Structure and quality of outpatient care for people living with an HIV infection

ABSTRACT

Policy-makers and clinicians are faced with a gap of evidence to guide policy on standards for HIV outpatient care. Ongoing debates include which settings of care improve health outcomes, and how many HIV-infected patients a health-care provider should treat to gain and maintain expertise. In this article, we evaluate the studies that link health-care facility and care provider characteristics (i.e., structural factors) to health outcomes in HIV-infected patients. We searched the electronic databases MEDLINE, PUBMED, and EMBASE from inception until 1 January 2015. We included a total of 28 observational studies that were conducted after the introduction of combination antiretroviral therapy in 1996. Three aspects of the available research linking the structure to quality of HIV outpatient care were evaluated: (1) assessed structural characteristics (i.e., health-care facility and care provider characteristics); (2) measures of quality of HIV outpatient care; and (3) reported associations between structural characteristics and quality of care. Rather than scarcity of data, it is the diversity in methodology in the identified studies and the inconsistency of their results that led us to the conclusion that the scientific evidence is too weak to guide policy in HIV outpatient care. We provide recommendations on how to address this heterogeneity in future studies and offer specific suggestions for further reading that could be of interest for clinicians and researchers.

KEYWORDS:

• HIV
• quality of health care
• treatment outcome
• ambulatory care facilities
• health personnel

Introduction

The availability of combination antiretroviral therapy (cART) has resulted in remarkable decreases in HIV-related morbidity and mortality (Thompson et al., 2012). Unfortunately, the complexity of the management of HIV infection has grown along with the advances in treatment (Hecht, Wilson, Wu, Cook, & Turner, 1999). Individuals infected with HIV face a lifetime of treatment with cART and monitoring of disease progression. Health-care providers are confronted with challenges related to cART including toxicities, drug–drug interactions and drug resistance, and comorbidities among the aging HIV-positive population.

In order to achieve optimal health outcomes, care for HIV-infected patients should be provided at health-care facilities and by care providers with sufficient expertise (Hecht et al., 1999). Previous studies have evaluated the impact of health-care facility and care provider characteristics on the quality of care for HIV-infected patients. However, questions regarding qualifications to treat HIV-infected patients remain. Ongoing debates include which settings of care (e.g., primary/secondary/tertiary facilities; facility volume) improve health outcomes, and how many HIV-infected patients a health-care provider should treat (provider HIV caseload) to gain and maintain expertise.

The first studies addressing these issues took place before the introduction of cART. Studies showed lower rates of mortality in facilities with greater volumes of HIV-infected inpatients and in facilities with units dedicated to HIV/AIDS care (Bennett et al., 1989, 1992, 1995; Cunningham, Tisnado, Lui, Nakazono, & Carlisle, 1999; Hogg et al., 1998; Stone, Seage, Hertz, & Epstein, 1992; Turner & Ball, 1992). In addition, greater clinician HIV experience was positively associated with survival (Kitahata et al., 1996; Laine et al., 1998).

With the advent of cART, the focus shifted from measurement of survival to assessment of virologic success and physician adherence to guidelines. Studies showed that patients treated by health-care providers with more training/expertise in HIV/AIDS care had greater plasma viral load control and were more likely to be on cART (Landon et al., 2003, 2005).

Currently, there are a growing number of studies assessing care delivery models and interventions in outpatient HIV care. The aims of the interventions, as well as the way they are operationalized, are very diverse. “Adherence support,” “case management,” “task shifting,” and “integration of care” are examples of widely studied interventions. They are beyond the scope of this review and have been reviewed separately elsewhere (Callaghan, Ford, & Schneider, 2010; Kredo, Adeniyi, Bateganya, & Pienaar, 2014; Lazarus, Safreed-Harmon, Nicholson, & Jaffar, 2014; Mdege, Chindove, & Ali, 2013; Soto, Bell, & Pillen, 2004).

In this article, we give an overview of the scientific literature linking health-care facility and care provider characteristics to the quality of HIV outpatient care. It is the first to focus exclusively on studies that took place after the introduction of cART. We used the structure–process–outcome framework (Donabedian, 1988) to distinguish between the characteristics and outcomes that have been evaluated. Structure refers to stable characteristics of health systems (e.g., staff and facility volume). Process describes the activities of care providers and patients in health care. Outcome refers to the effects of care on the health status of patients.

We aim to provide a critical evaluation of three aspects of the available research linking the structure to quality of HIV outpatient care: (1) assessed structural characteristics (i.e., health-care facility and care provider characteristics); (2) measures of quality of HIV outpatient care; and (3) reported associations between structural characteristics and quality of care. For this third part of the review, we focused exclusively on studies that measured virologic success.

Methods

We searched the electronic databases MEDLINE, PUBMED and EMBASE from the dates of their inception until 1 January 2015, using the search terms presented in Box 1. Titles, abstracts, and full text studies were screened. Reference lists of relevant studies were screened for additional relevant citations.

Box 1. Search terms

In order to be eligible, articles had to report an original observational research study with an adult HIV-infected population. We excluded studies that (i) were conducted before the introduction of cART in 1996 or (ii) focused exclusively on interventions, due to the great variety in their aims and operationalization.

We extracted general parameters (study design, country, and population); structural characteristics on the health facility level (setting, volume, specialization, level of care, and disciplines in care team), and characteristics on the health-care provider level (caseload, training, and experience). We distinguished clinical outcome measures (HIV-RNA levels, mortality, AIDS events, CD4 cell counts, and other) and process outcome measures (cART exposure; hepatitis B and C, tuberculosis (TB), and cervical cancer screening; pneumocystis jiroveci pneumonia (PCP) prophylaxis; vaccination for influenza and hepatitis B; and monitoring of CD4 cells and other). Finally, we extracted patient reported outcome measures (PROMs) and measures of adherence, retention in care, and care utilization.

Findings from the literature review were grouped into (1) structural characteristics; (2) quality of care outcome measures; and (3) reported associations between structural characteristics and the quality of care.

During the search process we found a big variation in health outcome measures. This complicated summarization and comparison of studies. We therefore performed a final selection and only evaluated associations among studies that measured virologic success, as this was the most consistently used outcome measure.

Results

Our search strategy identified 770 potentially relevant citations. Forty additional citations were identified through manual screening of the references. We excluded 751 citations after screening of titles and abstracts (not relevant, measured inpatient mortality, or had a study period in the pre-cART era). We retrieved 59 articles for full text screening. Subsequently, 31 articles were excluded because they were only descriptive or focused exclusively on interventions or patient characteristics.

In total, 28 articles were included. The majority were retrospective cohort (n = 12) (Assefa et al., 2012; Backus et al., 2010; Chan et al., 2011; Chu, Umanski, Blank, Grossberg, & Selwyn, 2010; Fatti, Grimwood, & Bock, 2010; Kerr et al., 2012; Kitahata et al., 2003; Landon et al., 2005; Odafe, Idoko, et al., 2012; Odafe, Torpey, et al., 2012; Sherr et al., 2010; Solomon, Flynn, & Lavetsky, 2005) or cross-sectional studies (n = 7) (Anaya, Yano, & Asch, 2004; Boyer et al., 2011; Horberg, Hurley, Silverberg, Kinsman, & Quesenberry, 2007; Saha et al., 2013; Shet et al., 2011; Wilson, Landon, Ding, et al., 2005; Wilson, Landon, Hirschhorn, et al., 2005). One study had a prospective cohort design (Sangsari et al., 2012), and eight studies did not explicitly state their design strategy. Most of the studies had taken place in the USA (n = 12) (Anaya et al., 2004; Backus et al., 2010; Chu et al., 2010; Horberg et al., 2007; Kerr et al., 2012; Kitahata et al., 2003; Landon et al., 2005; Rodriguez, Marsden, Landon, Wilson, & Cleary, 2008; Saha et al., 2013; Solomon et al., 2005; Wilson, Landon, Ding, et al., 2005; Wilson, Landon, Hirschhorn, et al., 2005) and countries in sub-Saharan Africa (n = 13) (Assefa et al., 2012; Balcha & Jeppsson, 2010; Boyer et al., 2010, 2011; Chan et al., 2011; Ekouevi et al., 2012; Fatti et al., 2010; Fatti, Grimwood, Mothibi, & Shea, 2011; McGuire et al., 2012; Odafe, Idoko, et al., 2012; Odafe, Torpey, et al., 2012; Reidy et al., 2014; Sherr et al., 2010). The remaining studies were situated in Brazil (Nemes et al., 2009), India (Shet et al., 2011), and Canada (Sangsari et al., 2012) (Figure 1).

Figure 1. Countries in which the identified studies were based (n = 28).

Structural factors

The most commonly evaluated health facility characteristic (Table 1) was number of HIV-infected patients in care (‘facility volume’, n = 7). Facility volume was generally defined as the number of HIV-infected patients in outpatient care. The number of categories and definitions varied: four categories: <25; 25–99; 100–299; and >300 patients (Backus et al., 2010); three categories: <15; 16–100; and 101–500 (Solomon et al., 2005) or <100; 100–500; and >500 patients (Nemes et al., 2009); and two categories: <950 and >950 patients (Fatti et al., 2011). The number of beds (Boyer et al., 2011) and number of outpatient visits per year (Wilson, Landon, Ding, et al., 2005) were also used to define facility volume.

Table 1. Summary of assessed structural characteristics (n = 28).

Structural characteristics	Sources (Author, year, countries)
Health facility characteristics
Facility volume (number of patients in care).	Anaya et al. (2004), Backus et al. (2010), Boyer et al. (2011), Fatti et al. (2011), Nemes et al. (2009), Solomon et al. (2005), Wilson, Landon, Ding et al. (2005) (Brazil, Cameroon, South Africa, the USA)
Facility specialization (exclusively treating HIV patients).	Nemes et al. (2009), Wilson Landon, Ding et al. (2005) (Brazil, the USA)
Health facility setting: including private vs. public funding, level of care, community-based vs. hospital care, and urban vs. rural setting.	Anaya et al. (2004), Assefa et al. (2012), Balcha and Jeppsson (2010), Boyer et al. (2011, 2011), Chan et al. (2011), Ekouevi et al. (2012), Fatti et al. (2010), McGuire et al. (2012), Odafe, Idoko, et al. (2012), Odafe, Torpey, et al. (2012), Reidy et al. (2014), Shet et al. (2011) (Cameroon, Ethiopia, India, Kenya, Malawi, South Africa, other countries in SSA)
HIV care team composition.	Horberg et al. (2007), Rodriguez et al. (2008) (the USA)
Care provider characteristics
Health-care provider expertise: specialization/training.	Assefa et al. (2012), Kerr et al. (2012), Landon et al. (2005), Rodriguez et al. (2008), Sherr et al. (2010), Wilson, Landon, Ding et al. (2005), Wilson, Landon, Hirschhorn et al. (2005) (Ethiopia, Mozambique, the USA)
Health-care provider expertise: HIV experience/ caseload.	Kitahata et al. (2003), Landon et al. (2005), Sangsari et al. (2012), Wilson, Landon, Ding et al. (2005), Wilson, Landon, Hirschhorn et al. (2005) (the USA, Canada)
Other characteristics (self-reported cultural competence).	Saha et al. (2013) (the USA)

Notes: SSA, sub-Saharan Africa.

Other health facility characteristics of interest were specialization (Nemes et al., 2009; Wilson, Landon, Ding, et al., 2005) and team composition (Horberg et al., 2007; Rodriguez et al., 2008). Facilities that only treated patients with Sexually Transmitted Diseases (STD) and HIV infection (Nemes et al., 2009), or that were regarded as a “specialized HIV practice” (Wilson, Landon, Ding, et al., 2005) by the medical director, were defined as specialized facilities. Studies focusing on team composition assessed the presence of a clinical pharmacist with specialization in HIV disease (Horberg et al., 2007), and treatment team diversity (7 categories based on the number of disciplines) (Rodriguez et al., 2008).

Thirteen studies evaluated the setting of the health facility. These studies assessed the impact of private versus public funding (Boyer et al., 2011; Ekouevi et al., 2012; Shet et al., 2011), level of care (primary/secondary/tertiary; hospital/clinic; or centralized/decentralized) (Assefa et al., 2012; Balcha & Jeppsson, 2010; Boyer et al., 2010, 2011; Chan et al., 2011; Ekouevi et al., 2012; Fatti et al., 2010; McGuire et al., 2012; Odafe, Idoko, et al., 2012; Odafe, Torpey, et al., 2012; Reidy et al., 2014), and location of care site (urban/rural) (Anaya et al., 2004; Balcha & Jeppsson, 2010; Boyer et al., 2010; Ekouevi et al., 2012).

The most commonly evaluated health-care provider characteristic was HIV-specific specialization (n = 7). The studies generally compared the performance of infectious disease specialists versus expert generalists and non-expert generalists. In addition, the quality of care provided by non-physician clinicians (nurse practitioners and physician assistants) was assessed.

HIV-related experience was another commonly studied provider characteristic. Experience was expressed in terms of number of HIV-infected patients previously treated by, or the number of patients currently under the care of the health-care provider (provider HIV caseload).

One study assessed self-reported cultural competence (CC) of health-care providers (Saha et al., 2013). Interventions to improve CC (the ability to understand and respect values, attitudes, and beliefs that differ across cultures) have gained attention, particularly in high-income countries with diverse population groups (Horvat, Horey, Romios, & Kis-Rigo, 2014).

Measures of quality of care

Table 2 shows that the most frequently used clinical outcome measures were virologic success (n = 11) and mortality (n = 9). Virologic success was generally defined as one undetectable HIV-RNA measurement with a threshold ranging from 50 to 500 copies/ml. A minority of studies looked at two consecutive measurements, or specified a time period after the initiation of cART. Other studies measured time to viral suppression. Mortality was assessed after a specific follow-up time (Balcha & Jeppsson, 2010; McGuire et al., 2012) or during the entire study period.

Table 2. Summary of assessed quality of care measures (n = 28).

Quality of care measures	Operationalization and definitions	Sources (Author, year, countries)
	Clinical outcomes
Virologic success	Predominantly one HIV-RNA measurement, with or without a specified period of time after cART initiation. Thresholds ranging from 50 to 500 copies/ml.	Backus et al. (2010), Chu et al. (2010), Fatti et al. (2010, 2011), Horberg et al. (2007), Landon et al. (2005), McGuire et al. (2012), Saha et al. (2013), Sangsari et al. (2012), Shet et al. (2011), Wilson, Landon, Hirschhorn et al. (2005) (Canada, India, Malawi, South Africa, the USA)
Mortality	Mortality after 12 or 24 months of follow-up, death during study period.	Assefa et al. (2012), Balcha and Jeppsson (2010), Chan et al. (2011), Fatti et al. (2010, 2011), Kitahata et al. (2003), McGuire et al. (2012), Odafe, Idoko, et al. (2012), Reidy et al. (2014) (Ethiopia, Kenya, Malawi, Nigeria, South Africa, the USA)
Immunologic success	Mean CD4 cell change (gains of 100 cells/µL with periods ranging from 12 to 24 months after starting cART), average monthly CD4 cell gain.	Assefa et al. (2012), Boyer et al. (2010), Chu et al. (2010), Horberg et al. (2007); Kerr et al. (2012), McGuire et al. (2012) (Cameroon, Ethiopia, Malawi, the USA)
Other clinical outcomes	Side effects, number of drug-resistant mutations.	Chan et al. (2011), Shet et al. (2011) (India, Malawi)
	Process measures
cART exposure	Rates of cART exposure among eligible population, time on cART.	Backus et al. (2010), Ekouevi et al. (2012), Kerr et al. (2012), Landon et al. (2005), Reidy et al. (2014), Rodriguez et al. (2008), Solomon et al. (2005), Wilson, Landon, Ding et al. (2005), Wilson, Landon, Hirschhorn et al. (2005) (Kenya, other countries in SSA, the USA)
PCP prophylaxis	Rates of appropriate prophylaxis.	Backus et al. (2010), Kerr et al. (2012), Landon et al. (2005), Rodriguez et al. (2008), Solomon et al. (2005), Wilson, Landon, Hirschhorn, et al. (2005) (the USA)
TB, hepatitis B, hepatitis C, cervical cancer screening	Rates of screening (among eligible population).	Backus et al. (2010), Kerr et al. (2012), Landon et al. (2005), Rodriguez (2008), Solomon et al. (2005), Wilson, Landon, Hirschhorn et al. (2005) (the USA)
CD4 cell count	Rates and frequency, depending on followed guidelines.	Backus et al. (2010), Kerr et al. (2012), Sherr et al. (2010), Solomon et al. (2005) (Mozambique, the USA)
Influenza vaccination	Rates of vaccination.	Landon et al. (2005), Wilson, Landon, Hirschhorn et al. (2005) (the USA)
HBV vaccination	Rates of vaccination.	Backus et al. (2010), Solomon et al. (2005) (the USA)
Other process measures	Laboratory measurements (complete blood cell count, creatinine, ALT or AST, cholesterol, glucose), syphilis screening; MAC prophylaxis; measurement of HIV-RNA; quality improvement interventions; service levels (composite score).	Anaya et al. (2004), Backus et al. (2010), Kerr et al. (2012), Nemes (2009), Solomon et al. (2005) (Brazil, the USA)
	Patient reported outcome measures (PROMS)
Patient satisfaction with care	Rating based on one question.	Rodriguez et al. (2008) (the USA)
Other PROMS	Problem Index score, number of unmet needs, perceived access to care, “7 items patient trust,” “interpersonal quality of care,” willingness to recommend care.	Rodriguez et al. (2008) (the USA)
	Other outcome measures
Loss to follow-up (retention in care)	Predominantly: absence from outpatient clinic for at least 3 months after the last missed appointment date.	Assefa et al. (2012), Balcha and Jeppsson (2010), Chan et al. (2011), Fatti et al. (2010, 2011), Odafe, Idoko, et al. (2012) Odafe, Torpey, et al. (2012), Reidy et al. (2014), Sherr et al. (2010) (Ethiopia, Kenya, Malawi, Mozambique, Nigeria, South Africa)
Care utilization	Primary care, specialty care and emergency room visits, hospitalization, home health visit, outpatient visits in 3 or 4 quarters.	Kitahata et al. (2003), Landon et al. (2005), Rodriguez et al. (2008), Sherr et al. (2010), Wilson, Landon, Hirschhorn et al. (2005) (Mozambique, the USA)
Adherence	Pharmacy refill compliance data, self-reported adherence (VAS), “mean adherence rate.”	Boyer et al. (2011, 2011), Saha et al. (2013), Sherr et al. (2010) (Cameroon, Mozambique, the USA)

Notes: ALT, alanine aminotransferase; AST, aspartate aminotransferase; cART, combination antiretroviral therapy; MAC, Mycobacterium avium complex; PCP, Pneumocystis jiroveci pneumonia; SSA, sub-Saharan Africa; TB, Tuberculosis; VAS, Visual analogue scale.

Changes in mean CD4 cell count, average monthly gain, and the proportion of patients with CD4 cells <200 cells/µL were used to define immunological success. Additional clinical outcomes were side effects (clinical diagnosis related to first-line treatment) (Chan et al., 2011) and number of drug-resistant mutations (Shet et al., 2011).

The most common process measure was cART exposure among eligible patients (n = 9). cART eligibility depended on the guidelines used by the health facility. Time on cART was also used to measure quality of care.

One study assessed processes that are not directly HIV-related in the measurement of quality of care (Kerr et al., 2012). In this study, safety laboratory assessments were measured (≥3 complete blood count and serum creatinine measurements, and liver function tests). In addition, HIV-related primary care measures were included (annual lipid assessments, annual fasting glucose, and baseline hepatitis A, B, and C laboratory assessment).

One study used (unstandardized/unvalidated) PROMs to measure quality of care (Rodriguez et al., 2008). Patient satisfaction was measured with one question. Number of “unmet needs” (income assistance, housing, home health care, counseling, and drug/alcohol treatment), perceived access to care, and patient trust and willingness to recommend care were also assessed.

Retention in care was assessed by measuring loss to follow-up (LTFU, n = 9). Most of the studies defined LTFU as absence from treatment for at least 3 months after the last missed appointment date (Fatti et al., 2010, 2011; Odafe, Idoko, et al., 2012; Odafe, Torpey, et al., 2012). In two studies, failed attempts at tracking the patients was included in the definition of LTFU (Odafe, Idoko, et al., 2012; Odafe, Torpey, et al., 2012).

The two remaining outcomes were care utilization and adherence to cART. Adherence in the previous 3 or 4 days was measured with standardized questionnaires (Boyer et al., 2010, 2011; Saha et al., 2013). In two of these studies, treatment interruptions in the 4 preceding weeks were also assessed (Boyer et al., 2011, 2010). One study used pharmacy refill data in the 180 days after starting cART to measure adherence (Sherr et al., 2010).

Associations between structural factors and quality of care

A total of 11 studies were included in the assessment of structural characteristics associated with quality of HIV outpatient care (Table 3). Two of the studies showed an association between facility volume and the likelihood of achieving viral suppression. In the first study, patients from large facilities (>300 patients in care) were more likely to achieve viral suppression than patients from small facilities (Backus et al., 2010). In the second study, based in South Africa, rates of viral suppression were lower in large care sites (>950). This association only applied to the first 12 months after starting cART (Fatti et al., 2011).

Table 3. Summary of fully reviewed studies (n = 11).

Source (Author, year)	Study design, population	Investigated factors	Outcome HIV-RNA (copies/ml)	Results
Backus et al. (2010)	RCS; 73 sites; 21,564 veterans; the USA	Facility volume: <25; 25–99; 100–299 or >300 patients in care.	<400	Patients from care sites with smaller numbers of HIV-infected patients in care (< 300) were less likely to have HIV-RNA control.
Chu et al. (2010)	RCS; 423 patients; the USA	Health facility setting: hospital-based specialty center vs. community-based primary care network.	“Undetectable”	No significant differences in the likelihoods of achieving viral suppression (16–32 weeks after starting cART) between patients from community-based vs. hospital care sites.
Fatti et al. (2010)	RCS; 59 sites; 29,203 cART naïve patients; South Africa	Health facility setting: primary health-care facilities, district hospitals or regional hospitals.	<400	Patients from district and regional hospitals were less likely to have viral suppression than patients in care at primary health-care facilities.
Fatti et al. (2011)	RCS; 54 sites; 40,861 cART naïve patients; South Africa	Facility volume: <950 or >950 patients in care.	<400	Patients from care sites with smaller numbers of HIV-infected patients in care were more likely to achieve viral suppression within 12 months of starting cART. No significant differences in the proportion of patients achieving viral suppression at all time points.
Horberg et al. (2007)	CSS; 14 sites; 1571 patients; the USA	Team composition: HIV clinical pharmacists.	<500	Patients exposed to an HIV clinical pharmacist were more likely to achieve viral suppression at 12 months.
Landon et al. (2005)	RCS; 64 sites; 5247 patients; the USA	Health-care provider expertise: Infectious disease specialists, expert generalist or non-expert generalist; and current caseload: low (0–19), medium (20–299), or high (>300 patients).	<400	Patients treated by Infectious disease specialists were most likely to have a controlled viral load, followed by those treated by expert generalists and non-expert generalists. No significant differences in viral suppression between patients with physicians with low, medium or high caseloads.
McGuire et al. (2012)	CSS; 1 hospital vs. 10 rural health sites; 702 patients; Malawi	Health facility setting: centralized vs. decentralized.	<50	No significant differences in viral suppression (one year after starting cART) between patients followed at centralized vs. decentralized facilities.
Saha et al. (2013)	CSS; 4 sites; 45 providers; 437 patients; the USA	Other health-care provider characteristics: self-rated cultural competence.	≤75	Greater racial disparity in viral suppression among patients with providers with a low cultural competence.
Sangsari et al. (2012)	PCS; 267 drug users; Canada	Health-care provider expertise - current caseload: number of patients enrolled in HIV registry.	<500	Patients treated by physicians with greater HIV experience were more likely to have viral suppression.
Shet et al. (2011)	CSS; 471 patients; India	Health facility setting: public, private or public–private.	<100	Patients visiting public and public–private health facilities were more likely to have viral suppression compared to those in private settings.
Wilson, Landon, Hirschhorn et al. (2005)	CSS; 68 sites; 6651 patients; the USA	Health-care provider expertise: physician vs. nurse practitioner/physician assistant.	<400	Patients treated by nurse practitioners and physician assistants were more likely to have viral suppression than those treated by non-expert generalists. Rates of viral suppression among patients treated by these non-physician clinicians were similar to that of those treated by infectious disease-trained physicians and expert generalists.

Notes: cART, combination antiretroviral therapy; ID, Infectious diseases; CSS, Cross-sectional study; PCS or RCS, Prospective or retrospective cohort study

In three studies, a positive association was found between physician expertise and virologic success. Two studies used a composite variable with three levels of expertise: infectious disease physicians, expert generalists, and non-expert generalists. Physicians were asked whether they considered themselves to be specialists in HIV (“experts”) (Landon et al., 2005; Wilson, Landon, Hirschhorn, et al., 2005). A significant association between current HIV caseload (an increase of 100 patients in care) and viral suppression was found in one study (Sangsari et al., 2012), and not found in another (categories: <19; 20–299; and >300 patients in care) (Landon et al., 2005). In one study assessing performance by non-physician clinicians, rates of viral suppression among patients treated by nurse practitioners and physician assistants were similar to those among patients treated by infectious disease-trained physicians and generalist HIV experts. Moreover, the rates of viral suppression among patients treated by non-physician clinicians were higher than those among patients treated by generalist non-HIV experts (Wilson, Landon, Hirschhorn, et al., 2005).

Four studies focused on health facility setting. There was no difference in virologic success among patients in care in centralized (hospital-based) versus decentralized (community-based) facilities (Chu et al., 2010; McGuire et al., 2012) (the USA and Malawi). In one study, patients treated in private health facilities were more likely to have treatment failure compared to patients in public/public–private health facilities (India) (Shet et al., 2011). This also applied to patients in district/regional hospitals versus primary health-care facilities (South Africa) (Fatti et al., 2010).

Finally, provider CC (assessment of cultural skills) (Saha et al., 2013) and the involvement of a clinical pharmacist in HIV outpatient care (Horberg et al., 2007) were positively associated with viral suppression.

Discussion

In this review, we evaluate the studies that have investigated the impact of health-care facility and care provider characteristics on health outcomes in HIV outpatient care. We point out a number of findings that are of particular importance when interpreting the results of the identified studies.

First, a disproportionate amount of studies is based in the USA and in countries in sub-Saharan Africa, making it difficult to generalize findings to other parts of the world (Figure 1).

Another important finding involves the diversity in choice and definition of structural variables across the different studies. This makes it difficult to summarize results and translate the results into practice.

A third key finding is the scope of measures of quality of care used in the studies. More than 20 outcomes were measured, with exact definitions often differing between studies (e.g., cART exposure and time on cART). Mortality was measured relatively often, evidently more often in low-resource countries. In contrast, screening of TB, hepatitis B and C, and cervical cancer was only assessed in the USA.

Interestingly, almost all of the studies only used measures that are directly related to HIV infection. Despite the fact that the importance of management of non-HIV comorbidities has been recognized (Chu et al., 2011), no studies measured non-HIV outcomes (e.g., blood pressure, cholesterol, and glucose) and only one measured process outcomes related to comorbidity (Kerr et al., 2012).

Also surprising is the fact that only one study used PROMs. PROMs can help in gaining an insight into whether patients’ health-care experiences match their expectations. The importance of including patients' views when evaluating health care has been widely recognized. A positive physician/patient relationship may lead to better HIV treatment adherence and improved clinical outcomes (Land, Nixon, & Ross, 2011). PROMS can also be used to evaluate the impact of both disease and treatment as perceived by the patient, also referred to as health-related quality of life (HRQOL) (de Boer-van der Kolk et al., 2010). Evidence suggests that better HRQOL is associated with survival in HIV-infected patients (de Boer-van der Kolk et al., 2010; Cunningham, Crystal, Bozzette, & Hays, 2005).

Consistent with previous reviews with a similar focus (Handford, Rackal, Tynan, Rzeznikiewiz, & Glazier, 2012; Rackal et al., 2011), we found positive associations between health-care provider expertise and health outcomes among HIV-infected patients. However, our results regarding facility volume were less convincing than those in previously published reviews (Handford et al., 2012; Handford, Tynan, Rackal, & Glazier, 2006). These reviews differ in the sense that they included many studies that took place before the introduction of cART. Furthermore, in-hospital mortality was often measured, whereas our review focusses on outpatient care.

For our assessment of the results of the studies, we only selected the studies that measured viral suppression. We did this for the sake of comparability of studies but are aware of the fact that we have not presented associations with other outcomes of care. We do, however, believe that HIV-RNA measurement is an appropriate quality of care measurement, more so perhaps than mortality (in this post-cART era) and CD4 cell count.

Implications for practice

Our findings suggest that health-care provider experience improves outcomes among HIV-infected patients. Additional research is needed to specify the number of patients required to gain and maintain expertise. We cannot make recommendations regarding facility volume requirements for outpatient care on the basis of the identified studies (Backus et al., 2010; Fatti et al., 2011). The studies took place in very different settings (the USA and South Africa), used different definitions of large hospitals (>300 and >950), and had contrasting results. The increased probability of virologic failure at large clinics in the South Africa-based study could be explained by heavy workloads, strain on the hospital infrastructure, and longer waiting times (Mugavero et al., 2007).

The study by Wilson et al., in which the quality of care by non-physician clinicians was evaluated, provides additional support for the effectiveness of task shifting (Wilson, Landon, Hirschhorn, et al., 2005). Task shifting is the assigning of particular tasks to health-care workers with shorter training and fewer qualifications (e.g., cART prescription by nurses). This model of care can result in substantial cost savings and has been recommended as an approach to reduce resource shortages in low-income countries (WHO, 2008). Reviews on task shifting are available (Callaghan et al., 2010; Kredo et al., 2014; Mdege et al., 2013).

Only one study evaluated the involvement of disciplines other than physicians, non-physician clinicians, or nurses (Horberg et al., 2007). The results support the involvement of a clinical pharmacist in HIV outpatient care. More evidence regarding this issue can be found in the review article by Saberi et al. (Saberi, Dong, Johnson, Greenblatt, & Cocohoba, 2012).

Implications for research

Research needs to be extended to regions outside the USA and sub-Saharan Africa. Existing cohort studies may create opportunities to expand this research field. In Europe for example, there are multiple cohort studies (Mary-Krause et al., 2014; Omland, Ahlstrom, & Obel, 2014; Schoeni-Affolter et al., 2010; van Sighem, Smit, Stolte, & Reiss, 2014), some of which participate in collaborations with other cohort studies (Bohlius et al., 2009; May et al., 2014). The collected data can be an excellent source for retrospective research into the impact of health facility characteristics (such as facility volume) on health outcomes.

Furthermore, researchers should align their methods of measuring quality of care. An initial step in this process is to define what good quality of HIV care is (Hung & Pradel, 2014). Since the role of non-HIV comorbidity is increasing, researchers and health-care providers need to go beyond HIV-related morbidity in the evaluation of health outcomes (Chu et al., 2011).

As retention in care has been recognized as a crucial step in maximizing patient outcomes, we believe it should continue to be measured in quality of care studies. Mugavero et al. present a number of released core indicators for retention in care (2012). In addition, patient preferences and satisfaction should play an important role in the evaluation of quality of care. In a systematic review, Land et al. provide an overview of methods for measuring patient satisfaction in HIV care delivery (Land et al., 2011). The authors concluded that there is no gold standard, and findings were used to develop a validated questionnaire (Land, Sizmur, Harding, & Ross, 2013).

There are several sources for HIV-related measures, including the article by Catumbela et al., in which the authors propose a core set of HIV-related quality of care measures based on a literature review (2013).

In conclusion, the studies that demonstrate associations between the quality of HIV outpatient care and facility and care provider characteristics are not scarce, but too diverse and inconsistent to guide policy regarding standards of care. Important reasons for this are the heterogeneity of study designs and the underrepresentation of studies outside the USA and sub-Saharan Africa. Patient views and non-HIV comorbidities should be included in the assessment of quality of HIV outpatient care in future studies.

Acknowledgements

E.A.N.E. collected all data and drafted the article. All authors edited, commented, and approved the final version of the article.

P.R. through his institution has received independent scientific grant support from Gilead Sciences, ViiV Healthcare, Merck & Co, Janssen Pharmaceutica, and Bristol-Myers Squibb (BMS). In addition, he serves on a scientific advisory board for Gilead Sciences and on a data and safety monitoring board for Janssen Pharmaceutica, for which his institution has received remunerations. K.B. serves on advisory boards for MSD, Gilead, BMS, Viiv, and Janssen for which he has received remuneration. All other authors report no potential conflicts.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by Aids Fonds [2011015].