“Distorted Into Clarity”: A Methodological Case Study Illustrating the Paradox of Systematic Review

Abstract

Systematic review is typically viewed in the health sciences as the most objective—that is, rigorous, transparent, and reproducible—method for summarizing the results of research. Yet, recent scholarship has shown systematic review to involve feats of interpretation producing less certain, albeit valuable, results. We found this to be the case when we tried to overcome the resistance to synthesis of a set of qualitative and quantitative findings on stigma in HIV-positive women. These findings were difficult to combine largely because of fuzzy conceptualizations of stigma and the volume of unique quantitative findings. Our encounter with findings resistant to synthesis heightened our awareness of the extent to which all systematic reviews are accomplished by practices that paradoxically “distort [research findings] into clarity.”

Systematic review is typically viewed in the health sciences as the most objective—that is, rigorous, transparent, and reproducible—method for summarizing the results of research. The increasing incorporation of qualitative research findings into systematic reviews (e.g., Dixon-Woods, Agarwal, Jones, Young, & Sutton, 2005; Harden & Thomas, 2005; Sandelowski, Voils, & Barroso, 2006), and intensive study of the systematic review process (Hammersley, 2001; MacLure, 2005; Moreira, 2007; Mykhalovskiy, 2003; Sandelowski, Voils, & Barroso, 2007), however, have shown systematic review to entail feats of interpretation producing less certain, albeit valuable, results. The assumptions and work practices of systematic reviews have been increasingly challenged. Moved to center stage are the backstage judgments and compromises required to make research reports “docile” to the review process (Moreira, p. 181), the “comparability work” required to make incomparable findings combinable (Sandelowski et al., 2007); and, the reading and writing practices that undermine claims of the greater objectivity of systematic review over other kinds of reviews of research (Sandelowski, in press). As we illustrate in this methodological case study, reviewers are compelled paradoxically to “distort [research findings] into clarity” (Law, 2004, p. 2) to render them malleable to synthesis. The viability of the systematic review process depends on these distortions.

Method

This article is an account of our efforts to synthesize findings on stigma contained in 38 reports of studies conducted with HIV-positive women. These reports were retrieved in an ongoing study, the purpose of which is to develop methods to synthesize qualitative and quantitative research findings in health sciences research. These 38 reports are part of a larger and still growing collection that includes also reports of studies of antiretroviral adherence and of studies explicitly examining the link between HIV-related stigma and antiretroviral adherence. We selected the topics of HIV-related stigma and of antiretroviral adherence because of their significance to the advancement of healthcare for HIV-positive women, their primacy in HIV/AIDS research, and their fit with our respective backgrounds in HIV/AIDS care (JB), social psychology (CV), and women’s health (MS). Because of the primacy of these topics in HIV/AIDS research, we assumed also that we would have studies with enough methodological diversity (i.e., varieties of qualitative, quantitative observational, and intervention studies) to achieve the central aim of our project, namely, to develop methods to synthesize qualitative and quantitative research findings.

The 38 reports featured in this paper are of empirical qualitative and quantitative studies focused on HIV-related stigma conducted with HIV-positive women living in the United States of any race/ethnicity, class, or nationality. We accepted reports of studies conducted with HIV-positive women that included HIV-positive men if an explicit research purpose—as indicated in the report title and abstract—was to examine sex/gender differences as they related to HIV-related stigma and if the findings featured those differences. We accepted also reports of studies conducted with HIV-positive women that included HIV-negative women, if comparisons of these two groups of women were explicitly featured in the findings on stigma. To accommodate the way research purposes are conceived and questions are asked in both qualitative and quantitative studies, we defined studies of HIV-related stigma broadly to include research reports featuring findings addressing how HIV-positive women understood, experienced, or managed the social consequences of being HIV-positive, and predictors and covariates of, or interventions to manage, stigma in women. Among these reports are those addressing disclosure of HIV status. We searched for reports published in or after 1997 to account for the advent of antiretroviral therapy, as it was instrumental in changing HIV infection from a mortal to a chronic disease and, thereby, likely altered social attitudes toward HIV infection and HIV-positive persons. All of the delimitations we set were primarily to obtain a sample size and composition with enough methodological diversity yet topical uniformity to accommodate the detailed analysis and experimentation required by a methods study, not to produce a comprehensive synthesis of findings per se.

The 38 HIV-related stigma reports accepted for inclusion were published between 1997 and 2006. They were retrieved in July of 2005 and in June and August of 2006 using electronic databases housing citations to literature across the health, behavioral, and social sciences. (Further information about databases and search terms used is available from the authors on request.) We augmented electronic searching with the use of ancestry and descendency approaches (Cooper, 1998), in addition to hand-searching the print collections of our university libraries for books, anthologies, and other documents on HIV/AIDS likely to include relevant reports. These 38 reports of 15 qualitative and 23 quantitative studies are profiled in Table 1 and cited (*) in the reference list.

Table 1.

Profile of Reports (N=38)

Ref. #	Report, publication year1	Method2	Sample size & racial/ethnic composition3
Qualitative Reports (n=15)
1	Black & Miles, 2002	Qualitative description	48 (all AA)
2	Carr & Gramling, 2004	Ethnographic	9 (all W)
3	Chin & Kroesen, 1999	Qualitative	9 (all A/PI)
4	Gielen, O’Campo, Faden, & Eke, 1997	Qualitative	50 (43 AA; 5 W; 1 H; 1 O)
5	Grove, Kelly, & Liu, 1997	Ethnographic	22 (18 W; 4 O)
6	Ingram & Hutchinson, 1999	Grounded theory	18 (9 W; 8 AA; 1 H)
7	Lekas, Siegel, & Schrimshaw, 2006	Thematic analysis	158 (68 AA; 46 W; 44 PR)
8	Murphy, D. A., Roberts, & Hoffman, 2002	Ethnographic	47 (23 AA; 10 M/O; 6 H; 4 W; 2 NA/AL; 2 A/PI)
9	Murphy, D. A., Roberts, & Hoffman, 2003	Qualitative description	47 (23 AA; 10 M/O; 6 H; 4 W; 2 NA/AL; 2 A/PI)
10	Ortiz, 2005	Qualitative description	19 (all H)
11	Schrimshaw & Siegel, 2002	Thematic analysis	45 (15 AA; 15 PR; 15 W)
12	Serovich, Kimberly, & Greene, 1998	Grounded theory	13 (7 W; 3 AA; 3 H)
13	Siegel, Lekas, & Schrimshaw, 2005	Thematic analysis	158 (68 AA; 46 W; 44 PR)
14	Stanley, 1999	Fieldwork	15 (all W)
15	Vallerand, Hough, Pittiglio, & Marvicsin, 2005	Qualitative description	35 (31AA; 4 NS)
Quantitative Reports (n=23)
16	Abel, Rew, Gortner, & Delville, 2004	Pilot intervention	11 (4 AA; 6 W; 1H)
17	Armistead, Morse, Forehand, Morse, & Clark, 1999	Descriptive/correlation	100 (all AA)
18	Armistead, Tannenbaum, Forehand, Morse, & Morse, 2001	Descriptive/correlation	87 (all AA)
19	Clark, Lindner, Armistead, & Austin, 2003	Descriptive/correlation	98 HIV+ & 146 HIV-negative women (all AA)
20	Comer, Henker, Kemeny, & Wyatt, 2000	Descriptive/correlation	176 (72 AA; 57 H; 47 W)
21	Derlega, Winstead, Greene, Serovich, & Elwood, 2002	Descriptive/correlation	145 total: 105 HIV+ men (59 W; 38 AA; 3 O; 2 H; 3 MD) & 39 HIV+ women (27 AA; 9 W; 2 H; 1 O) &1 sex-unknown HIV+ person
22	Gielen, Fogarty, et al., 2000	Descriptive/correlation	257 (236 AA; 21 O?)
23	Gielen, McDonnell, Burke, & O’Campo, 2000	Descriptive/correlation	310 (293 AA; 17 O)
24	Jeffe et al., 2000	Descriptive/correlation	202 total: 106 HIV+ men (55 AA; 51 W) & 96 HIV+ women (75 AA; 21 W)
25	Kalichman, DiMarco, Austin, Luke, & DiFonzo, 2003	Descriptive/correlation	331 total: 233 HIV+ men & 98 HIV+ women (238 AA; 76 W; 17 O)
26	Kirschenbaum & Nevid, 2002	Descriptive/correlation	56 (28 AA; 23 H; 5 W?)
27	Letteney & LaPorte, 2004	Descriptive/correlation	88 (37 AA; 37 H; 11 W; 1 M?)
28	Mizuno, Moneyham, Sowell, Demi, & Seals, 1998	Descriptive/correlation	256 (209 AA; 46 O; 1 MD)
29	Murphy, D.A., Steers, & Dello Stritto, 2001	Descriptive/correlation	135 (53 AA; 35 H; 28 W; 11 NA/AL; 8 M/O)
30	Murphy, L. M., Koranyi, Crim, & Whited, 1999	Descriptive/correlation	40 (22 W; 16 AA; 2 M)
31	Ostrom, Serovich, Lim, & Mason, 2006	Descriptive/correlation	45 HIV+ women (36 AA; 8 W; 1 H) & 45 children
32	Pilowski, Sohler, & Susser, 2000	Descriptive	29 (21 AA; 7 H; 1 W)
33	Sayles, Wong, & Cunningham, 2006	Descriptive/correlation	1910 total: 1378 HIV+ men & 532 HIV+ women (?race distribution)
34	Shaffer, Jones, Kotchick, Forehand, & The Family Health Project Research Group, 2001	Longitudinal/descriptive/Correlation	99 HIV+ women (all AA) & 99 HIV- children
35	Simoni, Demas, Mason, Drossman, & Davis 2000	Descriptive/correlation	143 (106 AA; 37 B/H)
36	Sowell, Lowenstein, et al., 1997	Descriptive/correlation	82 (56 AA; 26 W)
37	Sowell, Seals, Phillips, & Julious, 2003	Descriptive/correlation	322 (266 AA; 35 O?)
38	Vanable, Carey, Blair, & Littlewood, 2006	Descriptive/correlation	221 total: 124 HIV+ men & 97 HIV+ women (93 AA; 101 W; 27O)

Notes.

¹Upublished reports were not included as they offered no methodological variation not found in published reports.

²We found only one pilot intervention study conducted with HIV-positive women to address stigma.

³All women were HIV-positive unless otherwise noted.

AA=African American; B/H-Black/Hispanic; A/PI=Asian/Pacific Islander; H/PR=Hispanic/Puerto Rican; M=Mixed race; NA/AL=Native American/Alaskan Native; MD=Missing data reported; NS=Not stated; O=Other or race/ethnicity not stated; W=White; ?=error or unclear in report

We extracted from each of these reports information on research purpose, theoretical framework, design, and methodology. We used metasummary techniques and taxonomic analysis (Sandelowski & Barroso, 2007) to extract, group, and conceptually order findings from the qualitative reports. We extracted from the quantitative reports all relationships in which an empirical indicator of stigma or disclosure appeared as an independent or dependent variable, and computed their effect sizes (Cohen’s d, or the standardized difference in means; Cohen, 1988). Each report was reviewed by the first three authors and all subsequent analytic work was conducted by at least two members of the research team. Rules for extraction and computation of findings were developed in meetings of the entire research team. Extractions generating uncertainty or disagreement were addressed also in team meetings. We excluded no report that met our selection criteria, as we considered every report potentially valuable to delineate some methodological point. As Pawson (2006) observed, the value of a report for any systematic review can be determined only in the course of conducting the review. We are here drawing a distinction between quality and utility, as even the methodologically weakest studies may have value for knowledge development. Moreover, our lean was toward inclusion of reports as the long-term aim of our study is to expand the methodological options available to accommodate the diversity in health sciences research.

Challenges to Synthesis of the Stigma Findings

The 38 reports of stigma presented sizeable challenges to our synthesis efforts involving the conceptualizations of stigma and disclosure, respectively, and the prevalence of unique over shared findings.

Conceptualization of Stigma

The most important of these challenges involved the use of the term stigma. Stigma functions in these reports as a “‘sensitizing concept’… the character or existence of which is not itself subject to scrutiny” (Manzo, 2004, p. 404). Reports contain passing, or “ceremonial” (Wickes & Emmison, 2007, p. 312), references to Goffman’s (1963) master work on stigma. Readers are presumed to know what stigma means and stigma is presumed to mean just about anything negative associated with HIV infection in the experiential or psychosocial realm.

In addition to the challenge of “conceptual inflation” (Deacon, 2005, p. ix), whereby a term comes to mean almost anything, was the problem of conceptual conflation. Here findings addressing what are usually referred to as women’s “perceptions of stigma” blur three different domains: (a) general knowledge that HIV-positive persons are often treated poorly; (b) fear of being treated poorly; and, (c) actual personal experiences with poor treatment. This blurring reflects the largely atheoretical nature of the studies reviewed, as neither data collection nor analysis were typically directed toward differentiating between felt and enacted stigma, or between what the affected persons fear or feel and what others actually do to the affected person, respectively (Jacoby, 1994). Such events as social rejection, discrimination, and violence, are similarly blurred as they are variously treated as both empirical manifestations and outcomes of stigmatization.

The way stigma appears in the reports reviewed reflects the generally “fuzzy” (Grönvik, 2007) conceptual status of research on stigma. Since Goffman (1963) introduced it as a subject for study, stigma has become one of the most used and, arguably, overused concepts in the health and social sciences (Link & Phelan, 2001; Manzo, 2004). Widely deployed as an “interpretive…explanatory…(and) discursive resource” (Manzo, p. 402), stigma has become a “social scientific gloss” (p. 402) revealing less about the experiences of the persons interpreted or explained by it and more about the work practices of the researchers who rely on it for interpretation or explanation. Reidpath, Chan, Gifford, and Allotey (2005, p. 470) referred to the conceptual development of stigma as “patchy” and proposed that stigma research may have “fallen into a stagnant and unpromising form of… Kuhnian…normal science.”

Nowhere is this more apparent than in the research on living with HIV infection. Stigma has become virtually synonymous with the infection itself (Parker & Aggleton, 2003). Manzo (2004, p. 408) proposed that HIV/AIDS was the condition most studied because it meets virtually all of the “stigma-identifying qualities” researchers have deemed appropriate for the study of stigma, including visibility, severity, contagiousness, culpability, difference, and deviance. Indeed, HIV/AIDS is considered to have such “unique fittingness as a stigma-relevant topic” that it has “earned (its) own form of stigma, (namely) HIV-related stigma.”

So conflated is HIV infection with stigma that a search for literature on stigma in the context of HIV infection will exclude few reports of studies of HIV infection in the experiential, psychological, or sociocultural domains. The virtual fungibility of HIV and stigma challenged us even to decide what studies would constitute the body of research on stigma. The delimitations we set for the selection of stigma studies were, in part, to address this problem. For example, we debated whether to include reports of studies conducted with HIV-positive women with findings on social support and interaction, emotions and mood, and psychological functioning, as these findings were often presented as implying the operation of stigma. Yet, to include these studies would have meant excluding virtually no study in the experiential and psychosocial realm. In a previous systematic review we had conducted of all published and unpublished qualitative studies addressing the experience of living with HIV infection in HIV-positive women, 93 of the 114 reports retrieved contain findings pertaining to stigma, yet only 16 of the studies featured in these reports had been targeted toward the study of stigma (Sandelowski, Lambe, & Barroso, 2004).

Conceptualization of Disclosure

A second challenge was that although stigma is central to the 38 reports reviewed, it is actually disclosure of HIV status that is empirically center stage. Disclosure is seen to occur against the background of stigma, which is assumed to drive or inform all thoughts, feelings, and actions related to whether, whom, when, and what to tell. Disclosure is treated as a proxy for or index of stigma even when it may not be related to stigma. This is best illustrated in reports of studies addressing mothers’ disclosure of their own HIV infection to their children. In these reports, mothers’ disclosure is tied primarily to the developmental capacity of the child to handle any serious illness in or mortal threat to their mothers. Disclosure here may, therefore, be better understood within a child development, as opposed to stigma, framework. Even the finding that children’s perceived inability to keep a secret is a typical reason mothers did not disclose their HIV infection to them may have been related more to the women’s views of privacy and habits of self-disclosure than they were to stigma per se.

Like stigma, disclosure in these reports is difficult to pin down conceptually. Classifications of disclosure are not in the same semantic plane. For example, terms like selective and full simultaneously refer, in the same report and across reports, both to the (a) type or number of persons to whom HIV infection was disclosed and to (b) what or how much of something about HIV was disclosed. Although disclosure is recognized as a process occurring over time, in most of the reports, it is treated as if it were a discrete yes/no event in time. Most of the findings were from cross-sectional analyses, including those from longitudinal parent studies.

Prevalence of Unique Over Shared Findings

A third challenge, due largely to the way stigma and disclosure were treated conceptually in the 38 reports, was extracting findings that actually address the same thing. After extracting 488 relationships in the quantitative reports (393 bivariate and 95 multivariable), we searched for instances in which the same relationship was measured in at least two studies with the intention of synthesizing them using meta-analysis. Yet, we found only six instances in which the same relationship was measured more than once and, in all of these, the relationship was measured only twice. Moreover, because of the different ways in which the independent variables were measured and treated, we were able to pool the effect sizes for only 2 of the 6 instances. In short, although the quantitative reports all ostensibly address the same topic (i.e., stigma and disclosure in HIV-positive women), their findings were not synthesizable either by meta-analysis or by a modified vote counting procedure we had used in another study (Voils, Sandelowski, Barroso, & Hasselblad, 2008).

Meta-analysis and vote counting are based on an assimilation logic requiring that at least two relationships obtained from at least two different samples be deemed the same for them to be combined statistically. Yet merely having any one finding repeated in only two studies does not produce an informative quantitative synthesis of quantitative findings. This is in contrast to the treatment of qualitative findings, where a finding that appears in only one report and in one case in that report, will be accounted for in a research synthesis. The mandate in qualitative research synthesis is that findings not be differentially weighted or treated on account of sample size. Qualitative research findings do not derive their significance from repetition, but rather from their capacity to illuminate or explain.

Because we could not produce a meaningful assimilation of quantitative findings, we attempted to create a configuration of findings. As we have described elsewhere (Voils et al., 2008), in contrast to assimilations of findings entailing the actual incorporation of recurring findings into each other, configurations of findings are coherent arrangements of findings that do not require repetition to be included in the arrangement. Whereas assimilated findings are merged, configured findings are meshed into taxonomies, theories, conceptual maps, meta-narratives, lines of argument, or other coherent arrangements. The taxonomy we created from the qualitative findings is an example of such an arrangement.

We, therefore, tried to apply a configuration logic to the quantitative findings, whereby findings with at least a small effect size (d ≥ .20; Cohen, 1988) could be assembled into a coherent arrangement without regard to whether they appeared in only one report. But this effort was difficult too because the quantitative reports were characterized by large numbers of unique findings even within reports. The quantitative reports contain more unique findings (i.e., appearing in only one report) than the qualitative reports, which, in turn, contain more sets of repetitive findings (i.e., appearing in two or more reports) than the quantitative reports. Even taking into account the difference in numbers of qualitative (15) and quantitative reports (23), this is contrary to expectations. Qualitative studies in a shared domain of research are expected to have more unique findings and, therefore, to produce a less comparable dataset than quantitative studies because of the open-ended nature of data collection in qualitative research. In our dataset, all 45 bivariate relationships addressing various forms of abuse, by intimate partner or other perpetrator, before and after HIV diagnosis, appear in only one report (Gielen, McDonnell, Burke, & O’Campo, 2000). All 87 relationships addressing reasons for (non)disclosure to persons other than children appear in only one report (Derlega, Winstead, Greene, Serovich, & Elwood, 2002), which contributed almost 18% of the relationships across reports (87 of the total 488). These relationships consist of various combinations of 5 reasons for disclosing, 6 reasons for not disclosing, and 3 types of persons disclosed to (parent, intimate partner, and friend), in both women and men together and in women in comparison to men.

Overcoming the Challenges to Synthesis

To overcome these challenges, we worked to find ways to make the findings more comparable to make them cohere. An excerpt of the end-product of our labors is shown in Table 2.

Table 2.

Alignment of Related Qualitative and Quantitative Findings (Excerpts)

HIV-Related Stigma

Qualitative findings

Strategies to manage stigmatization included normalization,5 education of others about HIV,7,14 helping others prevent infection,1 public advocacy and support of other HIV-positive persons,1,14 attributing HIV-stigma to sexism and/or racism,5,7 identity management,2,5,6,14 and information management.all

Quantitative findings

Among Black women, knowledge of HIV as stigmatizing, felt stigma, and enacted stigma were positively associated with psychological functioning (.847mv) and number of persons disclosed to (−1.036mv).19

Felt Stigma

Qualitative findings

Felt stigma consisted of HIV-positive women’s own feelings of shame, devaluation,1,2,3,7,13,14 fear of rejection and social ostracism,1,4,5,6,9 fear for personal safety,4 fear of discrimination,5 and negative Redefinition of self as HIV-positive.2,3

Quantitative findings

Women with (versus without) felt stigma had greater odds of identifying a support person (.614mv).28

A greater percentage of Black women than White women feared rejection by their family (.453).36

Enacted Stigma

Qualitative findings

Enacted stigma consisted of others’ rejecting behaviors3,4,6,11 discrimination,7,11 and violence.4

Quantitative findings

The number of persons disclosed to increased the odds of enacted stigma (.279mv), but having instrumental social support decreased the odds of enacted stigma (−.300mv).22

Being abused by a perpetrator prior to HIV diagnosis increased the odds of being abused after diagnosis (.476mv).23

Timing of Disclosure

Qualitative findings

Timing of disclosure was immediately after diagnosis,10 delayed until women accept/adjust to diagnosis4 or child is developmentally ready1,10,11,15 or until assess impact on others,4 before imminent disclosure by others (preemptive disclosure),10 or at any opportune moment, when it felt right.1

Quantitative findings

Among Black women, the number of months between diagnosis and disclosure to father (−.222), to extended family (−.239), to children (−.263), to partner (−.433), and to mother (−.449) was shorter than the number of months between diagnosis and disclosure to friends.17

Target of Disclosure

Qualitative findings

Women variously chose to restrict or not to restrict disclosure to children, partners, family, friends, acquaintances, employers, co-workers, religious leaders, and other HIV-positive persons.all

Quantitative findings

A greater percentage of women with AIDS (.524) or who were symptomatic (1.490) disclosed to their sexual partners compared to asymptomatic women, and a smaller percentage of women with AIDS disclosed to their sexual partners compared to symptomatic women (1.490).36

A greater percentage of women than men disclosed to extended family (.215) and to mother (.326).25

Benefits of, Reasons for Disclosure (excluding children)

Qualitative Findings

Conditions, or rules, for disclosure included the need to know,1 right to know/duty to inform,13 capacity or readiness to know, 1,10 likelihood of burden to others from knowing,1 degree of acceptance of HIV, trustworthiness, ability to keep a secret,1 physical proximity, intimacy, and quality of relationship with other,3,10 and perceived benefit versus risk of harm to self or others.1,10

Perceived benefits of disclosure, or reasons to disclose included positive social support,1,3,4,5,6,10,11,13 reconciliation with family,10 avoidance of stress/burden of secrecy, catharsis, relationship authenticity,2,11,13 control of disclosure process,10 minimization of felt stigma,2,14 and healing, wellbeing, improved quality of life, growth, renewed interest in life.3,4

Quantitative Findings

Testing other’s reactions was endorsed to a greater extent by women than men as a reason for disclosing to an intimate partner (.389) and to a friend (.562).21

Catharsis was endorsed to a greater extent by women than men as a reason for disclosing to a parent (.401), friend (.563), and to an intimate partner (.637).21

Similarity (having something in common) was endorsed to a greater extent by women than men as a reason for disclosing to a parent (.487).21

Risks of, Reasons for Nondisclosure (excluding children)

Qualitative Findings

Perceived risks of disclosure, or reasons not to disclose included loss of positive or negative social support,3,13 stigmatization,2,11 disappointment, burden, suffering, stigmatization of loved ones,1,3,6 culturally incongruent roles (adult child cared for by parent instead of reverse),3 loss of privacy,4 and uncontrolled (unplanned) disclosure.1,7,10,13

Quantitative Findings

Protecting the other as a reason for not disclosing to a friend was positively associated with knowledge of HIV as stigmatizing among women (.402) and among men (.246); as a reason for not disclosing to a parent was positively associated with knowledge of HIV as stigmatizing among women (.928) and among men (.723); and as a reason for not disclosing to an intimate partner was positively associated with knowledge of HIV as stigmatizing among women (.254).21

Self-blame as a reason for not disclosing to an intimate partner was positively associated with knowledge of HIV as stigmatizing among women (.270) and among men (.272); as a reason for not disclosing to a friend was positively associated with knowledge of HIV as stigmatizing among women (.523) and among men (.415); and as a reason for not disclosing to a parent was positively associated with knowledge of HIV as stigmatizing among women (1.317) and among men (.711).21

Content of Disclosure

Qualitative Findings

Disclosure was full (everything), partial, or totally concealed.2,5,6

Partial disclosure or total concealment was enabled by virtue of not being seen as at risk for HIV,5 having no visible signs of illness, covering, passing, and by circumventing the need to disclose.6

Quantitative Findings

(All quantitative findings on content of disclosure are in relation to children.)

Disclosing Maternal HIV to Children

Qualitative Findings

Reasons for disclosing to children included perceived ability of older children to understand or cope with the effects of maternal HIV,1,8,9,10,11,15 perceived ability of older children to keep HIV a secret,8,11 desire to protect children from transmission and from courtesy stigma,8,11,15 felt obligations to children (right to know, to tell the truth, should hear it from mother),9,11,15 desire to preserve good relations with children and identity as good mother,8,11 and mothers’ own health and need for help.9,11,15

Quantitative Findings

The odds of needing more physical assistance or social support were greater among mothers who disclosed to children than mothers who did not (.366mv); mothers who disclosed to children had greater personal feelings of social support than mothers who did not (.544mv).29

A greater percentage of mothers with AIDS than symptomatic mothers (.293)27 and asymptomatic mothers (.386) 27 & (.330)18 disclosed to children; a greater percentage of symptomatic mothers than asymptomatic mothers disclosed to children (.444).18

Not Disclosing Maternal HIV to Children

Qualitative Findings

Reasons for not disclosing to children included perceived inability of young children to understand or cope with the effects of maternal HIV,1,9,10,11,15 perceived inability of young children to keep HIV a secret,1,8,11 desire to protect children from transmission and from courtesy stigma,8,11,15 desire to preserve good relations with children and identity as good mother,8,11 and disclosure seen as too threatening.11

Quantitative Findings

Not disclosing to children because of not wanting to burden them (1.036), fear that child will tell others (1.124), and to keep information from them (1.036) were positively associated with felt stigma.31

Not disclosing to children because HIV was deemed personal (1.008) was positively associated with knowledge of HIV as stigmatizing (1.008).31

Note. Superscript numbers refer to reports listed in Table 1. All relationships are bivariate, except where indicated as multivariable (mv).

We translated similar ideas into a common language. For example, we used the words selective and full consistently to refer to just one idea, namely, the content of disclosure, and restricted and unrestricted to refer to the persons to whom HIV infection was disclosed. We distinguished among instances of felt stigma, enacted stigma (Jacoby, 1994), and general knowledge of HIV-related stigma whenever simply stigma or perceived stigma were used to refer to instances of two or more of these manifestations of stigma. We inferred from the way variables were operationalized and positioned in relation to each other the actual concepts addressed. This is another instance of translation, but one that addresses the problem of using one term to tap different aspects of a phenomenon (as opposed to more than one term to tap the same aspect). For example, in reports in which perceived stigma was stated as the concept but the measurement tools used to operationalize it indicated that knowledge of stigma was actually being tapped (as opposed to felt or enacted stigma), we renamed the concept as knowledge of stigma and then grouped it with other findings addressing knowledge of stigma. Had we not engaged in this form of “distortion” of the original text, we would have been summing highly divergent findings about stigma, albeit all named perceived stigma. In the studies of abuse mentioned previously, we conceived abuse, losing a job, losing a home, and family rejection as empirical instances of enacted stigma in an effort to impose some reasonable conceptual order on the findings and, thereby, to make them more amenable to combination with other findings.

Such translations exemplify distortions, or re-writings, of reports. As more fully described elsewhere (Sandelowski, in press), no matter whether a review includes reports solely of qualitative, quantitative, or of both qualitative and quantitative studies, data are never simply “extracted” as given in these reports, but rather they are “transformed, transposed, converted, tabulated…funnel- and forest-plotted… or otherwise manipulated, modified, and reconfigured” to permit the comparison of the previously incomparable and the combination of the previously uncombinable. Other examples of such transformations include the reciprocal translation of metaphors and concepts into each other in meta-ethnography (Noblit & Hare, 1988), and the calculation of effect sizes from different statistical expressions of findings in meta-analyses (Cohen, 1988), both of which also require reviewers to make individual judgments.

Conclusion

Any one body of literature selected for and subjected to systematic review presents reviewers with a distinctive set of challenges that will require a distinctive set of solutions, or may even resist solution. Overcoming the resistance to synthesis of the qualitative and quantitative findings in this particular dataset was challenging because of the different ways stigma and disclosure were conceptualized and treated. Synthesis was difficult also because of the volume of unique findings within and across individual quantitative studies that precluded our linking them in a meaningful way to the qualitative findings (e.g., to create a conceptual model). That said, there is some evidence that the developmental capacity of children influenced mothers’ disclosure of their own HIV status to their children, and that the more HIV symptomatology, the more likely mothers were to disclose. The synthesized qualitative finding—that disclosure decision-making was a dynamic, taxing, and continuous evaluative process—encompass all of the quantitative findings explicitly linking persons disclosed to, content and timing of disclosure, and reasons for (non)disclosure with a host of demographic, clinical, psychological, and other variables. Whether these conclusions constitute a meaningful or even worthwhile synthesis of qualitative and quantitative findings on stigma in HIV-positive women is for users of reviews to decide.

Given the familiar lament in reviews of the literature and discussion sections of research reports about the need for conceptual clarity and consistency in measurement in research, reviewers of research findings will likely encounter more or less the same challenges we encountered, no matter what the domain of research. This raises the questions yet to be fully answered of other challenges that reviewing qualitative and quantitative research may present and of whose responsibility it is to make research findings more docile to the systematic review process: reviewers of research findings or the researchers who create them.

Any coherence the results of our work have is because of the considerable effort we put into rewriting, modifying, and re-assembling the results in the reports reviewed, that is, paradoxically distorting them: not to misrepresent, but rather more faithfully to represent, them. Although typically hidden in claims for the rigor and transparency of systematic review, the process is inescapably a highly subjective, albeit disciplined, encounter whereby reviewers “deconstruct,” in order to “reconstruct,” the information in reports into a form useable for review (Harden et al. 2004, p. 796). This is the case even in meta-analyses of the results of clinical trials in which what ends up being compared and combined, and the manner in which this is accomplished, are outcomes of reviewer and user judgments about what constitute meaningful comparisons and combinations (Cooper, 1998; Deeks, Higgins, & Altman, 2006; Sandelowski et al., 2007). Highlighting the subjectivity of systematic review is in no way an argument against it, but rather an exercise in the transparency toward which it is directed.