Recent HIV surveillance data suggest that approximately 33% of HIV-infected persons in the United States present for HIV testing late and have AIDS (CD4+ cell count <200 cells/mL or an AIDS-defining illness) within one year after HIV diagnosis [1, 2]. Patients are less likely to experience the full benefits of highly active combination antiretroviral (cART) therapy if they enter HIV care and initiate treatment at a CD4 count <350 cells/mm³ [3, 4]; the clinical cost is even more profound when the CD4 count is <200 cells/mm³ or the patient has already developed clinical AIDS [5–8]. In addition, persons who remain unaware of their HIV-positive status (estimated 21% to 25% of infected persons in the USA in recent years) [9, 10] may not only miss the benefits of earlier cART treatment, but are also more likely to remain chronically viremic and are thereby more likely to transmit HIV to their sexual and needle-sharing partners [9].

The CDC has been promoting strategies to encourage more widespread HIV screening to diagnose infected persons earlier in the course of their illness, including by releasing in 2006 the guidelines for implementing routine universal opt-out testing in healthcare settings [11]. Yet, the latest HIV surveillance data [1, 2] and epidemiologic studies in multiple US populations indicate that the proportion of persons who are diagnosed late in the course of HIV infection [2, 12, 13] or present late for HIV care [14, 15] remains unacceptably high. Stable or worsening trends in the proportion of patients HIV diagnosed with low CD4 counts have also been observed internationally [16–18]. However, encouraging trends have been seen in select US jurisdictions which have dramatically expanded their HIV testing programs [19, 20]. With the recent shift in antiretroviral treatment guidelines toward earlier therapy initiation to benefit patients' health [21] and the growing interest in HIV “test and treat” and “test and link to care-plus” strategies to limit the spread of the US epidemic [22], it is important to understand the burden of late HIV diagnosis among contemporary US patients as it affects the likelihood of success of these interventions.

We studied participants in the HIV Outpatient Study (HOPS) who were recently diagnosed with HIV infection to examine the trends in median CD4 count at diagnosis and the proportion of patients diagnosed with CD4 count <200 cells/mm³ (also termed “late HIV diagnosis”) during 2000–2009.

Of 3,670 new HOPS participants seen at 10 HIV clinics during 2000–2009, 1,223 (33%) had a recent HIV diagnosis (≤6 months before entry into care at a HOPS site). Compared with the 2,447 patients who were diagnosed with HIV infection >6 months prior to entry into care at a HOPS site (and may have been in care elsewhere), recently diagnosed patients were significantly (P < 0.05) more likely to be ARV naïve (86% versus 16%), younger (median age of 38 versus 40 years), to have heterosexual activity as their sole risk for HIV infection (37% versus 25%), to be privately insured (61% versus 51%), less likely to be male (75% versus 80%), less likely to be of white race (40% versus 47%), and less likely to have injection drug use (IDU) activity as their sole risk for HIV infection (6% versus 11%).

Of the 1,223 patients with recent diagnoses, 936 (77%) had a CD4 count documented within 3 months following HIV diagnosis while still antiretroviral naïve (a median of 15 days, IQR: 4–34 days after diagnosis). Of these 936 patients (median age = 38, range: 17–77 years), 77% were male, 43% were white, 39% were black, 54% were men who had sex with men (MSM), 36% had heterosexual HIV risk, and 5% were injection drug users (Table 1). The 23% of patients whom we excluded because they did not have a CD4 count documented in the relevant timeframe were significantly more likely to have had a history of IDU (9% versus 5%), to have public insurance (41% versus 29%), and to have been antiretroviral-experienced (23% versus 7%) at entry into care at a HOPS site.

Among HOPS patients recently diagnosed with HIV infection, we found no statistically significant improvement in the median CD4 count at diagnosis during 2000–2009. Overall, 36% of patients were diagnosed with a CD4 count <200 cells/mm³ and 58% with a CD4 count <350 cells/mm³. Persons whose risk for HIV infection was other than being MSM, persons aged ≥35 years, and persons of nonwhite race/ethnicity were more likely to be diagnosed with a CD4 count <200 cells/mm³ and thus more likely to have missed an opportunity for timely access to HIV care and initiation of ARV therapy; the correlates of HIV diagnosis with CD4 <350 cells/mm³ were largely similar. Our finding that MSM were less likely to be diagnosed with advanced HIV infection than some other risk groups (e.g., IDUs) is consistent with the findings from US HIV surveillance [1, 2] and data from other HIV cohorts reporting on late HIV diagnosis [23] and presentation for care [15]. The association of younger age (<35 years) with a lower likelihood of late HIV diagnosis may be partially explained by the fact that younger persons de facto have had less lifetime opportunity, if they became HIV-infected, to progress to CD4 <200 cells/mm³; older age has been associated with late HIV diagnosis previously [2, 15].

In a recent study by Althoff and colleagues of 44,491 HIV-infected patients enrolled in the North American-AIDS Cohort Collaboration on Research and Design (NA-ACCORD), the median CD4 count at first presentation to HIV care increased from 256 cells/mm³ (interquartile range, 96–455 cells/mm³) to 317 cells/mm³ (interquartile range, 135–517 cells/mm³) from 1997 to 2007 (P < 0.01). The median CD4 count at HIV diagnosis for HOPS patients whom we studied (who all entered care within 6 months of diagnosis, per inclusion criteria) was remarkably similar to median CD4 count at first presentation to HIV care among NA-ACCORD participants. Our findings of a high (approximately 34%) prevalence of late HIV diagnosis during the period 2004–2007 correspond well with the estimate from national HIV surveillance (34%) for 2005–2007, and the national surveillance system also detected no marked increases in CD4 counts at diagnosis over time [1]. The reason why the median CD4 counts in the HOPS and NA-ACCORD population were higher as compared with those in HIV surveillance may be related to several factors. First, compared with the nationwide HIV epidemic, the HOPS is enriched in whites and MSM, populations who tend to be diagnosed at higher CD4 cell counts. Second, HIV testing capacity and therefore the opportunity for timely diagnosis vary by jurisdiction. HOPS clinics are located in major urban centers with large HIV epidemics where testing opportunities might be greater leading to a greater likelihood of indentifying infected persons earlier in the course of their disease. Third, the CD4 count findings from national HIV surveillance are derived from 37 states, some of only report CD4 counts <200 cells/mm³ for HIV-infected persons, a practice that would tend to bias the median CD4 cell count downwards [2].

Our results should be interpreted in light of some additional important caveats. First, we restricted our analyses to patients diagnosed with HIV within 6 months prior to entry into care at a HOPS site. We used this criterion because HOPS patients who initiated HIV care earlier elsewhere often lack documentation of their initial CD4 count value. The timeframe of ≤6 months between HIV diagnosis and entry into care at a HOPS site, however, appeared to capture well the patients who newly entered care (only 7% of these patients had any previous ARV exposure). Although we have thus excluded a subset of patients who delay entry into HIV care (who may also be more likely to be diagnosed late in HIV infection) a recent meta-analysis suggests that approximately 72% of patients enter care within 4 months of HIV diagnosis [24]. Second, we studied a self-selected group of recently diagnosed patients who, by definition, entered HIV care and were also willing to consent to enrollment in the HOPS; these patients could potentially differ from patients not entering care [25, 26]. Available resources constrain some HOPS sites from enrolling the entire clinic population, and although attempts are made to balance the representativeness of the enrollees there is some opportunity for convenience sampling bias. Additionally, the HOPS does not enroll patients who present to a HOPS clinic for care but die prior to the opportunity to consent for the study. Both of these effects are greater at public facilities compared with private facilities. Consequently, there might be a tendency for the HOPS to enroll healthier patients, and, therefore, our findings would not represent the entire clinic population served by HOPS sites or be generalizable to all patients in cities where HOPS sites are located. Third, the 23% of patients with recent HIV diagnoses who were excluded due to missing CD4 count data were more likely to be IDU and publically-insured, subgroups that tended to have lower CD4 counts at diagnosis. Therefore our finding that approximately one-third of patients were diagnosed with HIV infection at a CD4 count <200 cells/mm³ might be an underestimate due to this information bias. Finally, because of small sample sizes in some sociodemographic subgroups (e.g., Hispanics, high-risk heterosexuals, persons with a history of IDU) we could not accurately estimate median CD4 counts and proportions of patients with CD4 counts <200 at diagnosis by calendar period for these subgroups; furthermore, estimates for some groups (e.g., women) could be subject to considerable random variability over time.

The CDC recommendations for universal opt-out HIV testing were published in September 2006; in light of inherent reporting delays with HIV surveillance data and data incompleteness, the evidence as to whether these guidelines have led to earlier HIV diagnosis (i.e., at higher CD4 counts) nationwide is not likely to be apparent for some time. Nonetheless, encouraging trends have been seen in jurisdictions that have dramatically expanded HIV testing [19, 20]. In our study, we did not detect a statistically significant trend to higher CD4 cell counts at HIV diagnosis overall, although the pattern of change was in the direction of increasing CD4 cell counts, particularly for patients with heterosexual exposure as a risk factor for HIV infection and those presenting to public facilities.

In conclusion we have found that even among HOPS participants who successfully entered care, over one-third were diagnosed with CD4 counts <200 cells/mm³ and over one-half were diagnosed with CD4 count <350 cells/mm³, the threshold at which initiating cART is unequivocally recommended (up to CD4 count of 500 cells/mm³, above which level it is to be considered) [21]. These findings should raise concern. Prompt HIV diagnosis, entry into care, and timely initiation of cART are critical for reducing the risk of both opportunistic and nonopportunistic disease, prolonging survival, and reducing onward HIV transmission. Our findings also suggest that expanding testing and reducing late HIV diagnosis need to be a priority, if the programs related to improving linkage to care and earlier antiretroviral treatment initiation [22] are to reach patients and potentially alter the trajectory of the US HIV epidemic [27]. An estimated 21 to 25% of HIV-infected persons in the USA remain undiagnosed [9, 10], with the prevalence of unrecognized HIV infection approaching 50% among some urban MSM [28]. Many untested persons either perceive themselves to be at low risk of HIV infection or are fearful of learning their HIV status [28]. Such persons are more likely to have a late stage or illness-triggered diagnosis. Our findings reinforce the need to establish universal routine HIV testing as standard of care for all adolescents and adults seen in private and public care settings, regardless of patient-reported HIV risk [11]. It is only under such circumstances that late-stage or illness-triggered HIV diagnoses will be reduced and that sociodemographic disparities in stage of HIV disease at diagnosis be eliminated.

The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.

None of the authors have conflict of interests.