Modest Public Health Impact of Nontargeted Human Immunodeficiency Virus Screening in 29 Emergency Departments FREE

During the last 15 years, human immunodeficiency virus (HIV) screening combined with early treatment has effectively reduced HIV-related mortality, and some authors have postulated that this strategy plays a key role in controlling the epidemic.¹ France has an estimated 140 000 persons living with HIV, and 7000 individuals are newly infected annually.^2- 3 Despite easy access to free HIV testing services and numerous HIV tests (5 million in 2009, for 65 million inhabitants), late diagnosis remains common, with one-third of HIV infections diagnosed in conjunction with AIDS or CD4 lymphocyte counts of less than 200/μL.⁴

To lower the number of undiagnosed infections and to improve early detection, nontargeted HIV rapid test (RT) screening in health care settings has been promoted by national health agencies in the United States, the United Kingdom, and, more recently, France.^5- 7 Studies showing that many persons who were unaware of their infection or who were diagnosed at late stages did not belong to traditional high-risk groups^8- 10 supported this strategy, as did models indicating its cost-effectiveness,^11- 12 but its efficacy remains controversial. Recently, a Denver, Colorado, emergency department (ED) study compared nontargeted opt-out HIV-RT screening with physician-directed diagnostic HIV-RT screening and showed the former to have modest benefit.¹³ However, to our knowledge, no large-scale study has yet assessed the relevance of this strategy to improve early HIV infection diagnosis in general populations, and its public health prevention benefit remains unknown.

Urban EDs represent an important health care source for the population, including low-income, uninsured, and other subgroups that might not be reached in other health care settings,¹⁴ while being at higher risk of undiagnosed HIV infection.^15- 17 In France, because an estimated 25% of inhabitants (14 million) visit an ED annually,¹⁸ EDs appear to be an ideal setting¹⁹ to assess nontargeted HIV-RT screening of the general population. In 2009, before the French guidelines were published, we evaluated the public health impact of nontargeted HIV-RT screening among ED patients in the metropolitan Paris region (11.7 million inhabitants), where the highest number of HIV infections (263/million inhabitants) are newly diagnosed annually, ie, half of France's new HIV diagnoses.

The study was approved by Île-de-France XI Committee for Patient Protection (No. 08053, October 9, 2008) and by the Commission Nationale de l’Informatique et des Libertés (French Data-Protection Authority).

This interventional study was conducted in 29 EDs from May 2009 through September 2010. In each ED, the intervention lasted 6 consecutive weeks, randomly allocated. The HIV RTs were performed in EDs on a 24-hour basis, involving existing caregivers and members of the investigation team when needed.

The 31 adult EDs selected for this study participate in the French acute syndrome surveillance network OSCOUR (Organisation de la Surveillance Coordonnée des Urgences),²⁰ which accounts for 60% of all patients who are seen at EDs in the region²¹ and reflects the diversity of French ED settings. The heads of EDs who belong to the OSCOUR network were contacted personally and asked to participate by the principal investigator (A.-C.C.) and the president of the French Association of Emergency Medicine (D.P.), who was also a study investigator. Before implementing the intervention, the principal investigator arranged several meetings with the ED teams to prepare the intervention, to reinforce their motivation, and to reassure them about the possible work overload and the help provided in such cases by the epidemiological research assistant (ERA). In each hospital, a group composed of an ED physician, an ED nurse, an infectious disease physician, and a virologist was responsible for the study. A total of 29 EDs agreed to participate and completed the intervention.

All 18- to 64-year-old patients who were able to give written informed consent were eligible to undergo HIV RT. Exclusion criteria were self-reported HIV infection; inability to provide consent because of neuropsychiatric disorders, substance abuse, language barrier, or being under arrest; unstable medical illness; and being seen at the ED for prophylaxis after sexual exposure to HIV.

Before the intervention at each site, every ED team participated in a 60-minute training session, which included an educational lecture, an HIV RT demonstration, hands-on practice, and information about how to disclose test results. On ED arrival, the patients received an information sheet explaining the study. At first contact with triage nurses, eligible patients were asked if they agreed to participate in this research by undergoing HIV RT free of charge (opt-in approach). After written consent was obtained, nurses performed a rapid HIV antibody test (OraQuick Advance Rapid HIV-1/2 Antibody Test; Orasure Technologies Inc, Bethlehem, Pennsylvania). This particular test was selected because of its good performance and ease of use.^22- 23 Within 20 to 40 minutes, the test results were interpreted as negative, reactive, or invalid. The nurses delivered negative results to the patients before discharge and recommended repeated testing in the case of recent exposure. The HIV RT reactivity was disclosed by an emergency care physician who arranged for a follow-up visit with an on-site infectious disease specialist within the following 72 hours, and nurses drew blood for standard enzyme-linked immunosorbent assay and Western blot confirmation. For invalid test results (interpreted outside the 20- to 40-minute window or uninterpretable), the HIV RT was repeated or blood was drawn for standard enzyme-linked immunosorbent assay and Western blotting.

Patients who agreed to be tested were assigned a code linking HIV RT results to their completed, anonymous, self-administered questionnaire, which included demographics (age, sex, country of birth) and additional information such as health insurance, sexual behavior, HIV testing history, and self-perception of HIV risk (Table 1). The questionnaire was constructed to enable data comparison with the French National population census 2006, metropolitan Paris region (National Institute of Statistics and Economic Studies), and French general population surveys.^24- 25

An ERA was present on site for 8 hours every day to monitor the study and, during that shift, helped the local team by offering and/or performing HIV RTs in EDs when necessary. However, HIV RTs were available and performed on a 24-hour basis by the local team in all EDs. During follow-up visits with the infectious disease specialist, after HIV infection had been confirmed, along with usual counseling, information about HIV testing history, HIV infection–related signs and symptoms, probable mode of transmission, and CD4 lymphocyte counts were recorded for comparison with the French national HIV case surveillance.⁴

To determine whether HIV RT refusal might bias the study results, a complementary study was conducted to identify covariates of test refusal in 7 EDs reflecting the diversity of the 29 participating EDs (inner Paris, suburbs, and more remote areas; private and public hospitals; and low and high patient flows). An ERA recorded the reasons for refusal among all consecutive eligible patients during randomly selected 40-hour observation periods of ED activity within 5 consecutive days. Age, sex, country of birth, history of HIV testing, and perceived HIV risk were also included for comparisons with the participants' characteristics that were reported in their questionnaire.

The outcome measures to evaluate the public health impact of nontargeted HIV-RT screening in EDs were the number of patients tested for HIV during the study period and their characteristics vs those of the general metropolitan Paris population as well as the percentage of newly diagnosed HIV patients among those tested and their characteristics vs those from the national HIV case surveillance in the same geographic area.

For each ED, we created a database that included the number of patients who were seen, patients who were eligible, tests offered and performed, and confirmed positive results collected from the ERA's and the nurses' daily records and a test register. The database was used to calculate the eligible patient acceptance, proportion of tested patients, and proportion of new HIV diagnoses. Information on all patients with reactive HIV RT results was examined by an expert panel that was composed of 2 emergency care physicians, 2 epidemiologists, 2 public health specialists, and 2 infectious disease specialists. Each 2-specialist group independently examined those data to identify the patients who were included in the study despite exclusion criteria and to determine HIV infection–associated signs and symptoms at ED consultation.

All analyses were conducted with SAS version 9.2 (SAS Institute Inc, Cary, North Carolina) and Stata version 11.0 (Stata Corp, College Station, Texas). Comparative analyses used a χ² test or a Fisher exact test, and multivariate analyses applied logistic regressions. The descriptive analyses, conducted exclusively on the database resulting from the merger of the 29 databases obtained through the data capture of each ED's questionnaires, were restricted to the patients with an HIV RT result and a completed questionnaire. Because some variables collected were incomplete, with 0.35% to 11.1% missing values, we applied a multiple imputation method²⁶ with chained equations²⁷ to estimate those values and to improve descriptive accuracy using Stata user's ICE program.²⁸ To optimize power and reproducibility, 50 multiple imputation data sets were analyzed individually and jointly to obtain overall variance estimates and confidence intervals (CIs).

Using national data,²⁹ we estimated the HIV infection prevalence in metropolitan Paris (11 million inhabitants) to 0.88%, with 0.18% undiagnosed HIV infections. As a conservative assumption, persons who were unaware of their HIV infection were considered to have the same ED use as the general population. According to American studies,^30- 35 we expected approximately 15% of eligible patients to be tested. It was estimated that 12 000 HIV RTs were required to achieve 95% power for the estimation of the percentage of newly diagnosed HIV infections with ±0.1% precision. We used the mean number of patients who were seen at each ED per week (OSCOUR database) to calculate the number of weeks necessary to reach the number of participants needed. This calculation led us to propose the study to all 31 adult OSCOUR-network EDs in metropolitan Paris for a randomly selected, 6-week-long, field intervention in each.

Of 138 691 patients who were seen at EDs (Figure), 78 411 (56.5%) were eligible. Of those who were eligible, 20 962 (26.7%) were offered HIV RT and 13 229 (63.1%) accepted. Finally, 12 754 patients (16.3%) were tested, and descriptive analyses were applied to 11 356 who had completed questionnaires. The HIV-RT screening was performed by the ED team alone in 14 of 29 EDs and required ERA support in the other 15, with similar percentages of patients accepting for the 2 groups (62.8% vs 63.5%, respectively). However, because the ED teams working alone offered more HIV RTs (37.5% vs 19.6%, respectively), their percentage of eligible patients tested was higher (27% vs 12.5%, respectively; P < .001).

The patients who were tested (52.2% male, 76.3% French) had a mean age of 35.8 (interquartile range [IQR], 26-45) years; 98.2% had national health insurance; two-thirds were living as a couple; 22.5% of men and 17.7% of women reported having more than 1 sexual partner during the past 12 months; 4.5% of men reported having sex with men (MSM) at least once in their lifetime; and 57.2% reported having undergone previous HIV testing, with 39.7% having been tested within the past 5 years. Overall, the population tested reflected the general population distribution of the Paris metropolitan region, with a slight overrepresentation of foreign-born persons (particularly sub-Saharan Africans), women with more than 1 sexual partner during the last 12 months, and persons who had undergone more frequent HIV screening (Table 1).

Among the 12 754 HIV RTs performed, the results of 38 were reactive, with HIV infection confirmed in 37 (0.29%; 95% CI, 0.20%-0.40%). Of the patients with confirmed HIV infection, 3 were excluded from the analyses because they had an altered state of consciousness and were tested for diagnostic purposes, irrespective of the study protocol, and 16 were excluded because they subsequently admitted being aware of their HIV positivity.

Therefore, 18 HIV infections were newly diagnosed (0.14%; 95% CI, 0.08%-0.22%) (Table 2). The mean age of the patients with newly diagnosed infections was 32.9 (IQR, 27-40) years. Of them, 12 (66.7%) reported previous HIV testing (median time since the last HIV test, 1 year), 7 (39.0%) were MSM, and 10 (55.0%) were heterosexuals from sub-Saharan Africa. Eight patients (44%) were seen at the ED for HIV-related symptoms: 7 with advanced-stage disease and 1 with HIV primo-infection; 10 infections were asymptomatic. Of those 18 patients, 6 (33.3%) did not return for their first follow-up visit, despite repeated recalls (4 women and 2 men, all born in Africa); 4 (22.2%) were hospitalized for HIV infection–related signs and symptoms; and 8 (44.4%) came to the first follow-up visit. Among the 12 patients (66.7%) linked to care, the CD4 lymphocyte counts were less than 200/μL in 5 (41.6%), 200 to 350/μL in 3 (25%), and higher than 350/μL in 4 (33.3%).

The proportions of newly identified HIV infections in EDs were the highest in MSM (2.61%; 95% CI, 1.06%-5.31%) and African-born heterosexuals, particularly women (1.57%; 95% CI, 0.63%-3.20%) (Table 3). The demographics of the 18 HIV-infected patients newly diagnosed in EDs and the 3008 cases involving 18- to 64-year-old individuals reported in the metropolitan Paris region by the national HIV case surveillance in 2009 were comparable (Table 4).

The complementary study to evaluate the covariates of HIV RT refusal involved 1404 patients who were seen at an ED. Of them, 655 (46.7%) were eligible and were offered HIV RT screening: 404 (61.7%) accepted (vs 63.1% for the whole study). Of the 251 refusers, 102 (40.7%) declined because they had been tested in the last year, 87 (34.7%) did not think that they were at risk for HIV, and 62 (24.6%) declined for other reasons, eg, not the right time or fear of the result. Age, sex, and country of birth were comparable among accepters and refusers. Refusal of HIV RT was associated with low self-perceived risk for HIV infection and having been tested previously (respective odds ratios [95% CI], 12.02 [2.60-55.58] and 2.04 [1.37-3.04]).

Although our study intervention reached a large sample (12 754) of individuals consulting at a metropolitan Paris region ED, the benefit of nontargeted screening was only modest. The prevalence of newly diagnosed infections (0.14%) was close to our initial prevalence estimate, ie, that of the general population (0.17%). However, unexpectedly, all but 1 of the 18 newly diagnosed HIV-infected patients belonged to a traditional high-risk group (MSM and persons born in a high HIV-prevalence country, namely, sub-Saharan Africa). Furthermore, most of them had late-stage disease, and one-third could not be linked to care. The major strength of this investigation is that, to our knowledge, it is the first study designed to evaluate the public health impact of nontargeted ED HIV-RT screening on such a large scale, involving an entire region of 11.7 million inhabitants, where half of the country's new HIV diagnoses are made annually.

Of the 138 691 patients who were seen at these EDs during the study, 56.5% were eligible, and 16.3% of the eligible patients were actually tested; the latter percentage is close to the percentages reported in most ED studies^30- 35 and close to our initial hypothesis. A potential study limitation is that HIV RT refusers might be at higher risk of HIV infection than accepters. However, our subanalysis of reasons for refusal showed that most of both groups' demographics were similar, with the only covariates of refusal being low perceived risk and having previously been tested for HIV, as in other studies.^36- 37

One of the objectives of implementing nontargeted HIV screening in the ED is to reach a large segment of the population, including the economically underprivileged and those with poorer access to health care. Moreover, ED recruitment was expected to reach infected persons who did not belong to the main high-risk groups, ie, heterosexual men and couples, who were not targeted by specific screening programs and who are thought to rarely undergo voluntary testing. Indeed, our results showed that the ED-tested population closely resembled the general population, albeit with slight overrepresentation of sub-Saharan African–born persons. Despite the diversity of the population tested through our intervention, no HIV infection was detected among French-born heterosexuals. Their low infection incidence² could explain this absence of unknown infections. In contrast, being an MSM and being sub-Saharan African born were associated with new ED-diagnosed HIV positivity, and their respective frequencies of unknown infection were 2.61% (95% CI, 1.06%-5.31%) and 1.12% (95% CI, 0.54%-2.08%), values consistent with data reported by the national HIV case surveillance.

Pertinently, new ED-detected infections were not found at earlier stages, compared with those reported by national HIV case surveillance. Most were already at late stages, as in the US study,¹³ which can partly be explained by the fact that HIV infection–related symptoms prompted 8 patients' ED visits. However, the unavailability of CD4 lymphocyte counts for 6 patients who were unavailable for follow-up limits the strength of this finding. Finally, one-third of new diagnoses, mainly involving patients from sub-Saharan Africa, could not be linked to further care, despite the on-site availability of follow-up and several ERA's attempts at contact. This issue must be considered when evaluating the benefit of proposing nontargeted HIV screening in health care settings. Our entry-into-care rate is in the lower range of such rates observed in the United States, where the average was 76% (95% CI, 66%-84%) after an HIV-positive test result in an ED, which is slightly above the average of 67% (95% CI, 54%-70%) when testing took place in community locations.³⁸

To our knowledge, our study is the first to measure unknown HIV infection prevalence in a large sample of the French general population. Our 0.14% prevalence is close to the percentages reported by some EDs in low HIV-prevalence countries.^13,37 Nontargeted HIV screening was cost-effective in the United States, even when the unknown infection prevalence was as low as 0.1%.¹¹ Applying a French-adapted model¹² to our results showed that ED nontargeted HIV-RT screening was at the lower limit of cost-effectiveness. Considering that 8 of 18 persons who were seen for HIV-related symptoms could have been tested for diagnostic purposes or that at least the 4 admitted patients would have been tested during their hospitalizations, this strategy falls below that limit. An ED-based screening strategy limited to men 18-to-45 years old and African-born persons would have identified all new HIV infections for 50% fewer HIV RTs than were actually performed and thus would have enhanced cost-effectiveness. Another way to heighten ED HIV-RT screening effectiveness would be to limit HIV testing to EDs with the highest HIV prevalences and/or located in areas with high-risk populations. Such a frequency-based screening policy, first recommended by the Centers for Disease Control and Prevention in 1993, is difficult to apply because accurate local estimates are not readily available.⁵

Our study did not compare nontargeted HIV-RT screening with diagnostic testing. In France, HIV testing is rarely performed in EDs, as symptomatic patients are hospitalized or referred for specialized consultations and HIV testing. The modest public health impact of nontargeted HIV screening demonstrated herein applies to EDs and could be different in other health care settings, eg, acute care, primary care center,³⁷ or general practice, where patients present with less severe disease, potentially at earlier stages. However, considering that a substantial segment of the general population visits EDs for a wide range of medical conditions and all grades of disease severity, we think that the main conclusions of our study, concerning the absence of a hidden epidemic in a low-risk population and the low efficiency of nontargeted HIV-RT screening, would have been maintained. Finally, despite some similarity to some data obtained in other low-prevalence countries, eg, the United States and the United Kingdom, our results cannot be extrapolated to other countries. Nonetheless, they clearly highlight the need for additional studies on this new strategy's effectiveness in those countries, where health authorities have recommended nontargeted screening.

In conclusion, ED-based HIV RT screening is feasible and can reach large numbers of patients. However, unexpectedly, nontargeted screening identified only a few new diagnoses, often already at late stages, and most newly diagnosed patients belonged to a high-risk group and had been tested previously. Therefore, our observations do not support the implementation of nontargeted HIV screening of the general population in EDs.