Searching for the Kinkeepers: Historian Gender, Age, and Type 2 Diabetes Family History

Abstract

Kinkeepers facilitate family communication and may be key to family medical history collection and dissemination. Middle-aged women are frequently kinkeepers. Using type 2 diabetes (T2DM) as a model, we explored whether the predicted gender and age effects of kinkeeping can be extended to family medical historians. Through a U.S. telephone survey, non-diabetic Mexican Americans (n = 385), Blacks (n = 387), and Whites (n = 396) reported family histories of T2DM. Negative binomial regressions used age and gender to predict the number of affected relatives reported. Models were examined for the gender gap, parabolic age effect, and gender-by-age interaction predicted by kinkeeping. Results demonstrated support for gender and parabolic age effects, but only among Whites. Kinkeeping may have application to the study of White family medical historians, but not Black or Mexican American historians, perhaps due to differences in family structure, salience of T2DM, and/or gender roles.

Introduction

The term kinkeeper refers to individuals who are responsible for “initiating, solidifying, and maintaining family contact with relatives” (Rice, 2001, p. 416) and kinkeepers are most frequently women in middle age (Brown & DeRycke, 2010). Explanations for the gender and age of kinkeepers vary, but a popular theory points to matrifocal tilt (Brown & DeRycke, 2010; Hagestad, 1985), which places mothers and grandmothers at the metaphorical center of the family due to their stereotypical roles as caregivers and resource exchangers within the family network. Because of the association between age and maternity, middle-aged women (40–60 years of age) are presumed to draw on the family network more than men or other age groups do, and are thus more invested in maintaining the network through kinkeeping (Haxton & Harknett, 2009; Marks & Mclanahan, 1993).

The kinkeeper literature has ramifications for health educators, particularly regarding family medical history. Kinkeepers are presumed to be the “custodian[s] of the family’s medical information” (Bennett, 2010, p. 51) and are expected to have greater knowledge about their family’s medical history than other family members. Thus, kinkeepers are prime targets for collection and dissemination of family history information. However, we have little actual evidence for the interaction of gender and age in kinkeeping as related to family health history. Existing studies examine these characteristics individually, or otherwise had sample designs that precluded such an analysis by, for example, focusing on diseases that primarily affect one gender or limited recruitment of one gender or one age group (Ashida, Goodman, Stafford, Lachance, & Kaphingst, 2012; Corona, Rodriguez, Quillin, Gyure, & Bodurtha, 2013; Kaphingst et al., 2012; Koehly et al., 2009; Lindenmeyer, Griffiths, & Hodson, 2011).

Along with age and gender, characteristics associated with knowing or sharing one’s family health history include individuals’ own risk factors and/or disease status (Kaphingst et al., 2012) and comprehension of the disease (van den Nieuwenhoff, Mesters, Gielen, & de Vries, 2007). For example, research suggests that individuals with diabetes are more accurate family historians about the condition than individuals without diabetes (Bensen et al., 1999) and that individuals with greater comprehension of a given disease are more likely to communicate that health information within the family (van den Nieuwenhoff, Gielen, & De Vries, 2007). In this study, we investigated respondents’ information concerning a condition that affects men and women equally—type 2 diabetes (T2DM)—and their reported family history of this condition. We also assessed, as two of the controls, respondents’ perceived T2DM knowledge and their body mass index (BMI), a T2DM risk factor.

Based on the aforementioned literature, we developed four hypotheses: After controlling for BMI, perceived T2DM knowledge, and demographic variables, (1) middle-aged participants will have greater knowledge of their T2DM family history (report the most relatives with T2DM), (2) women will report more relatives with T2DM than will men, and (3) gender will moderate the relationship between age and number of reported relatives with T2DM, such that the effect of middle age on reported family history will be more pronounced among women. Additionally, T2DM prevalence (Centers for Disease Control and Prevention, 2011) and family health history collection behaviors (Yoon, Scheuner, Gwinn, Khoury, Jorgensen, Hariri, & Lyn, 2004) differ across ethnic/racial groups. Further, there is evidence that gender roles function differently across race/ethnicity, at least with respect to family socialization (Leavell, Tamis-LeMonda, Ruble, Zosuls, & Cabrera, 2012). Because of the above factors and the need for culturally-sensitive health education (Knoerl, Esper, & Hasenau, 2011), we further hypothesized that age and gender effects would vary across the three ethnic/racial groups investigated in this study (non-Hispanic White, non-Hispanic Black, and Mexican Americans).

Our paper makes three main contributions to the kinkeeping literature. First, much of the existing literature on kinkeeping is based on a qualitative methodological standpoint. This brief report attempts to bring quantitative methods into the kinkeeping literature. Second, we contribute to the existing literature by exploring whether gender and age effects vary by race. Much of the kinkeeping literature focuses on non-Hispanic White populations; therefore, our research investigates whether kinkeeping models based on non-Hispanic White samples are generalizable to other races/ethnicities. Third, previous work has been geographically isolated and non-representative, whereas our study includes nationally representative samples of three racial/ethnic groups within the U.S.

Method

Participants

Participants (N = 1,168) in this study (1) resided in the contiguous U.S., (2) were 18–75 years old, (3) did not have T2DM, (4) and were non-Hispanic Black (n = 387), non-Hispanic White (n = 396), or Mexican American (n = 385). Data were weighted to be nationally representative of the U.S. population meeting inclusion criteria.

Measures

Demographics

Respondents self-reported race, ethnicity, gender, age, education level, nativity (U.S.-born or foreign-born), height, and weight. Using height and weight, we calculated (kg/m²) BMI. Table 1 presents demographics for each racial/ethnic group.

Table 1.

Gender Comparisons of Participant Characteristics within Racial/Ethnic Groups^*

	Mexican American			Non-Hispanic Black			Non-Hispanic White
Characteristic	Men (n = 157)†	Women (n = 228)†	p value	Men (n = 163)†	Women (n = 224)†	p value	Men (n = 225)†	Women (n = 171)†	p value
Nativity (Foreign-born)	41	53	0.24	12	3	0.08	1	2	0.61
Age (mean, SD)	37 (24)	33 (26)	0.34	37 (20)	43 (24)	0.06	44 (11)	51 (9)	0.002
Education			0.008			0.04			0.47
No diploma	19	45		13	10		4	5
High school	49	27		46	28		41	30
Some college	20	11		17	42		21	26
Bachelor’s or higher	12	17		24	19		33	39
BMI (mean, SD)	29 (11)	28 (15)	0.35	28 (8)	32 (12)	0.03	27 (4)	27 (3)	0.72
Perceived Knowledge			0.29			0.08			0.02
A lot	6	7		18	17		12	16
A fair amount	27	19		28	33		21	42
Some	33	37		15	33		36	23
A little	14	28		28	12		18	14
Almost nothing or nothing	19	9		11	5		13	5

^*Percentages are weighted as described in Research Design and Methods. Values may not sum to 100 due to rounding.

^†Unweighted sample size. Weighted to population in millions: Mexican Americans N = 6.6 men, 8.9 women; non-Hispanic Blacks N = 10.5 men, 13.6 women; non-Hispanic Whites N = 62.3 men, 58.3 women

Reported Family History of T2DM (Reported Fhx)

Interviewers read aloud the following: “[Type 2 diabetes] is the most common type of diabetes, and most people who have type 2 diabetes get it when they are adults. They might take pills for it, but they don’t usually need insulin shots.” Interviewers asked whether participants had any living or deceased relatives with T2DM. Computer software allowed participants to list up to seven blood relatives with diabetes; most reported six or fewer (97.7% Blacks; 99.8% Whites; 96.6% Mexican Americans).

Perceived T2DM knowledge (Perceived Knowledge)

Interviewers asked, “How much do you know about diabetes?” Responses were on a 5-point scale: “A Lot,” “A Fair Amount,” ”Some,” “A Little,” or “Almost Nothing.” Higher scores indicate higher Perceived Knowledge.

Procedure

The data were collected as part of the Genetic Explanations for Type 2 Diabetes: Prevention Implications (GEDI) project—an IRB-approved, list-assisted, random-digit-dialed, U.S. landline telephone survey that included measures of T2DM risk factors, perceived risk, and causal attributions. After obtaining oral informed consent, trained interviewers from the University of Michigan Institute for Social Research surveyed respondents in English or Spanish, based on respondent preference. The overall, unweighted, American Association for Public Opinion Research Response Rate 3 was 40.8%. Interviews averaged 45 minutes and respondents received at least $20 for participation.

Data Analyses

Analyses were conducted using Stata/SE Version 12. Population estimates were computed by weighting to adjust for non-coverage error, non-response, unequal selection probability, and sample stratification. Missing data were removed listwise.

Within each racial/ethnic group, statistics summarizing respondent demographics, Perceived Knowledge, and Reported Fhx were calculated as weighted means and standard deviations (continuous variables) or weighted percentages (categorical variables). For all variables we computed gender comparisons within racial/ethnic group (Table 1). Continuous variables were compared using general linear models. Categorical variables were compared using design-adjusted, second-order Rao-Scott F-tests of independence.

Regressions were used to investigate how gender and age relate to Reported Fhx in each of the three racial/ethnic group, controlling for education, nativity, BMI, and Perceived Knowledge. Because Reported Fhx was a count variable demonstrating a right-skewed distribution with overdispersion, negative binomial regressions were employed. The kinkeeper literature predicts a non-linear relationship between age and Reported Fhx, with more affected relatives reported in middle age than any other time (suggesting a negative quadratic relationship). To test for this relationship, age was centered about its mean in each racial/ethnic group and entered into the model as both linear and second-order (squared age) variables. All regression statistics were adjusted for sample design.

Results

As expected from our weighted, representative sampling design, and consistent with national trends, there were statistically significant differences between the three racial/ethnic groups in age, education, nativity, and BMI, but not gender. The three groups also did not differ significantly on Perceived Knowledge. All of these factors were controlled for during regression analyses. Many participants reported zero affected relatives (32% Blacks; 34% Mexican Americans; 43% Whites). Among participants with a positive family history, the majority reported one to three affected relatives (79% Mexican Americans; 83% Blacks; 90% Whites). In each racial/ethnic group, the median number of affected relatives was one, but the mean number was shifted upwards (Mexican Americans: M = 1.76, SD = 3.4; Blacks: M = 1.56, SD = 2.36; Whites: M = 1.06, SD = 0.85).

The hypothesized interaction between age and gender was non-significant in all racial/ethnic groups (Blacks p = 0.31; Mexican Americans p = 0.11; Whites p = 0.06) and was removed from the final model (Table 2). Gender was a significant predictor of Reported Fhx only among Whites. White women reported 49% more affected relatives than their male counterparts. Age was a significant predictor in all groups. Among Blacks and Mexican Americans, each additional year of life predicted a 1% or 2% decrease (respectively) in the participant’s Reported Fhx. Among Whites, the effect of age was more complex, with evidence for a negative quadratic relationship between participants’ age and Reported Fhx. The model showed a gradual increase in White participants’ Reported Fhx for each additional year from age 18 (women reported 0.98 relatives; men reported 0.66) until age 39 (women reported 1.32 relatives; men reported 0.89), with a gradual decrease in Reported Fhx for each additional year thereafter up to age 75 (women reported 0.58 relatives; men reported 0.39).

Table 2.

Incidence Rate Ratios (IRR) for Predicting Reported Fhx within Racial-Ethnic Group^*

	Mexican American		Non-Hispanic Black		Non-Hispanic White
	IRR (95% CI)	P-value	IRR (95% CI)	P-value	IRR (95% CI)	P-value
Gender (Ref: men)	1.25 (0.94, 1.68)	0.13	1.07 (0.77, 1.48)	0.70	1.49 (1.10, 2.01)	0.009
Age (centered)	0.98 (0.97, 0.99)	0.007	0.99 (0.98, 1.00)	0.049	0.99 (0.98, 1.00)	0.04
Squared Age	1.00 (1.00, 1.00)	0.76	1.00 (1.00, 1.00)	0.22	1.00 (1.00, 1.00)	0.03
Education (Ref: No diploma)		0.34		0.02		< 0.001
High school	1.05 (0.72, 1.51)	0.81	0.59 (0.37, 0.94)	0.03	0.50 (0.32, 0.79)	0.003
Some college	1.15 (0.79, 1.67)	0.46	0.48 (0.30, 0.78)	0.003	0.33 (0.21, 0.54)	< 0.001
Bachelor’s or higher	1.53 (0.95, 2.49)	0.08	0.52 (0.28, 0.96)	0.04	0.41 (0.26, 0.65)	< 0.001
Nativity (Ref: U.S.-born)	1.11 (0.80, 1.52)	0.53	0.58 (0.23, 1.47)	0.25	0.14 (0.01, 1.63)	0.12
BMI	1.03 (1.00, 1.05)	0.02	1.03 (1.01, 1.06)	0.002	1.01 (0.98, 1.03)	0.58
Perceived Knowledge (Ref: Almost nothing or nothing)		< 0.001		0.03		0.007
A little	1.67 (0.78, 3.57)	0.18	1.15 (0.54, 2.46)	0.71	1.07 (0.48, 2.40)	0.87
Some	2.50 (1.29, 4.86)	0.007	2.13 (1.04, 4.34)	0.04	1.25 (0.59, 2.64)	0.56
A fair amount	2.64 (1.36, 5.13)	0.004	2.15 (1.01, 4.60)	0.048	2.10 (1.00, 4.39)	0.049
A lot	4.38 (2.23, 8.58)	< 0.001	2.42 (1.18, 4.96)	0.02	1.88 (0.85, 4.18)	0.12
Constant	0.29 (0.11, 0.74)	0.01	0.45 (0.16, 1.27)	0.13	1.20 (0.39, 3.73)	0.75

^*Statistics are weighted as described in Methods

Each control variable, except nativity, significantly predicted Reported Fhx in at least one of the three regression models. Among Whites and Blacks, higher education predicted reporting fewer affected relatives. Higher BMI predicted reporting more affected relatives among Mexican Americans and Blacks, but not Whites. Among all racial/ethnic groups, higher Perceived T2DM Knowledge predicted reporting more affected relatives.

Discussion

We expected (1) middle-aged participants, compared to other age groups, would report more relatives with T2DM, (2) women, compared to men, would report more affected relatives, and (3) gender would moderate the relationship between age and Reported Fhx. Our findings were mixed, with support within one racial/ethnic subgroup for our first two hypotheses, but no support for our third hypothesis. Among Whites, we found evidence for the hypothesized gender gap and for a negative quadratic relationship between age and Reported Fhx. However, the number of relatives reported by Whites peaked just below (not within) the middle-age bracket, at age 39. Among Blacks and Mexican Americans, we found no gender difference and a negative linear relationship between age and Reported Fhx.

The kinkeeper literature provides a sociological explanation for the gender effect among Whites. It suggests that women report more affected relatives than men because they are more connected with their family medical history due to gender-role-consistent kinkeeping behaviors (e.g., providing family care). Kinkeeping provides a similar explanation for the negative quadratic relationship between age and Reported Fhx among Whites. Emerging adulthood (ages 18 to 25) is period of self-exploration and transition to independence. Combined with a pattern of delayed child-bearing (Mathews & Hamilton, 2009) and physical/emotional separation from family due to increased college enrollment (Snyder & Dillow, 2013), emerging adults may report fewer affected relatives because of less connection with kin. Older young adults (26- to 40-year-olds) and middle-aged adults (40- to 60-year-olds) are more often involved in child-rearing and elder support, and thus report more affected relatives because they are enmeshed in kin networks. Among adults over 60, isolation from family is a prominent theme (Cornwell, Laumann, & Schumm, 2008), which could disrupt the flow of medical history information to families’ eldest members.

However, our findings provide no support for extending the kinkeeper literature to Mexican American or Black family medical history. Aggregate differences in childbearing and education timelines, T2DM prevalence, family structure, and gender-roles may explain the lack of gender or quadratic age effects within these two groups. Similar to Whites, Mexican Americans and Blacks have experienced delayed childbearing and higher college enrollment, but to a lesser degree (Mathews & Hamilton, 2009; Snyder & Dillow, 2013). Because T2DM is more prevalent among Blacks (13%) and Mexican Americans (13%) compared to Whites (7%; Centers for Disease Control and Prevention, 2011), T2DM family history may also be more salient to Blacks and Mexican Americans. Additionally, extended family networks are particularly important among Black and Mexican Americans, who are more likely than Whites to live in multi-generational households (Pew Research Center, 2010). Furthermore, variations between racial/ethnic groups in gender-related behaviors may explain differing findings by race/ethnicity. For example, both Black and Hispanic fathers engage in more social-visiting activities with their children compared to White fathers, including visiting relatives (Leavell, Tamis-LeMonda, Ruble, Zosuls, & Cabrera, 2012). Finally, within Mexican American families, key male figures are often central to medical decision-making (Galanti, 2003); thus, these men may be more aware of relatives’ diabetic status. All these aforementioned factors may lead to greater awareness of family medical history across genders and age groups for Blacks and Mexican Americans.

Overall, our results support the extension of kinkeeper literature to White T2DM family historians. White women in the 30 to 60 age range may be preferred targets for the collection and dissemination of T2DM family history. Among Blacks and Mexican Americans, however, gender may be less useful in identifying T2DM family historians, with other sociological mechanisms likely at work instead.

Limitations and Future Directions

This foundational work has several limitations. Our findings are specific to T2DM and may not be transferrable to other disorders; thus, future research should explore whether our results are replicated for other diseases. This study focused purely on the number of reported relatives with T2DM and could not take into consideration geographic or psychosocial factors affecting family history knowledge. Future research could investigate the effect of family networks on communication. Additionally, we cannot definitively explain the gender difference found for Whites. This difference may be caused by male historians under-reporting affected relatives or by female historians over-reporting. Although the kinkeeper literature supports the former interpretation, further research is necessary to provide clarification.