Takes Two to Tango: Cognitive Impairment and Sexual Activity in Older Individuals and Dyads

Abstract

Objectives

This article examines the association between cognitive impairment, including mild cognitive impairment (MCI) and early dementia (ED), and sexual activity in a large, longitudinal sample of community-dwelling older adults. We focus here on sexual activity, which includes both sexual activity with a partner and masturbation.

Methods

We analyzed 3,777 older individuals and members of 955 intimate dyads using the National Social Life, Health, and Aging Project (2010 and 2015). We used ordered probit regression, cross-lagged panels models, and probit regression.

Results

We find that older adults with cognitive limitations, either MCI or ED, are about as likely to be sexually active with a partner as those with normal cognitive function. This is the case both in the cross-section and longitudinally. Both men and women with ED are less likely to have masturbated, however. Among married and cohabiting couples, we find no association between cognitive impairment in either the husband, the wife, or in both partners and their joint sexual activity. Women whose measured cognitive function is lower than their husbands are less likely to report any masturbation.

Discussion

Sex with a partner, a fundamentally social activity, seems to be conserved in the face of cognitive limitation but masturbation, a solitary activity, does not. We argue that the intimate dyad plays an important role in conserving partnered sexual activity. Results can inform strategies to maintain the sexual well-being of older adults with cognitive impairment and their partners as part of overall strategies to improve quality of life.

Cognitive impairment, including Alzheimer’s disease and related dementias, carries enormous medical and financial costs (Allen, 2018; Hurd et al., 2013); and Alzheimer’s disease is now the sixth leading cause of death in the United States (Murphy et al., 2017). The severity of cognitive impairment is traditionally defined based on an individual’s ability to independently manage daily activities (Petersen, 2004), including self-care, household chores, managing medications, and handling finances. However, increasing attention has been focused on social activities, including the ability to socialize with friends or maintain networks of social connections, which can be affected at earlier stages of cognitive decline, such as mild cognitive impairment (MCI) or early dementia (Amano et al., 2020; Kotwal et al., 2016). Social functioning is strongly associated with quality of life, independence in the community, functional impairment, and death among older adults. Yet, we know relatively little about the impact of different stages of cognitive impairment on a domain of social and emotional functioning central to the health of many older adults: sexuality (Kotwal et al., 2016).

Virtually all older men and the majority of older women consider sex an important part of their life and about the same share think about sex relatively frequently (Waite et al., 2009). Among those with a partner, the substantial majority of older men and about half of older women are sexually active, with the proportion active declining with age and poor health (Lindau et al., 2007). Sexuality among older adults varies by gender. This is true regardless of the dimension of sexuality considered; men are more likely to have a sexual partner, more likely to be active with that partner, express greater interest in sex, and are more likely to masturbate than are women of the same age (Iveniuk & Waite, 2018; Waite et al., 2009). Older men also show higher levels of both proceptive and receptive sexuality than women do. Proceptive sexuality leads a person to seek out a sexual partner, whereas receptive sexuality increases willingness to have sex when asked (Galinsky et al., 2014). Sexual scripts that include passive roles for women in sexuality may be a part of the explanation for this difference (Waite & Iveniuk, 2021).

Sexual function is a critical, often neglected, domain of social and emotional functioning among older adults. Sexuality is a source of emotional intimacy and connection, which are desirable in and of themselves, and positively affects marital quality and general well-being (Galinsky & Waite, 2014; Lee et al., 2016; Schnarch, 1991). In addition, there are indirect benefits of sexual activity. Sex with a partner is a form of exercise, which involves stretching, flexibility, increased heart rate, and release of hormones (Frappier et al., 2013; Liu et al., 2016). Exercise physiologists also point to the laudatory effect of sex on circulation and metabolism (Butt, 1990). Sex may also act to reduce stress or to aid recovery from stress, through the mechanisms just mentioned. Taken together, a better understanding of the link between sexual activity and early stages of cognitive impairment may inform efforts to understand well-being and preserve quality of life among older adults with cognitive impairment.

Cognitive Limitations and Social Functioning: Previous Empirical Evidence

A large body of literature over the last several decades has examined the link between social and cognitive functioning. Most research done on this relationship is either cross-sectional or examines the association of social behavior with incident dementia. This literature generally shows a positive association of some social behavior, including community engagement and socializing with friends and family, with at least some dimensions of cognition in the cross-section (Kotwal et al., 2016; Lindau et al., 2018; Wright & Jenks, 2016; Wright et al., 2019). A meta-analysis of longitudinal cohort studies (Kuiper et al., 2015) concludes that low social participation, less frequent social contact, and higher levels of loneliness are significantly associated with incident dementia. Similarly, a recent article by Liu et al. (2020) finds that being married reduces the risk of incident dementia.

Those with MCI and early dementia often retain the ability to carry out many ordinary day-to-day activities (Domínguez-Chávez et al., 2019), but changes to social function can be subtle. Jiang and Xu (2014) find very small differences in frequency of contact with family and friends between those with MCI and those with normal cognitive function. The substantial majority of older adults with MCI frequently talk on the phone and do activities with children (Amano et al., 2020), but two thirds of these adults show very low participation in formal volunteering or meetings of groups.

Within intimate relationships, Wong and Hsieh (2019) find, in a cross-sectional study, the interdependence of older partners’ social lives when confronted with a spouse’s functional or cognitive limitations, and that patterns of disruption vary by gender. For example, cognitive impairment was not associated with marital strain among either partner. However, cognitive impairment among husbands was associated with wives having more perceived friend and family support. Hawkley et al. (2020) found partners’ cognitive limitations to be associated with increased anxiety and perceived stress among noncognitively impaired spouses as they transition to a caregiver role. Taken together, prior literature demonstrates different associations of cognition with social relationships depending on the domain of relationship and gender. The relationship of cognition to sexuality may most closely mirror prior findings within intimate relationships; close family relationships, including partnered sexual behavior between spouses, may be maintained in the face of cognitive limitations. To provide an account of this, we now turn to a theoretical model that grounds sexuality and health in a dyadic context.

Theoretical Framework Linking Cognitive Impairment and Sexuality

The Interactive Biopsychosocial Model (Lindau et al., 2003) posits that well-being is produced in the dyad or social group as a function of biophysical, psychocognitive, and social resources of the actors. In this model, shown in Figure 1, these resources can be exchanged across domains and between partners and can be combined to produce individual and dyadic outcomes. This general model can be applied to specific dimensions of each of these resources and specific components of health. We focus here on cognitive function, a component of psychocognitive capital that constitutes a resource that each partner brings to the dyad, and sexuality, a dimension of social well-being.

Figure 1.

Biopsychosocial model of cognitive function and sexuality.

In the Interactive Biopsychosocial Model, cognitive capital of one member of the dyad is shown in the orange circle in the lower-left corner of the figure and cognitive capital of the partner, subscript p, is shown in dark blue in the lower-right corner of the figure. Each partner’s overall health is represented in the semicircle segments on each side of the figure. In this article, our focus on individual health is the sexual function of each person (Waite, 2018). Resources are combined to produce partnered sexual activity, as represented in the circle in the middle of the figure. Sexual activity and functioning are determined by the interaction of each partner’s characteristics and are further shaped by the quality and condition of the intimate dyad, the social relationships that surround the dyad (Iveniuk & Waite, 2018), and more broadly, the social environment in which the relationship occurs (Waite & Iveniuk, 2021).

All these processes are likely to be profoundly affected by gender. Gender script theory asserts that sexual scripts assign different roles to men and women in intimate dyads, with men more often taught to express their sexuality and women to suppress theirs (Wiederman, 2005). These scripts may make sexual activity, including continued sexual activity at older ages, important for the maintenance of a masculine self-image but less central to successful aging for a woman (Carpenter, 2015). Some evidence suggests, for example, that older women place more importance on remaining attractive to their spouse than men do, whereas unpartnered men put more effort into being attractive for a potential new partner than do women (Waite & Iveniuk, 2021).

In the Interactive Biopsychosocial Model, cognitive impairment in either partner may impair the sexual functioning of the dyad—shown inside the middle circle. For example, this could occur through the association of cognitive impairment with reduced biophysical capital, including physical mobility (Montero-Odasso et al., 2012; Pedersen et al., 2014), or sensory functioning (Zhong et al., 2018). Cognitive limitations can increase difficulty with planning, sequencing, and remembering tasks (Jiang & Xu, 2014; Pedersen et al., 2014), such as carrying out a sexual encounter. Perceived stress associated with subclinical cerebrovascular disease reduces psychocognitive capital (Aggarwal et al., 2014). Apathy and other emotional changes, common among older adults with developing cognitive impairment, might affect sexual interest (Apostolova & Cummings, 2008). Cognitive impairment could also affect sexuality by influencing social functioning (Amano et al., 2020; Kotwal et al., 2016) or the dynamics of the partnered relationship. For example, partners may feel less comfortable with engaging in sexual activity with a spouse experiencing cognitive impairment due to changes in how they engage with one another or questions of consent. Such discordances in cognitive health among partners may lead unimpaired partners to engage in masturbation.

Importantly, the Interactive Biopsychosocial Model accounts for the central role of partners in sexual health. Partners, through their own biophysical, psychocognitive, or social capital, may help individuals compensate for cognitive decline to maintain sexual function in the dyad. Intimate relationships may build longstanding feelings of trust, help with planning, overcome emotional changes, and help to maintain enjoyment in sexual activities. Moreover, various levels of cognitive impairment may affect social functioning differently, so that sexual interest is preserved for many individuals, especially at earlier stages of impairment such as MCI. At later stages of cognitive impairment, sexual interest may be more affected, along with the partners’ ability to help individuals compensate. Partnered sexual activity, in contrast to masturbation, is fundamentally social and may protect against poor health outcomes (Galinsky & Waite, 2014). A sizeable literature shows that intercourse is more physiologically and psychologically satisfying than masturbation. Brody and Krüger (2006) report that the release of prolactin, an indicator of sexual satiety, is 400% greater following intercourse than following masturbation. Brody et al. (2003) link frequency of sexual intercourse—but not other sexual activities—to better psychological well-being. Cognitive impairment may consequently affect partnered versus nonpartnered forms of sexuality differently, which may lead to different downstream health consequences.

The Present Study

We examine the association between cognitive impairment, specifically MCI and early dementia, and partnered versus solitary sexual activity in older adults. While severe cognitive impairment may compromise older adults’ sexual lives, older adults with early or mild cognitive deficits may still be able to carry out many activities, which may include sexual functioning. In addition, we consider sex with a partner and masturbation separately. Sex with a partner requires inputs from two people with an intimate, often longstanding connection and sexual relationship. Such partnerships may be able to continue to function well in spite of deficits in some of the resources of one of the partners. Comparison between partnered sex and masturbation allows us to test our hypothesis that couples draw on the resources of both partners to produce a shared sexual experience, whereas resources of only one person are drawn on for masturbation. Put differently, we would expect to see partner effects when examining partnered sex, but none when considering masturbation. This analysis provides insight into the role of a long-functioning dyad in preserving sexual function in older adulthood. This article therefore examines the association between cognitive impairment, including MCI and early dementia, and sexuality in a large, longitudinal sample of community-dwelling older adults. We focus here on sexual activity, which includes both sexual activity with a partner and masturbation. We ask the following research questions:

• (1) Are older adults with cognitive impairment less sexually active than those with normal cognition? Or, is sexuality conserved in the face of cognitive decline? Does this differ for partnered versus solitary sexual activity? Does this differ by gender?

• (2) Is this true for both dyadic sexuality and individual sexual activity (i.e., masturbation)? Is it true for both men and women?

• (3) Does normal cognition in one partner protect older couples from lower sexual activity in the face of cognitive limitations by the other partner? Does the gender matter?

Method

Data

The data from this study come from two waves of the National Social Life, Health, and Aging Project (hereafter, NSHAP), a nationally representative probability sample of older Americans. In the 2010/2011 wave (W2), and 2010 hereafter, and the 2015/2016 wave (W3) primary respondents and their spouses or cohabiting partners were administered the same survey protocol. This allows us to link the two partners to each other and to examine the effect of each person’s characteristics on own and partner’s health and well-being. Furthermore, the longitudinal nature of NSHAP allows both couples and individuals to be followed up over time.

NSHAP also benefits from a large sample size both cross-sectionally and longitudinally. In 2010, 3,777 persons took part in the study, and 1,910 of these persons were in coupled relationships so that 955 couples with both partners were represented. From W2 to W3, 2,409/3,377 (71.3%) persons were interviewed in both waves. Individuals judged by the interviewer to be unable to complete the interview, either because of physical or cognitive limitations, were not interviewed. The study also included oversamples for key groups of interest, including non-White older adults, the oldest old, and rural populations, in order to ensure adequate statistical power for making generalizations to these groups. Individuals ranged in age from 36 to 99 at W2, although 90% of the sample was between the ages of 62 and 86, where the youngest and oldest respondents were partners to primary respondents. At W3, the returning sample ranged in age from 41 to 95, with 90% of the sample between 66 and 88. All respondents provided written informed consent, and the protocol was approved by the institutional review boards at the University of Chicago and NORC at the University of Chicago.

We conduct analyses of both partnered sexual activity and masturbation on all respondents with values on the outcome of interest. We also conducted analyses on the effects of characteristics of the dyad on a sample of all heterosexual dyads (N = 953). There was one same-sex male couple and one same-sex female couple, and these were excluded because of the difficulties with generalizing to the population of same-sex couples. Supplementary Table A presents the means of the demographic variables for the sample.

Measures

Cognitive function is measured using a version of the Montreal Cognitive Assessment (MoCA). The MoCA is a clinical screening tool designed to differentiate cognitive changes of normal aging from MCI and early dementia in clinical populations (Nasreddine et al., 2005). Following extensive testing, an 18-item survey adaptation of the Montreal Cognitive Assessment (MoCA-SA) was included in NSHAP W2 and W3 (Shega et al., 2014). The MoCA-SA is designed for administration by nonmedical personnel using Computer-Assisted Personal Interview, and items were selected from the MoCA to reduce respondent burden within the context of a large, time-limited national survey while preserving the MoCA’s sensitivity to a range of cognitive abilities. MoCA-SA scores are highly correlated with the full MoCA and show excellent psychometric properties (Kotwal et al., 2015). NSHAP W2 and W3 assessed cognitive function in every respondent; there are no missing data on this measure. Scores range from 0 to 30. Following others (Kotwal et al., 2016; Lindau et al., 2018), we divide the MoCA-SA into three clinically relevant categories based on suggested cut points for community samples: “Normal” function (23–30), MCI (18–22), and early dementia (0–17). We employ these divisions for comparability to other work using the MoCA.

Sexual activity consists of sex with a partner and sex by oneself (Galinsky et al., 2014). Older ages are characterized by changes in some of the resources for sexuality, as described in the Interactive Biopsychosocial Model, including one’s own health, the presence of a partner, and the partner’s health.

Partnered sexual activity. Respondents who had a spouse, cohabiting partner, or other partner at any point during the year were asked how often in the past 12 months they had sex with their spouse/partner. This could include petting, fondling, sexual touching, oral sex, or intercourse—whatever the respondent considers to be partnered sex. We focus here on any sex with a partner in the past year compared to none, as 39% of older adults with a partner reported “none at all” when asked about the frequency of partnered sex (Supplementary Table B).

Masturbation. Masturbation was measured in NSHAP with the following question: “Masturbation is a very common practice. By masturbation, we mean stimulating the genitals (sex organs) for sexual pleasure, not with a sexual partner. On average, in the past 12 months, how often did you masturbate?” As is the case for partnered sex, there is a large share of respondents who report never masturbating in the past year. This is especially the case for women (Waite et al., 2009). For this reason and similar to partnered sex, we code masturbation dichotomously, as any masturbation in the past year. Supplementary Table B presents the distributions of these variables for the sample of partnered respondents.

Analytic Strategy

The analyses for this article proceed in several stages. First, we examine the association between any sexual activity with a partner and cognitive impairment, as well as any masturbation and cognitive impairment, both cross-sectionally at W2, using a probit regression. Probit regression is a multivariable regression analysis where the outcome is categorical (probit regression for dichotomous; ordinal probit for ordinal). Note because the substantive meaning of each coefficient in a probit can be difficult to interpret (corresponding to a change in a z-score for a 1-unit change in a predictor), we focus on significance and direction of association (Long, 1997). We use probit rather than logit because other models, below, require a probit link function for estimation, and we wish to ensure comparability across steps on our analysis. These models are presented in Table 1.

Table 1.

Cross-Sectional Models, Estimating Associations Between Cognitive Function, Any Partnered Sex, and Any Masturbation, 2010 (W2): Probit Regressions

Model predicting	Model 1: Any partnered sex	Model 2: Any masturbation
Cognitive function
Mild cognitive impairment	−0.05	−0.02
Early dementia	0.03	−0.30*
Female (Ref. = male)	−0.31***	−0.58***
Age	−0.23***	−0.17***
Age, squared	0.00	0.00
Education (Ref. = less than high school)
High school	0.16	0.53**
Voc Cert	0.42***	0.60***
BA or more	0.45***	0.90***
Depressive symptoms	−0.06	0.27**
Functional limitations	−0.08*	−0.02
Race/ethnicity (Ref. = White)
Black	0.18	−0.12
American Indian/Pac. Islander	0.87*	0.17
Asian	−0.18	−0.12
Else	−0.11	−0.13
Hispanic	0.23	−0.01
Constant	0.04	−0.57**
N	2,472	2,226

*p < .05, **p < .01, ***p < .001.

Second, we fit Cross-Lagged Panel Models (CLPMs) across 2010 and 2015, using a bivariate ordinal probit specification. CLPMs assist researchers with estimating the direction of associations between two variables by fitting two equations simultaneously, one regressing some Y at time t on some X at time t − 1, and a second regressing that same X at time t on that same Y at time t − 1. CLPMs also simultaneously estimate a correlation between error terms in these two equations (Kenny, 1975; Kenny & Harackiewicz, 1979). Because these models include lagged dependent variables, they allow us to look at changes in any sexual activity and changes in cognitive function over time. They therefore provide a detailed view of any reciprocal associations between the variables over time and are suited to a longitudinal analysis of these two variables. Note that current software requires the use of a probit, rather than a logistic, link function. In order to account for selection across waves, sample weights were multiplied by the inverse probability of retention in the sample over time, produced using a logistic regression. The regression predicted retention using demographics, depression, functional health, global self-rated physical health, global self-rated mental health, and how difficult the case was to obtain, according to the interviewer. Sample weights were then multiplied by the inverse of the predicted probability from this regression, essentially up-weighting those least likely to return over time (Morgan & Todd, 2008). These models are presented in Table 2.

Table 2.

Longitudinal, Reciprocal Associations Between Partnered Sex, Masturbation, and Cognitive Impairment Using Cross-Lagged Panel Models With Probit (Sex/Masturbation) and Ordinal Probit (Cognition) Link Function 2010–2015

	Model 3		Model 4
	Cognitive impairment at W3	Any partnered sex at W3	Cognitive impairment at W3	Any masturbation at W3
W2 variables
Cognitive impairment
Mild cognitive impairment	0.92***	−0.02	0.94***	−0.02
Early dementia	1.67***	−0.17	1.75***	−0.22
Any partnered sex	0.02	1.22***
Any masturbation			−0.09	1.35***
Female (Ref. = male)	0.03	−0.15	−0.08	−0.24**
Age	0.22***	−0.10***	0.23***	−0.07*
Age squared	0.02	0.02	0.02**	−0.01
Education (Ref. = less than high school)
High school	−0.27*	0.12	−0.21	−0.01
Some college/Assoc/Voc cert.	−0.41***	0.14	−0.37**	0.16
BA or more	−0.77***	0.15	−0.73***	0.53***
W3 variables
Depressive symptoms	0.20*	−0.11	0.20*	0.08
Functional limitations	0.09***	−0.07*	0.09***	−0.09**
Partnered	0.06	0.46***	0.04	−0.24**
Background characteristics
Race/ethnicity
Black	0.65***	0.10	0.68***	0.20
American Indian/Pacific Isl.	0.08	0.23	−0.08	0.73
Asian	0.23	−0.39	0.27	−0.77*
Else	0.07	0.37*	0.11	0.10
Hispanic	0.42***	0.04	0.39**	0.00
Intercept/s
Cut 1	0.54***	1.40***	0.46**	1.12***
Cut 2	1.69***		1.66***
N	2,198		1,951
Rho (correlation of error terms)	−0.02		−0.14*

*p < .05, **p < .01, ***p < .001.

Third, we investigate whether the cognitive limitation in either partner affects their sexual activity at the dyadic level, by regressing any partnered sex, according to both partners, on characteristics of matched male and female partners in the same dyad. Because the outcome here is dyadic, the unit of analysis changes from the individual to the dyad, and predictor variables become characteristics of the husband and the wife in each dyad. These models are also estimated using probit regression and are displayed in Table 3.

Table 3.

Dyadic Analysis; Any Partnered Sex as Reported by Both Partners in a Couple at 2010 (W2), Predicted by W2 Characteristics of Both Partners: Probit Regressions

	Any partnered sex at W3
	Model 5		Model 6		Model 7
	Wives	Husbands	Wives	Husbands	Wives	Husbands
Cognitive function
Mild cognitive impairment	0.01	0.08
Early dementia	0.17	0.04
Age	−0.12*	−0.18***	−0.12*	−0.18***	−0.12*	−0.17***
Age squared	0.05*	0.02	0.05*	0.02	0.04*	0.02
Education (Ref. = less than high school)
High school	0.59***	−0.05	0.56***	−0.05	0.55***	−0.06
Some college/Assoc/Voc cert.	0.59***	0.05	0.56***	0.06	0.54***	0.04
BA or more	0.70***	0.21	0.66***	0.22	0.62**	0.19
Depressive symptoms	−0.12	−0.47**	−0.11	−0.47**	−0.11	−0.46**
Functional limitations	−0.07*	−0.08*	−0.07*	−0.08*	−0.07*	−0.08*
Couple-level variables
Race (Ref. = White couple)
Black couple	0.06		0.07		0.13
Hispanic couple	0.63**		0.66**		0.73***
All other ethnic combinations	0.38		0.38		0.40
Cognition interactions (Ref. = neither impaired)
Husband impaired			0.06
Wife impaired			0.02
Both impaired			0.12
Cognition disparity
Husband’s cognitive advantage (raw difference)					−0.01
Husband’s cognitive advantage (squared)					0.00
Constant	−0.36		−0.33		−0.23
Num dyads (unweighted)	953		953		953

*p < .05, **p < .01, ***p < .001.

Fourth and finally, we examine whether the cognitive function in either partner in a dyad affects masturbation by either member of that same dyad. Here, we fit Actor–Partner Interdependence Models (APIMs), also using a bivariate ordinal probit specification (Cook & Kenny, 2005). Similar to CLPMs, the APIM is essentially two regression equations fit simultaneously, with correlated error terms to aid in obtaining consistent regression estimates. The APIM allows a researcher to see the association of one partner’s characteristics with another partner’s outcomes, within the same dyad. Because masturbation may be correlated within dyads, this is an appropriate modeling strategy for this stage. We display these results in Supplementary Table C.

All models control for age, age squared, education, partnership status (except the dyadic analysis where all respondents are partnered), respondent’s race/ethnicity, depressive symptoms (from NSHAP’s version of the Center for Epidemiology Studies—Depression scale; see Payne et al., 2014), and functional health (activities of daily living), and gender. We chose these variables because various demographic factors may confound the association between sexual activity and cognition, and both depression and functional health may affect sexual function (Waite et al., 2017). We did not enter variables separately, because we were not interested in bivariate associations, but only associations in the context of controls. In a series of robustness checks, reported separately, we add interactions with gender and with partnership status. In dyadic models, we also look at the interaction between husbands’ and wives’ impairment in the same dyad, first by cross-tabulating any impairment in the husband with any impairment in the wife, and then by taking the difference between husbands’ and wives’ continuous MoCA-SA scores. All analyses were carried out in Stata version 15 (StataCorp, 2017), using survey weights produced by the NSHAP survey team to account for oversampling, stratification, clustering, and nonresponse. Analyses using bivariate ordinal probit specifications (i.e., CLPM, APIM) used Stata’s bioprobit package; regular probit regression (i.e., cross-sectional regressions, dyadic sex as an outcome) used the probit command.

Missing data ranged from none at all (e.g., education, gender, age) to 10.3% (masturbation). Missing data were handled using listwise deletion; however, we employed techniques to minimize missing data on the cognitive status measure. Detail on the imputation of missing values can be found in detail in the work of Kotwal et al. (2015), but in brief, missing values are predicted using a series of imputation models, summed, and rescaled to produce final cognitive scores.

Taken together this set of analyses provides a highly comprehensive view of when—and whether—cognitive impairment affects partnered sexual activity or individual sexual activity.

Results

Supplementary Table A presents descriptive statistics for the demographic characteristics of NSHAP respondents considered in this analysis. These results are presented separately for all respondents, for those who are partnered, and members of dyads. We can see that the partnered sample contains a smaller share of females than the sample as a whole, as well as being slightly less likely to be White, and more likely to be Hispanic. Supplementary Table B displays the means of the measures of any partnered sex, any masturbation, and cognitive limitations for members of dyads, for those who are partnered, and for all respondents. The majority of those with partners were sexually active in 2010 and 2015, but the share of those active fell from 60% to 51%. In the sample of all respondents, a minority reported partnered sex in the past year, and the share fell from 49% to 40% across waves. Note that cross-wave comparisons include only those interviewed in both waves.

The majority of both the partnered and the total sample reported no masturbation in the past years, although the share sexually inactive on this measure increased from about 60% to about 70% across waves. In contrast to the declines in sexual activity given in Table 2, we see virtually no difference in the share with either MCI (24% vs. 23%) or early dementia, in either the sample of those with partners or the sample of all respondents.

Table 1 displays a cross-sectional analysis of respondents, using an ordinal probit specification to predict any partnered sexual activity in the past year and any masturbation in the past year with cognitive impairment while controlling for demographic and health characteristics of each person. We find no association between either MCI (b = −0.05, n.s.) or early dementia (b = −0.03, n.s.) and any partnered sex. However, the results show that those who scored as having early dementia were substantially less likely than others to report that they masturbated at all in the previous year (b = −0.30, p < .01), but we see no association for MCI and masturbation (b = −0.02, n.s.).

Table 2 (Models 3 and 4) presents results from longitudinal models in which sexuality at 2010 (W2) predicts each dimension of sexuality and cognitive impairment at 2015 (W3), and cognitive impairment at W2 predicts both impairment and sexuality at W3. These CLPMs (Kenny & Harackiewicz, 1979) allow us to investigate any longitudinal, reciprocal associations between cognitive impairment and sexual activity. Here as well we can see no prospective association between cognitive function and any partnered sex (b_MCI = −0.03, n.s.; b_ED = −0.06, n.s.). Sex in the past year in 2010 also had no prospective association with cognitive impairment in 2015 (b = 0.03, n.s.). Note as well that the residual correlation between cognitive impairment and frequency of sex is −0.02 and is not significant, suggesting that there is also no cross-sectional association at W3 once the factors in this regression are taken into account.

Next, we investigated the association between cognitive impairment and any partnered sex at the dyadic level, examining how husbands’ and wives’ characteristics affect the likelihood that the couple is sexually active. These results are given in Table 3. We focus here on the link between the cognitive resources of the dyad and their joint production of partnered sexual activity. First, we estimated a model predicting any partnered sex in the dyad, where husbands’ and wives’ characteristics were independently allowed to predict dyadic sex; none of these coefficients were significant. We then divide couples into those in which the wife is cognitively impaired and the husband is not, those in which the husband is impaired and the wife is not, those in which both are impaired, and those in which neither is impaired. This allows us to answer the question posed above about the ability of older couples to maintain partnered sexual function in the face of cognitive limitation in either partner, and to see if this differs by the gender of the impaired partner. Note that there was a high degree of concordance between husbands’ and wives’ reports of sexual activity (Pearson’s r = 0.67, p < .001), with 798 out of 953 (83.7%) of dyads reporting agreement in whether or not they had partnered sex in the past year. Results are robust to inclusion or exclusion of the dyads and to alternative coding of sexual activity in these couples. In terms of cognition, 417 dyads showed no impairment (43.8%), in 212 only the husband was impaired (22.3%), in 120 only the wife was impaired (12.3%), and in 204 both were impaired (21.4%).

We find that older couples in which one or both partners were cognitively impaired do not differ from couples in which neither showed cognitive impairment in the probability that they are sexually active (all coefficients n.s.). We also estimate models where we take the difference between husbands and wives’ cognition scores, adding a squared term to account for a parabolic shape in the score; neither the raw difference (b = −0.01, n.s.) nor the squared term (b = 0.00, n.s.) were significant. This strongly supports the argument that partnered sex as a social activity is conserved by couples in the face of cognitive limitations in either or both partners.

Finally, we fit APIMs examining masturbation in either partner within dyads at W2; these results are given in Supplementary Table C. Similar to Model 7, these models show the difference between husbands’ and wives’ cognition scores, where higher indicates that the husband has greater cognitive scores than his wife. For wives, if their husband has better cognitive function than them, they are less likely to report masturbating. We also fit models analogous to Models 5 and 6 and found no significant associations between cognition and masturbation or partner’s cognition and masturbation.

In this article, we found no associations with cognitive impairment and partnered sexual activity, but that early dementia was negatively associated with masturbation in cross-sectional analyses; there were no gender interactions. We found no longitudinal associations between cognition and masturbation or partnered sexual activity—again, with no gender interactions. Finally, in couples, there seemed to be no “protection” evident in the patterns of association, where sexuality was conserved as cognition declined, if there was no reduction in the cognitive function of a person’s partner. However, if husbands scored higher than their wives in cognitive function, then their wife was less likely to report masturbating. These results are summarized in Table 4 and Supplementary Table D.

Table 4.

Research Questions in Light of Results

Question	Results
Are older adults with cognitive impairment less sexually active than those with normal cognition? Or, is sexuality conserved in the face of cognitive decline? Does this differ for partnered vs solitary sexual activity? Does this differ by gender?	Sexuality seems largely conserved in the face of cognitive decline. No gender differences.
Is this true for both dyadic sexuality and individual sexual activity (i.e., masturbation)? Is it true for both men and women?	Some inconsistent evidence that masturbation is less common among people with early dementia. Again no gender differences.
Does normal cognition in one partner protect older couples from lower sexual activity in the face of cognitive limitations by the other partner? Does the gender matter?	No evidence of a protective association from partner’s cognition. Cognitive advantage of husband over wife is associated with a lower likelihood that the wife reports any masturbation.

Discussion

This article examined the relationship between cognitive limitations and sexual health, as reflected in partnered sexual activity and masturbation. Although we found some support for the hypothesis that masturbation is negatively associated with cognitive impairment, we also found that individuals in intimate dyads seem to maintain a key dyadic activity—partnered sex—in spite of MCI or early dementia in either partner. This was the case regardless of whether the husband or the wife had cognitive limitations, and perhaps more surprisingly, even in the face of cognitive limitations in both partners. We also found no effect of impairment at one point on sexuality 5 years later or of sexuality at one point and cognitive impairment 5 years later. In summary, we found little evidence of any link between cognitive impairment and partnered sexuality. Taken together, our results suggest that within long-term dyads of community-dwelling older adults, partnered sexual activity is a highly resilient form of social interaction.

We build on the Interactive Biopsychosocial Model as a framework to understand the ways in which the dyad as a unit influences partnered health outcomes like sexuality. Results support the argument that partnered sexual activity draws on the resources of both partners in a dyad to maintain a shared, likely longstanding behavior even if one partner is experiencing cognitive limitations. This suggests that some social functions may be resilient to mild cognitive decline and even early dementia, perhaps by drawing on a deep reservoir of shared practices and understandings which are preserved even when short-term memory, for example, becomes less reliable (Almkvist, 1996). Our findings point to the need to think more carefully about relationships and the resources that exist within them. We tend to focus more on the characteristics of individuals, perhaps because they are easier to measure. Recent work on the role of the marital dyad, for example, in helping couples cope with functional limitations (Warner & Kelley-Moore, 2012), is an important step toward this expanded perspective.

Our findings on the links—or lack thereof—between cognition limitations and sexual activity are especially compelling given the strengths of the data on which they are based. The data come from a large and nationally representative longitudinal sample of older persons, we have dyadic data to compare people at the individual and couple levels, and our measure of cognitive function is a version of a highly validated and widely used clinical tool used to differentiate MCI from early dementia. Thus, the lack of association between cognitive function and partnered sex is unlikely to be related to problems arising from low power, poor validity and reliability of our measures, or issues of causal direction. Also, note that NSHAP has no missing data on cognitive function; every respondent completed the MoCA-SA (Kotwal et al., 2015). Missing data on the questions on partnered sexual activity and masturbation are also quite low.

In our cross-sectional findings, we documented that although cognitive impairment had no association with partnered sexual activity, partnership status itself had a highly significant and large association. This is unsurprising, because living with a romantic/sexual partner would of course increase the potential for partnered sex. However, it also suggests that being embedded in the dyad itself is a structural precipitant of sexual activity, over and above individual characteristics (i.e., cognitive impairment) of the two partners. This account may not apply to all resources—as we saw, husbands’ physical health was a resource for partnered sexual activity in a way that wives’ physical health was not. Nevertheless, it suggests that the relationship is not just a conduit for each couple’s resources to contribute to each other’s well-being; rather, the relationship itself is a resource. Cohabitation, and the social expectations that surround the relationship, is a mechanism producing sexual activity over and above the individuals who constitute it.

Our study examined partnered sexual activity and masturbation separately based on the Interactive Biopsychosocial Model, which suggests, conceptually, that individual behaviors like masturbation may be less influenced by a partner’s cognitive impairment. Our results on masturbation indicate lower self-reported masturbation among older men and women with early dementia, net of other factors. So sex by oneself does seem to be negatively affected by poor cognitive function, at least in the cross-section. In addition, wives were less likely to report any masturbation in dyads where husbands had a higher MoCA score than their wives. Taken together, these results suggest that these dyads discordant on cognition include some women with early dementia, who are less likely to masturbate. Because partnered sexual activity requires another person but masturbation does not, and because many older adults, especially older women, are unpartnered, this difference has implications for findings on the links between sex and cognition, for both theory and for interventions that might slow cognitive decline.

Results shed light on a key question with implications for quality of life: For people living with cognitive impairments, what is their sexual life like? In the light of these findings, it is not tenable to assume that no sex occurs when one or both partners have some form of cognitive impairment, and it may be valuable to better understand how sex occurs and remains meaningful under conditions of cognitive impairment. Unfortunately, sexual function is often unaddressed as a health concern among older adults (Schaller et al., 2020) and may be an especially neglected health domain among older adults with early cognitive impairment. This is concerning as studies indicate older adults continue to value their sexual lives as they age (Gott & Hinchliff, 2003; Lindau et al., 2007), and optimizing sexual health can positively affect individual health, including happiness, pain, and cardiovascular health (Liu et al., 2016; Wade et al., 2018), and the relationship quality, stress reduction, and emotional intimacy of couples.

It is important that clinicians including medical doctors, nurses, psychologists, social workers, and relationship therapists explore how to support couples trying to navigate their sex lives in the face of cognitive decline. Although results indicate that many older adults are having sex regardless of their cognitive impairment or the impairment of their partner, these couples may face challenges to an active and satisfying sex life beyond the presence or absence of sex. Rather than make assumptions about sexual health, clinicians should ask older adults what matters to them and provide space to discuss sensitive topics like sexual health which an individual may not bring up on their own due to fear, embarrassment, or stigma. If sexual health is a priority to individuals or couples, clinicians might take immediate steps to address related medical needs or provide referrals to couples’ counselors or other specialists. Based on the Interactive Biopsychosocial Model, we further suggest that clinicians discuss sexual health both at an individual level and with couples; counseling that leverages shared health behaviors at a dyadic level might facilitate health behavior change and has cascading individual-level effects on each spouse (Lewis et al., 2006).

It is also important to consider the negative dimensions of partnered sex. These include sexual problems, such as lack of desire or erectile dysfunction, or having sex more often or less often than one would like (Laumann et al., 2008). Having sex because one feels obligated may also increase stress in older adults (Shen & Liu, 2021). In particular, concerns have been raised about the potential for sexual abuse of those with cognitive limitations, especially dementia. These concerns are especially acute for cognitively impaired residents of nursing homes, who may be exposed to risk of abuse by both staff and other residents (Malmedal et al., 2015; Burgess & Phillips, 2006), although the incidence of such abuse may be low (Malmedal et al., 2015). Our results can be generalized to community-dwelling older adults only and do not apply to those with a dementia diagnosis or to residents of nursing homes. Future in-depth quantitative and qualitative work may help to shed light on these topics, especially projects that utilize mixed methods to allow researchers to hear the voices of people with cognitive impairments and their partners (Laumann et al., 2008).

In conclusion, the results illustrate the role of the intimate dyad in conserving a key aspect of well-being in older adulthood: sexual activity. Our results also point to the importance of understanding when and under what conditions close, longstanding social relationships provide a structure within which individuals are able to maintain functioning in the face of cognitive limitations. Supporting the role of these relationships may offer a path forward as we and societies around the world seek to meet the complex and intersecting medical and social needs of individuals experiencing cognitive declines.

Funding

The National Social Life, Health, and Aging Project is supported by the National Institute on Aging and the National Institutes of Health (R01AG021487, R37AG030481, R01AG033903, R01AG043538, and R01AG048511). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Conflict of Interest

None declared.

Author Contributions

L. J. Waite, J. Iveniuk, and A. Kotwal conceptualized the article and designed and oversaw the analyses. J. Iveniuk carried out the analyses. L. J. Waite wrote the initial draft of the manuscript. L. J. Waite, J. Iveniuk, and A. Kotwal all participated in revising the article.