation

Variable	Study Sample	Distance Vision Impairment (VI)				Depressive Symptoms
	N =226	No VI n=87	Mild VI n=121	Moderate to Severe VI n=18	P	No n= 170	Yes n=56	P
Age (years), Mean (SD)
	50.0 (2.6)	49.9 (2.7)	50.0 (2.5)	49.8 (3.2)	0.95a	49.9 (2.5)	50.1 (2.8)	0.65d
Body mass index (kg/m 2 ), Mean (SD)
	33.2 (8.6)	31.6 (8.4)	34.5 (8.5)	32.0 (9.6)	0.05a	33.9 (8.8)	31.0 (7.7)	0.03d
Race, n (%)					0.36b			0.21b
African American	129 (57.1%)	47 (54.0%)	69 (57.0%)	13 (72.2%)		93 (54.7%)	36 (64.3%)
White	97 (42.9%)	40 (46.0%)	52 (43.0%)	5 (27.8%)		77 (45.3%)	20 (35.7%)
Economic strain, n (%)					0.25b			0.005b
High stain	28 (12.4%)	6 (7.0%)	20 (16.5%)	2 (11.1%)		16 (9.5%)	12 (21.4%)
Some strain	78 (34.7%)	33 (38.4%)	37 (30.6%)	8 (44.4%)		54 (32.0%)	24 (42.9%)
No strain	119 (52.9%)	47 (54.7%)	64 (52.9%)	8 (44.4%)		99 (58.6%)	20 (35.7%)
Education, n (%)					0.008b			0.25b
≤ High School	67 (30.7%)	18 (21.7%)	40 (34.2%)	9 (50.0%)		46 (27.9%)	21 (39.6%)
> High School	89 (40.8%)	31 (37.3%)	53 (45.3%)	5 (27.8%)		69 (41.8%)	20 (37.7%)
College/Post College	62 (28.4%)	34 (41.0%)	24 (20.5%)	4 (22.2%)		50 (30.3%)	12 (22.6%)
Smoking status, n (%)					0.16b			0.005b
Non-Smoker	175 (77.8%)	71 (81.6%)	93 (77.5%)	11 (61.1%)		139 (82.2%)	36 (64.3%)
Smoker	50 (22.2%)	16 (18.4%)	27 (22.5%)	7 (38.9%)		30 (17.8%)	20 (35.7%)
Menopausal status, n (%)					0.69c			0.15b
Pre-menopausal	12 (5.3%)	5 (5.8%)	5 (4.1%)	2 (11.1%)		11 (6.5%)	1 (1.8%)
Peri-menopausal	126 (55.8%)	51 (58.6%)	66 (54.6%)	9 (50.0%)		92 (54.1%)	34 (60.7%)
Post-menopausal	52 (23.0%)	16 (18.4%)	32 (26.4%)	4 (22.2%)		36 (21.2%)	16 (28.6%)
Unknown (hysterectomy/HT)	36 (15.9%)	15 (17.2%)	18 (14.9%)	3 (16.7%)		31 (18.2%)	5 (8.9%)
Diabetes, n (%)					0.009b			0.40c
Yes	18 (8.0%)	2 (2.3%)	12 (9.9%)	4 (22.2%)		12 (7.1%)	6 (10.7%)
No	208 (92.0%)	85 (97.7%)	109 (90.1%)	14 (77.8%)		158 (92.9%)	50 (89.3%)
Hypertension, n (%)					0.52b			0.90b
Yes	62 (27.4%)	21 (24.1%)	37 (30.6%)	4 (22.2%)		47 (27.6%)	15 (26.8%)
No	164 (72.6%)	66 (75.9%)	84 (69.4%)	14 (77.8%)		123 (72.4%)	41 (73.2%)
Osteoarthritis, n (%)					0.33b			0.50b
Yes	53 (23.4%)	16 (18.4%)	33 (27.3%)	4 (22.2%)		38 (22.4%)	15 (26.8%)
No	173 (76.6%)	71 (81.6%)	88 (72.7%)	14 (77.8%)		132 (77.6%)	41 (73.2%)
Depressive Symptoms, n (%)					0.71b
Yes	56 (24.8%)	22 (25.3%)	31 (25.6%)	3 (16.7%)		n/a	n/a	n/a
No	170 (75.2%)	65 (74.7%)	90 (74.4%)	15 (83.3%)		n/a	n/a	n/a
Distance Vision Impairment, n (%)								0.71b
No VI	87 (38.5%)	n/a	n/a	n/a		65 (38.2%)	22 (39.3%)
Mild VI	121 (53.5%)	n/a	n/a	n/a		90 (52.9%)	31 (55.4%)
Moderate-Severe VI	18 (8.0%)	n/a	n/a	n/a		15 (8.8%)	3 (5.4%)

VI, vision impairment; P, P-value; SD, standard deviation; kg, kilograms; m, meters; n, number; HT, hormone therapy; n/a, not applicable

^aAnalysis of variance F-test

^bChi square test

^cFisher exact test

^dStudent’s t-test

Nearly one-quarter of women (n=56, 24.8%) reported depressive symptoms at the analytic baseline. The prevalence of depressive symptoms was more common among women with high levels of economic strain (p=0.005) and among women who smoked (p=0.005). The average BMI was lower for women with depressive symptoms (31.0 kg/m², standard deviation (SD) = 7.7) as compared to women without depressive symptoms (33.9 kg/m², SD=8.8) (p=0.03).

In unadjusted analyses, women with mild VI had approximately 50% higher odds and women with moderate-severe VI had more than 2-fold higher odds of subsequent depressive symptoms as compared to women with no VI, but these findings were not statistically significant (Table 2). However, after adjustment for age and pre-existing depressive symptoms the relationship between VI and depressive symptoms strengthened and became statistically significant. Women with mild VI had 78% higher odds (odds ratio (OR)=1.78, 95% confidence interval (CI) 1.04–3.05) and women with moderate-severe VI had more than two and a half fold higher odds (OR=2.71, 95% CI 1.28–5.73) of depressive symptoms as compared to women without VI. Further adjustment for race, level of education, economic strain, body mass index and smoking status attenuated the estimates only slightly but only the relationship between moderate-severe VI and depressive symptoms remained statistically significant (OR=2.55, 95% CI 1.13–5.75). However, additional adjustment for diabetes, hypertension and osteoarthritis further attenuated the effect estimate and became only marginally statistically significant (p=0.07) (Table 2).

Table 2:

Association between vision impairment and subsequent depressive symptoms, Michigan Study of Women’s Health Across the Nation

	Mild Vision Impairment		Moderate-Severe Vision Impairment
	Odds Ratio	95% Confidence Interval	Odds Ratio	95% Confidence Interval
Model 1 a	1.54	(0.99–2.42)	2.03	(0.95–4.34)
Model 2 b	1.78	(1.04–3.05)	2.71	(1.28–5.73)
Model 3 c	1.68	(0.97–2.90)	2.55	(1.13–5.75)
Model 4 d	1.63	(0.95–2.82)	2.22	(0.94–5.27)

^aUnadjusted

^bAdjusted for age and pre-existing depressive symptoms

^cAdjusted for Model 2 + race, level of education, economic strain, body mass index, and smoking status

^dAdjusted for Model 3 + diabetes, hypertension, and osteoarthritis

In the fully adjusted model, age was inversely associated with depressive symptoms; every year increase in age was associated with 6% lower odds of subsequent depressive symptoms (p=0.04). History of depressive was highly predictive of future depressive symptoms (p<0.0001). None of the other covariates were statistically significant predictors of subsequent depressive symptoms.

To examine the impact of follow-up time on the relationship between vision impairment and subsequent depressive symptoms, we conducted analyses restricted to 2, 4, and 6 years of follow-up. As shown in Table 3, the effect estimates were 34% larger for mild VI and 6% larger for moderate-severe VI in analyses restricted to 2 years of follow-up as compared to 6 years of follow-up. Further, mild VI was statistically significantly associated with higher odds of subsequent depressive symptoms in time-restricted analyses but not in analyses including the full 6 years of follow-up.

Table 3:

Association between vision impairment and subsequent depressive symptoms at 2, 4 and 6 years of follow-up, Michigan Study of Women’s Health Across the Nation

	Mild Vision Impairment		Moderate-Severe Vision Impairment
	Odds Ratio	95% Confidence Interval	Odds Ratio	95% Confidence Interval
2 years of follow-up a	2.47	(1.22–5.01)	2.36	(0.85–6.55)
4 years of follow-up	1.90	(1.07–3.39)	2.03	(0.74–5.59)
6 years of follow-up	1.63	(0.95–2.82)	2.22	(0.94–5.27)

^aAdjusted for age, pre-existing depressive symptoms, race, level of education, economic strain, body mass index, smoking status, diabetes, hypertension, and osteoarthritis

DISCUSSION

This study demonstrates that midlife VI has a significant impact on depressive symptoms. Distance VI was independently associated with depressive symptoms during the midlife years after adjustment for age and pre-existing depressive symptoms, the only covariates to be statistically significant predictors in this analysis. Presence of a significant association of mild VI with depressive symptoms indicates that impaired vision impacts depressive symptoms even at milder levels. This finding emphasizes the importance of considering and treating mild VI, which is often ignored but may considerably impact the quality of life.^29,30 Moderate to severe VI was more strongly associated with depressive symptoms, and this finding remained robust after adjustment for demographic and health behavior covariates. However, the magnitude of the association between moderate-severe VI attenuated and became non-significant after adjustment for other chronic health conditions. This may reflect the importance of coping strategies demonstrated by those with multiple health conditions. Coping strategies, such as accommodative coping involving flexible goal pursuit and goal readjustment, are believed to be beneficial to prevent a decline in mental well-being following VI, and this has been observed among middle aged adults with substantial VI.³¹

The attenuation of the effect following addition of diabetes, hypertension and osteoarthritis into the models may also reflect commonalities in the burden and pathophysiologic underpinnings between depression and other co-morbid health conditions. Globally, the prevalence of depression is twice as high in those with diabetes as compared to the general population³² and evidence supports shared etiologic factors and biological mechanisms between depression and diabetes and hypertension³³. For example, low socioeconomic status and shared behavioral factors such as low physical activity, poor diet, and poor sleep are all risk factors for depression^34,35, diabetes^36,37, and hypertension^38,39. Given this, stress may be one mechanism leading to disruption of key physiologic pathways including the hypothalamic-pituitary-adrenal (HPA) axis⁴⁰ and immune responses. Dysregulation of the HPA axis can lead to increasing levels of cortisol and adrenalin, and, like immune dysregulation, can increase the production of pro-inflammatory cytokines – all of which are associated with incident depression^41,42, diabetes^43,44, and hypertension⁴⁵, as well as other conditions including obesity and metabolic syndrome. Given the shared epidemiologic and physiologic relationships between depression, diabetes and hypertension, inclusion of these chronic health outcomes in models when depression is the outcome would be expected to attenuate the effect size. Future work can further explore potential mediation of the relationship between vision and depression with these behavioral risk factors, mechanistic markers or key pathophysiologic pathways, or the chronic health conditions themselves.

Previous cross-sectional studies reported higher odds of depressive symptoms among those who were visually impaired, and results of this study confirm that finding.^{15–18,21,46} A population-based cohort study of midlife and older adults reported distance VI being associated with higher odds of depressive symptoms and an attenuation of the effect size over the 10-year study period.¹⁹ Our analyses did not demonstrate a statistically significant interaction between VI and follow-up time, but our exploratory analyses suggested that the magnitude of the relationship between VI and subsequent depressive symptoms was strongest when restricting the analysis to the first two years of follow-up. However, the sample size in our study declined over time, so comparisons of early to extended follow-up are not completely analogous.

A recent longitudinal study that assessed vision and depressive symptoms over a period of 10 years, using a nationally representative sample, also reported VI as a significant risk factor for depressive symptoms.⁴⁷ Age stratified analyses reported in that study revealed that the only age group that had a significant association of VI and depression was midlife, defined as 30–59 years old. Three other longitudinal studies among older adults with follow-up data over a period of 2–4 years, identified VI as a significant risk factor for depression.^5,8,48 The current study confirms the longitudinal association of VI and depression and reports an association earlier in the aging process. Further, our finding that increasing age is a protective factor for depressive symptoms highlights the importance of examining predictors for depressive symptoms during the mid-life, a period of time often characterized by high levels of stress and mental health concerns. Given some suggestion that depressive symptoms may decline during the mid-life period^49,50, coupled with evidence that the prevalence of depression and chronic conditions are highly correlated³², future studies should examine the impact of VI on depressive symptom trajectories during the critical transition from mid- to late adulthood, across which time period the burden of comorbid conditions increases⁵¹.

Strengths of this study include its multi-racial population-based sample, focus on the midlife years prior to onset of advanced or severe vision loss, and use of objective measures to assess vision. The longitudinal study design allowed us to observe the relationship between VI and the outcome of depressive symptoms over several years of follow-up and to understand the temporality of the relationship. This study used presenting visual acuity (as opposed to best-corrected visual acuity) thus reflecting the true burden of VI in the community. Examining parameters pertinent to midlife VI assisted us in understanding issues unique to this life stage.

Limitations of this study include the use of the Titmus vision screener rather than commonly used clinical methods to assess vision. Further, because this was a secondary data analysis within an ongoing cohort study not designed to answer this specific question, and thus we only had 18 women with moderate to severe VI, we may have been underpowered. A larger sample size may have improved the precision of our estimates. Finally, a study sample consisting of midlife women limits the study’s generalizability. However, VI and depression are both more prevalent in women compared to men, hence an exclusive focus on women’s health is relevant to understand the relationship of VI and depressive symptoms.^1,9

CONCLUSION

In conclusion, this study reports a significant longitudinal association of mild and moderate-severe VI with subsequent depressive symptoms among midlife women. Midlife depression has far reaching consequences not only in terms of concurrent poor health outcomes, but also as a deterrent to a healthy aging process.^52–55 In addition, poor vision worsens the effects of comorbid conditions, including depression, and the combined presence of VI and depression has a profound negative impact on functioning and social participation.⁵⁶ The complex and bi-directional nature of the relationship between depression and VI could potentially lead to a vicious cycle.^4,7 Thus, early identification and timely correction of vision problems is an important step in preserving mental and physical health among midlife women.

Supplementary Material

Supplemental Data File (.doc, .tif, pdf, etc.)

Supplemental Digital Content: Depression and Vision Paper_Supplementary Table 1.docx