Structural and Contextual Patterns in Family Health History Knowledge among African American Adults: A Mixed-Methods Social Network Analysis Study*

The family is a functional unit, serving as a complex relational system that allows the exchange of resources, such as information (Koehly et al., 2003). The extent of these exchanges is highly dependent upon the family capital (i.e. availability of resources). Within families, “key” individuals may serve in specific functions that influence the actions and outcomes of their relatives. For example, findings from several studies of familial networks have suggested that mothers are key influential figures in family networks (Koehly et al., 2003). Additionally, past studies have identified parents and older adults as key figures, often serving in the role as gatekeepers for communication and decision-making (Ashida, Kaphingst, Goodman, & Schafer, 2013; Koehly et al., 2009). The close involvement of extended family members (i.e. non-first degree) is a uniquely important aspect of the African American family structure and function (Hecht, Jackson, & Ribeau, 2003). In particular, African American family members have a high degree of interaction across multiple generations.

The family/kinship system offers a unique intergenerational network, where preventive and hereditary health information can be shared intergenerationally among all members of a family - men, women, and children. To date, most research on hereditary risk has been limited to first- or second-degree relatives. However, relevant to the African American family network structure, recent research suggests that there is utility in assessing hereditary risk in the context of the extended family (Solomon, Whitman, & Wood, 2016).

Limited research has been done regarding the assessment of individual disease risk based on multi-generational family health history (FHH). In clinical practice, providers typically only obtain the health history of patients’ first-degree relatives – a process that primarily occurs during patients’ initial visit with providers and is rarely updated during follow-up appointments (Daelemans et al., 2013; Rich et al., 2004). However, scholars have increasingly emphasized the importance of conducting “comprehensive” FHH, which extends beyond first-degree relatives and includes as many family members as possible (Maradiegue and Edwards, 2006). Within genetic counseling and medical genetics, the recommended FHH is a 5-generation pedigree (Solomon et al., 2016). Information obtained is important for developing a more accurate appraisal of FHH risk, in order to determine individuals’ risk for developing conditions based on their FHH. A recent study on the contribution of extended family history assessment in cancer risk by Solomon, Whitman, and Ward (2016) found that limited FHH information can have a detrimental impact on determining patients’ eligibility for screenings. In particular, their results indicated that over 70% of patients eligible for breast cancer screening is missed if extended family history is not utilized.

FHH is an important, but often underestimated, aspect of disease risk. FHH includes any health conditions or illnesses that a person’s biological relatives have been diagnosed with or that run in a person’s family. Individuals are more likely to develop certain chronic diseases if they have a family history of those conditions, especially among close relatives (Annis, Caulder, Cook, & Duquette, 2005). For example, having one first-degree female relative (sister, mother, daughter) diagnosed with breast cancer doubles an individuals’ risk of developing breast cancer, and this risk is increased by five times if an individual has/had two first degree relatives with diagnoses (Breastcancer.org, 2018).

A variety of benefits have been associated with having increased knowledge about one’s FHH. When FHH information is shared, family members are informed about health problems and possibly even risk behaviors related to those health problems, that may have occurred generation after generation in their family. By connecting this information with adverse health outcomes they have observed in their family (i.e. lingering illness and premature death), individuals may be more motivated to engage in preventive behaviors. In particular, awareness of one’s FHH has been associated with increased practice of preventive behaviors, such as physical activity, healthy diet, and participation in chronic disease health screenings (Baptiste-Roberts et al., 2007). Individuals’ knowledge and awareness of their FHH also has important clinical implications. For example, increased awareness of FHH information can improve accuracy of information shared with clinicians (Kaphingst et al., 2012). Consequently, healthcare providers have more information available to guide clinical decisions, such as targeted disease screenings, genetic counseling recommendations, and preventive health behavior recommendations. Several studies have also emphasized that the sharing of health history information in familial networks is especially useful for younger generations of family members (Ashida & Schafer, 2014; Forrest et al., 2003; Newcomb, Raudonis, Snow, & Cauble, 2012), who often still have time to engage in health behaviors and decision-making that can prevent or delay the onset of conditions that run in their families.

In contrast, lack of FHH information can have a detrimental impact on individuals. Among some racial and ethnic minority groups (Black, Hispanic, Asian), low FHH knowledge has been associated with a decrease in perceived risk of disease (Orom, Kiviniemi, Underwood, Ross, & Shavers, 2010). Previous intervention research has demonstrated a strong positive association between individuals’ knowledge of their FHH risk and their perceived risk and worry over developing common diseases (coronary heart disease, stroke, diabetes, and breast, ovarian, and colon cancers) (Acheson et al., 2010). Risk perception, as an important determinant of individual health decision-making, plays a vital role in preventive health decisions. Consequently, individuals who have a low perception of FHH risk may also have a decreased practice of preventive and/or screening and early detection behaviors (Sivell et al., 2008; Ward et al., 2008; Yoon et al., (2003). Scholars also note that individuals’ non-disclosure about their health conditions limits the decision-making autonomy of at-risk relatives who would greatly benefit from the information (Forrest et al., 2003; Wilson et al., 2004).

Despite its importance, individuals among the public have been found to have a low awareness of their FHH (Catz et al., 2005). Results from a former national study found that though most respondents (96.3%) acknowledged that FHH is important, only 29.8% of respondents indicated that they had ever actively attempted to collect their FHH (Yoon, Scheuner, Gwinn, Khoury, & Jorgensen, 2004). Some studies have found that ethnic minority groups, including Latinas and Latinos, have a lower likelihood of having collected their FHH than other populations (Chen, Li, Talwar, Xu, & Zhao, 2016; Yoon et al., 2004). In a recent qualitative study on African American women’s perspectives and experiences regarding FHH collection and communication, few participants reported that anyone in their family kept formal FHH records (Thompson et al., 2015). However, several characteristics have been associated with increased likelihood of FHH collection, including being female (Halbert et al., 2016), being an older adult, and individuals having a higher income (Case, 2008). Among these groups, FHH information was primarily gathered by designated family historians, and was obtained from documents, such as death certificates, and obituaries (Case, 2008). Qualitative studies have also found that FHH information is learned via other, less formal approaches including “word of mouth” (Newcomb et al., 2012; Pettey et al., 2015; Yamasaki & Hovick, 2015). Degree of relation has also been identified as an important factor in FHH knowledge. In particular, studies have found that individuals’ knowledge is more accurate about the health history of their most proximal relatives (i.e. first-degree) rather than other relatives (Mai et al., 2011; Theis, Boyd, Lockwood, & Tritchler, 1994; Wideroff et al., 2010). Overall, individuals’ access to, and knowledge of, their FHH is influenced by a variety of factors, including family structure and norms, as well as cultural characteristics, particularly as it relates to health communication.

Individuals were eligible to participate in the African American Family Networks and Health study if they self-identified as African American and were at least 18 years of age. An additional criterion for participation was residence in the [Name Removed for Blind Review] metropolitan area, so that individuals could complete the study interview in person. Participants were recruited from a variety of community venues, including local churches, health fairs, universities, African American community events, and African American family reunions. At these venues, an announcement was made by a research team member about the opportunity to participate in the study and/or an information table was available for participants to speak with research team members and sign up to be contacted for an interview appointment in the near future. Additionally, participants were recruited via electronic flyers and word of mouth. Prior to being enrolled, all prospective participants were screened for eligibility by the study project manager (L.R.). Data collection for the study began in September 2016 and was completed in December 2016. The study received Institutional Review Board approval from the Indiana University Human Subjects Office (protocol # 1502818898).

The African American Family Networks & Health study employed a convergent parallel mixed methods study design (Creswell and Clark, 2017) to gather information about participants’ familial networks. Convergent parallel mixed-methods study designs afford researchers the opportunity to simultaneously collect quantitative and qualitative data from each participant during his or her interview, thus allowing for more comprehensive information to be gathered. In the African American Family Networks and Health study, quantitative data was collected about participants’ familial networks and corresponding qualitative data was collected to contextualize the quantitative findings.

Each participant was asked to complete a “family tree interview” in order to gather information about his or her familial network. Specifically, the interview was conducted to elicit information regarding participants’ knowledge of their FHH, as well as to elicit information about their interpersonal exchanges with their family members, including health communication and social support. Interviews were scheduled to accommodate each participant’s convenience and availability. The family tree interviews were conducted in a private conference room at the Indiana University Fairbanks School of Public Health, and were administered by trained research assistants (K.B., H.S.), who were racially matched with the study sample (i.e., African American). Upon each participant’s arrival to his or her interview, the research assistant reviewed the informed consent form with the participant and answered any questions he or she had.

Next, the research assistant gathered the participant’s demographic information and constructed a family tree for the participant, using Progeny Genetic Pedigree Software (Progeny Genetics LLC, 2018), a secure web-based program. The pedigree was constructed to function as a visual aid during each participant’s interview. Information provided by the participant to construct the pedigree was gathered prior to the interview via a structured workbook provided by the research team. Specifically, the workbook gathered information about the participant and all relatives in his or her generation (i.e., siblings and cousins). The workbook also gathered information about the participants’ parents’ generation (i.e., parents, aunts, and uncles), as well as gathered information about the participants’ grandparents. In this family pedigree, information about younger generations (i.e., children, nieces and nephews, and grandchildren) was also collected, but we excluded these younger generations from our analyses to focus on individuals’ knowledge of health history, based on the sharing of personal health information (health history) from relatives in previous generations and within the ego’s current generation. For each participants’ family network members (alters), the following information was gathered: 1) familial status (full, half, step, or adopted); 2) vital status (living or deceased); 3) age (current or at death); and 4) current city and state of residence (if living).

Following the construction of the participants’ family pedigree, his or her family tree interview was administered. The qualitative component of the interview was audio recorded. Interviews lasted approximately 1.5 hours on average, and participants were compensated with a $65 gift card and a copy of their family tree for completion of the pedigree construction and network interview process.

Demographic characteristics for each participant were collected and recorded using Research Electronic Data Capture (REDCap), a secure web-based electronic data management software. General demographic variables collected include age, gender, household size, marital status, and education level. Health-related demographic variables were also collected for each participant, including personal health history and self-rated health status.

The Colored Eco-Genetic Relationship Map (CEGRM) (Kenen & Peters, 2001) was used to collect information about each participant’s family network. The CEGRM approach is highly interactive and engages participants throughout the entire data collection process. First, the participant generates a family pedigree (prior to the interview). Next, during an in-person interview, the participant’s family tree is used as a visual aid to guide his or her discussion of health information exchanges with his or her family members on the pedigree. Throughout the interview, participants are asked to apply color-coded symbols to represent characteristics of specific “key” family network members, such as relatives who function in specific health communication and social support roles (See figure 1). In the African American Family Networks and Health study, participants were asked to identify biological relatives (alters) who: 1) talk to the participant about the participant’s health; 2) who talk to the participant about their own health; 3) who avoid having discussions about health; 4) are sources for general family health information; 5) whose health the participant helps to monitor and track; 6) for whom the participant helps with managing their personal health; and 7) who share helpful health facts or information with the participant. Our primary outcome of interest for this analysis was whether a family member was named as someone who talks to the participant about their own (i.e., the family member’s) personal health information. Throughout this paper, we refer to family members named in this role as “personal health informants.”

Figure 1.

In addition to providing quantitative characteristics about participants’ familial networks, the CEGRM approach also allows the participant to provide qualitative contextual information about the key family members who, to the participants’ knowledge, were diagnosed with specific conditions. For the African American Family Networks and Health study, participants were asked to discuss how he or she learned about the health diagnosis the identified family members, including those who were identified as “personal health informants.” Specifically, participants were asked, “How do you know that your family member has or had “X” condition? For example, what were the circumstances of you learning this information?”

Quantitative analysis of participants’ familial network data from their CEGRMs functioned to identify structural network patterns regarding African Americans’ interpersonal communication exchanges regarding their relatives’ personal health history. In particular, the network analysis sought to identify patterns in FHH information sharing based on characteristics of the family members who were named by the participant as personal health informants, such as their gender, geographic location, or their degree of relation to the participant. All quantitative data were analyzed using SPSS 24.0 and SAS 9.4 statistical software.

Standard descriptive statistics (percentage, mean, standard deviation, and range) were generated for the participant sample (egos) and the resulting sample of family network members (alters). Descriptive variables reported for the participant sample (egos) include age, gender, marital status, household size, educational level, health status, personal history of specific health conditions, and history of work in the health care field (see Table 1). Descriptive variables reported for the sample of family network members (alters) are presented in Table 2, and include alter gender, age, health conditions, and geographic homophily (operationalized as whether the network member lives in the same state as the participant). Table 2 also presents relational (tie) characteristics, including alter relationship to the participant (e.g., parent, sibling), the familial generation of the alter (i.e. whether the familial alter was in the same generation, a generation above, or two generations above the participant), as well as health communication and support roles of the familial alters. Finally, network-level descriptive statistics were generated based on the aggregate sample of family network members provided by participants (see table 3). Network-level statistics calculated include network size, gender proportions, homophily characteristics (gender, generation, and state of residence), and proportions of networks by relationship (e.g., the proportion of cousins in an average network), health communication roles (e.g., the proportion of network members named as a health discussant of the ego’s health information or a personal health informant of the alters’ health information), and generation (i.e., same generation, parents’ generation, or grandparents’ generation).

For our primary quantitative analysis, a generalized estimating equation (GEE) model was used to identify significant predictors associated with the likelihood of an alter (family network member) being identified as a personal health informant (i.e. a family member who talks to the participant about the family member’s personal health) during the family tree interview (see Table 4). Predictors of interest included alter gender, gender homophily (whether the alter identified as the same gender as the participant), geographic homophily (whether the alter identified lives in the same state as the participant), alter generation (whether the alter was in the same generation as the participant, a generation above, or two generations above), alter family side (maternal or paternal), and whether the alter was reported to have one or more chronic health conditions. The model also took into account whether the alter was identified as participating in any of the following specific health communication and support roles in relation to the participant: familial alter talks to the participant (ego) about the participant’s health; familial alter avoids having discussions about health; familial alter is a source for general family health information; the participant (ego) helps to monitor and track the alter’s health; the participant helps the alter in managing the alter’s health; and the alter shares helpful health facts or information with the participant. In our analysis, standard errors were clustered at the participant level to account for correlation within family units.

Analysis of qualitative data collected during each participant’s family network interview functioned to gain in-depth contextual information regarding how African Americans learn about their relatives’ health history. Audio-recorded data collected during each participant’s family tree interview was professionally transcribed, and later analyzed using Dedoose qualitative and mixed-methods software. Emergent themes were identified in the transcripts using inductive content analysis (Dedoose version 8.0.35, 2018). Specifically, open codes were applied to participants’ narrative statements, where statements with similar content were grouped together in categories to reveal primary themes (Saldaña, 2015). Qualitative analysis and coding was conducted by two research members who were trained in qualitative methods. Each research team member independently coded the data and they later convened to discuss and agree upon final emergent themes occurring in participants’ narrative data.

A total of 37 African American adult participants (egos) were included in the analytic sample, with a mean age of 43.8 years (SD = 16.8 years; range = 19-68 years) (see Table 1). About half of the participant sample was comprised of males (n = 19; 51.4%). Over half of the participants were non-married (i.e., single, divorced, or widowed; 48.6%), while 45.9% were married or in a long-term partnership. The sample was highly educated, with a majority of participants having completed at least some college or more (94.6%). Nearly two-fifths of the sample (40.5%) reported previous experience working in the health care field. A majority of participants rated their health status as “excellent” (13.9%) or “good” (61.1%). Specific health conditions indicated among participants included hypertension (35.1%), high cholesterol (27.0%), asthma (13.5%), and diabetes or high blood sugar (10.8%). No significant demographic differences were observed by ego gender, with the exception of marital status, where male participants were significantly more likely to be married than female participants were.

A total of 1,078 familial alters were generated from the participants’ family tree interviews (see Table 2).

The network sample contained an even proportion of male and female alters (48.2% male alters and 51.8% female alters). On average, familial alters were 49.6 years of age (SD = 23.5). One-quarter of alters (25.5%) were reported to have at least one of the six chronic conditions that were inquired about during family tree interview, with hypertension (10.8%) and cancer (6.9%) the most common conditions reported. Approximately one-third of alters (34.3%) were reported to live in the same state as the participant. Regarding familial alters’ ties to participants, the majority of alters in the network data sample were cousins (43.7%) and aunts or uncles (24.8%). Concerning health communication roles, nearly 15% of alters were identified as personal health informants who talk to the ego about his or her (the alter’s) health, while 11% were identified as discussants of the ego’s health. Only 8.5% of alters were acknowledged as functioning in both roles (i.e., as a reciprocal communicator about health). Almost one-tenth (9.2%) of the alter sample was identified as someone who tends to avoid discussions about health. Approximately 13.2% of alters were acknowledged as sources of FHH information, while 7.4% were reported to share helpful health facts or information with the participant. Finally, 12.3% of alters were identified as someone whose health the participant monitors or tracks and 5.6% of alters were identified as someone whose health the participant helps to manage.

Figure 2 (pictured above) illustrates an example familial network of a 33-year old African American male participant from the African American Family Networks and Health study.

Figure 2.

Characteristics of participants’ familial networks are presented below in Table 3. The average size of a participant’s family network was approximately 29 relatives (SD = 15.5), with network sizes ranging from 10 to 67 alters among the participant sample. On average, half of a participant’s familial network was homophilous to the participant with respect to gender (mean = 0.50) and generation (mean = 0.52), respectively, and about one-third of the average family network lived in the same state as the participant (mean = 0.34). Networks, on average, were proportionately comprised of 36% cousins, 25% aunts or uncles, 15% siblings, 15% grandparents, and 8% parents. On average, about one-half of networks were comprised of family members in the same generation as the participant and another one-third from the participant’s parents’ generation. Participants, on average, reported FHH information for about one-quarter (mean=0.26) of their familial network. In terms of health communication roles, the largest proportion of family network members was comprised of personal health informants, or alters who talk to the ego about their (the alter’s) health (mean = 0.17), followed by alters whose health the participant monitors and tracks (mean = 0.15), alters who talk to the ego about the ego’s health (mean = 0.13), and alters who know FHH information (mean = 0.13). Only one-tenth of networks, on average, were comprised of alters who avoid or “block” health discussions. Average network proportions of other health communication and support are listed in Table 3.

In our multivariate analysis (Table 4), we found that participants were more likely to name an alter as a personal health informant if the alter was female (OR = 2.14, p = 0.0519), from the maternal side of the participant’s family (OR = 1.12, p = 0.0006), had one or more chronic health conditions (OR = 2.41, p = 0.0041), is someone who had discussions with the participant about the participant’s health (OR = 16.28, p < 0.0001), was a source of family health information (OR = 3.46, p = 0.0072), and was someone whose health the participant helped to monitor or track (OR = 5.93, p = 0.0002).

Qualitative data collected during the Family Tree Interviews gathered important contextual information regarding how participants learned about the health history of relatives in their family network (via direct interpersonal communication and other mechanisms). The data yielded several emergent themes, including open family communication, observation of illness, caregiving, post-death knowledge (see Table 5).

During the integrated analysis of our quantitative network data and our qualitative contextual data, it was observed that participants who reported more “personal health informant” alters during the network interview portion typically learned about their family’s health history from open discussions, where participants described having direct conversations with specific family members about the family members’ health. In these instances, information was described as being shared for the purposes on informing the ego about his or her potential risk, or for gaining social support from the ego. For example, one female participant discussed how her cousin disclosed her hypertension diagnosis to inform her that she may also be at risk. Another female participant described receiving open communication from her sister about her sister’s breast cancer diagnosis, emphasizing that “we communicate, there is no secrets and all that.” Similar to our quantitative finding that “personal health informant” alters were more likely to be female, most participants’ description of open-discussions via direct communication with relatives were focused on their open communication exchanges with female family members.

In contrast, our participants who reported less FHH knowledge (based on fewer reported personal health informants) typically described their family as having closed communication patterns. In such instances, these participants described how they learned about the health history of their relatives via indirect methods, such as personal observations of their relatives’ illness, hearsay, gaining information during caretaking activities, as well as learning about a relative’s health issues following the relatives’ death. Similar to the quantitative network analysis findings, many of the relatives who did not disclose their own health history tended to be males. Limited disclosure about personal health information, including among first-degree relatives, as well as relatives that are in close geographic proximity to participants, was often perceived as being the result of relatives wanting to maintain their privacy due to being embarrassed by their diagnosis or not wanting to appear as being “weak” or “needy.” Additionally, participants indicated that they perceived that their more proximal relatives might not want to be burdensome and cause them to worry. The exhibition of closed communication behavior was especially discussed in the context of male family members, including fathers and brothers, and was oftentimes linked to masculinity and males’ continuous efforts to be viewed as “strong.”

Among discussions about closed communication, diabetes, in particular, was a condition that participants frequently discussed that they learned about by observing their relatives’ self-management behaviors, especially administering insulin. While participants directly described how they learned of relative’s diagnosis of some conditions, such as diabetes, hypertension, and cancer, their reports of a relative’s diagnosis of mental illness were discussed in a speculative context, rather than as a confirmed diagnosis. This was primarily directed toward female family members. In very few instances did participants discuss learning about their FHH through their own intentional efforts to collect the information. Table 5 presents a list of emergent themes observed in the qualitative contextual data, which are accompanied by exemplary quotes from study participants.

The African American Family Networks and Health study possesses several strengths. While the Colored Eco-Genetic Relationship Map (CEGRM) approach has been used to study a variety of conditions, including ovarian-breast cancer (Peters, Hoskins, Prindiville, Kenen, & Greene, 2006; Peters et al., 2004) and testicular cancer (Peters et al., 2012), this familial network data collection method has only been used with White participant samples. To our knowledge, the African American Family Networks and Health study is the first study to utilize the CEGRM approach with an African American participant sample, thus diversifying the application of its use. Additionally, the use of a mixed-methods research design is a significant strength of our study, as the qualitative data collected complemented and contextualized our quantitative network findings. Mixed-methods studies are considered to be particularly rigorous, because they afford the opportunity to gather comprehensive information about the topic of study, and thereby tell a more complete story when answering research questions. Regarding our study, the collection of quantitative network data facilitated our understanding of “what” FHH information sharing patterns look like within African American families. The rich, complementary qualitative data that was collected as part of our participant interviews enhanced our understanding of “when,” “where,” “how,” and even “why” FHH information is shared and/or obtained within African American families. While many studies have sought to assess the extent of individuals’ health history knowledge, and identify patterns of FHH knowledge consistency among relatives, very few have gathered information about how individuals learn about their family’s health history, and which contexts facilitate this knowledge. This contextual information provides valuable insight for future efforts to enhance the transmission of health history information in this population.

While our study consists of various strengths, there are several limitations worth noting. First, we recognize the limitation of our single-informant design, as it limited the ability to confirm our participants’ reports of their FHH. Closely related, our use of health communication exchanges with relatives identified as “personal health informants” as a proxy for FHH participants’ is a limiting factor in the depth of our results, as it pertains to understanding FHH knowledge among African Americans. Additionally, our study sample size may limit the generalizability of our findings. Finally, our decision not to include younger generations in our analytic sample may be a limiting factor in our reported study results. However, it is important to note that the number of family members from younger generations named as personal health informants (the outcome of interest) was extremely small in our total sample of familial alters (less than 3%).