Strategies for improving the lives of US women aged 40 and above living with HIV/AIDS: an evidence map

We assembled a panel of eight Key Informants that included experts in the care of HIV patients, nationally recognized researchers, policy makers, state government (Department of Health) officials, and nationally recognized advocates for women who live with HIV/AIDS to share their perspective on the issues that the guiding questions address. Based on these discussions, we refined study eligibility criteria and data extraction items for the evidence map of empirical studies that apply to women over 40 who live with HIV/AIDS. We grouped the studies into three areas of interest: (Area 1) studies that measure the impact of or describe barriers to engaging with existing resources; (Area 2) studies that measure the impact of insurance coverage on outcomes related to engaging with a program or accessing care; and (Area 3) studies that assess the performance of diagnostics for co-occurring disease or assess the effects of treating co-occurring diseases or risk factors.

Eligible studies reported results stratified by age and included at least 75% of the participants were women 40 years and older who live with HIV, an operational age cutoff to identify premenopausal women who are less likely to bear children, and peri- and postmenopausal women. All patient-level and system-level outcomes were eligible. Eligible services provided medical, behavioral, or social support or were programs that bundled several such services.

All study designs were eligible, including qualitative studies, which informed on barriers to accessing care. We excluded studies that analyzed fewer than 10 women who are older than 40 and live with HIV/AIDS, because they were unlikely to yield precise or broadly applicable conclusions. We also excluded studies completed longer than 10 years ago, because older empirical data are less likely to be relevant to today’s setting. Because the main focus of the evidence map is to inform about the US setting, we did not include studies conducted exclusively in other countries. Finally, we excluded studies not reporting empirical data.

A medical librarian designed and implemented searches in PubMed, PsycINFO, CINAHL, and SocINDEX for terms related to HIV or AIDS, crossed with terms on interventions, policies, services, or programs. We limited the search to English language reports published after January 01, 2005, having at least one author with a US affiliation. The search was last run on February 10, 2016, and is reported in full in Additional file 1: Appendix 1 in the full report of the project [32]. After a pilot phase to ensure that the eligibility criteria was being correctly and uniformly applied, abstracts were single-screened by a human reviewer and double-checked against the predictions of the machine learning algorithms implemented in Abstrackr [33]. All potentially eligible citations were retrieved and screened in full text for eligibility. For these papers, we recorded reasons for exclusion. For papers on studies that were excluded only because they did not report subgroup-specific information or pertinent interaction analyses, we recorded the area of interest the paper fell into.

All data were extracted in a predefined electronic form by a single investigator. The form recorded bibliographic information; demographic details for women over 40 where available, and for all women in the study where not; along with details about the programs, barriers, insurance impact, and comorbidities. Due to the nature of the project, we did not extract results, nor did we make strength of evidence assessments for individual studies or for the whole evidence base.

Thirty-two papers were eligible in Area 1; 16 quantified the impact of strategies for engaging resources, and 18 were about barriers to accessing resources (one paper was about both). Additional file 1: Appendix B, Table S1 summarizes the characteristics of the women enrolled in these studies. Most studies were conducted in urban centers, though a few focused on rural populations or examined differences between rural and urban populations. Studies included an average of 70 women (nine included more than 200 women). Mean ages ranged between 36 and 57 years, and no study provided results in the elderly. No study reported the women’s menopausal status or the age at which they were diagnosed with HIV.

Among the 28 studies that reported information on the race/ethnicity of participants, the median proportion of African American, Latino/Hispanic, and White women was 78, 15, and 14%, respectively. Three studies enrolled a small proportion of Native American women, and one included Pacific Islander women. No study enrolled Asian women.

On average, 20% of women reported having a spouse or partner, 58% perceived having adequate social support, and 31% cared for a dependent, but such information was provided in less than a third of the studies. Among studies listing such information, the average proportion of women who had annual income less than approximately $10,000 were unemployed, had not finished high school, used illegal substances, or faced mental health challenges ranged between 31 and 75%. Only one or two studies reported on the proportion of women who were undocumented immigrants (one study), experienced violence (two studies), or had a history of incarceration (one study).

Cumulative data from seven studies reported that a median of 73% of women had some form of public insurance (Medicare or Medicaid), 15% had private insurance, and 27% had no insurance. However, this may have changed or be changing as the Affordable Care Act is implemented.

Finally, the vast majority of studies did not explicitly report information on sex of women’s sexual partners or gender self-identification. No study reported the proportion of women who have more than one concurrent sexual partner, four studies included women who were identified in the study as lesbian or bisexual (ranging from 6.6 to 33% of the study population), and one study enrolled only transgender women.

Sixteen papers on 13 unique studies measured the impact of engaging with various services or with programs bundling several services. Eight RCTs (described in 10 papers) randomized between 30 and 184 participants to receive a specific program or an alternative (no program or a different program). An additional six studies qualitatively assessed the impact of engaging with programs. Three were based on interviews of 15 to 21 participants [34‐36], a fourth examined a cohort of 364 participants in a stress-management intervention [37], and two examined programs aimed at reducing substance use in 46 and 76 participants, respectively [38, 39]. Details of all programs and measured outcomes are listed in Table 1.

Eighteen observational studies evaluated barriers to care in populations ranging from 17 to 1701 women. Almost all of the barriers studied involved engaging or retention in HIV care, though three studies exclusively assessed outcomes not related to HIV, including cancer screening (one study), accessing program services online (one study), and using the internet (one study) (Table 2). Of the 14 studies that evaluated barriers to accessing care or retention in care, the barriers were sociodemographic in 12, cultural in 6, psychosocial in 11, having to do with experience with incarceration in 1, having to do with medical history in 12, and having to do with mental history in 6. Two studies reported on barriers to cancer screening that were sociodemographic, psychosocial, and medical in nature [40, 41], and a third evaluated barriers to adherence to treatment in the context of a pharmacy program that were psychosocial and medical [35]. Two studies, both by Blackstock et al., evaluated barriers having to do with using the Internet, either in general or to access social support. These studies identified barriers from nearly every aspect (sociodemographic, cultural, psychosocial, medical history, and mental history) [42, 43], which may inform the design of Web-based programs, as well as strategies to enhance their uptake and use.

The most commonly cited system- or provider-level factors were lack of transportation and logistical challenges, including navigating, wait times, clinic hours, insurance-related bureaucracy (six studies), and relationship with the provider (four studies).

Seven eligible papers evaluated the overall effect of insurance coverage levels on outcomes related to engaging with a program or accessing care. Six of these are included among the studies of barriers described in the previous section. Full details for the population are in Additional file 1: Appendix B, Table S2. Of these, six studies reported the impact of different insurance coverage on engagement with or retention in care, satisfaction with care, antiretroviral adherence, and Internet use. The seventh study evaluated the association between insurance coverage and successful management of major depressive disorder [44].

In no study did more than 18% of the women have private insurance. Where it was reported, between 60 and 100% of women had public insurance and between 7 and 35% had no insurance. In all studies, insurance coverage was one of several factors in a regression model for engaging with or accessing care. Although we did not perform risk of bias assessments, we deem it unlikely that such analyses can yield good estimates of the causal effect of insurance on outcomes. These studies were not designed to estimate the causal effect of insurance, and each study considered and controlled for different variables.

Eight papers (seven studies), with sample sizes between 46 and 1234 participants, were deemed eligible. Full details for the population are in Additional file 1: Appendix B, Table S3. No study informed on diagnosis or screening for comorbidities among older women living with HIV/AIDS; six reported information on the management of comorbidities; and two reported on predictive models of HIV progression or overall mortality in the presence of comorbid conditions.

Of the six studies that included information on management of comorbidities, two pertained to depression, three to alcohol and other substance abuse, and one to cardiovascular disease. Of the two studies on depression, one examined whether psychopharmacologic and psychotherapeutic treatment of depressed HIV-women met standards defined in the best practice literature and assessed predictors of standard-concordant care [44], and the other examined whether antidepressant use by HIV-positive patients results in better employment rates [45]. Among the three studies of drug and alcohol abuse, one evaluated a family-based intervention (Structural Ecosystems Therapy) compared to a psychoeducational intervention [46, 47]. A second looked at associations between heavy alcohol consumption and antiretroviral treatment initiation and mortality [48], and a third evaluated a brief intervention to reduce or eliminate alcohol consumption in seropositive women with heavy drinking problems [49]. Finally, a cohort study reported rates of acute myocardial infarction among HIV-positive patients with cardiovascular risk factors [50].

Two eligible cohort studies described models predicting HIV progression or overall mortality among older women living with HIV/AIDS who also have comorbidities. The first predicted clinical progression of HIV disease based on illicit drug use and various sociodemographic factors (such as age, race, and sexual history) and medical history factors (such as antiretroviral use, CD4 cell counts, and CDC category of HIV disease at baseline) [51]. The second predicted all-cause mortality among older women living with HIV/AIDS who also have substance and alcohol abuse problems based on sociodemographic factors (age and race), medical history factors (including using various drugs, alcohol and tobacco, having hepatitis C, prior antiretroviral therapy, CD4 counts, and viral load), and mental health factors (including a history of depression, anxiety, or psychosis) [48].

This work has several limitations. We developed an evidence map to describe the amount and type of practically available evidence, but did not summarize findings of eligible studies. For feasibility, we did not consider research completed longer than 10 years ago, because older empirical data are less likely to be relevant to today’s setting and for informing the future research agenda. We decided not to search EMBASE or any other large, international general database, because we deemed that the added benefit of searching multiple targeted databases was more cost and time efficient and would yield more relevant studies than a second large general database search [60].

We did not conduct a detailed analysis of the 37 eligible papers, prioritize outcomes by their importance, or do a detailed analysis of the risk of bias or the strength of evidence of individual studies or of the entire evidence base, because these tasks were not relevant to the goal of this project, which was to describe the literature in broad terms.

Because the main focus of the evidence map was to identify studies that inform about the US setting, we did not consider studies conducted exclusively in other countries. To operationalize this, we limited our search to papers in which at least one author had a US affiliation. In screening, we excluded a study only if it reported that it had been done exclusively outside the USA.

Because of limited time and resources, screening and data extraction were done by a single experienced reviewer. We believe that the overall picture of the research is accurate.

The 37 eligible studies do not represent a mature evidence base. The evidence base is even more sparse if one considers that the identified studies focused on different research questions and cannot possibly cover the range of practically implementable interventions that should be examined. By their nature such interventions are complex, in that they have many versions, and their implementation and deployment can depend on the targeted population.

More generally, when the range of questions of interest is so large, and the directly applicable evidence base is so limited, most, if not all, questions can be characterized as under-researched. The following are some general observations, which may help those looking to prioritize this research: