Grief, John Henryism & Suicide Loss in African American Mothers

Research Methodology and Philosophy

This methodology chapter seeks to accomplish two goals: (1) explain the methodology this research adopts, and (2) establish the validity of the adopted methodology for this particular research work. To address these aims clearly, this section is divided into three sub-sections: (1) Research Philosophy, (2) Research Approach, and (3) Research Type and Timeline.

Through this segmentation, the methodology will be understood as carefully planned, cogent, and applicable to all issues under examination, while allowing exploration of all topics relevant to the study (Cohen, Manion, and Morrison, 2007). The segmentation also allows the researcher to explain the research process along the lines suggested by the "Research Process Onion" methodology framework (Saunders et al., 2012, p. 83), such that the systematic, progressive steps taken during the course of the study are illustrated clearly.

Trochim (2006) found in his exploration that most research work focuses on the perception that, by achieving the objectives of the work, the academic world will benefit toward understanding and making sense of our environment, ambient conditions, and processes — and in so doing, understand how people arrive at varying inferences about the same phenomenon, whether in individual or collective capacities. Two main thought processes drive this assumption. The first is Positivism, a perspective that seeks to conduct measurable performance appraisal by experiencing a process or phenomenon in a quantified manner: positivist scientists seek to establish facts based on encounters that can be quantified and measured. The second is Post-Positivism, alternatively known as the "constructivist" framework, which holds that singular experiences and observations lead us to believe that living beings and the world they inhabit follow certain patterns. Both these frameworks are vital when considering traditional approaches to research studies.

According to positivist thought, science helps scientists assimilate the real value of things and the logic and rationality underlying them, making it simpler to control actions and activities. The positivist is inclined to deduce rules, regulations, and laws through practical analysis and measurement. For the majority of those adhering to positivism, objectivity of purpose and inference is an ingrained attitude toward research, hence the inclination toward practical affirmation of facts. The post-positivist, by contrast, contends that absolute objectivity is not possible at all times. All experiences, according to this framework, invariably involve perception that is subjective to some extent. This, in turn, means a certain amount of bias enters any research study. Fair-mindedness, according to post-positivists, is a social construct (Trochim, 2006). For the purposes of this study, the post-positivist approach is the best fit and has been accepted as the guiding methodology for arriving at rational inferences.

According to Trochim (2006), there are essentially two methodologies used for exploratory work: inductive and deductive. Deductive thinking, alternatively understood as a "top-down" procedure, starts from broader concepts and experiences and moves progressively toward narrowed, specific questions — the objective of the study. Inductive thinking, commonly called the "bottom-up" strategy, approaches research from a specific point and expands outward to reach the full range of thoughts and practices relevant to the singular origin of the research work. This research has chosen to adopt the inductive methodology to reach its objectives.

The path taken by a research study bears direct correspondence with its framework. The essential frameworks for an exploratory work are: (1) cross-sectional and (2) longitudinal studies. In the cross-sectional method, the researcher focuses on a determined frame and time period — meaning only a portion of the whole concept or objective is examined. By contrast, in a longitudinal study, the researcher spreads focus over larger time scales; independent variables of considerable scope, as well as more dependent scales, are incorporated, and variations encountered as the study progresses over the time scale are accounted for when assembling the final outcome (Trochim, 2006). In this particular study, given the available research time and the broad potential influence of the findings on the population at large, a longitudinal configuration has been adopted.

The research design of this study is quantitative, with a correlational cross-sectional approach. Certain cities were chosen, and from these, data were obtained with respect to religious orientation, professional and educational status, death certificates, and current census information. Data from surveys were gathered and coded numerically for statistical analysis (Creswell, 2012).

Research Design and Sampling Procedure

According to the method of Baron and Kenny (1986), conclusions for the four research questions were obtained by hierarchical multiple linear regression (HMLR) and by multiple linear regression. HMLR was selected because these analyses permit moderation; demographic covariates were entered in the HMLR model at the first step, while study variables were entered in the subsequent two steps to yield the second and third models (Baron and Kenny, 1986).

Subsequent to the encoding of findings and finalized data analysis, all materials were placed in a secure facility, where they will be stored for a period of no less than five years. At the end of that period, all items related to this work, including notes and data, shall be destroyed.

According to Heiman (2002), correlational design can usefully show how one variable predicts another. In this work, the dependent variable was "severity of grief"; independent variables were: (a) John Henryism; (b) interaction of John Henryism and stigma; (c) interaction of societal cause of suicide with John Henryism; (d) belief that suicide is derived from societal causes; and (e) perceived stigma due to suicide. A goal of the study was to determine the predictive effects of independent variables (a) through (e) upon severity of grief, the dependent variable.

Within the African American community, research on suicide is limited, as are studies on the effects of a son's suicide on mothers. Furthermore, because this population is not one that is typically open, it may be more difficult to recruit study participants (Barnes, 2006; Crosby and Molock, 2006; Day-Vines, 2007; Perry, Pullen, and Oser, 2012). In order to perform snowball sampling, Sydor (2013) discussed the importance of trust, suggesting that the development of researcher-respondent rapport might lead to additional referrals. Given the highly sensitive research topic, such trust is a prerequisite. Furthermore, Sydor (2013) indicated that because these populations may be more emotionally vulnerable, they may not wish to come forward openly and should be accessed as early as possible for data acquisition.

For this study, the population was assessed using a snowball and conventional random survey. Specifically, all possible channels connected to the treatment and care of African American mothers who had lost male children to suicide were utilized. These included educational, religious, social, and health facilities, as well as death certificates, Boys Club, Girls Club, YMCA, YWCA, community organizations, churches, support groups, social service organizations, and community leaders and outreach offices. In all cases, these organizations were approached and asked for assistance and referrals to African American mothers who had lost a male child to suicide. All participants and institutions contacted were provided with a statement from the institutional review board (IRB) including details concerning confidential and ethical precautions.

Snowball sampling, defined as a method where participants refer others, is a non-probability sampling approach with advantages in that it may assist in leading researchers to less obvious or "hidden" populations. However, as Browne (2005) notes, snowball sampling may also introduce bias in the participant group. While some researchers criticize this approach, others such as Noy (2008) state that the snowball method can deliver unique data and is quite informative. To an extent, the researcher cedes some study control to the participants, as expansion of the project requires new referrals from existing participants (Noy, 2008). Given the very limited data available on the population studied here, the snowball sampling method appears to be the most efficacious choice.

Respondent-driven sampling is a modification of snowball sampling based on the work of Heckathorn (as discussed by Treiman, Lu, and Qi, 2012). Considering the potential subjective bias of the snowball method (Creswell, 2009), respondent-driven sampling was established using biased network theory and Markov chain theory — both mathematical frameworks. Applying these assumptions, the respondent-driven sampling method can yield population estimates that are unbiased (Treiman, Lu, and Qi, 2012).

Mothers included in this study are those whose sons committed suicide between the ages of 15 and 24. There are five predictors for this work: John Henryism, beliefs about suicide, and stigma-perception of suicide, as well as two interaction terms — John Henryism and beliefs about suicide, and John Henryism and stigma-perception of suicide. There are also five covariates: family history of suicide, relationship status, depression, age, and the mother's socioeconomic status. Using the methodology of Judd and Kenny (2010) and Faul et al. (2007), sample determination for the study using multiple regression indicates significant power at 0.80 and a medium effect size of f² = 0.15, with significance set at p < 0.05. Thus, the sample size was set at N = 92.

For this work, urban areas with an elevated African American population were chosen. The targeted cities included Hampton Roads, VA; Richmond, VA; Charlotte, NC; Greensboro, NC; Durham, NC; Raleigh, NC; and Atlanta, GA. Using a targeted mailing list, census data were obtained and death certificates were reviewed for African American men deceased by suicide. Support groups such as churches, community groups, social work facilities, and healthcare organizations potentially connected to the treatment or care of African American women who had lost sons to suicide were reached via personal visits, telephone calls, and email. Mental health professionals and African American churches were also contacted, as well as any organization known to handle issues of suicide, trauma, and grief. Local support groups were contacted where identifiable. Once survey responses were received, they were coded after two months; data were analyzed and follow-up thank-you letters were sent to all participants.

Variables and Measurement Instruments

The surveys described below were used for measurement of the study variables. Alignment with the study's constructs guided instrument selection.

Fiegelman and colleagues (2011) outline a Suicide Stigmatization Scale (SSS) used in this work to determine the extent to which mothers of sons who committed suicide perceived that they had been stigmatized. The SSS instrument comprises two assessments: (a) an 11-item subscale addressing social and family harm and/or help, and (b) an 11-item subscale addressing social and family strain (Fiegelman et al., 2011).

For this study, subscale questions for harm/help are coded as follows: 1 = harmful/very harmful; 0 = no effect/not harmful. Subscale questions for strain are coded as follows: 1 = became strained/weaker; 0 = not applicable/remained the same, or became closer/stronger. The SSS scale of 22 questions is computed by summing the two subscales, yielding a ratio ranging from 0 to 22, with lower scores indicating non-strained relationships.

According to Jordan (2012), Fiegelman et al. (2011), and Fiegelman and Fiegelman (2011), inter-item reliability for the SSS scale is good, with Cronbach's alpha in the mid 0.70s to mid 0.80s, and good two-week test-retest reliability with rs = 0.50s, ps < 0.01. There is also criterion validity for the SSS scale, with significant correlation for measures of complicated grief, rs = .70s, ps < .001 (Fiegelman et al., 2011; Fiegelman and Fiegelman, 2011; Jordan, 2012).

The 6-item societal causes of suicide subscale from the interval-coded Attitude Towards Suicide Scale (ATTS) was used to determine the extent of individual attribution of societal causes for suicide, following the work of Knight and colleagues (2000) and Lester and Bean (1992). This subscale uses a Likert-type scale ranging from 6 = strongly agree to 1 = strongly disagree. For 6 items, an individual's score may range between 6 and 36, with larger numbers indicating a stronger belief that society is the cause of suicide. Reliability is measured by test-retest methodology at a one-month interval; Loibl and colleagues (2008) report r = 0.56, p < 0.001 on the societal causes subscale. As shown in many studies (Lester and Bean, 1992; Knight et al., 2000; Kodaka et al., 2010; Loibl et al., 2008), inter-item reliabilities for this subscale range from the low 0.70s to low 0.90s. Kodaka et al. (2007) and Vatan and colleagues (2010) report the subscale has a significant correlation with indicators for hopelessness and external locus of control, demonstrating criterion-related validity with rs from 0.14 to 0.20 and ps < 0.01.

John Henryism is defined as a coping strategy used by Black males involving overwork and resulting physical and physiological health issues (James et al., 1992). The scale rates agreement with statements such as "When things don't go the way I want them to, that just makes me work even harder," and correlates with hypertension (James et al., 1992). James (1994) provided a 12-item scale for the John Henryism Active Coping method, which is used here to assess the extent of John Henryism. This scale has an interval basis including: 5 = completely true; 4 = true; 3 = neither false nor true; 2 = false; and 1 = completely false. For 12 items, scores can range from a high of 60 to a low of 12, with lower scores indicating weaker John Henryism. As demonstrated by many research groups (James, 1994; Neighbors et al., 2007; Rosenfield and Mouzon, 2013; Subramanyam et al., 2013; Wang, Trivedi, Treiber, and Snieder, 2005), there is strong inter-item reliability using this scale: Cronbach's alpha ranges from the mid 0.70s to the mid 0.90s. According to James (1994), with a two-week test-retest period, reliability was r = 0.50, p < 0.01. This approach also has excellent criterion-related validity, with rs ranging from 0.50s to 0.70s and ps < 0.001, using measures of perceived stress and self-reliance (Matthews et al., 2013; Merritt et al., 2011; Rosenfield and Mouzon, 2013).

The intensity of grief as a result of suicide was measured using the Inventory of Complicated Grief (ICG: Prigerson et al., 1995, 1996). This instrument is a 19-item, interval-coded scale with responses ranging from 5 = always to 1 = never; potential scores range between 5 and 95, with lower scores indicating grief that may be less complicated (Prigerson et al., 1995, 1996). Cronbach's alphas range from the high 0.80s to the mid 0.90s for the ICG, which has excellent inter-item reliability (Boelen and Prigerson, 2013; Boelen, van den Bout, and de Keijser, 2003; Lobb et al., 2010; Mitchell, Kim, Prigerson, and Mortimer-Stephens, 2004; Prigerson et al., 1995, 1996). Pearson correlations are in the 0.80s for one-month test-retest reliability (Lobb et al., 2010; Prigerson et al., 1995, 1996). A particular advantage of the ICG is that it can discriminate between clinical depression and prolonged grief disorder (Boelen and Prigerson, 2013; Sung et al., 2011); additionally, the ICG has demonstrated criterion-related validity for measures of spiritual grief, with rs = 0.40s and ps < 0.05 (Burke et al., 2014).

To ascertain maternal socioeconomic status, the question "How would you rate the financial well-being of your family?" was asked, coded from 5 = very well-off to 1 = not at all well-off — a question used nationally in demographic research (Caro and Cortés, 2012). Maternal age was determined using the birth date provided by respondents, as previous studies have shown that participants more readily provide a birth year than their actual age (Settersten and Mayer, 1997; Skoog and Ciecka, 2010). Maternal relationship status was assessed using the single-item question recommended by Gates (2011): "What is your current relationship status?", a categorical variable coded as follows: 5 = widowed; 4 = divorced; 3 = in a committed relationship but not married; 2 = married; and 1 = single.

The CESD-R (Revised Center for Epidemiologic Studies Depression Scale), provided by Eaton and colleagues (2004), is a 20-item ratio-coded instrument measuring depression severity as self-rated by the individual. Scoring ranges from 4 = "nearly every day for two weeks" (severe symptoms) to 0 = "not at all or less than one day"; with 20 items, the possible scale ranges from 0 to 80. According to Wood and colleagues (2010), any score over 16 suggests higher levels of depression. The CESD-R is widely used with diverse populations including minorities, women, and the elderly (Wood, Taylor, and Joseph, 2010; Pigeon et al., 2011), and has demonstrated strong criterion-related validity with rs from 0.20 to 0.70 across multiple depression measures (Aalto et al., 2012; Dozois et al., 1997; Gloria et al., 2012; Smarr and Keefer, 2011). A validation study by Van Dam and Earleywine (2011) confirmed the CESD-R's convergent and divergent validity, strong factor loadings, internal consistency, and excellent psychometric properties.

To address family history of suicide, respondents were asked "Have any of your blood relatives committed suicide?" — a question from the 2013 University Family History instrument — with response codes of 1 = yes and 0 = no.

The biopsychosocial contextual model of stress defined by Clark and colleagues (1999) is used throughout this study. This framework essentially holds that, at some point, regardless of individual resilience, stress becomes a physical health issue in which the individual no longer functions optimally. According to Boss (2002), stressors for a given individual are derived in part from that individual's system of beliefs and values, and each event's personally correlated significance depends upon sociodemographic factors — class, ethnicity, gender, race, and age (Boss, 2002). In the context of this study, it is thus likely that stigma attached to depression and suicide may prevent African American mothers of sons who have committed suicide from seeking therapeutic assistance in dealing with their grief (Kneeland, 2006).