|Home | About | Journals | Submit | Contact Us | Français|
Lack of power in relationships has been offered as an explanation for why women do not always engage in safer sex behavior with male partners. However, few studies have empirically tested the association between power and self-protective behavior. Causal modeling procedures were used to examine the interrelations of seven measures of power and to examine the effects of power on women’s HIV-related communication and condom use with male partners. Power was measured by education, employment, decision making, perceived alternatives to the relationship, commitment to the relationship, investment in the relationship, and absence of abuse in the relationship. The sample comprised 187 Puerto Rican women, aged 18–35 years, attending a comprehensive health clinic in the Bronx, New York, who were at risk for heterosexual transmission of HIV. Women who were currently employed and those who were more committed to their relationships reported less HIV-related communication. Condom use was predicted by shorter length of the relationship, more education, current employment, less investment in the relationship, and more HIV-related communication.
Trends show that AIDS poses a serious threat to the health of American women. In 1994 AIDS became the third leading cause of death among women 25–44 years of age (Centers for Disease Control and Prevention [CDC], 1996b). Additionally, the percentage of women exposed to HIV through heterosexual contact increased from 34% in 1990 to 40% in 1996 (CDC, 1991, 1996a). An examination of data on number of sex partners suggests that women may be at risk for heterosexually transmitted HIV not because they have multiple partners, but because they are having unprotected sex with male partners who do (Diaz et al., 1994; Seidman et al., 1992; Smith, 1991). Hispanic women may be at especially high risk of acquiring HIV heterosexually because Hispanic men may be more likely than men of other racial and ethnic groups to have multiple sex partners (Choi et al., 1994; Marín et al., 1993a,b; Sabogal et al., 1993).
Currently, the most effective method for preventing the transmission of HIV from men to women is the male condom, which requires the cooperation of male partners. The literature suggests that condom use among women with high-risk partners is low and inconsistent (Campbell and Baldwin, 1991; Catania et al., 1992; Harrison et al., 1991; Soskolne et al., 1991) and that some subgroups of women, such as Hispanic women, have particularly low rates of condom use with male partners (Harrison et al., 1991). Qualitative research has suggested that many high-risk women do not feel efficacious at implementing safer sex practices because they believe they have little power to negotiate sexual behavior change with their partners (Fullilove et al., 1990; Maxwell and Boyle, 1995; Nyamathi and Lewis, 1991; Pitts et al., 1995; Shervington, 1993; Singer et al., 1996). Many researchers and women’s health advocates have called for interventions to empower women to negotiate safer sex with their partners (Amaro, 1995; Flaskerud and Uman, 1993; Fullilove et al., 1990; Kline et al., 1992; Levine et al., 1993; Singer et al., 1996; Ulin, 1992; Williams, 1995; Worth, 1989).
Despite the hypothesis that women’s lack of power is associated with high-risk sexual behavior in many heterosexual couples, few quantitative studies have directly tested whether women with more power are more likely to engage in protective activities (e.g., requesting safer sex from their partner, actual condom use). Given the importance ascribed to the power construct, additional research is needed to determine whether power among women increases levels of condom use with their male partner, thus potentially decreasing their heterosexual risk for HIV. We explored the effect of power on self-protective behavior among a group of Puerto Rican women who are at risk for HIV because of the sexual behavior of their male partners.
The first step in examining the importance of power in the prevention of HIV among women is to operationalize the construct. The literature is replete with various definitions and measures of the power construct. In the literature on women and HIV/AIDS, as well as the literature on power in general, discussions of the power construct appear to fall into two groups: power as a function of resources and power as a function of various interpersonal reactions. Thus, we classified the power variables in this study into two categories: resource power and relationship power.
Resource power was measured with two variables: level of education achieved by the women and whether they were currently employed. Relationship power was measured with five variables: (1) more decision making by the women, (2) more perceived alternatives to the relationship, (3) less commitment to the relationship, (4) less investment in the relationship, and (5) absence of abuse in the relationship. Grouping the power measures in this way allowed us to explore whether resource power or relationship power might be more important in predicting self-protective behavior.
In the literature on women and HIV, relationship power is often discussed as the ability to make sexual decisions within the relationship (Amaro, 1988; Fullilove et al., 1990; Kline et al., 1992; Mays and Cochran, 1988; Worth, 1989, 1990). Marriage and family researchers also have equated power and decision making, arguing that within the family, power is exhibited in the making of decisions that affect the entire family unit; therefore, the person who makes most of the decisions is the most powerful (Blood and Wolfe, 1960; Madden, 1987). Thus, for the purposes of this study, we equated high levels of decision making by women with high levels of relationship power.
Three of our measures of relationship power—commitment, investment, and perceived alternatives to the relationship—come from the close-relationship literature. Rusbult and colleagues (Rusbult and Buunk, 1993; Rusbult et al., 1991) demonstrated that persons who are more committed to their relationships are more likely to stay with their partners, engage in more relationship maintenance behaviors, and are more willing to make sacrifices for the good of the relationship. Closely related to commitment, according to Rusbult, are the constructs perceived alternatives to the relationship and investment in the relationship. Those who are more committed are more invested and they also perceive fewer good alternatives to their current relationship. Certainly, in many cases, commitment and investment may describe a strong bond between intimate partners. However, when facing a potential conflict, such as asking a partner to use a condom, a highly committed person may have difficulty being assertive, given the need to ensure that the relationship continues and the perception that good alternatives to the relationship do not exist. This could be especially challenging for highly committed and invested women who know or suspect that their partners are having sex outside of the relationship; these women may fear that their partner will choose the other relationship over them if they insist on safer sex. Therefore, in our study, we defined high commitment to the relationship, high investment in the relationship, and low perceived alternatives to the relationship as low power.
Other researchers have defined physical or emotional abuse as an indication of coercive power in a relationship (Dobash and Dobash, 1979; French and Raven, 1960; Holland et al., 1990). Wingood and DiClemente (1997) demonstrated that abusive male partners may indeed affect HIV-protective behaviors in heterosexual relationships. In their study, battered women reported lower levels of condom use and were more likely to report threats of physical violence and emotional abuse in response to requests for condom use than did women who were not in physically abusive relationships. Thus, for the purposes of our study, we defined abuse as an inverse of power.
We had two major objectives. First, we undertook a closer examination of the power construct by exploring the interrelations of several different measures of power. Is power a large, unidimensional construct for which all of these measures are merely equal, interchangeable indicators? Or is power a more complex, multidimensional construct? Our second objective was to assess the impact of power on self-protective behavior among a group of women at high risk for contracting HIV through heterosexual contact. We hypothesized that women with more power would engage in more self-protective behavior.
Self-protective behavior against the transmission of HIV is a difficult construct to define for women because they do not actually wear the male condom. Thus, the use of this prevention method requires the cooperation of the male partner. Several researchers have suggested that communication with the male partner about condom use is an integral part of HIV self-protection for women. In other words, because of the nature of the prevention method, women must first engage in discussions or negotiations about condom use with their male partners (Campbell and Barnlung, 1977; Dembo and Lundell, 1979; Moore and Erickson, 1985). Some studies have shown that HIV-related communication is indeed a predictor of condom use (Catania et al., 1992, 1994; Moore et al., 1995). In our study, we operationalized self-protective behavior as both HIV-related communication and use of the male condom. Because several researchers have argued that communication about HIV, AIDS, and safer sex may mediate the effects of power on sexual practices (Campbell and Barnlung, 1977; Dembo and Lundell, 1979; Moore and Erickson, 1985), we tested both the direct effects of power on condom use and the indirect effects through HIV-related communication.
Study staff screened women who were waiting to be seen at a comprehensive health clinic in the Bronx, New York. These women were visiting the clinic for various reasons, including annual gynecologic exams, pediatric visits for their children, general health concerns, and to collect certificates from the Special Supplemental Program for Women, Infants, and Children (WIC). To participate in the study, women had to be of Puerto Rican ancestry, 18–35 years of age, currently in a heterosexual relationship of at least 1 year’s duration, and to know or suspect that their primary male partner had other sex partners during the past year. The following were excluded: (1) women who were pregnant or had given birth during the past 6 months, (2) women who might be at risk for HIV due to drug use (i.e., ever injected drugs), and (3) women whose sexual decision making might be influenced by the desire or need to obtain drugs (i.e., any noninjecting use of crack, cocaine, amphetamines, heroin, or methadone during the past 6 months; daily use of marijuana during the past 6 months).
Interviewers approached 259 women to ask them to participate in the study. Of those approached, 9% refused to be screened for the study. Of the women who agreed to be screened, 17% did not meet the eligibility criteria and 4% gave incomplete information during the interview, resulting in a final sample size of 187.
Several quantitative measures of power were used. Resource power was measured in two ways. Employment was a dichotomous variable, scored as 0(not working) or 1 (working). Education was scored 1 (eighth-grade or less), 2 (some high school), 3 (high school diploma or general equivalency diploma), 4 (some college), and 5 (college graduate).
Five relationship power constructs were measured with seven observed variables. The first relationship power construct was decision making within the relationship, which was measured on a 24-item scale. The scale was patterned after one developed by Madden (1987), which was an adaptation of the scale by Blood and Wolfe (1960, pp. 39–40). Participants were asked who made the decision about 24 activities, with possible answers 1 = I always make this decision, 2 = I make this decision more than he does, 3 = He and I make this decision about the same amount of time, 4 = He makes this decision more than I do, and 5 = He always makes this decision. Items were first reverse-scored so that a higher number indicated more decision making by the woman. Then, a mean score on the 24 items was computed for each participant rather than a sum, since some of the items did not make sense for all women to answer (e.g., decisions about children for women without children). The Cronbach alpha for these items was .90.
The next three relationship power constructs (i.e., commitment, perceived alternatives, and investment) were measured by four scales and a single item patterned after those developed by Rusbult et al. (1991; Rusbult and Buunk, 1993). A participant’s score was the sum of responses on the items for each scale. Commitment was measured with two scales, commitment and satisfaction, each having four items with Cronbach alphas of .77 (e.g., “How committed do you feel to continuing the relationship with your husband/boyfriend?”). Perceived alternatives was measured with one three-item scale on quality of alternatives (Cronbach alpha = .75) and a single item tapping the respondent’s belief in the possibility of finding another relationship (e.g., “How would getting involved in another relationship compare to your current relationship with your husband/boyfriend?”). Investment in the relationship was measured with a three-item scale with a Cronbach alpha of .84 (e.g., “Are there other relationship which are important to you that would end or become strained if your relationship with your husband/boyfriend ended, i.e., shared friends, his family?”). Higher scores on these scales indicated more commitment, more investment, and more perceived alternatives to the relationship.
The final relationship power construct, abuse in the relationship, was represented by measures of physical and verbal abuse. Three physical abuse items were combined into a severity index that was scored for the most severe type of abuse experienced by the participant. Uniformed Crime Report (UCR; U.S. Department of Justice, 1984) codes were used as a guide for ranking the yes responses. Women who had experienced sexual assault were given a score of 3; women who had experienced aggravated assault, but no sexual abuse were assigned a 2; women who had experienced simple assault alone were given a score of 1; and those who answered no to all three physical abuse items were given a score of 0. Scores for verbal abuse during the past 12 months were 0 (no verbal abuse), 1 (1–9 incidents of verbal abuse), 2 (10–19 incidents), or 3 (20 or more incidents).
HIV-related communication between the participant and her sex partner was measured by two subscales. The items on the first subscale referred to talking about HIV-related topics (e.g., “Have you ever talked to your husband/boyfriend about using a condom to protect you from getting HIV, the virus that causes AIDS?”), and the second subscale referred to requests by the participant that her sex partner behave in a certain way (e.g., “Did you ask him to use condoms when having sex with you, in order to protect you from HIV?”). The two HIV-related communication subscales were created by summing participant responses on the 8 talk items (Cronbach alpha = .80) and the 10 request items (Cronbach alpha = .74). Higher scores on the communication subscales indicated higher levels of talking and requesting. The second aspect of self-protective behavior, condom use, was measured by asking women whether they had used condoms for protection against HIV during the past 12 months and, if so, how often. Items on self-reported condom use were combined to create one indicator of condom use: 0 = never, 1 = less than half the time, 2 = about half the time, 3 = more than half the time, and 4 = always.
Due to the extensive inclusion and exclusion criteria for enrollment into this study, it was not necessary nor even possible to include several of the background variables that are often used as covariates. More specifically, there was no variability in our sample on gender, ethnicity, and type of HIV risk. However, all of our models did control for age and relationship length, both measured in years.
The associations between the indicators of power and self-protective behavior were explored via the Linear Structural Relations program (LISREL, version 8), which was developed by Joreskog and Sorbom (1993). LISREL was selected as the best method for this study because it tests both a measurement model and a structural equation model. The measurement model indicates how observed variables are related to underlying or latent constructs. The structural equation model specifies the relations between the latent constructs, including both direct and indirect effects.
Maximum likelihood methods were used to examine the fit of the hypothesized model to the observed data. The first measure of fit was the ratio of chi square to the degrees of freedom, which was used rather than the level of significance of chi square because chi square is very sensitive to sample size (Carmines and McIver, 1981). The second measure of fit was the Adjusted Goodness of Fit Index (AGFI), where 1.0 indicates a perfect fit and any value of .90 or above is considered adequate. Also considered was the criticial-N value computed by the LISREL program; the model is considered to provide a good fit to the data if this value is greater than 200 (Hoelter, 1983). T values were used to test the significance of specific parameters such as regression coefficients.
Approximately one fourth of the sample (27.3%) was married to and living with their primary partner, and the mean length of participants’ relationships was 4 1/2 years. Almost three fourths of the study participants (73.8%) had children who were living with them and who were dependent upon them for financial support at the time of the interview. The intercorrelations, means, and standard deviations on all the observed variables that were used in the LISREL model are shown in Table I.
The final measurement model specified the relations between 15 observed variables and 11 latent variables. This model differed somewhat from the measurement model that was first envisioned; however, it is the model that provided the best fit to the data. The major difference involved the need to treat employment and education as separate latent variables rather than as two indicators of the same latent variable: resources. Education had pervasive relations with other variables in the model, whereas current employment did not; thus, education was treated as a background variable to allow its effects on other variables in the model to be expressed. However, both education and employment were still considered resource variables for purposes of interpretation. No measurement errors were correlated. The measurement model fit the data well, χ2(45, N = 175) = 46.87, p > .05, χ2/df = 46.87/45 = 1.04, AGFI = .91, critical N = 261.
The structural equation model was composed of eight simultaneous equations predicting condom use, HIV-related communication, and the power constructs. Higher levels of employment resources, decision making, and perceived alternatives to the relationship were hypothesized to predict higher levels of HIV-related communication and condom use. Because commitment to the relationship, investment in the relationship, and abuse were viewed as inverses of power, a negative association was expected between these variables and self-protective behavior.
The structural equation model shown in Figure 1 has been trimmed, that is, only the statistically significant paths(p < .05) were retained from the full model. This model depicts the effects of the background variables as well as the effects of the power variables on self-protective behavior. Overall, the trimmed model fit the data quite well, χ2(60, N = 175) = 52.53, p > .05; χ2/df = 52.53/60 = .88, AGFI = .93, critical-N = 294.
The trimmed model accounted for 23% of the variance in HIV-related communication and 31% of the variance in condom use. The variance accounted for in the trimmed model was further broken down to assess how much variance was uniquely accounted for by two sets of variables, the resource aspects of power (i.e., education and employment) and the relationship aspects of power (i.e., decision making, commitment, investment, alternatives, and abuse). For HIV-related communication, the resource power variables uniquely accounted for 8% of the variance and the relationship power variables uniquely accounted for 12%. Conversely, resources accounted for 16% of the variance in condom use above and beyond that accounted for by all other variables in the model, including communication, and the relationship power variables uniquely accounted for only 2% of the variance in condom use. Thus, relationship power was more important in predicting HIV-related communication, whereas resource power was more important in predicting condom use.
Only the significant correlations among the power variables are presented in Figure 1. Women who were abused reported less commitment to their relationships. In addition, low commitment to the relationship was associated with higher levels of decision making by the women as well as higher perceived alternatives to the relationship. More decision making by the woman was related to less investment in the relationship.
As expected, HIV-related communication had a direct effect on condom use: women who reported higher levels of communication also reported higher levels of condom use, t(1, 166) = 2.54, P < .05. Two power variables, employment and commitment to the relationship, predicted HIV-related communication. Contrary to our hypothesis, women who were currently employed reported lower levels of HIV-related communication, t(1, 166) = −3.32,p < .01. However, our prediction regarding the relationship between commitment and communication was supported; women with higher levels of commitment to their relationship reported lower levels of HIV-related communication, t(1, 166) = −3.67, P < .01. Because of their effects on HIV-related communication, employment and commitment had indirect effects on condom use through HIV-related communication. The direct, indirect, and total effects of employment resources and commitment are shown in Table II. Although employment had a positive direct effect on condom use (i.e., employed women reported higher condom use), the indirect effect was negative because employed women reported lower levels of communication. Thus, the total effect was also positive, but less strong than the direct effect. Commitment, on the other hand, had no direct effect on condom use, and only a negative indirect effect through decreasing levels of communication.
Three variables, education, relationship length, and investment in the relationship, had only direct effects on condom use, and all were in the expected directions. Higher education was associated with more condom use, t(1, 165) = 3.58,p < .01. However, longer relationships were associated with less condom use, t(1, 165) = −3.42, p < .01, as were higher levels of investment in the relationship, t(1, 165) = −2.57, P < .05.
We had also hypothesized that decision making, perceived alternatives to the relationship, and the existence of abuse in the relationship would affect self-protective behavior. However, these variables, as well as age, had no effect on either HIV-related communication or condom use.
Lack of power in relationships has been viewed as an important barrier to women’s self-protection against HIV/AIDS (Amaro, 1995; Flaskerud and Uman, 1993; Fullilove et al., 1990; Holland et al., 1990, 1992; Kline et al., 1992; Thomas and Holland, 1994; Ulin, 1992; Worth, 1989). The present study had two objectives: to examine associations between several measures of power and to assess the impact of power on self-protective behavior. Overall, our findings support a multidimensional view of power in relationships and suggest that some aspects of power may be more important than others for the prevention of HIV transmission to women.
In response to the first objective of this study, two resource power constructs and five relationship power constructs were used based on theoretical definitions of power in the literature: education, employment, commitment to the relationship, perceived alternatives to the relationship, investment in the relationship, decision making, and abuse in the relationship. The patterns of relations among the power variables and between these variables and others in the model support a multidimensional view of power in relationships. If power were unidimensional, one would expect all the power indicators to be strongly associated with one another and to have similar patterns of association with outcome variables: however, this was not the case. Low to moderate associations were found among the measures of power, and the measures were differentially related to the dependent variables. This finding has implications for future research; researchers should be specific in defining how power is being operationalized and measured. In addition, the various measures of power cannot be viewed as interchangeable, given their differing relations with outcome variables.
The second purpose of this study was to examine the impact of resource and relationship power on women’s ability to protect themselves against HIV. Again, the findings on the effects of power clearly show that the construct cannot be understood in a unidimensional sense. Rather, different types of power, as measured by our various indicators, were related to the self-protective behaviors in several different ways, including direct effects on communication (i.e., employment, commitment), direct effects on condom use (i.e., education, length of relationship, investment), indirect effects on condom use through HIV-related communication (i.e., employment, commitment), and no effects on either outcome (i.e., decision making, alternatives, abuse). Relationship power was more important in predicting HIV-related communication, whereas resource power was more important in predicting condom use. As expected, HIV-related communication predicted condom use in this study: women who reported higher levels of communication also reported more condom use with their male partners.
We had hypothesized that resources would positively affect both types of self-protective behaviors; however, being employed predicted less HIV-related communication, but more condom use, and a higher level of education was related to more condom use, but not to more HIV-related communication. Women who are employed and have higher levels of education may simply be involved with partners who are more amenable to condom use; therefore, once condoms are used, discussions and negotiation about them may no longer be necessary.
Alternatively, this pattern of results may be an artifact of the way in which we measured HIV-related communication. It may be that the power offered by resources such as education and employment is so strong that women who possess these resources bypass talking and requesting strategies and instead simply declare that condoms will be used. Future research should attempt to measure a wider range of communication and influence strategies (Sagrestano, 1992) that women may be using to ensure that self-protective behaviors take place.
Higher levels of commitment to the relationship predicted less HIV-related communication, and HIV-related communication mediated the effect of commitment on condom use. Additionally, higher levels of investment in the relationship predicted less condom use. These were the expected effects; highly committed and invested women were hypothesized to be less powerful and therefore to make fewer demands of their male partners and consequently to have lower condom use. Women may be afraid that their partner will react negatively to requests for condom use (Ehrhardt et al., 1992; Moore et al., 1995). Highly committed and invested women who know or suspect that their male partners are having sex outside of the relationship may be unwilling to discuss HIV-related issues and to request that their male partners use condoms for fear of disrupting their relationship. This interpretation is consistent with that proposed by Misovich et al., (1997), who assert that members of a committed relationship may be less motivated to engage in AIDS preventive behavior because as commitment increases, the focus shifts from self-protection and enhancement to relationship maintenance and enhancement. Thus, a highly committed woman may, at the expense of self-protection, avoid discussions of condom use for fear that it will disrupt the relationship, especially if she fears that her male partner may already be involved in another relationship to which he might retreat if pushed too hard.
Researchers concerned with preventing the heterosexual transmission of HIV should look more closely at relationship issues such as stability, trust, and intimacy. We also need a better understanding of what happens when women attempt to initiate condom use. Do men refuse to use condoms and threaten to end the relationship, or do they agree to use condoms? These questions should be asked of both women and their male partners so that the information can be used in developing and implementing interventions to prevent HIV. Creative strategies such as interventions for the male partners of high-risk women and couple interventions should be tested to ascertain if the involvement of male partners alleviates feelings that safer sex may threaten the relationship.
One of the power measures that did not predict condom use deserves further discussion: abuse in the current relationship. This finding is in contrast to those of Wingood and DiClemente (1997) in which women in abusive relationships reported less condom use than women who were not in abusive relationships. The levels of verbal and physical abuse were high within our sample; almost 30% reported verbal abuse, and approximately 20% reported physical abuse; 18% of Wingood and DiClemente’s sample reported physical abuse. A major difference between the two studies was the race/ethnicity of the samples: our sample was Puerto Rican women, whereas Wingood and DiClemente’s sample was African-American. Future research exploring this potential association should include women from different racial and ethnic backgrounds and explore potential cultural differences in the effects of abuse on condom use. Regardless of the inconsistency of findings, specialized prevention messages should be developed for women in violent relationships so that we are not insensitive to their situation and inadvertently cause it to worsen. Additionally, women in these situations may benefit greatly from methods that they could more easily initiate or that they could use without the knowledge or consent of their male partner, such as topical microbicides and the female condom.
Since the participants who were enrolled in this study were all Puerto Rican women, one important issue to consider is that of acculturation. Specifically, it could be argued that some of the associations we found between the power variables and self-protective behavior might be explained by acculturation. However, adding acculturation (i.e., language spoken in various situations) to the structural equation model did not change any of the associations presented in Figure 1.
Although our study offers some information about the effect of power on self-protective behavior, it also has several limitations. Models such as the one we tested assume that women want their male partners to use condoms: that condom use is important to women. Additionally, our study lacked the perspective of the male partner. In fact, most research on relationship issues and HIV/AIDS risk has focused on women rather than on men or couples.
Certainly, there may be more precise ways to measure the power constructs that we propose to be important (e.g., economic history rather than current employment to measure resource power). Additionally, the operationalization of power tested with our model is clearly not exhaustive. Thus, there may be other power constructs that could be important predictors of self-protective behavior, such as dependency and social isolation. Also, in order to focus on the effect of power on self-protective behavior, we did not include other established predictors of risk behavior in the model (e.g., desire to become pregnant, childhood sexual abuse). Future research should explore the effects of power relative to these other predictors.
In addition to implications for research and prevention, the findings from this study have implications for the theoretical models traditionally employed to explain HIV risk behavior. Since condom negotiation and use must be carried out in the context of a couple, some researchers have questioned the use of social psychological theories that focus on individual behavior, such as the health belief model (HBM; Becker, 1974), the theory of reasoned action (TRA; Fishbein and Ajzen, 1975), social–cognitive theory (Bandura, 1986), the AIDS risk reduction model (AARM; Catania et al., 1990), and the information–motivation–behavioral skills model (IMB; Fisher and Fisher, 1992). Instead of declaring that the traditional theories are inappropriate, we propose that future research consider power as a potential moderator of these models. Using the IMB model as an example, we hypothesize that women with high levels of power would indeed be able to take advantage of knowledge, motivation, and behavioral skills in order to engage in AIDS-preventive behaviors. However, women with very low levels of power might have difficulty engaging in protective behavior regardless of their level of knowledge, motivation, and skills. Thus, research to integrate concepts from traditional social psychological theory and the construct of power is crucial.
This work was funded through CDC Cooperative Agreement #U64/CCU200714. The authors wish to acknowledge the interviewers on the project, Carmen Ferraro and Minnie Ferraro. In addition, we wish to thank the health clinic which assisted in our recruitment of participants for this project.