1. Intimacy and Social Isolation In the
National Social Life, Health and Aging Project
(NSHAP)
Linda J. Waite
July 14-15, 2008
ESRC/NIA London Workshop

2. Intimacy Protects & Social Isolation Kills… but How and Why?
Variety of indicators of isolation
Different types of isolation affect particular aspects of health
Possible mechanisms?
2001 paper by James House (Psychosomatic Medicine) was entitled, “Social Isolation Kills… but How and Why?”
Noted that we know well that social isolation has harmful effects – particularly among older adults.
The effects of isolation can be equated with the effects of smoking and other well-known health risks – but the research in this area is lacking.
Inconsistent indicators across studies
Possibility that different types of isolation affect particular aspects of health
Variety of possible mechanisms, based on what we know about social connectedness (it buffers stress, increases self esteem, provides social control, and allows access to resources) – is isolation the opposite/absence of these benefits, or does it affect health through some other mechanism 2

3. National Social Life, Health, and Aging Project
Interview 3,005 community-residing adults ages 57-85
Population-based sample with minority over-sampling
120-minute in-home interview
Questionnaire
Biomeasure collection
Leave-behind questionnaire 3

4. Mental Health Measures
Anxiety HADS (7 items) modified to match those of the Center for Epidemiological Studies Depression Scale
Bereavement Four items
Depression CES-D (11 items)
Happiness Single item
Loneliness Three-item scale
Self esteem Single item
Stress Four item modification of Cohen’s Perceived Stress Scale

5. Biological or biologically-derived indicators of disease
Biomeasures
Biological or biologically-derived indicators of disease
Anthropometrics
Height, weight, BMI
Waist circumference
Medications
Blood pressure
Get up and go

6. Blood Spots
C-reactive protein
Epstein-Barr Virus Antibody Titers
Hemoglobin A1c

7. Indicators of Social Isolation Forms of Social Isolation
Objective Social Isolation
Living alone
Social network size and characteristics
Social inactivity
Perceived lack of social support
Emotionally distant relationships
Loneliness
Physical separation from others
Subjective Social Isolation
Various well-known indicators of social isolation
Problems:
measures/assessed in different ways
Very few studies have examined more than 1-2 of these indicators – we don’t know their relative importance
Unlikely that any one aspect of isolation tells the full story about the effect of isolation on health
In fact, these factors probably work in combination – so that the effect of isolation on health is more likely the result of a combination of factors
But, we can’t simply put all of these factors into a regression model – need something more parsimonious
We suggest that these measures can be grouped together – they actually comprise two forms of isolation:
Objective social isolation
Subjective social isolation
Objective = physical separation from others
Subjective = the perception of low quality or emotionally distant relationships, or dissatisfaction with one’s relationships with others
Overlap, but distinct
Perception that relationships with others are low quality, emotionally distant, or unsatisfying

8. Objective Social Isolation Scale (α = .74)
Social network characteristics
Size
Range
Contact volume
Proportion of network members in the household
Social involvement
Participation in organized groups
Religious attendance
Volunteering
Number of Friends
First, we created a scale measuring objective social isolation, or physical separation from others
Items in this scale come from three general areas:
Social network characteristics
2) Social involvement
3) Number of Friends
Total of 8 items.
5 other items were initially included based on their content validity, but removed in order to maximize the scale’s internal consistency (removing them increased Chronbach’s alpha).
- Living alone
- Partnership status
- Neighborhood socializing
- Number of children
- Number of grandchildren
Item-scale correlations were moderate to strong.
PCA shows one eigenvalue of 2.82, which accounts for 35% of the variance.
(Erin -- Be sure to have summary statistics and measurement information for all of these variables!)

9. Subjective Social Isolation Scale (α = .71)
Emotional closeness with network members
Perceived social support
Spouse or current partner
Family members
Friends
Feelings of loneliness and not belonging
Loneliness Scale (Hughes, Waite, Hawkley, and Cacioppo 2004)
Lack companionship
Feel left out
Feel isolated
“I was lonely”
Next, we constructed a scale measuring subjective social isolation, or dissatisfaction with one’s relationships, or the perception that they are low quality, emotionally distant relationships
Items in this scale come from three different areas:
Average emotional closeness with network members
Perceived social support (6 items, open up to/rely on for each)
Feelings of loneliness or not belonging
Scores for items from the first two groups were reversed.
All scores were averaged and then standardized.
11 items total – all items were retained.
Item-total correlations were in the weak to moderate range
PCA shows one eigenvalue of 2.87, which accounts for about 26% of the variance.
(Erin -- Be sure to have summary statistics and measurement information for all of these variables!)
NOTE that we don’t mean to say that these are the BEST scales, but that this is an early example of how these indicators can be combined.

10. Figure 2. Predicted Probability of Very Good or Excellent Physical Health, by Levels of Objective and Subjective Isolation
Predicted probabilities of having very good or excellent self-rated physical health.
Both objective and subjective isolation reduce the probability of having very good or excellent physical health,
Even after we’ve controlled for age, gender, race, education, marital status, living alone, and co-morbidities.

11. Figure 3. Predicted Probability of Very Good or Excellent Mental Health, by Levels of Objective and Subjective Isolation
Here we see that objective isolation has very little effect on the probability of having very good or excellent self-rated mental health.
But, the effect of subjective isolation on mental health is very strong.
As subjective isolation increases, the probability of reporting good mental health decreases pretty drastically – from almost .9 (at the lowest levels of subjective isolation) to about .1 at the very highest levels of subjective isolation.
We wanted to verify this result using another indicator of mental health, for two reasons:
Violation of proportional odds assumption in Model 4 – EXPLAIN
Self-rated mental health indicator is new
So… we ran a set of OLS regressions predicting depressive symptoms (as measured by our shortened 11-item version of the CES-D scale, minus the item about loneliness)

12. Social Isolation and Self-Rated Health
Objective and subjective isolation are distinct forms of social isolation
Both objective and subjective isolation are independently related to self-rated physical health
The effect of objective isolation on mental health may operate through the very strong link between subjective isolation and mental health
So, to summarize our findings…
These early analyses indicate that objective and subjective isolation are, indeed, different forms or dimensions of social isolation.
Considering them separately has analytic advantages (create more parsimonious models) and substantive advantages (think about combinations of facets of isolation)
And, objective and subjective isolation seem to have different effects on physical and mental health. …
Of course, we’re limited by the cross-sectional nature of the data. Supplemental analyses of endogeneity and instrumental variables regression analyses engender confidence in our findings and suggest that isolation does have a causal effect on health. However, we’re looking forward to being able to look at these relationships longitudinally, after a second wave of data collection.
Most important and most immediate next step = look at how objective and subjective isolation are related to more objective measures of health, including biomarkers, and systems such as endocrine function, inflammation, cardiovascular function, and metabolism. This will help us to think about mechanisms through which isolation affects health among older adults.

13. How Do Intimacy & Isolation Affect Health?
Through damage done by stress

14. Increased + frequent rapid changes in blood pressure
Sympathetic Nervous System Activation and Atherosclerosis and Hypertension
Stress
Increased + frequent rapid changes in blood pressure
Increase in turbulent flow + shear stress on artery walls
Endothelial injury
Atherosclerosis and narrowed vessels
Hypertension
Atherosclerosis—hardening of the arteries. SNS=sympathetic nervous system; SAM=sympathetic adreno medullary system

15. Regressions Predicting Systolic Blood Pressure by Gender
Male
(n=1067)
Female
(n=1147)
Objective isolation
-0.779
1.969
Co-resident partner
-1.823
-5.004**
Very Low Loneliness (omitted) -
Low Loneliness (4-5)
-2.418
-1.313
Moderate Loneliness (6)
-4.377*
-4.021†
High Loneliness (7-9)
-2.745
-1.895
Lacks Social Support
2.437*
1.034
Obese
1.913
1.019
Exercises
1.406
-0.174
Smokes cigarettes
0.939
-1.614
Drinks heavily
5.731**
3.429
Net of: Age, education, race/ethnicity, presence of spouse/partner, antihypertensive drugs, depression, anxiety, stress, and sleep quality.

16. Regressions Predicting CRP by Gender
Male
(n=689)
Female
(n=726)
Objective isolation
0.034
0.153†
Co-resident partner
-0.194†
0.104
Loneliness (standardized)
-0.013
0.020
Lacks Social Support
-0.004
-0.045
Obese
0.462**
0.622**
Exercises
-0.239**
-0.089
Smokes cigarettes
0.493**
0.197
Drinks heavily
0.145
-0.241
Good sleep quality
-0.233*
0.002
Net of: Age, education, race/ethnicity, presence of spouse/partner,SBP, antihypertensive drugs, depression, anxiety, and stress. 
† significant at 10%; * significant at 5%; ** significant at 1%

17. National Social Life, Health, and Aging Project
NSHAP data publicly available
NACDA at ICPSR
University of Michigan

18. Extra Slides

19. Figure 1. The Relationship between Objective Isolation and Subjective Isolation
First, let’s look at the relationship between the two scales.
r = .27, p < .001
(weak to moderate in strength, statistically significant)
Answers our first question -- Objective and subjective social isolation are related, but they’re distinct.
Capturing different dimensions of isolation.
Physical separation from others does not necessarily bring loneliness
Loneliness can be experienced even by those who have frequent social interactions or a wide variety of social connections
So, it’s possible that their effects on health are distinct.
Next, we move to regression analysis to assess their contributions to physical and mental health (second question)

20. Table 1. Ordered Logistic Regressions Predicting Self-Rated Physical Health (n = 2978)
Ordered logistic regression models predicting self-rated physical health
Not going to review this in detail – the main point here is that both objective and subjective isolation seem to have significant, negative effects on self-rated physical health
Interaction term – objective x subjective – not significant, no increase in predictive power
Better illustrated with a figure…

21. Table 2. Ordered Logistic Regressions Predicting Self-Rated Mental Health (n = 2980)
[MAY CUT THIS SLIDE AND TRANSITION DIRECTLY FROM FIGURE 2 to FIGURE 3… ALTHOUGH SHOWING THIS SLIDE VERY BRIEFLY MAKES THE TRANSITION A LITTLE MORE SMOOTH?]
Next, we wanted to look at the effects of objective and subjective isolation on mental health.
We use a single item indicator of self-rated mental health that mirrors that for physical health – “How would you rate your mental or emotional health?” – excellent, very good, good, fair, or poor.
Again, we ran ordered logistic regression models, predicting self-rated mental health. I will show these models only very briefly.
Both objective and subjective isolation are negative predictors of self-rated mental health, when entered separately, but when we enter them both into the model together, we find that the effect of objective social isolation disappears. That is, the effect of objective social isolation on mental health appears to be operating through subjective assessments of isolation.
This is more clear in the figure, based on the fourth model.

22. Table 3. OLS Regressions Predicting Depressive Symptomsa

23. Figure 4. Predicted Depressive Symptoms, by Levels of Objective and Subjective Isolation
I won’t show the models, but the findings are very consistent with the models predicting self-rated mental health.
Here, we see basically the same relationship as in the previous figure.
Objective isolation has little effect on depressive symptoms…
But as subjective isolation increases, the predicted depressive symptoms increase from almost 0 (at lowest levels of subjective isolation) to nearly 14 (at highest levels of subjective isolation).
(While the CES-D is not predictive of clinical depression, research indicates that the 80th percentile on the CES-D scale is usually a good benchmark for a high likelihood of being diagnosed with clinical depression. The 80th percentile for this modified version of the CES-D is about [SHOW DOTTED LINE] – so we see that the predicted level of depressive symptoms crosses this 80th percentile mark at about 1 standard deviation above the average level of subjective social isolation.)