Negative Social Interactions and Incident Hypertension among Older Adults

OBJECTIVE
To determine if negative social interactions are prospectively associated with hypertension among older adults.


METHOD
This is a secondary analysis of data from the 2006 and 2010 waves of the Health and Retirement Study, a survey of community-dwelling older adults (age > 50 years). Total average negative social interactions were assessed at baseline by averaging the frequency of negative interactions across 4 domains (partner, children, other family, friends). Blood pressure was measured at both waves. Individuals were considered to have hypertension if they reported use of antihypertensive medications, had measured average resting systolic blood pressure of 140 mmHg or higher, or measured average resting diastolic blood pressure of 90 mmHg or higher. Analyses excluded those who were hypertensive at baseline and controlled for demographics, personality, positive social interactions, and baseline health.


RESULTS
Twenty-nine percent of participants developed hypertension over the 4-year follow-up. Each 1-unit increase in the total average negative social interaction score was associated with a 38% increased odds of developing hypertension. Sex moderated the association between total average negative social interactions and hypertension, with effects observed among women but not men. The association of total average negative interactions and hypertension in women was attributable primarily to interactions with friends, but also to negative interactions with family and partners. Age also moderated the association between total average negative social interactions and hypertension, with effects observed among those ages 51-64 years, but not those ages 65 or older.


CONCLUSION
In this sample of older adults, negative social interactions were associated with increased hypertension risk in women and the youngest older adults.

The effects of negative aspects of social relationships on cardiovascular outcomes, however, have received less attention.By negative social interactions, we mean exchanges or behaviors that involve excessive demands, criticism, disappointment, or other unpleasantness.Here we focus on the role of negative interactions in risk for hypertension.Until now, support for an association between negative interactions and elevated blood pressure has been limited to a cross-sectional study (de Gaudemaris et al., 2011), a prospective study predicting self-reported hypertension (Wickrama et al., 2001), and several experimental studies (e.g., Ewart, Taylor, Kraemer, & Agras, 1991).;Kiecolt-Glaser & Newton, 2001;Smith et al., 2013).Cross-sectional studies provide evidence for an association between negative interactions and blood pressure, but leave the temporal ordering uncertain.The study of self-reported disease suffers in that, at best, self-report is a weak marker of objectively verified hypertension.Finally, experimental studies are limited in that they do not reflect negative interactions as they are experienced in natural social networks and assess short-term changes in blood pressure that quickly return to baseline.
The purpose of the current study was to examine the effects of negative social interactions on the incidence of hypertension, a major risk factor for cardiovascular disease, stroke, and mortality among older adults.Negative social interactions may be especially relevant for older adults, because they have smaller social networks and fewer types of social relationships (Fung, Carstensen, & Lang, 2001), as well as greater (age-related) vulnerability to cardiovascular disease.The study design was prospective, used objective assessments of blood pressure, pursued a range of potential mechanisms that may link negative interactions to hypertension, tested whether the association of negative interactions and onset of hypertension are moderated by sex or by age, and evaluated whether associations are independent of stable individual differences in social personality traits (e.g., extraversion, agreeableness, hostility, neuroticism) or levels of positive interaction.

Effects of Negative Interactions May Be Modified by Sex and Age
We were particularly interested in the possibility that negative social interactions might be most harmful for women.Women are thought to be more sensitive to the quality of their social interactions, particularly to negative ones.For example, women have more negative psychological responses to social stress than men (Bakker, Ormel, Verhulst, & Oldehinkel, 2010;Rudolph, Ladd, & Dinella, 2007;Shih, Eberhart, Hammen, & Brennan, 2006), are more bothered by negative social exchanges than men (Newsom, Rook, Nishishiba, Sorkin, & Mahan, 2005), demonstrate more cortisol and cardiovascular reactivity to interpersonal laboratory stress (Kiecolt-Glaser & Newton, 2001;Stroud, Salovey, & Epel, 2002), and have greater parasympathetic withdrawal in response to interpersonal conflict (Bloor, Uchino, Hicks, & Smith, 2004;Smith et al., 2009;Smith et al., 2011).
We also expected that age might moderate the association between negative social interactions and hypertension risk.As people age, they decrease the size of their social networks, possibly to devote more time, attention, and emotional resources to relationships with close friends and family (Carstensen, 1992(Carstensen, , 1993;;Carstensen, Gottman, & Levenson, 1995;Fung, Carstensen, & Lang, 2001).Having fewer social relationships may exacerbate the effects of negative interactions, because there would be fewer other network members with whom one may have positive interactions to buffer the effects of aversive ones.
Alternatively, increasing age may provide protection from the deleterious effects of negative interactions.As people get older, they adapt to negative aspects of their relationships and perceive them as less problematic (Akiyama, Antonucci, Takahashi, & Langfahl, 2003;Hansson, Jones, & Fletcher, 1990).There are also age differences in the strategies that individuals use for handling interpersonal difficulties that make older adults less vulnerable to the adverse effects of negative interactions.For example, Birditt and Fingerman (2005) found that older adults were more likely to report loyalty strategies (e.g., doing nothing) in response to interpersonal conflict, while younger adults were more likely to report exit strategies (e.g., yelling).Similarly, Diehl, Coyle, and Labouvie-Vief (1996) observed that older people demonstrated more impulse control, less outward aggression, and more positive appraisals of conflict situations than younger people.

Alternative Explanations
We included a group of standard control variables in this study because of the possibility that they may cause or contribute to both the occurrence of negative interactions and risk for hypertension.These included demographic characteristics (age, sex, marital status, race and ethnicity, education, employment status), and markers of baseline health (history of chronic illnesses, baseline systolic and diastolic blood pressure).We also considered the possibility that associations between negative interactions and health may be attributable to personality characteristics that contribute both to the quality of interactions and to health outcomes.For example, low extraversion and agreeableness and high hostility and neuroticism have been associated with both higher levels of conflictive social relationships (Berry, Willingham, & Thayer, 2000;Brondolo et al., 2003;Lincoln, 2008) and dysregulated cardiovascular function (extraversion : Miller, Cohen, Rabin, Skoner, & Doyle, 1999;Shipley, Weiss, Der, Taylor, & Deary, 2007;agreeableness: Miller et al., 1999;hostility: Suls & Bunde, 2005;Chida & Steptoe, 2009;Tindle et al., 2009;neuroticism: Shipley et al., 2007).
Finally, those with higher levels of negative interactions are likely to also have lower levels of positive ones (e.g., Okun & Keith, 1998).This inverse correlation leaves the possibility that what appears to be an association with more negative experiences may in fact be attributable to having fewer positive ones.Thus, we conducted additional analyses to evaluate the association between negative social interactions and hypertension controlling for positive interactions.

Participants and Design
This study was a secondary analysis of data from the 2006 and 2010 waves of the Health and Retirement Study (HRS), a largescale longitudinal study of community-dwelling older adults (age Ͼ 50 years).The HRS sampling methods and study design have been previously documented (Heeringa, & Connor, 1995;Juster & Suzman, 1995).Briefly, HRS uses a national area probability sample of U.S. households; the sample includes individuals aged greater than 50 years and (when applicable) their partners.A total of 18,469 individuals provided data at baseline (2006 wave).Fifty percent of this sample was randomly selected to participate in enhanced face-to-face interviews, including questions assessing demographics, health status, health behaviors, negative interac-tions, and psychological well-being.The interview period also included a blood pressure assessment.Of participants invited to be interviewed, 7,144 provided interview and blood pressure data at baseline (2006 wave).Of these, 6,817 also provided blood pressure data at the 4-year follow-up.The mean follow-up time was 50.18 months (SD ϭ 4.06; range: 39 -61).
From the sample with blood pressure data at both assessments, we excluded all participants who were hypertensive at baseline, which included those using antihypertensive medications (n ϭ 3,778) and those with baseline blood pressure readings in the hypertensive range (average resting systolic blood pressure Ն140 mmHg or average resting diastolic blood pressure Ն90 mmHg; n ϭ 1,023).Then, we excluded 371 individuals with missing data on at least one of our control variables and 1 individual with incomplete negative interaction data.Finally, we excluded 142 individuals who died during the follow-up period.Our final sample included 1,502 participants (see Table 1) who were 84.8% non-Hispanic White, 6.8% Hispanic, 6.5% non-Hispanic Black, and 1.9% other racial or ethnic backgrounds.Participants were 59.8% female and ranged from 51-91 years in age at baseline (M age ϭ 64.28 years, SD ϭ 8.95).When comparing our sample with those who provided blood pressure data during the 2006 wave but were excluded, our sample tended to be younger, employed, more educated, more likely to be nonsmokers, and married (see Table 1).These variables are all well-established associates of hypertension, the screening variable responsible for over 90% of those not meeting criteria for inclusion in our analysis.Our sample included a mixture of individuals (80.6%) and of couples (19.4%) who both participated in the study.

Negative Social Interactions
Negative social interactions were assessed in the selfadministered psychosocial questionnaire across four domains: relationships with spouse/partner, children, other family, and friends (adapted from Krause, 1995).Four questions were used to evaluate negative interactions in each domain: (a) "How often do they make too many demands on you?";(b) "How much do they criticize you?"; (c) "How much do they let you down when you are counting on them?"; and (d) "How much do they get on your nerves?"Responses for each question were coded on a 4-point scale ranging from 1 (not at all) to 4 (a lot).We calculated mean scores for each of the four domains (friends: n ϭ 1,437, M ϭ 1.43, SD ϭ 0.48, range: 1-4; partner: n ϭ 1,141, M ϭ 1.98, SD ϭ 0.66, range: 1-4; children: n ϭ 1,360, M ϭ 1.75, SD ϭ 0.63, range: 1-4; other family: n ϭ 1,441, M ϭ 1.62, SD ϭ 0.62, range: 1-4) by averaging across item scores within each domain.A domain score was set to missing if more than two items had missing values in accordance with the scoring guidelines established by the HRS coordinating center (Clarke, Fisher, House, Smith, & Weir, 2008).To create a total average negative interaction score (range: 1-4), we averaged the scores across the four domains.Total scores were calculated using only scored domains, and only for those with a score for at least one of the four domains.This method for calculating an average score for this questionnaire has been commonly used in studies of negative social interactions and health (Friedman, Karlamangla, Almeida, & Seeman, 2012;Newsom, Mahan, Rook, & Krause, 2008;Seeman, Berkman, Blazer, & Rowe, 1994;Tun, Miller-Martinez, Lachman, & Seeman, 2013) and reflects the average level of negativity across interaction domains.

Outcome Measures
Blood pressure assessments were performed in both 2006 and 2010 by study staff who underwent repeated training (Crimmins et al., 2008).Resting blood pressure values were based on the average of the three blood pressure measurements, 45 s apart, taken on the respondent's left arm using an automated blood pressure monitor (Crimmins et al., 2008).Hypertension was defined as either self-reported use of antihypertensive medications, or an average resting systolic blood pressure of 140 mmHg or higher, or an average resting diastolic blood pressure of 90 mmHg or higher (Chobanian et al., 2003).

Control Variables
Demographics, baseline blood pressure and health, personality variables, and positive social interactions were included as "standard" controls in all analyses.Hours of volunteering (related to hypertension in previous analysis of this data set; Sneed & Cohen, 2013), also assessed at baseline, were added in separate analyses.All categorical breakdowns were based on the original categories from the HRS dataset and were dummy coded in the analyses.
Baseline blood pressure assessments are described in the section Outcome Measures.
Demographic controls included age (continuous), sex (male, female), self-reported race (non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic other), education (less than high school, high school diploma, general equivalency diploma, some college, college and above), marital status (married, annulled, never married, divorced, separated, widowed), and employment status (employed, not employed).
Self-reported history of illness included seven separate dummycoded variables evaluating history of diabetes (yes, no), cancer (yes, no), heart problems (yes, no), stroke (yes, no), memory problems (yes, no), arthritis (yes, no), and lung problems (yes, no).
Personality characteristics included cynical hostility, neuroticism, extraversion, and agreeableness.All were continuous variables assessed at baseline using a self-report questionnaire.Cynical hostility was measured by five items from the Cook-Medley Hostility Inventory (Cook & Medley, 1954) as per HRS protocol.Participants rated the extent to which they agreed with each item on a scale from 1 (strongly disagree) to 6 (strongly agree).An index of cynical hostility was created by averaging the scores across all items.Neuroticism, extraversion, and agreeableness were assessed using 16 adjectives (six for neuroticism, five for extraversion, five for agreeableness) from the Midlife Development Inventory Personality scales (Lachman & Weaver, 1997).Participants rated the extent to which each adjective described them on a scale ranging from 1 (not at all) to 4 (a lot).Scores for each of these personality characteristics were created by averaging the scores across the corresponding items.
Positive social interactions were assessed as social support and measured in each of the four social domains (partner, children, other family, friends).Within each domain, participants answered the following three items on a scale from 1 (never) to 4 (a lot): (a) "How much do they really understand the way you feel about things?";(b) "How much can you rely on them if you have a serious problem?"; and (c) "How much can you open up to them if you need to talk about your worries?"We created a total average positive social interaction score by averaging responses to these items across domains.Total scores were calculated using only scored domains, and only for those with a score for at least one of the four domains.
Finally, we assessed self-reported hours of volunteer work in the 12 months prior to baseline (none, 1Ϫ49, 50Ϫ99, 100Ϫ199, Ն 200 hr).The categories for volunteerism were preestablished by HRS study staff.

Potential Mediating Variables
The following variables were assessed at baseline and follow-up and were tested as potential mediating variables: alcohol use, tobacco use, physical activity, body mass index, and nine measures of psychological well-being.
Alcohol use (Witteman et al., 1990), tobacco use (Bowman et al., 2007;Halperin et al., 2008), and physical activity (Paffenbarger et al., 1983) are all typically associated with blood pressure.The measures used for evaluating these variables were based on standard instruments typically used in large-scale epidemiological studies.Alcohol use was based on participant responses to the following three questions: (a) "Do you ever drink any alcoholic beverages such as beer, wine, or liquor?";(b) "In the last three months, on average, how many days per week have you had any alcohol to drink?"; and (c) "In the last three months, on the days you drink, about how many drinks do you have?"The average number of drinks per week for each participant was determined by multiplying average number of days per week drinking by average number of drinks per day.Similarly, tobacco use was based on participant responses to the following two questions: (a) "Do you smoke cigarettes now?"; and (b) "About how many cigarettes or packs do you usually smoke in a day now?"Responses were used to calculate the average number of cigarettes per day for each participant, with each pack of cigarettes corresponding to 20 cigarettes.
Physical activity was assessed with questions about frequency of participation in vigorous or moderate sports or activities using these categories: more than once a week, once a week, one to three times a month, or hardly ever/never.Vigorous activities included the following examples: running, jogging, swimming, cycling, aerobics, a gym workout, tennis, or digging with a spade or shovel.Moderate activities included the following examples: gardening, cleaning the car, walking at a moderate pace, dancing, or floor or stretching exercises.We used a summary physical activity variable previously used in the English Longitudinal Study of Aging (Mc-Munn, Hyde, Janevic, & Kumari, 2003) to organize vigorous and moderate physical activity into five ordinal categories ranging from 0 (sedentary) to 4 (active).
We also included body mass index as a potential mediator because it is a marker of adiposity (Keys, Fidanza, Karvonen, Kimura, & Taylor, 1972) and a risk factor for adverse cardiovascular outcomes (Wilson, D'Agostino, Sullivan, Parise, & Kannel, 2002).We hypothesized that negative social interactions may cause changes in health behaviors (e.g., poor diet, physical inactivity) that might lead to greater adiposity.Body mass index was calculated as weight in kilograms divided by height in square meters.
Standard scales representing nine individual psychological constructs (see Table 2) were used to measure psychological wellbeing at baseline and 4-year follow-up: personal control, purpose in life, life satisfaction, positive affect, optimism, loneliness, hopelessness, negative affect, and pessimism.Scores on each measure were determined by averaging the scores across the individual items.Scores were set to missing if more than 50% of individual items for each respective measure had missing values.

Statistical Analyses
Our primary analytic strategy focused on evaluating the association between total average negative social interactions and hypertension risk, and on determining the extent to which age and sex modified the association between negative social interactions and hypertension.A secondary objective was to identify factors that might mediate the association between negative social interactions and hypertension.Finally, we performed additional exploratory analyses based on our initial primary analyses to better understand associations between total average negative social interactions and hypertension risk.
Logistic regression was used to evaluate the relationship between total average negative social interactions (across all four domains) and hypertension.Odds ratios (ORs) and 95% confidence intervals (CIs) were used to estimate the risk for each 1-unit increase in the total average negative interaction score.We also calculated domain-specific ORs and 95% CIs to determine if effects were driven by any particular domain.All ORs reflect a change in the likelihood of developing hypertension for every 1-unit increase in the negative interactions scale (range: 1Ϫ4).
To evaluate the extent to which the potential mediating variables linked negative social interactions with hypertension, we used the PROCESS macro (Hayes, 2013), a computational tool for path analysisϪbased mediation analysis with dichotomous outcomes.This analytical method uses a logistic regressionϪbased approach to estimate direct and indirect effects in multiple mediator models.Bootstrap methods are used to draw inferences about indirect effects.We standardized the average scores for all of the psychological variables, and tested for mediation in three ways: (a) using baseline values of each individual mediator, (b) using values of each individual mediator at follow-up, and (c) evaluating residual change (entering both baseline and follow-up) in each potential mediator from baseline to follow-up.With each approach, the potential mediators were entered into the model both individually and simultaneously.Mediation was supported if addition of these covariates substantially reduced the association of negative social interactions with hypertension (Sobel, 1982).To determine if either age or sex moderated the associations observed between negative social interactions (across all four domains and within each domain) and hypertension, we used first-order cross-product terms for total average negative social interactions and these proposed modifier variables.Interaction terms were entered into individual regression equations with the corresponding main effects and control variables.
Finally, because our sample included a mixture of individuals and couples, it is possible that interdependence of data for partners might bias the results.To address this, we reanalyzed our data, including all individual participants in addition to a randomly selected member of each couple (selected using a random sequence generator).This resulted in a reduction of power (decreased sample size from 1,502Ϫ1,356), but eliminated the possibility of correlated partner responses.We then repeated the analyses using the other member of each couple in the sample.

Control Variables and Hypertension
Of the 1,502 study participants included in our analyses, 445 (29.6%) were hypertensive at follow-up.When entered into the logistic regression model simultaneously, the following standard control variables were related to increased hypertension risk: older age (B ϭ 0.03, p ϭ Ͻ .001),Hispanic ethnicity compared with non-Hispanic White ethnicity (B ϭ 0.79, p ϭ .001),self-reported history of diabetes (B ϭ 0.52, p ϭ .008),greater average baseline systolic (B ϭ 0.04, p Ͻ .001) and diastolic blood pressure (B ϭ 0.03, p ϭ .003),and higher levels of agreeableness (B ϭ 0.38, p ϭ .03).Having less than a high school education was associated with decreased hypertension risk (B ϭ Ϫ0.75, p ϭ .04).The other covariates were not related to hypertension risk.

Negative Interactions and Hypertension
The mean total average negative interaction score for study participants was 1.67 (SD ϭ 0.45).In a regression including the standard covariates, total average negative interactions across the four domains predicted greater hypertension risk, OR ϭ 1.38; 95% CI [1.00, 1.89] (see Table 3).Given that hours of volunteer work were related to hypertension risk in a previous analysis of this data set (Sneed & Cohen, 2013), we performed additional analyses simultaneously entering volunteerism and total average negative social interactions into a regression model, adjusting for our standard controls.Here, total average negative social interactions were independently associated with greater hypertension risk, OR ϭ 1.39; 95% CI [1.01,1.91].Further, even after adjusting for negative interactions, those who volunteered at least 200 hours per year were less likely to develop hypertension than nonvolunteers, OR ϭ 0.57; 95% CI [0.36,0.89].
We were also interested in whether the association between negative interactions and hypertension was driven by negative interactions in any particular social domain.Domain-specific analyses demonstrated that negative interactions were associated with increased hypertension risk when examining relationships with one's friends, OR ϭ 1.36; 95% CI [1.04, 1.78], and family other than children or partner, OR ϭ 1.33; 95% CI [1.07, 1.65] (see Table 3).There were no main effects of negative interactions with children, OR ϭ 1.01; 95% CI [0.80, 1.28], or partner, OR ϭ 1.15; 95% CI [0.90,1.48] on hypertension (see Table 3).

Sex and Age as Modifiers
We also evaluated how sex and age might modify the association between total average negative interactions and hypertension risk.We found that sex moderated the association between the total average negative interaction score and hypertension (B ϭ 0.7, p ϭ .02)(see Figure 1).The relationship between the total average negative interaction score and hypertension was not observed among men, OR ϭ 0.88; 95% CI [0.50, 1.53], but was pronounced among women, OR ϭ 1.87; 95% CI [1.25,2.79].In domainspecific analyses, we observed that sex moderated the association between negative interactions with friends and hypertension (B ϭ 0.93, p Ͻ .001).Among men, there was no association between negative friend interactions and hypertension, OR ϭ 0.69; 95% CI [0.43, 1.12].Among women, however, negative interactions with friends predicted increased hypertension risk, OR ϭ 2.15; 95% CI [1.51, 3.08].Similar, but nonsignificant patterns were found for the interactions of sex and negative interactions with partner (B ϭ 0.31, p ϭ .15)and family (B ϭ 0.37, p ϭ .07),but not children (B ϭ 0.11, p ϭ .61).
Age also moderated the association between total average negative social interactions and hypertension (B ϭ Ϫ0.03, p ϭ .02).There was no association between total average negative social interactions and hypertension among participants aged 65 years or Note.All models adjust for age, race and ethnicity, sex, employment status, marital status, education, baseline systolic blood pressure, baseline diastolic blood pressure, extraversion, agreeableness, cynical hostility, neuroticism, and self-reported history of diabetes, cancer, heart problems, arthritis, memory problems, lung problems, or stroke.OR ϭ odds ratio; CI ϭ confidence interval.
older, OR ϭ 1.16; 95% CI [0.72, 1.86].Among those ages 51-64 years, however, total average negative social interactions were associated with increased hypertension risk, OR ϭ 1.67; 95% CI [1.06, 2.62].In domain-specific analyses, we observed that age moderated the association between negative partner interactions and hypertension (B ϭ Ϫ0.03, p Ͻ .01).Further analyses suggested that the effects of negative partner interactions were most potent for the youngest participants.Similar, but nonsignificant patterns were found for the interactions of sex and negative interactions with children (B ϭ Ϫ0.02, p ϭ .08)and family (B ϭ Ϫ0.12, p ϭ .18),but not friends (B ϭ Ϫ0.006, p ϭ .71).Finally, we tested the three-way interaction of age, sex, and negative interactions by entering it into a model with all the main effects and two-way interactions.There was no three-way interaction of these variables in predicting hypertension (B ϭ 0.009, p ϭ .31).
We considered the possibility that blood pressure measurements from study participants might have been artificially inflated due to the white coat effect (i.e., spuriously high blood pressure readings in response to the clinical environment; Mancia et al., 1983;Mancia et al., 1987).To account for this, we repeated our main analyses by dropping the first blood pressure reading and averaging the last two readings.In doing so, there was no longer a main effect of total average negative interactions, OR ϭ 1.15; 95% CI [0.83, 1.59], or friend interactions, OR ϭ 0.81; 95% CI [0.61, 1.06].The remaining results were the same for negative family interactions, OR ϭ 0.80; 95% [0.64, 1.00], and there was still no association between negative partner, OR ϭ 0.87; 95% CI [0.68, 1.11], or child, OR ϭ 1.07; 95% CI [0.85, 1.36], interactions.The effects of negative social interactions were still similarly modified by sex (B ϭ 0.66, p ϭ .03)and age (B ϭ Ϫ0.03, p ϭ .03).

Accounting for Partners in Same Analysis
To address the issue of interdependence of data for partners, we reanalyzed our data including all individual participants in addition to a randomly selected member of each couple.We observed virtually identical results, with total average negative interactions associated with increased hypertension risk, OR ϭ 1.42; 95% CI

Mediation
Finally, we explored factors that might mediate the association between negative interactions and hypertension risk, conducting separate mediation analyses for the entire sample, for women, and for participants ages 51-64 years.There were no indirect effects of negative social interactions on hypertension through any of our potential mediators (body mass index, alcohol use, tobacco use, physical activity, or the nine individual measures of psychological well-being).

Discussion
We found that total average negative social interaction scores were associated with increased hypertension risk in a sample of community-dwelling adults over 50 years of age.Specifically, each 1-unit increase in an individual's score resulted in a 38% increased odds of hypertension over a 4-year follow-up.This association persisted even after controlling for demographics, personality variables, and positive social interactions.Of note, this association was independent of volunteerism, which had been associated with decreased hypertension risk in a previous analysis of this dataset (Sneed & Cohen, 2013).Both negative social interactions and volunteerism exerted distinct, independent effects on hypertension risk when evaluated simultaneously.Our findings are consistent with literature linking acute negative social interactions to short-term elevated blood pressure responses in younger adults (Ewart et al., 1991;Kiecolt-Glaser & Newton, 2001;Smith et al., 2013) and suggest that negative interactions may contribute to long-term alterations in blood pressure regulation.
Total average negative interaction scores predicted hypertension risk among women, but not among men.This association was driven primarily by women's negative interactions with friends and, to a lesser degree, partners and family.There are several possible reasons why women but not men were more impacted by negative relationships.There is evidence that women care more about and pay more attention to their social interactions (Coriell & Cohen, 1995) and have greater expectations of their social relationships than men.In a study of older adults ages 50 -97 years, Felmlee and Muraco (2009) observed that women demonstrated greater disapproval of behavior that violated friendship rules, such as betrayal of confidence, failure to confide in them, or not standing up for them when someone criticized them.These greater expectations may lead to greater distress among women when such relationship expectations are not met.Further, women have reported more intimacy, closeness, and self-disclosure in their social relationships than men (Sheets & Lugar, 2005;Singleton & Vacca, 2007).This greater intimacy may intensify reactions to conflict when it arises (Crick, 1995;Crick & Grotpeter, 1995).Finally, women and men often have different strategies for responding to and resolving conflict.In laboratory marital conflict discussions, for example, women have been more likely to confront the conflict directly, whereas men tended to withdraw (Carstensen et al., 1995).This difference in conflict management style may also have implications for cardiovascular response to aversive social situations.
We also found that age moderated the association between negative social interactions and hypertension risk, with effects observed among younger (51Ϫ64 years) but not older (Ն 65 years) participants.These findings suggest that negative social interactions are particularly potent for the "sandwiched generation," those within the later years of midlife who typically bear responsibility for both their own children as well as aging parents.Negative social interactions may only exacerbate the adverse effects of existing life stressors among individuals in this age group.Our observation that negative social interactions were not related to hypertension risk among the oldest participants (Ն 65 years) is consistent with evidence that, as people get older, they may adapt to negative aspects of their relationships and perceive them as less problematic (Akiyama et al., 2003).It is also possible that the oldest adults have more adaptive strategies for handling interpersonal difficulties that render such interactions less potent with respect to hypertension risk.
We also found no main or interaction effects linking negative interactions with children to hypertension.This is consistent with recent evidence from the MacArthur Study on Successful Aging that regular interactions with spouse and other family members were associated with pulmonary health in older adults, but interactions with children were not (Crittenden, Pressman, Cohen, Smith, & Seeman, 2013).It is possible that individuals in this age group have lower expectations of their children than they do of other members of their social networks.Studies have shown that adult children are willing to provide more support to their parents than their parents expected (Hamon & Blieszner, 1990;Silverstein, Chen, & Heller, 1996).Parents may also have schemas about the nature of the parentϪchild relationship that lead them to expect more difficult relationships with their children.They may also have greater insight into their children's behaviors and use this to justify negative interactions in ways that reduce psychological distress.Finally, parents may simply have less face-to-face contact (or contact of any kind) with their children than with other members of their social networks, thus limiting the potential impact of any negative social interactions.
We found no psychological or behavioral explanations for the association between negative interactions and hypertension risk.It is possible that our findings may be explained by factors that were not measured in this study.For example, negative social interactions may increase maladaptive health behaviors linked to hypertension risk, such as poor sleep quality (Gangwisch et al., 2006) or unhealthy nutritional habits (Reddy & Katan, 2004).Our measure of physical activity also may not have been ideal for evaluating activity levels among participants.Self-report questionnaires gauging moderate and vigorous activity may not be adequate measures of physical activity among elderly individuals, because most physical activity performed by older adults involves walking in the context of regular daily activities rather than a formal exercise regimen (Walsh, Pressman, Cauley, & Browner, 2001).More sensitive, objective measures of physical activity (e.g., accelerometry) might be more useful in this population.There may also be physiological explanations for our findings.For example, acute conflictive social interactions are associated with short-term increases in blood pressure, greater parasympathetic withdrawal, activation of the sympathetic nervous system and the hypothalam-icϪpituitaryϪadrenal axis, and increased production of inflammatory cytokines (Kiecolt-Glaser & Newton, 2001).Sustained negative social interactions may cause long-term wear and tear in these physiological systems (allostatic load), leading to an inability to effectively regulate blood pressure (McEwen, 1998).Future research should explore how additional behavioral factors (e.g., sleep habits, physical activity, diet) and physiological indicators (e.g., increased sympathetic nervous system activity) may mediate the association between negative social interactions and hypertension.
This study is not without limitations.Although the study sample is drawn from a larger, nationally representative sample of older adults, the individuals who met the inclusion criteria for this study were, in some ways, demographically different from the larger HRS sample.Consequently, we did not employ the weighting and sampling design adjustments often used to analyze HRS data (Heeringa & Connor, 1995).Thus, these findings should not be interpreted as being representative of the U.S. population of older adults.Our findings, however, are consistent with other work linking negative social interactions to adverse health outcomes (De Vogli, Chandola, & Marmot, 2007;Krause, 2005).Further, the main effects of negative social interactions were no longer significant when we reanalyzed the data by dropping the first blood pressure reading.These analyses, however, reinforce the importance of negative social interactions with family (irrespective of participant sex) and suggest that negative interactions with friends mostly matter for women.
The negative interaction questionnaire assesses four social domains (partner, friends, children, other family members) that are all traditionally considered to be close relationships.People maintain fewer peripheral relationships as they grow older (e.g., Fung, Carstensen, & Lang, 2001), devoting more attention to relationships with close friends and family.Hence, the measure used here reflects the largest and closest components of the social network in this age group, making negative social interactions in these domains particularly salient.
This study is the first to prospectively evaluate negative social interactions as a predictor of the onset of hypertension.The negative interactions reflect experiences in natural social networks.The work includes controls for multiple alternative explanations for the results, including the first test of the role of stable individual differences that could contribute to both the occurrence of negative interactions and the onset of hypertension.It is also the first study of negative interactions and hypertension to control for the potential overlap with positive social interactions.Finally, it suggests that a range of predicted mediators do not play a role in this association for this older sample.
Our sample included individuals ages 51-91 years at baseline (M age ϭ 64.28 years, SD ϭ 8.95).Hypertension risk increases with age, with rates reaching 70% by age 65 (McDonald, Hertz, Unger, & Lustik, 2009).Thus, individuals from our sample were particularly at risk for developing hypertension during the follow-up period.In fact, 29% of the sample developed hypertension over the 4-year follow-up.Because hypertension is a major risk factor for diseases of aging, including cardiovascular disease (the number one cause of death in the United States), stroke, and mortality, the association of negative social interactions with increased hypertension risk has important public health significance.Further, interpersonal conflict is one of the most frequently reported types of chronic stress; thus, it is particularly important to understand the role of interpersonal strain in health outcomes such as hypertension.

Figure 1 .
Figure 1.Association of total average negative social interaction scores with hypertension by sex.

Table 1
Descriptive Statistics of the Study Sample a Values are mean (standard deviation).b Values are n (percent).

Table 3
Logistic Regression for Association of Negative Social Interactions With Hypertension Across Domains and Within Each Social Domain