Contributions to early HIV diagnosis among patients linked to care vary by testing venue

The HIV epidemic continues despite decades of aggressive prevention efforts and advances in effective treatment. An estimated one-quarter of those infected remain unaware of their HIV status, and late diagnosis remains common. [1‐3] Traditional HIV screening that relies on resource-intensive outreach, or individuals to seek testing, has failed to fully access at-risk populations. Early diagnosis is essential to enable opportunities to halt further transmission [4] and to obtain medical therapy that improves morbidity and mortality for infected patients. [5] The CD4+ T cell count is the accepted marker of disease progression, and is a surrogate measure of how early in the disease the diagnosis has been made. [2, 3, 6, 7] CD4+ count declines variably in HIV patients but typically decreases by an average of 60–100/μL per year [8‐10].

Strategies to improve early diagnosis necessarily depend on who is tested and under what circumstances. Generally, later diagnosis is associated with diagnostic testing as a result of illness, while earlier diagnosis is more typical of patient self-request for testing, self-perceived risk, or universal screening. [2, 11] Modifications to expand universal screening within health care settings have been proposed. [12] The CDC initiative particularly encourages screening in non-traditional settings such as emergency departments, substance abuse treatment centers, and corrections facilities. [7, 11‐14] The resources required to implement these changes broadly have not yet been allocated.

Understanding the differences between test venues in terms of numbers of new diagnoses successfully linked to care and the extent to which those diagnoses are early or late would inform health care planners on current practice patterns and resource use. For example, patients who are seeking HIV testing may self-refer to different venues than patients who are seeking medical care because of symptoms. Prior reports have examined the contributions by different screening venues to HIV testing. [1, 7, 13] However, to our knowledge, variation in stage of diagnosis by screening venue has not been assessed. To improve rates of early HIV diagnosis on a regional basis, policy makers would benefit from an improved understanding of diagnosis patterns to inform future allocation of scarce prevention resources.

We hypothesized that diverse regional testing venues would vary significantly in the number of new diagnoses made with successful linkage to care and the initial CD4 count of those patients at the time of diagnosis. This study was designed with the following two objectives, 1) to determine how many newly diagnosed persons were successfully referred by each local test venue relative to the total number of newly diagnosed persons seen at the regional HIV treatment center, and 2) to assess the differences in initial stage of HIV by test venue. We were specifically interested in determining the relative contribution of a publicly funded emergency department testing program to regional patterns of diagnosis.

Of 1,226 records for patients newly attending the HIV treatment center during the study period, 277 unique patients were newly diagnosed and determined to meet study criteria. Reasons for exclusion are shown in Figure 1. Overall, included patients had a median age of 33 years (range 17 to 77 years), were 76% male, 46% African-American, and are described in Table 1. There were 143 patients identified through medical testing and 75 diagnosed by public health testing. Most medical outpatient settings referring newly diagnosed patients were either hospital affiliated primary care clinics or publically funded clinics and health centers in relatively close geographic proximity to the study site. Of those patients tested in these outpatient settings, 29 (34.5%) were privately insured, 4 (4.8%) had Medicare, 14 (16.7%) had Medicaid, 5 (6%) had some other form of public assistance, and 32 (38.1%) had no medical insurance. Other venues not classified as health care or public health testing included a health department funded HIV counseling and testing program in an emergency department [15, 16] (n = 23), a publicly funded partner testing program at the region's HIV treatment center (n = 8), and universal screening by the blood bank, military, and insurance pre-screening (n = 13). Two patients were identified outside of the region and 13 were from an unknown referral source.

Overall, there were 129 patients (46.6%) diagnosed early with a CD4 greater than 350, and 104 (37.5%) diagnosed late with a CD4 less than 200. The median values for initial CD4 count are shown in Figure 2. There were no differences in HIV viral load by site of diagnosis (data not shown). Partner testing by the IDC (median CD4 648/μL) and universal screening (median CD4 483/μL) identified patients earlier in their disease course than testing in other venues (partner testing v all others, p = 0.008 and universal screening v all others, p = 0.05). While inpatient testing contributed a large number of new diagnoses, it predominately identified patients with more advanced HIV disease (median CD4 24/μL, v all others p < 0.001).

Table 2 describes the reasons for testing for each referral source. Generally, testing because of symptoms or AIDS defining illness was more common at those venues that contributed predominately to later diagnoses. The emergency department testing program, in accordance with program design and setting, tested both for purposes of diagnosis and according to patient risk and patient self-request. Thus, amongst those with recorded reasons for testing in the emergency department; roughly half were for symptoms consistent with AIDS (7/9 had a CD 4 count < 350/μL) and half were based on risk factors or patient request (6/10 had a CD4 count > 350/μL).

The number of new cases identified by each setting was plotted against the median CD4 count of newly diagnosed patients in Figure 3. With the exception of health department and medical outpatient testing, the results show that settings identifying patients earlier contributed fewer total cases, whereas later diagnosis by hospital testing contributed significantly to the total number of newly identified patients.

To our knowledge this is the first regional assessment comparing testing venues by the number of newly diagnosed patients and the stage of illness at diagnosis for patients successfully linked to subsequent care. As hypothesized, the number and timing of new HIV diagnoses are variably distributed. Our data demonstrate a mismatch between stage of diagnosis and what is optimal for public health goals; many newly identified HIV patients were diagnosed later as a result of diagnostic testing and too few were identified early enough through screening. This conceptual framework affirms the recent changes to the guidelines for testing in healthcare settings. [12] On a regional and national basis, progress towards earlier patterns of HIV diagnosis would be expected to occur if available resources were shifted for maximal effect. Screening by those venues with a track record of earlier diagnoses could be increased provided the additional resources would be effectively translated into a similar pattern of success. In settings where diagnoses are typically later, it might be possible to modify testing methods in a manner that would promote earlier diagnosis. Alternatively, resources might be shifted from screening in late diagnosing venues in favor of those providing earlier diagnoses, although diagnostic testing in such settings would presumably still be encouraged. Serial evaluations such as the one provided here or as provided by name-based HIV case surveillance could directly assess the effects of strategic planning by local, state, and national health officials. Regardless of the source of such data, monitoring new HIV diagnoses, linkage to care, and CD4 count at the time of diagnosis by testing venue should be an increasingly emphasized and utilized component of prospective surveillance efforts. With such a strategy it becomes possible to assess the extent to which patterns of resource allocation for screening and testing match patterns of early diagnosis and linkage of HIV positive patients to care without detailed investigation or reporting from every possible testing facility.

Our setting and study design was advantageous for a variety of reasons. There is a low to moderate prevalence of HIV, making results generalizable to a wide number of settings that are likely to renew and expand HIV prevention efforts after the release of the 2006 CDC guidelines for HIV testing in healthcare settings. [12] While representing an urban area of considerable size, the vast majority of HIV positive patients in the region are cared for by a single HIV treatment center. There is a highly active emergency department based HIV screening program in the area, allowing preliminary assessment of the potential importance of recently expanded CDC emphasis on emergency departments. [11, 12] Our study design allowed a relatively robust sample size without the time and expense required to construct a prospective surveillance system. Also, duplication of new diagnoses by testing the same patient in different care settings was not a factor.

A primary limitation of our work is the use of records from an HIV treatment center to identify study subjects. Our analysis only indirectly equates testing venue with the testing strategies employed by those venues. Our methods did not assess the total number of tests performed or the costs of service provided for any venues. Similarly, we cannot account for diagnosed patients who never received their results or were not linked to the regional specialty care clinic. Some of these patients may have remained in, or been linked to, an alternative and appropriate patient-provider relationship. Because advanced illness would more reliably lead to contact with the HIV treatment center, our inclusion criteria would be biased in favor of those with advanced illness; the net result would be a greater representation of diagnoses arising from settings that test late in the course of illness, such as hospitals.

Another important limitation of our work is that direct assessment of screening versus diagnostic testing was not possible. Nonetheless, we have broadly grouped testing venues by the patterns of testing and patient populations that are likely to have been present during the study period. We suggest that these patterns mirror what is likely the case on a national basis. Testing by medical providers in the absence of dedicated screening efforts is likely to be primarily based on signs and symptoms of illness. This diagnostic testing has been consistently shown to lead to later diagnosis. [2, 11] Not surprisingly, testing by hospitals was especially likely to identify patients with symptomatic HIV and very low initial CD4+ counts. Testing by publicly funded programs is far more likely to be a result of screening based on patient request or recognized risk factors, and accordingly leads to earlier diagnosis than diagnostic testing [2, 11].

The inclusion criteria of our study may also be subject to scrutiny. We assumed all patients to be newly diagnosed if there was no direct indication to the contrary. Several patients could have tested positive months or even years prior to the positive test that ultimately led them to the infectious disease center. This too would bias our results in favor of later diagnosis. However, if there is no chart record of prior tests, patient awareness of those tests, or medical care based on those tests, then the importance of prior testing, even if it did occur, may be suspect. In this sense, a repeat positive test might still be considered a new diagnosis if it led to subsequent actions where there had been none previously.

The emergency department program was considered separately due to its unique approach. During the study period, this emergency department program was the only known healthcare setting with a large scale dedicated screening effort. Notably this program was funded by the local health department and involved health counselor screening for HIV risk. [15, 16] However, the setting also provided access to patients with acute medical illness and accepted referrals from medical providers for testing based on signs and symptoms potentially suggestive of HIV. Overall, the median CD4 for the emergency department screening program was lower than anticipated, both in absolute terms and relative to other regional screening efforts. While this finding prompts many questions, it should not be interpreted as inhibitory to the increased interest in emergency department based HIV testing. Because emergency departments see both acutely ill patients and disadvantaged populations that seek emergency department care for more routine health needs, it is likely that diagnosis will result from both diagnostic testing leading to late diagnosis and screening leading to early diagnosis. Indeed, among those patients for whom a testing indication was recorded, roughly half were for symptoms consistent with AIDS and half were based on risk factors or patient request. It is also notable that, despite the support of the local health department, available resources allowed only a small proportion of the eligible emergency department population to be tested [15, 16].

Patients diagnosed by the HIV treatment center had an exceptionally high median CD4 count. This testing program was targeted towards partners of HIV patients cared for at the clinic. At the time, this program was also funded by the local health department. When fully contextualized with markers of early diagnosis, the partner testing program is compelling. Our results further support the importance of partner testing programs.

There were several sources of universal screening during the study period, including insurance pre-screening, screening prior to entering military service, and blood bank surveillance. Blood bank surveillance is a unique contributor to the HIV prevention picture. The blood bank attempts to remove the majority of at-risk patients prior to donation, but then routinely tests all donations to ensure the safety of the blood supply. This is entirely different than other testing which is aimed at seeking undiagnosed HIV positive persons. It is therefore expected that the number of positive patients identified by the blood bank will be small in absolute number, but that those diagnosed will be diagnosed early and well before symptoms of illness develop, an expectation that our data supports.

Our findings should be interpreted in light of several limitations in addition to those already noted. Although we anticipated only 200 patients and actually identified 277, there were 362 of 1278 potential records that could not be located. If any missed records were randomly allocated between disease severity and referral source then it would have little effect on our primary analyses other than to reduce sample size. The low regional prevalence additionally contributes to the small sample size of some referral subgroups. Our chart review methods were rigorous, but data such as the original reason for testing or even the testing venue was frequently unavailable. There were also 14 patients excluded since they did not have a CD4+ count measured proximate to their time of diagnosis.

Our data show that among patients linked to care at this referral center, many newly identified persons with HIV are diagnosed late, and the extent to which they were likely to have been diagnosed late varied by the facility in which they were diagnosed. Many of these patients were tested in healthcare settings due to a medical suspicion for HIV infection. Screening programs with demonstrated potential to identify patients with HIV early in their disease course contributed less. This is the inverse of what is desirable. This suggests that monitoring of core outcomes, including initial CD4 and linkage to care by testing venue, is a critical consideration when allocating resources. We provide a conceptual paradigm for use when resources are insufficient to maximize testing in all possible ways; settings that generally diagnose HIV positive patients late in their disease course should either modify methods to augment earlier diagnosis or be de-emphasized as a focus of coordinated screening efforts in favor of other venues. Settings that typically provide early identification should receive additional support provided that support translates to an increased absolute number of new diagnoses. Emergency department testing can contribute significantly to regional diagnosis patterns, but the extent to which that contribution is early will depend upon the strategies utilized. Markers of early diagnosis must be tracked by screening programs and policy makers to quality assure prevention efforts and coordinate prevention activities on a regional level.