Introduction
The Centers for Disease Control and Prevention (CDC) estimated that in 2016, approximately 14.2% of persons living with HIV were unaware of their infection [1] and that persons unaware of their infection accounted for approximately 38% of HIV transmissions in the United States [2]. Early diagnosis of individuals with HIV is important for the prevention of transmission, as many after diagnosis would change sexual and drug use behaviors and take antiretroviral treatment (ART) to reduce the risk of transmitting HIV to their sex and needle-sharing partners [3–5]. Early HIV diagnosis is also important for the prevention of HIV-related illness, as early diagnosis is the essential first step toward early ART, which greatly reduces the risk of HIV-related illness [6].
Given the importance of early HIV diagnosis, the US Department of Health and Human Services (DHHS) has included early diagnosis as one of the four pillars of its new initiative, ‘Ending the HIV Epidemic: A Plan for America [7]’. Studies have reported the mean, median, and interquartile range of the time from HIV infection to diagnosis, but we need an indicator for early HIV diagnosis that can be easily understood by policy makers and other stakeholders [8–11]. Our objective was to develop such an indicator.
Methods
Probability of diagnosis within 1 year of HIV acquisition
The New York City (NYC) HIV surveillance data were used for the analysis. Modelled on the infant mortality rate (IMR), which is the number of infants who died during a given calendar year and were aged less than 1 year at death divided by the number of live births in that calendar year, to assess early deaths [12], we use the probability of diagnosis within 1 year of HIV acquisition, which is the number of cases diagnosed in a given calendar year for which diagnosis occurred within 1 year of HIV acquisition divided by the number of incident HIV cases in that calendar year [Equation (1)], to assess early HIV diagnosis. Similar to IMR that assumes stable birth rate in the calendar year and the one prior, the probability of diagnosis within 1 year of HIV acquisition assumes stable HIV incidence.
Both measures have the limitation that some persons included in the numerator are not part of the denominator. Modelled on the adjusted IMR [12], we also report the adjusted probability of HIV diagnosis within 1 year of HIV acquisition, which is the number of persons who acquired infection in a given calendar year and who were diagnosed within 1 year of acquisition divided by the number of incident HIV cases in that calendar year [Equation (2)]. Similar to the adjusted IMR (that assumes stable mortality rate in the calendar year and the one following and requires one more year of data than IMR), the adjusted probability assumes stable probability of diagnosis and requires one more year of data than the probability.
;)
Estimation method
The two equations have different numerators but the same denominator. The method to estimate the two numerators and the one denominator has been described previously [13,14]. Briefly, the trajectory of CD4+ count on a square root scale was modelled as a linear function of time since infection, with a different slope and intercept for each group by the combinations of sex (male and female patients), age (13–19, 20–24, 25–34, 35–44, 45–54, and 55+), and transmission risk (MSM, people who inject drugs, heterosexuals, and other) [Equation (3)] [15,16].
where CD4+ count (t) is an individual's CD4+ count at time t, which is the time since HIV infection, the intercept (ai) and slope (bi) are random variables following normal distributions, and eit is the error term.
To estimate the date of infection, we first updated previously published CD4+ depletion model parameters, using a subset of CASCADE (Concerted Action on Seroconversion to AIDS and Death in Europe) data pooled in September 2014 within EuroCoord Network of Excellence of selected cohorts in countries where most of the HIV infections are subtype B. We also redefined covariates to match the HIV risk and population stratifications commonly used in the United States.
The model and an individual's first CD4+ count, which is a required data element on the HIV case report form, were used to estimate the date of infection [Equation (4)]. Persons with CD4+ tests were assigned a weight to account for those without a CD4+ test based on the year of HIV diagnosis, sex, race/ethnicity, transmission category, age at diagnosis, and the status at the end of the study period – whether the person was living with HIV whose disease had never been classified as AIDS, died without ever having been classified as AIDS, or had progressed to AIDS regardless of whether living or dead.
where the first CD4+ count is an individual's first measurement of CD4+ count at or after diagnosis but before ART.
With the date of HIV infection, the distribution of diagnoses was estimated using standard survival analysis techniques for right truncated data and assuming that the distribution of diagnosis delay does not change over time [17]. We first estimated the probability, P(x), retrospectively -- that a person with diagnosed HIV would have been diagnosed within a certain duration (x units of time) after infection. A diagnosis delay weight, W(x) = 1/P(x), was then assigned to each person according to the infection date estimated for the person (x = time from infection to the ending date of the study period). The diagnosis delay weights were generated separately by strata determined by sex, transmission risk, and race/ethnicity. With the date of infection and diagnosis delay weight associated with each case diagnosed during the study period, we estimated HIV incidence in each calendar year by summing the diagnosis delay weights [13,14], and used the number as the denominator of Equations (1) and (2). The numerator of Equation (1) was obtained by counting persons who acquired infection in the given calendar year or the one prior and were diagnosed in that calendar year within 1 year of acquisition. The numerator of Equation (2) was obtained by counting persons who acquired infection in the given calendar year and were diagnosed in that calendar year or the one following within 1 year of acquisition.
Uncertainties associated with each estimation step were calculated and combined based on analytical formulas or approximations using the delta method [18]. As multiple imputation was used to assign missing transmission risk, and multiple dates of infection were simulated for each person, the standard procedure for combining results from multiple imputation datasets was followed to obtain the final estimates and their standard errors [19].
Results
All persons diagnosed with HIV in NYC from 1981 to 2017 (N = 221 995) were included in the analysis; their characteristics are shown in Table 1. Approximately 21% of persons did not have an identified transmission risk in the registry. Multiple imputation was used to assign a set of plausible values from 10 imputed datasets [13]. The estimated number of HIV diagnoses in the calendar year for which diagnosis occurred within 1 year of acquisition, that is, the numerator of Equation (1), decreased from 1300 in 2012 to 940 in 2016 (eTable 1, https://links.lww.com/QAD/B685), and the estimated HIV incidence, i.e. the denominator of Equations (1) and (2), decreased from 3100 in 2012 to 2200 in 2016 (eTable 2, https://links.lww.com/QAD/B685).
Table 1: Number of HIV diagnoses in New York City included in the analysis, by year of diagnosis and characteristics.
Overall, the estimated probability of diagnosis within 1 year of acquisition in NYC, 2012–2016, was stable, with 43.0% [95% confidence interval (CI): 37.9--48.2%) in 2012, 42.5% (95% CI: 36.8--48.3%) in 2013, 42.8% (95% CI: 36.3--49.2%) in 2014, 42.9% (95% CI: 35.4--50.3%) in 2015, and 42.2% (95% CI: 33.1--51.2%) in 2016 (Table 2).
Table 2: Probability of diagnosis within 1 year of HIV acquisition in New York City, 2012–2016.a
Among males, by race/ethnicity, whites had the highest probability of diagnosis within 1 year of acquisition (e.g. 53.8% in 2016), and persons of ‘Other’ race/ethnicity including American Indian/Alaska Native, Asian, Native Hawaiian/Other Pacific Islander, and persons of multiple races, had the lowest (e.g. 33.9% in 2016). By age, persons in the 13–24-year age group were the least likely to have their infection diagnosed within 1 year of acquisition, with 35.1% (95% CI: 27.8--42.3%) in 2012, 36.3% (95% CI: 27.5--45.0%) in 2013, 37.6% (95% CI: 26.6--48.6%) in 2014, 37.4% (95% CI: 24.1--50.7%) in 2015, and 36% (95% CI: 18.7--53.3%) in 2016. By transmission risk, over 42% of male individuals with infection attributed to male-to-male sexual contact had their infection diagnosed within 1 year of acquisition, with 43.8% (95% CI: 37.7--49.8%) in 2012, 42.7% (95% CI: 36--49.5%) in 2013, 43.4% (95% CI: 35.8--50.9%) in 2014, 43.9% (95% CI: 35.1--52.6%) in 2015, and 42.8% (95% CI: 32.0--53.5%) in 2016. For other transmission risk groups, all estimates had a wide 95% CI because of small sample sizes.
Among female individuals, all estimates for subgroups had a wide 95% CI because of small sample sizes, but persons in the 13–24 years’ age group were generally the least likely to receive a diagnosis within 1 year of acquisition, with 38.3% (95% CI: 16.3--60.3%) in 2012, 36.3% (11.9--60.8%) in 2013, 41.8% (7.2--76.4%) in 2014, 37.6% (95% CI: 0.0--79.2%) in 2015, and 44% (95% CI: 0.0--100.0%) in 2016.
eTable 3, https://links.lww.com/QAD/B685 shows the estimated number of persons who acquired infection in the calendar year and were diagnosed within 1 year of acquisition in NYC, 2012–2016, that is, the numerator of Equation (2). The adjusted probability of diagnosis within 1 year of acquisition in NYC overall and by subgroups are essentially the same as those in Table 2, with the overall probability being 43% (95% CI: 39.2--46.7%) in 2012, 42.4% (95% CI: 38--46.9%) in 2013, 42.7% (95% CI: 37.6--47.9%) in 2014, 42.9% (95% CI: 36.7--49.1%) in 2015, and 42.4% (95% CI: 34.5--50.2%) in 2016 (eTable 4, https://links.lww.com/QAD/B685).
Discussion
We developed a new indicator, the probability of diagnosis within 1 year of HIV acquisition, to assess early diagnosis, and found that over 42% of persons who acquired HIV in NYC during 2012–2016 had been diagnosed within 1 year of acquisition. The probability remained stable, suggesting that the effectiveness of diagnosing HIV infection in NYC may have reached a plateau. Moreover, disparities were observed: nonwhites were less likely to have their infection diagnosed in the first year than whites, and persons in the 13–24-year age group were the least likely compared with other age groups. Persons in their first year of infection, particularly those acutely infected have a high viral load [20,21], and high viral loads are associated with high transmission risk [4,22,23]. A lower probability of diagnosis within 1 year of acquisition can lead to more persons with high viral load remaining undiagnosed for a longer period of time, potentially unknowingly transmitting HIV to their sex or needle-sharing partners.
The measure and our findings have some limitations. First, the probability relies on the assumption of stable HIV incidence in the calendar year and the one prior, and the adjusted probability relies on the assumption of stable probability of diagnosis in the calendar year and the one following. When the assumption of stable HIV incidence is not met, the probability will be biased; when the assumption of stable probability of diagnosis is not met, the adjusted probability represents the average effectiveness of HIV testing in the calendar year and the one following and may not be appropriate for the evaluation of HIV testing in any calendar year. In NYC during the study period, 2012–2016, HIV incidence and the probability of diagnosis between two adjacent calendar years were relatively stable, and both assumptions were met.
Second, the accuracy of our estimates relies on the accuracy of the CD4+ count depletion model, and the precision relies on the variability in individuals’ CD4+ count. The main goal of our article is to introduce the indicator, not the estimates. In our analysis, we used the CD4+ count depletion model to estimate the numerator and denominator and calculate the probability. Other methods can also be used, if they provide more accurate and precise estimates.
Third, the overall and by subpopulation probabilities of diagnosis within 1 year of acquisition provide a complete picture of early HIV diagnosis in a jurisdiction, but when the probabilities are used to guide HIV testing efforts, other factors, such as HIV incidence, need to be taken into account.
One of the key strategies in the ‘Ending the HIV Epidemic: A Plan for America’ is to diagnose all people with HIV as early as possible [7]. We recommend that the nation and local health jurisdictions use this new indicator -- the probability of diagnosis within 1 year of HIV acquisition -- to monitor their progress towards ending the HIV epidemic in the United States. Like other measures, for example, the mean and median time from HIV infection to diagnosis, the probability of diagnosis within 1 year of HIV acquisition requires large numbers of infections/diagnoses to provide stable estimates. To avoid misinterpretation, jurisdictions may need to set a policy on whether to report estimates with a large relative standard error, for example, at least 30% [1].
Acknowledgements
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 (CDC).
The authors would thank Kent Sepkowitz, Oni Blackstock, and James Hadler of the New York City Department of Health and Mental Hygiene, and Anna Satcher Johnson, Azfar-E-Alam Siddiqi, Timothy Green, and James Carey of the Centers for Disease Control and Prevention (CDC), for their review and comments on an earlier version of the manuscript.
Conflicts of interest
There are no conflicts of interest.
References
1. Centers for Disease Control and Prevention
Estimated HIV incidence and prevalence in the United States 2010-2016.
HIV Surveill Suppl Rep 2019; 24:1–89.
2. Li Z, Purcell DW, Sansom SL, Hayes D, Hall HI.
Vital signs: HIV transmission along the continuum of care - United States, 2016.
MMWR Morb Mortal Wkly Rep 2018; 68:267–272.
3. Steward WT, Remien RH, Higgins JA, Dubrow R, Pinkerton SD, Sikkema KJ, et al.
Behavior change following diagnosis with acute/early HIV infection-a move to serosorting with other HIV-infected individuals. The NIMH Multisite Acute HIV Infection Study: III.
AIDS Behav 2009; 13:1054–1060.
4. Cohen MS, Chen YQ, McCauley M, Gamble T, Hosseinipour MC, Kumarasamy N, et al.
Prevention of HIV-1 infection with early antiretroviral therapy.
N Engl J Med 2011; 365:493–505.
5. Myers JE, Braunstein SL, Xia Q, Scanlin K, Edelstein Z, Harriman G, et al.
Redefining prevention and care: a status-neutral approach to HIV.
Open Forum Infect Dis 2018; 5:ofy097.
6. Grinsztejn B, Hosseinipour MC, Ribaudo HJ, Swindells S, Eron J, Chen YQ, et al. HPTN 052-ACTG Study Team
Effects of early versus delayed initiation of antiretroviral treatment on clinical outcomes of HIV-1 infection: results from the phase 3 HPTN 052 randomised controlled trial.
Lancet Infect Dis 2014; 14:281–290.
7. Fauci AS, Redfield RR, Sigounas G, Weahkee MD, Giroir BP.
Ending the HIV epidemic: a plan for the United States.
JAMA 2019; 321:844–845.
8. Xia Q, Torian LV, Shepard CW.
Limitations of indicators of HIV case finding.
Epidemiology 2015; 26:e6–e8.
9. Xia Q, Ning Z, Torian LV.
Should we report the proportion of late HIV diagnoses?.
AIDS 2017; 31:2559–2561.
10. Ndawinz JD, Costagliola D, Supervie V.
New method for estimating HIV incidence and time from infection to diagnosis using HIV surveillance data: results for France.
AIDS 2011; 25:1905–1913.
11. van Sighem A, Nakagawa F, De Angelis D, Quinten C, Bezemer D, de Coul EO, et al.
Estimating HIV incidence, time to diagnosis, and the undiagnosed HIV epidemic using routine surveillance data.
Epidemiology 2015; 26:653–660.
12. Becker S.
Mortality and its measurement. Baltimore, MD: Johns Hopkins Bloomberg School of Public Health; 2008.
13. Song R, Hall HI, Green TA, Szwarcwald CL, Pantazis N.
Using CD4 data to estimate HIV incidence, prevalence, and percentage of undiagnosed infections in the United States.
J Acquir Immune Defic Syndr 2017; 74:3–9.
14. Szwarcwald CL, Pascom ARP, de Souza PR.
Estimation of the HIV incidence and of the number of people living with HIV/AIDS in Brazil, 2012.
J AIDS Clin Res 2015; 6:430.
15. Lodi S, Phillips A, Touloumi G, Geskus R, Meyer L, Thiébaut R, et al. CASCADE Collaboration in EuroCoord
Time from human immunodeficiency virus seroconversion to reaching CD4+ cell count thresholds <200, <350, and <500 cells/mm3: assessment of need following changes in treatment guidelines.
Clin Infect Dis 2011; 53:817–825.
16. Touloumi G, Pantazis N, Pillay D, Paraskevis D, Chaix ML, Bucher HC< ET-AL>. CASCADE collaboration in EuroCoord
Impact of HIV-1 subtype on CD4 count at HIV seroconversion, rate of decline, and viral load set point in European seroconverter cohorts.
Clin Infect Dis 2013; 56:888–897.
17. Song R, Green TA.
An improved approach to accounting for reporting delay in case surveillance systems.
JP J Biostat 2012; 7:1–14.
18. Oehlert G.
A note on the delta method.
Am Stat 1992; 46:27–29.
19. Rubin DB.
Multiple imputation for nonresponse in surveys. New York: John Wiley & Sons, Inc; 1987.
20. Pilcher CD, Tien HC, Eron JJ Jr, Vernazza PL, Leu S-Y, Stewart PW, et al.
Brief but efficient: acute HIV infection and the sexual transmission of HIV.
J Infect Dis 2004; 189:1785–1792.
21. Fauci AS, Pantaleo G, Stanley S, Weissman D.
Immunopathogenic mechanisms of HIV infection.
Ann Intern Med 1996; 124:654–663.
22. Quinn TC, Wawer MJ, Sewankambo N, Serwadda D, Li C, Wabwire-Mangen F, et al.
Viral load and heterosexual transmission of human immunodeficiency virus type 1. Rakai Project Study Group.
N Engl J Med 2000; 342:921–929.
23. Attia S, Egger M, Muller M, Zwahlen M, Low N.
Sexual transmission of HIV according to viral load and antiretroviral therapy: systematic review and meta-analysis.
AIDS 2009; 23:1397–1404.
