Persons living with HIV (PLWH) must fulfill several steps along the care continuum to achieve optimal clinical outcomes (Figure 1).^1-3 Individuals should be screened for HIV; however, 20% of PLWH in the United States (US) are unaware of their diagnosis.⁴ Linkage to care is then necessary, but only 77% of individuals link to care.^{1-2, 5-6} Patients must remain in care; yet 50-75% of those linked to care meet the U.S. Health Resources and Services Administration (HRSA) retention criteria of ≥2 visits separated by 90 days in a year.^{2, 5, 7-10} Finally, PLWH need to receive and adhere to antiretroviral therapy (ART). Navigating all these steps is often unsuccessful:¹¹ only 28% of those with HIV achieve virologic suppression.¹²

Linkage to care is a critical step in this process.^13-15 However, no consistent definition of linkage to care exists. Prior studies have defined linkage to care as attending ≥1 clinic visit for HIV care within one to six months of diagnosis,^16-35 or ≥2 visits within six to twelve months of diagnosis.^36-38 Other researchers have used laboratory monitoring data--CD4 T-cell counts and HIV-1 RNA levels--to investigate linkage to care, defining linkage as the occurrence of laboratory testing within one to six months of diagnosis,^39-41 21 or more days after diagnosis,⁴² or within twelve months of diagnosis.⁴³ In addition, governmental agencies and professional organizations differ in how they measure linkage. The US National HIV/AIDS Strategy defines linkage as laboratory testing within 90 days of diagnosis,⁴⁴ while the Emergency Department National HIV Testing Consortium defines linkage as a clinic visit within 30 days of diagnosis.²⁷

Linkage rates have similarly varied among single-clinic or multi-site cohort studies, from 38% to 100%, depending on the linkage criteria used.^45-46 Few studies have been community-based. A King County, Washington study examined the timing of linkage to HIV care countywide,⁴⁷ but did not specifically address the predictive ability of linkage metrics. A recent study looked at the predictive ability of two linkage criteria (having either one or two laboratory tests within a year of diagnosis) for retention in care, but did not evaluate other linkage definitions.⁴³

To effectively monitor and improve linkage to care, a better understanding of the predictive accuracy of linkage to care measures for retention and virologic suppression is necessary. To determine the diagnostic accuracy of linkage measures to predict retention in HIV care and HIV virologic suppression, we compared clinic visit and laboratory testing based linkage measures using a city-wide cohort.

Our cohort included 1781 patients. Most patients were male (70.1%) and black (63.3%), and had a heterosexual risk exposure (61.0%). Approximately one-third had AIDS at HIV diagnosis (34.5%) (Table 1). Linkage rates for clinic-based measures were applicable for RWP-funded clinics only, and ranged from 17.5% for having two visits in 180 days separated by 90 days, to 39.5% for having any visit within 365 days of diagnosis. For laboratory-based measures, linkage rates ranged from 34.0% for having two tests 90 days apart within 365 days, to 81.6% for having one test in 365 days (Table 2). Data and diagrams representing progression through the HIV care cascade for each linkage metric are presented (Supplemental Digital Content [SDC]: Table 1 and Figures 1-28).⁵¹ Progression through the HIV care cascade, stratified by sociodemographic characteristics, are also shown (SDC Table 2).

Between 366 and 730 days after diagnosis, 780 patients (43.8%) successfully attended ≥1 clinic visit, and 503 patients (28.2%) met the HRSA retention measure (two clinic visits ≥90 days apart in one year). Similarly, 366-730 days after diagnosis, 23.5% (N=419) had virologic suppression. In total, 1108 patients (62.2%) were neither retained in care nor had virologic suppression 366-730 days after diagnosis.

Using retention in care as the outcome of appropriate linkage, the measure with the highest PPV for retention was attending two clinic visits ≥90 days apart within 365 days of diagnosis (74.2%; 95% CI: 70.8%-78.0%). The linkage measure with the lowest PPV for retention was laboratory monitoring within 30 days (32.0%; 95% CI: 29.1%-34.9%). PLWH who did not have a clinic visit between 21 and 365 days after diagnosis had the highest NPV for retention in care (94.5%, 95% CI: 0.928-0.962), while the absence of laboratory monitoring within 30 days of diagnosis had the lowest NPV for retention (77.0%, 95% CI: 74.0%-80.0%). The measure with the highest AUC for retention in care was attending a clinic visit between 21 and 365 days after diagnosis (Table 3).

Using virologic suppression as the outcome of appropriate linkage, the measure with the highest NPV for virologic suppression was lacking laboratory testing within 365 days of diagnosis (98.2%; 95% CI: 96.1%-99.3%) and the measure with the lowest NPV was the absence of a clinic visit within 30 days of diagnosis (87.0%, 95% CI: 85.0%-89.0%). The measure with the lowest PPV for virologic suppression was having two tests in 180 days separated by 90 days (41.0%, 95% CI: 37.0%-45.0%), while having two visits in 180 days separated by 90 days had the highest PPV for virologic suppression (49.0%, 95% CI: 43.0%-55.0%). Patients completing two laboratory tests separated by 90 days in 365 days had the highest AUC for virologic suppression (Table 4).

Sensitivity analyses were performed to account for care received outside of RWP-funded clinics. Linkage measures associated with the highest and lowest odds of retention in care, sensitivities, specificities, NPVs, and PPVs did not change. Sensitivity analyses were also performed to account for the possibility of unreported undetectable viral loads. NPVs and PPVs did not differ greatly. When sensitivity analyses were performed assuming both laboratory monitoring-based measures and the outcome measure of virologic suppression were underestimated, PPV changed little, but NPV decreased to as little as 11% for having any laboratory test sent within 365 days (SDC, Tables 3-6).

Additional analyses were performed to determine the effect of mortality on estimates. Point analyses of those alive at 730 days after diagnosis did not differ by 10% of their baseline. We also performed subanalyses examining only those patients who did not move out of Philadelphia during the study. Point analyses did not differ by 10% of their baseline.

This is one of the first studies comparing multiple laboratory and clinic-based measures of linkage to care. Clinic-based linkage measures, in particular completing ≥1 clinic visits between 21 and 365 days after diagnosis, best predict retention in care. On the other hand, completing two laboratory tests separated by 90 days within 365 days of diagnosis best predicts virologic suppression. These data suggest that both clinic and laboratory-based linkage measures have value. Selection of a linkage measure should be tailored to the outcome of interest.

Government agencies and professional organizations vary in criteria used to define linkage to care. The Emergency Department National HIV Testing Consortium metric is a clinic visit within 30 days of diagnosis.²⁷ In our study, only 62.3% of patients meeting this measure were retained in care, and only 40.5% achieved virologic suppression. The US National HIV/AIDS Strategy recommends linking 85% of persons to care within 90 days of diagnosis, using a laboratory-based measure.⁴⁷ While 72.0% of patients met this measure, only 32.5% were retained, and 27.7% achieved virologic suppression. Using the linkage metric most predictive of retention in care, 38.2% of patients completed a clinic visit between 21 and 365 days after diagnosis, with 65.0% of patients meeting this measure retained in care. Alternatively, using the linkage metric most predictive of virologic suppression, 56.3% of patients had two laboratory tests 90 days apart within 365 days of diagnosis, and 38.0% of these achieved virologic suppression. Our data suggest that agencies and organizations should consider which linkage metrics best meet their outcomes of interest when recommending linkage criteria.

Laboratory-based linkage measures had lower predictive abilities for retention in care than clinic-based measures. 81.6% of the population had a laboratory test within 365 days of diagnosis, while only 39.5% had a clinic visit within the Ryan White system within 365 days of diagnosis. Clinic visits were underreported to a greater extent than laboratory testing, as we used the HRSA definition for HIV clinic visits⁵⁰ (which excludes pharmacy, nursing, social work, and other visits to providers without prescribing privileges), and as we were unable to detect visits to non-RWP-funded clinics. Furthermore, laboratory testing may occur outside of the primary HIV care setting, including in non-HIV clinics, inpatient hospitals, and emergency departments. Our data differs from recent studies suggesting the use of laboratory testing as a proxy for clinic visits.⁵² While using laboratory tests as a proxy for clinic visits may be helpful in clinic cohort-based studies,⁵² our data suggest that laboratory tests may not be as predictive of clinic visits in all settings, such as surveillance.

Prior studies have evaluated metrics for retention in care^53-54 and predictors of virologic suppression.⁵⁵ However, few studies have examined how well different linkage metrics predict retention in care or virologic suppression. Among these, only one⁴¹ or two⁴³ linkage metrics were compared. We present a more complete picture of linkage to care, retention in care, and virologic suppression than prior clinic cohort-based studies.

Our study had limitations. First, generalizability was limited as we only studied patients in one US city. Also, linkage, retention, and virologic suppression are lower than reported elsewhere.^{2, 5, 45-46, 56-59} For example, a meta-analysis of 28 studies estimated that 77% of patients are linked to care, 51% are retained in care, and 35% achieve virologic suppression.^{2, 5} Linkage rates in our study ranged from 17.5% to 81.6%, 28.2% of patients were retained in care, and 23.5% achieved virologic suppression. Our inability to access clinic visit data on patients seen outside RWP-funded clinics would have underestimated linkage to and retention in care for the metrics that used this data. However, sensitivity analyses accounting for PLWH not attending RWP-funded clinics demonstrated that the linkage metrics most predictive for retention in care and virologic suppression did not change. If laboratories were not reporting undetectable HIV-1 RNA levels, we also may have underestimated virologic suppression. Sensitivity analyses accounting for this did not alter the results. We also performed sensitivity analyses to account for those who died or migrated out of the city. While it is possible that PLWH in care may lack laboratory monitoring, this is rare (unpublished data, Philadelphia DPH). Similarly, it is unlikely that PLWH not in care would achieve virologic suppression.

We focused on virologic suppression and retention in care between 366 and 730 days after diagnosis to differentiate linkage to care from retention in care, as some researchers defined linkage to care within the first 365 days after diagnosis.³⁸ Some patients may have dropped out of care prior to this time, potentially contributing to the lower retention rates seen in the study. In addition, we did not collect data on the timing of ART initiation, which has been associated with both improved retention in care and virologic suppression.^60-61 Low rates of ART initiation among PLWH in our cohort could further lower the proportion of patients not meeting retention in care and virologic suppression targets.

Despite these adjustments, linkage and retention metrics were still lower than reported elsewhere.^{2, 5, 45-46, 56-59} Our data came from a large city with many racial minorities and persons below the poverty line, which may be associated with lower rates of retention in care. Care provider characteristics may also differ from other cities. For example, providers, including non-HIV providers working in urgent care, emergency, inpatient, and primary care clinic settings, may be less likely to send laboratory tests than in other regions, lowering laboratory-based linkage rates. Further studies are needed to evaluate how linkage measures perform in other locals and settings.

Understanding the predictive ability of measures of linkage to care is necessary for improving the quality of HIV care and reducing HIV transmissions. Our data suggests that selection of the ideal linkage measure depends on the outcome of interest being evaluated. The clinic-based measure of completing a visit between 21 and 365 days after diagnosis best predicted retention in care, and may be useful for testing centers focused on referring PLWH to care. Meanwhile, the laboratory-based measure of completing two laboratory tests separated by 90 days within 365 days of diagnosis best predicted virologic suppression, and may be a helpful definition for test-and-treat strategies aimed at reducing community viral load. Researchers studying retention and virologic suppression in PLWH, testing agencies seeking to improve the quality of their work, and funding agencies deciding how to allocate resources should tailor linkage measures based on the outcome of interest.