Evaluation of Algorithms Used for PrEP Surveillance Using a Reference Population From New York City,July 2016

Abstract

Objective:

Daily tenofovir disoproxil fumarate/emtricitabine (TDF/FTC) use as HIV preexposure prophylaxis (PrEP) is monitored by identifying TDF/FTC prescriptions from pharmacy databases and applying diagnosis codes and antiretroviral data to algorithms that exclude TDF/FTC prescribed for HIV postexposure prophylaxis (PEP), HIV treatment, and hepatitis B virus (HBV) treatment. We evaluated the accuracy of 3 algorithms used by the Centers for Disease Control and Prevention (CDC), Gilead Sciences, and the New York State Department of Health (NYSDOH) using a reference population in Bronx, New York.

Methods:

We extracted diagnosis codes and data on all antiretroviral prescriptions other than TDF/FTC from an electronic health record database for persons aged ≥16 prescribed TDF/FTC during July 2016–June 2018 at Montefiore Medical Center. We reviewed medical records to classify the true indication of first TDF/FTC use as PrEP, PEP, HIV treatment, or HBV treatment. We applied each algorithm to the reference population and compared the results with the medical record review.

Results:

Of 2862 patients included in the analysis, 694 used PrEP, 748 used PEP, 1407 received HIV treatment, and 13 received HBV treatment. The algorithms had high specificity (range: 98.4%-99.0%), but the sensitivity of the CDC algorithm using a PEP definition of TDF/FTC prescriptions ≤30 days was lower (80.3%) than the sensitivity of the algorithms developed by Gilead Sciences (94.7%) or NYSDOH (96.1%). Defining PEP as TDF/FTC prescriptions ≤28 days improved CDC algorithm performance (sensitivity, 95.8%; specificity, 98.8%).

Conclusions:

Adopting the definition of PEP as ≤28 days of TDF/FTC in the CDC algorithm should improve the accuracy of national PrEP surveillance.

Keywords

algorithm HIV New York preexposure prophylaxis surveillance validation

Preexposure prophylaxis (PrEP) with daily oral tenofovir disoproxil fumarate combined with emtricitabine (TDF/FTC) prevents HIV infection in persons with sexual or injection drug use behaviors that place them at risk of becoming infected with HIV.^1

-5 The use of TDF/FTC as PrEP increased after its approval by the US Food and Drug Administration in 2012 and recommendation by the Centers for Disease Control and Prevention (CDC) clinical practice guidelines in 2014.¹ Although CDC estimates that 1.1 million persons who are at risk for HIV acquisition could benefit from PrEP, fewer than 10% of these persons are estimated to be using PrEP, according to data up to 2017.^6
-8 Furthermore, PrEP use among African American/black persons, Hispanic/Latinx persons, young adults, and persons living in the South is low even though these persons are at the highest risk of HIV.^6,8

-14 Surveillance of PrEP is needed to track overall PrEP use and identify disparities in use among racial/ethnic groups, sexes, and regions.

Program Description

PrEP surveillance is conducted by applying algorithms designed to identify PrEP by excluding TDF/FTC use for HIV treatment, HIV postexposure prophylaxis (PEP), and hepatitis B virus (HBV) treatment to large pharmacy databases.^6,8,15,16 The 2 pharmacy databases used in surveillance are from IQVIA and Source Healthcare Analytics. These pharmacy databases collect prescription data from commercial and retail pharmacies and use proprietary methods to match these data with health care databases for International Classification of Diseases, Clinical Modification (ICD-CM) codes and financial databases such as Experian for data on demographic characteristics. These pharmacy databases capture most (>90%) commercial and retail pharmacy prescriptions in the United States but do not capture prescriptions from closed health care systems (eg, Kaiser Permanente) or government systems (eg, Veterans Health Administration). Prescription-level data are linked with clinical data by using a unique patient identifier to generate person-level data that can be applied to PrEP algorithms.

CDC, Gilead Sciences, and the New York State Department of Health (NYSDOH) have each developed algorithms that use a combination of ICD-CM codes, National Drug Codes (NDCs), and prescription duration to identify TDF/FTC as PrEP.^6,8,15,17 CDC and Gilead Sciences use these numbers to generate national and state PrEP estimates, and NYSDOH estimates the number of Medicaid recipients on PrEP.

Algorithm Descriptions

The CDC, Gilead Sciences, and NYSDOH surveillance algorithms differ in the data source used; look-back periods for clinical data; sequence of excluding cases classified as representing HIV treatment, HBV treatment, and PEP; and ICD-CM codes and NDCs used for exclusion (Table 1). The CDC algorithm used the IQVIA pharmacy database, and the Gilead Sciences and NYSDOH algorithms used Source Healthcare Analytics as the data source. The CDC algorithm included all persons prescribed TDF/FTC, and the Gilead Sciences and NYSDOH algorithms excluded persons without linkages to ICD-CM codes associated with encounters. The Gilead Sciences and NYSDOH algorithms used clinical data from up to 2 years before the first TDF/FTC prescription in a given year, and the CDC algorithm included clinical data from 2 years before until 30 days after the first TDF/FTC prescription in a given year. The CDC and Gilead Sciences algorithms excluded HIV treatment first, then HBV treatment, and finally PEP, whereas the NYSDOH algorithm excluded PEP first, then HIV treatment, and finally HBV treatment.

Table 1.

Comparison of the preexposure prophylaxis surveillance algorithms used by the Centers for Disease Control and Prevention, Gilead Sciences, and New York State Department of Health, 2018

Characteristic	Centers for Disease Control and Prevention	Gilead Sciences	New York State Department of Health
Data source	IQVIA	Source Healthcare Analytics	Source Healthcare Analytics
Ages included	≥16	All ages	All ages
Linkages to clinical data	Includes all persons prescribed TDF/FTC with or without linkages to clinical data	Excludes persons without linkages to clinical data	Excludes persons without linkages to clinical data
Look-back period for clinical data	1 year before and up to 30 days after the first TDF/FTC prescription	2 years before the first TDF/FTC prescription	2 years before the first TDF/FTC prescription
Exclusion sequence	HIV, HBV, PEP	HIV, HBV, PEP	PEP, HIV, HBV
HIV treatment exclusion	• HIV antiretroviral drugs • HIV ICD-CM codes	• HIV antiretroviral drugs • HIV ICD-CM codes • Opportunistic infection ICD-CM codes	• HIV antiretroviral drugs • HIV ICD-CM codes
HBV treatment exclusion	• HBV antivirals • HBV ICD-CM codes	• HBV antivirals • HBV ICD-CM codes	• HBV antivirals • HBV ICD-CM codes
PEP exclusion	TDF/FTC prescription ≤30 or ≤28 days^a	Needle stick or prophylaxis ICD-9-CM and ICD-10-CM codes^b	TDF/FTC plus dolutegravir or raltegravir prescribed concurrently for ≤30 days

Abbreviations: HBV, hepatitis B virus; ICD-CM, International Classification of Diseases, Clinical Modification; ICD-9-CM, International Classification of Diseases, Ninth Revision, Clinical Modification; ICD-10-CM, International Classification of Diseases, Tenth Revision, Clinical Modification; PEP, postexposure prophylaxis; TDF/FTC, tenofovir disoproxil fumarate/emtricitabine.

^a The Centers for Disease Control and Prevention reported a preexposure prophylaxis estimate using both PEP definitions.

^b ICD-9-CM codes were used for analyses before October 1, 2015, and ICD-10-CM codes were used for analyses from October 1, 2015, and later.

All 3 algorithms used HIV-related ICD-CM codes and other antiretroviral NDCs to exclude for HIV treatment (Table 2). The CDC algorithm also excluded for HIV treatment by using ICD-CM codes for HIV-2 disease and HIV during pregnancy. The Gilead Sciences algorithm also excluded for HIV treatment by using ICD-CM codes for HIV-2 disease and opportunistic infections. All 3 algorithms excluded for chronic HBV infection by using a similar set of ICD-CM codes and HBV antiviral NDCs. The CDC and NYSDOH algorithms also excluded for acute HBV infection. Each algorithm excluded PEP differently. The CDC algorithm defined PEP as any TDF/FTC prescription covering ≤30 days. CDC guidelines recommend PEP for a 28-day course and, as such, CDC also reports a PrEP estimate by using a more specific PEP definition of TDF/FTC prescription covering ≤28 days.¹⁸ Gilead Sciences used ICD-CM codes associated with a needle stick or unspecified prophylaxis to exclude PEP. NYSDOH defines PEP as a prescription for TDF/FTC prescribed concurrently with dolutegravir or raltegravir covering ≤30 days.

Table 2.

ICD-9-CM and ICD-10-CM codes used in preexposure prophylaxis surveillance algorithms by the Centers for Disease Control and Prevention, Gilead Sciences, and the New York State Department of Health, 2018

ICD Revision	Centers for Disease Control and Prevention	Gilead Sciences	New York State Department of Health
HIV treatment exclusion
ICD-10-CM	B20: HIV disease B97.35: HIV-2 disease O98.71X: HIV complicating pregnancy O98.72: HIV complicating childbirth O98.73: HIV complicating puerperium R75: Inconclusive laboratory evidence of HIV Z21: Asymptomatic HIV infection	B20: HIV disease B97.35: HIV-2 disease Z21: Asymptomatic HIV infection	B20: HIV disease Z21: Asymptomatic HIV infection
ICD-9-CM	079.53: HIV-2 disease 042: HIV disease 795.71: Serologic evidence of HIV V08: Asymptomatic HIV infection	079.53: HIV-2 disease 042: HIV disease 795.71: Serologic evidence of HIV V08: Asymptomatic HIV infection	042: HIV disease V08: Asymptomatic HIV infection
HIV opportunistic infections exclusion
ICD-10-CM	Not applicable	A07.2: Cryptosporidiosis A31.0: Pulmonary mycobacterial infection A31.2: Disseminated mycobacterium avium B25.X: Cytomegaloviral disease B37.1: Pulmonary candidiasis B37.81: Candidal esophagitis B38.X: Coccidioidomycosis B45.X: Cryptococcosis B58.2: Toxoplasma meningoencephalitis B59: Pneumocystosis C46.X: Kaposi’s sarcoma	Not applicable
ICD-9-CM	Not applicable	007.4: Cryptosporidiosis 031.0: Pulmonary mycobacterial infection 031.2: Disseminated mycobacterium 078.5: Cytomegaloviral disease 112.4: Pulmonary candidiasis 112.84: Candidal esophagitis 114.X: Coccidioidomycosis 117.5: Cryptococcosis 130.X: Toxoplasmosis 136.3: Pneumocystosis 176.X: Kaposi’s sarcoma	Not applicable
HBV treatment exclusion
ICD-10-CM	B16.X: Acute hepatitis B B17.0: Acute hepatitis delta infection B18.0: Chronic hepatitis B with delta agent B18.1: Chronic hepatitis B no delta agent B19.10: Hepatitis B without hepatic coma B19.11: Hepatitis B with hepatic coma Z22.51: Hepatitis B carrier	B18.0: Chronic hepatitis B with delta agent B18.1: Chronic hepatitis B no delta agent B19.10: Hepatitis B without hepatic coma B19.11: Hepatitis B with hepatic coma	B16.X: Acute hepatitis B B18.0: Chronic hepatitis B with delta agent B18.1: Chronic hepatitis B no delta agent B19.10: Hepatitis B without hepatic coma B19.11: Hepatitis B with hepatic coma Z22.51: Hepatitis B carrier
ICD-9-CM	070.2X: Hepatitis B with hepatic coma 070.3X: Hepatitis B without hepatic coma V02.61: Hepatitis B carrier	070.2X: Hepatitis B with hepatic coma 070.3X: Hepatitis B without hepatic coma V02.61: Hepatitis B carrier	070.2X: Hepatitis B with hepatic coma 070.3X: Hepatitis B without hepatic coma V02.61: Hepatitis B carrier
PEP exclusion
ICD-10-CM	Not applicable	W46.0XXA: Hypodermic needle contact W46.1XXA: Contaminated needle contact Z29.9: Unspecified prophylaxis	Not applicable
ICD-9-CM	Not applicable	V07.8: Other prophylaxis or treatment V07.9: Unspecified prophylaxis E920.5: Hypodermic needle accident	Not applicable

Abbreviations: HBV, hepatitis B virus; ICD-9-CM, International Classification of Diseases, Ninth Revision, Clinical Modification; ICD-10-CM, International Classification of Diseases, Tenth Revision, Clinical Modification; PEP, postexposure prophylaxis.

Purpose of the Evaluation

The differences in each algorithm resulted in varying estimates of PrEP prescriptions. For example, CDC estimates that in the 50 states and the District of Columbia, 13 748 persons were prescribed PrEP in 2014, 38 879 in 2015, and 78 360 in 2016.⁶ By comparison, the Gilead Sciences algorithm estimated that 27 596 persons were prescribed PrEP in 2014, 59 427 in 2015, and 77 120 in 2016.⁸ Validation studies of these algorithms individually or compared with each other have not been published, leading to uncertainty about the accuracy of the surveillance systems. Furthermore, evaluation allows for the identification of high-performing components of each algorithm that could guide future algorithm refinement. The goal of this study was to evaluate the performance of these 3 algorithms by using electronic health records and medical record review of a reference population from the Montefiore Medical Center in Bronx, New York.

Methods

Setting

The Montefiore Medical Center is located in the Bronx borough of New York City and is the largest health care provider in the borough. The health system comprises 4 acute care hospitals and more than 40 primary care and specialty ambulatory clinics.¹⁹ PrEP is offered at several primary care clinics, an adolescent clinic, and a sexual health clinic that specializes in PrEP and PEP administration. The Einstein-Rockefeller-City University of New York Center for AIDS Research (ERC-CFAR) maintains a database that includes demographic, clinical, and prescription data from the Montefiore electronic health record system for all persons with HIV tests conducted at all of their sites. The ERC-CFAR database has been validated and used to classify HIV status, characterize antiretroviral use, and identify persons prescribed PrEP in the Montefiore system, and it undergoes continuous quality control.^19
-21

Patient Sampling

We identified persons aged ≥16 who were prescribed TDF/FTC in the Montefiore Medical Center from July 1, 2016, through June 30, 2018, through the ERC-CFAR database and included them in the analysis. We extracted data on demographic characteristics and HIV testing from this database. We also extracted diagnostic codes used in the algorithms and data on other HIV or HBV antiviral prescriptions for the date of first TDF/FTC prescription and in the preceding 2 years. We extracted both International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes because the review period for clinical data captured data before and after October 2015, when use of ICD-9-CM codes switched to ICD-10-CM codes nationally.

Medical Record Review

For persons identified by the ERC-CFAR database as being prescribed TDF/FTC, we conducted a manual medical record review of clinical notes, problem lists, laboratory results, and medications in the electronic health record to determine the indication of the first TDF/FTC prescription in the sampling time frame. We classified prescriptions of TDF/FTC as prescribed for PrEP, PEP, HIV treatment, HBV treatment, or unknown. We considered the TDF/FTC indication as determined by medical record review to be the criterion standard, and the medical record reviewer was blinded to the algorithms’ classifications.

The data extracted from the ERC-CFAR database could not accurately identify the coverage period for individual prescriptions, which was important for the PEP definition in the CDC algorithm. Therefore, medical record review captured the number of TDF/FTC tablets continuously prescribed by counting the first prescription and subsequent renewals across multiple visits using the assumption of daily TDF/FTC use and allowing for a gap of ≤14 days between prescriptions, which is consistent with the CDC algorithm. Medical record review also captured information on whether persons who were prescribed PEP then later transitioned to PrEP within 3 months of their initial TDF/FTC prescription as a part of PEP.

Exclusion Criteria

We identified 3203 persons with a TDF/FTC prescription during the study period. We excluded from analysis 336 persons who received TDF/FTC only as an isolated inpatient encounter and did not have an outpatient prescription for TDF/FTC, because inpatient TDF/FTC prescriptions are not considered in reported PrEP estimates. We also excluded 5 persons from analysis because the indication for TDF/FTC could not be determined through medical record review. Persons who were not excluded comprised the reference population derived from the ERC-CFAR database.

Statistical Analysis

ERC-CFAR data for the reference population were subjected to the HIV, HBV, and PEP exclusion parameters of each algorithm using the ICD-CM and antiretroviral medication data from the CFAR extract and the number of days of TDF/FTC continuously prescribed from the medical record review by using SAS version 9.4.²² For the CDC algorithm, we generated an additional estimate by using the definition of PEP as ≤28 days of TDF/FTC. We compared the classification of TDF/FTC prescription by each algorithm with the criterion TDF/FTC classification determined from medical record review. We then used these data to calculate the sensitivity, specificity, positive predictive value, and negative predictive value of each algorithm. This study was approved by the Albert Einstein College of Medicine Institutional Review Board and a project determination of routine surveillance from the CDC National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention.

Results

Of the final reference population (N = 2862), 694 (24.2%) persons were prescribed PrEP, 748 (26.1%) were prescribed PEP, 1407 (49.2%) were receiving HIV treatment, and 13 (0.5%) were receiving HBV treatment according to the medical record review (Table 3). Eight-six of 748 (11.5%) persons prescribed PEP later transitioned to PrEP after their initial TDF/FTC prescription. Persons who were prescribed PrEP were, on average, younger than persons receiving HIV or HBV treatment. Persons who were prescribed PrEP were also disproportionately male (76.2%) compared with persons who were prescribed TDF/FTC for PEP (48.8%), HIV (56.6%), or HBV (38.5%). Most persons were either Hispanic/Latinx or non-Hispanic African American/black.

Table 3.

Montefiore Medical Center reference population demographic characteristics by tenofovir/emtricitabine indication, July 2016–June 2018

Characteristic	PrEP (n = 694)	PEP (n = 748)	HIV (n = 1407)	HBV (n = 13)	All (N = 2862)
Mean (SD) age, y	34.8 (11.7)	33.4 (10.6)	50.9 (11.7)	55.8 (14.6)	42.5 (14.2)
Gender, no. (%)
Men	529 (76.2)	365 (48.8)	797 (56.6)	5 (38.5)	1696 (59.3)
Women	141 (20.3)	378 (50.5)	590 (41.9)	8 (61.5)	1117 (39.0)
Transgender women	22 (3.2)	5 (0.7)	20 (1.4)	0	47 (1.6)
Transgender men	2 (0.3)	0	0	0	2 (0.1)
Race/ethnicity, no. (%)
Non-Hispanic white	63 (9.1)	93 (12.4)	57 (4.1)	0	213 (7.4)
Non-Hispanic African American or black	228 (32.9)	246 (32.9)	614 (43.6)	3 (23.1)	1091 (38.1)
Hispanic of any race	284 (40.9)	241 (32.2)	588 (41.8)	6 (46.2)	1119 (39.1)
Other or unknown	119 (17.1)	168 (22.5)	148 (10.5)	4 (30.8)	439 (15.3)

Abbreviations: HBV, hepatitis B virus; PEP, postexposure prophylaxis; PrEP, preexposure prophylaxis.

At the HIV exclusion step, the CDC algorithm accurately excluded 1406 of 1407 (99.9%) persons receiving HIV treatment (Figure). A total of 715 of 748 (95.6%) persons who were prescribed PEP were also excluded at the HIV exclusion step because of their concurrent use of either dolutegravir or raltegravir. Finally, 2 of 13 persons receiving HBV treatment were excluded at the HIV exclusion step because of use of antiretroviral drugs also used in HIV treatment. At the HBV exclusion step, the CDC algorithm accurately excluded 10 of 11 persons receiving HBV treatment who remained after the HIV exclusion step, although 12 of 678 (1.8%) persons who were prescribed PrEP and remained after the HIV exclusion step were also excluded because of having an HBV-associated ICD-CM code. At the PEP exclusion step, the CDC algorithm using a definition of ≤30 days of TDF/FTC for PEP excluded 13 of 32 (40.6%) persons prescribed PEP who remained after the HIV and HBV exclusion steps but also excluded 109 of 666 (16.4%) persons prescribed PrEP who remained after the HIV and HBV exclusion steps, because they received TDF/FTC for only ≤30 days. By comparison, the CDC algorithm using a definition of ≤28 days of TDF/FTC for PEP resulted in only 1 of 666 (0.2%) persons receiving PrEP being excluded.

Figure.

Centers for Disease Control and Prevention (CDC), Gilead Sciences, and New York State Department of Health preexposure prophylaxis (PrEP) algorithm results: prescription classifications of the Montefiore Medical Center reference population based on medical record review indication, New York, 2018. HBV, hepatitis B virus; PEP, postexposure prophylaxis; TDF/FTC, tenofovir disoproxil fumarate/emtricitabine.

As a whole, the CDC algorithm using a definition of ≤30 days of PEP classified 578 persons as having been prescribed PrEP, of whom 557 (96.4%) were true receivers of PrEP prescriptions by medical record review. The algorithm using a definition of ≤28 days of PEP classified 692 persons as having been prescribed PrEP, of whom 665 (96.1%) were true receivers of PrEP prescriptions.

The Gilead Sciences algorithm performed similarly to the CDC algorithm at the HIV exclusion step, excluding 1406 of 1407 (99.9%) persons receiving HIV treatment and 715 of 748 (95.6%) persons who were prescribed PEP. The Gilead Sciences algorithm also performed similarly to the CDC algorithm at the HBV exclusion step, excluding 9 of 11 persons receiving HBV treatment who remained after the HIV exclusion step. At the PEP exclusion step, the algorithm excluded none of the 32 persons prescribed PEP who remained after the HIV and HBV exclusion steps. It also excluded 10 of 667 (1.5%) persons prescribed PrEP who remained after the HIV and HBV exclusion steps and had an ICD-CM code indicating needle stick or unspecified prophylaxis. The algorithm classified 692 persons as having been prescribed PrEP, of whom 657 (94.9%) were true receivers of PrEP prescriptions by medical record review.

The NYSDOH algorithm, which started with PEP exclusion first, excluded 612 of 748 (81.8%) persons prescribed PEP determined by medical record review and did not exclude any persons prescribed PrEP. At the HIV exclusion step, the NYSDOH algorithm excluded 1382 of 1383 (99.9%) persons receiving HIV treatment who remained after the PEP exclusion step and 103 of 136 (75.7%) persons prescribed PEP who remained after the PEP exclusion step. At the HBV exclusion step, the NYSDOH algorithm excluded 10 of 11 persons receiving HBV treatment who remained after the PEP and HIV exclusion steps. The NYSDOH algorithm classified 701 persons as having been prescribed PrEP, of whom 667 (95.1%) were true receivers of PrEP prescriptions by medical record review.

All 3 algorithms had high specificity in classifying the number of persons prescribed PrEP (range, 98.4%-99.0%; Table 4). However, the CDC algorithm that defined PEP as ≤30 days of prescription had significantly lower sensitivity for classifying persons prescribed PrEP (80.3%) than the Gilead Sciences algorithm (94.7%), the NYSDOH algorithm (96.1%), and the CDC algorithm using a definition of ≤28 days of PEP (95.8%).

Table 4.

Preexposure prophylaxis (PrEP) algorithm performance for PrEP detection using Centers for Disease Control and Prevention, Gilead Sciences, and New York State Department of Health algorithms on the Montefiore Médical Center reference population, 2018^a

Measure	Centers for Disease Control and Prevention^b		Gilead Sciences	New York State Department of Health
Measure	≤30 d	≤28 d	Gilead Sciences	New York State Department of Health
Sensitivity	80.3 (77.1-83.1)	95.8 (94.1-97.1)	94.7 (92.7-96.1)	96.1 (94.4-97.3)
Specificity	99.0 (98.5-99.4)	98.8 (98.2-99.1)	98.4 (97.8-98.8)	98.4 (97.8-98.9)
Positive predictive value	96.4 (94.5-97.6)	96.1 (94.4-97.3)	94.9 (93.0-96.3)	95.2 (93.3-96.5)
Negative predictive value	94.0 (93.0-94.9)	98.7 (98.1-99.1)	98.3 (97.7-98.8)	97.9 (97.3-98.3)

^a All values are percentage (95% confidence interval).

^b Centers for Disease Control and Prevention reports a PrEP estimate for both the ≤30-day and ≤28-day definition for postexposure prophylaxis.

Lessons Learned

PrEP is an important component of the multipronged effort to prevent new HIV infections in the United States.^23,24 PrEP surveillance is crucial in tracking both the overall use of PrEP and disparities in PrEP use among populations at the highest risk of acquiring HIV infection.^9,12
-14 Estimates of PrEP use are important for HIV prevention planning, resource allocation, program evaluation, and communication to the public. As such, several federal agencies, academic institutions, state and local health departments, and HIV community-based organizations depend on the accuracy of these data.

The PrEP algorithms are closely related but differ most in the parameters used in the PEP exclusion step. Although the differences in PEP definition in the NYSDOH algorithm resulted in fewer persons prescribed PEP being misclassified as being on HIV treatment (29.0% for the NYSDOH algorithm vs 95.6% for the CDC and Gilead Sciences algorithms), these differences had a minimal effect on the number of true PrEP prescriptions classified as PrEP prescriptions by the algorithm (667 for NYSDOH vs 657 for Gilead Sciences). At the PEP exclusion step in the CDC algorithm using a definition of ≤30 days, 109 of 666 (16.4%) true PrEP prescriptions were misclassified as PEP prescriptions because they had TDF/FTC prescriptions covering ≤30 days. The most common scenario for the misclassification of persons receiving PrEP as persons receiving PEP by the CDC algorithm using the ≤30-day definition of PEP was persons who initiated PrEP by being prescribed an initial 30-day course but then did not return for follow-up care. The misclassification of true PrEP prescriptions as PEP prescriptions resulted in the ≤30-day CDC algorithm having lower sensitivity than the Gilead Sciences and NYSDOH algorithms.

Because CDC recommends PEP for a 28-day course, CDC has deliberated whether a ≤30-day or ≤28-day definition was more specific for defining persons prescribed PEP.⁶ Our data show that defining PEP as ≤28 days of TDF/FTC significantly improves the sensitivity of the CDC algorithm for correctly classifying PrEP to the point that the performance of the 3 algorithms is equivalent. This finding supports the use of a definition of ≤28 days of TDF/FTC for PEP in the CDC algorithm.

For 2016, CDC estimated that 78 360 persons were prescribed PrEP using a definition of ≤30 days of TDF/FTC for PEP,⁶ and this estimate increased to 98 599 persons prescribed PrEP when a definition of ≤28 days of TDF/FTC for PEP was used. However, the estimate of persons prescribed PrEP in 2016 for Gilead Sciences was 77 120 persons.⁸ The discrepancy in the estimated number of persons receiving PrEP may be related to the differences in the data sources, review periods for clinical data, or exclusion of persons without linkages to clinical data. Prescriptions may vary by data source. Longer review periods may identify more ICD-CM codes for exclusion than the 1-2 years used currently. Furthermore, the CDC algorithm did not exclude persons without linked clinical data containing ICD-CM codes, whereas the Gilead Sciences algorithm did. Using the definition of ≤28 days of TDF/FTC for PEP and including only persons with linkages to clinical data, the CDC algorithm identified 71 814 persons receiving PrEP in 2016, similar to the Gilead Sciences estimate of 77 120 for 2016.⁸ These data suggest that the discrepancy in estimates of the number of persons receiving PrEP is largely a result of the decision to include or exclude persons without linkages to clinical data.

Limitations

Our study had several limitations. First, the medical record review validation relied on prescriptions recorded in the electronic health record. Prescriptions in the electronic health record are an imperfect estimate of patient use of PrEP, because the prescription in the electronic health record may not be filled by the patient, and it would not be counted in pharmacy databases. Second, this study relied on a reference population from the Bronx that may not be representative of national PrEP, PEP, HIV treatment, and HBV treatment prescribing practices. Finally, these data represent only the classification of the first TDF/FTC prescription within the sampling period. Twelve percent of persons who received initial prescriptions for PEP later transitioned to PrEP. Most of these PEP-to-PrEP transitions would have been classified as PEP in the year of their transition in the current algorithms, resulting in an underestimation of persons prescribed PrEP. The exclusion of persons who transition from PEP to PrEP represents a limitation of the current algorithms and indicates the need for further research into the frequency of PEP-to-PrEP transitions and how the surveillance algorithms can account for this practice.

Currently, the PrEP algorithms are built on the assumption of daily TDF/FTC increases in the number of PEP-to-PrEP transitions,²⁵ use of alternative non-daily dosing,^26,27 and use of other medications currently in clinical trials, such as tenofovir alafenamide/emtricitabine,²⁸ and could significantly affect the validity of the current analyses. Therefore, an iterative approach toward the refinement and evaluation of PrEP surveillance needs to be continued to accurately assess PrEP use nationwide.

Footnotes

Authors’ Note

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

Acknowledgments

The authors thank Kathryn M. Anastos, MD; Mindy S. Ginsberg, BA; Karen W. Hoover, MD, MPH; Franklin N. Laufer, PhD; Robertino Mera-Giler, MD, PhD; and Allan M. Spielman, MBA, for their contributions to this research.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by the Einstein-Rockefeller-City University of New York Center for AIDS Research (ERC-CFAR, P30-AI-124 414), K23-MH-102 118 (V. V. Patel), K01-HL-137 557 (D. B. Hanna), K23-MH-106 386 (U. R. Felsen), and funds from the Division of HIV/AIDS Prevention at CDC.

ORCID iD

Nathan W. Furukawa, MD, MPH

References

US Public Health Service. Preexposure prophylaxis for the prevention of HIV infection in the United States—2017 update: clinical practice guidelines. https://www.cdc.gov/hiv/pdf/risk/prep/cdc-hiv-prep-guidelines-2017.pdf. Published 2018. Accessed December 19, 2019.

Grant

Lama

Anderson

, et al. Preexposure chemoprophylaxis for HIV prevention in men who have sex with men. N Engl J Med. 2010;363(27):2587–2599. doi:10.1056/NEJMoa1011205

Baeten

Donnell

Ndase

, et al. Antiretroviral prophylaxis for HIV prevention in heterosexual men and women. N Engl J Med. 2012;367(5):399–410. doi:10.1056/NEJMoa1108524

Thigpen

Kebaabetswe

Paxton

, et al. Antiretroviral preexposure prophylaxis for heterosexual HIV transmission in Botswana. N Engl J Med. 2012;367(5):423–434. doi:10.1056/NEJMoa1110711

Choopanya

Martin

Suntharasamai

, et al. Antiretroviral prophylaxis for HIV infection in injecting drug users in Bangkok, Thailand (the Bangkok Tenofovir Study): a randomised, double-blind, placebo-controlled phase 3 trial. Lancet. 2013;381(9883):2083–2090. doi:10.1016/S0140-6736(13)61127-7

Huang

Zhu

Smith

Harris

Hoover

. HIV preexposure prophylaxis, by race and ethnicity—United States, 2014-2016. MMWR Morb Mortal Wkly Rep. 2018;67(41):1147–1150. doi:10.15585/mmwr.mm6741a3

Smith

Van Handel

Grey

. Estimates of adults with indications for HIV pre-exposure prophylaxis by jurisdiction, transmission risk group, and race/ethnicity, United States, 2015. Ann Epidemiol. 2018;28(12):850–857. doi:10.1016/j.annepidem.2018.05.003

Sullivan

Giler

Mouhanna

, et al. Trends in the use of oral emtricitabine/tenofovir disoproxil fumarate for pre-exposure prophylaxis against HIV infection, United States, 2012-2017. Ann Epidemiol. 2018;28(12):833–840. doi:10.1016/j.annepidem.2018.06.009

Hammack

Meyer

Krueger

Lightfoot

Frost

. HIV testing and pre-exposure prophylaxis (PrEP) use, familiarity, and attitudes among gay and bisexual men in the United States: a national probability sample of three birth cohorts. PLoS One. 2018;13(9):e02028 06. doi:10.1371/journal.pone.0202806

10.

Demeke

Johnson

Moonesinghe

Dean

. Unequal declines in absolute and relative disparities in HIV diagnoses among black women, United States, 2008 to 2016. Am J Public Health. 2018;108(S4):S299–S303. doi:10.2105/AJPH.2018.304641

11.

Singh

Song

Johnson

McCray

Hall

. HIV incidence, prevalence, and undiagnosed infections in U.S. men who have sex with men. Ann Intern Med. 2018;168(10):685–694. doi:10.7326/M17-2082

12.

Hamilton

Goodreau

Jenness

, et al. Potential impact of HIV preexposure prophylaxis among black and white adolescent sexual minority males. Am J Public Health. 2018;108(S4):S284–S291. doi:10.2105/AJPH.2018.304471

13.

Garner

Wilson

Beste

Maier

Ohi

Van Epps

. Gaps in preexposure prophylaxis uptake for HIV prevention in the Veterans Health Administration. Am J Public Health. 2018;108(S4):S305–S310. doi:10.2105/AJPH.2018.304788

14.

Jenness

Maloney

Smith

, et al. Addressing gaps in HIV preexposure prophylaxis care to reduce racial disparities in HIV incidence in the United States. Am J Epidemiol. 2019;188(4):743–752. doi:10.1093/aje/kwy230

15.

Laufer

O’Connell

Feldman

, et al. Vital signs: increased Medicaid prescriptions for preexposure prophylaxis against HIV infection—New York, 2012-2015. MMWR Morb Mortal Wkly Rep. 2015;64(46):1296–1301. doi:10.15585/mmwr.mm6446a5

16.

Mendoza

Huang

Hayes

Smith

Hoover

. Uptake of HIV preexposure prophylaxis among commercially insured persons—United States, 2010-2014. Clin Infect Dis. 2017;64(2):144–149. doi:10.1093/cid/ciw701

17.

Siegler

Mouhanna

Giler

, et al. The prevalence of pre-exposure prophylaxis use and the pre-exposure prophylaxis-to-need ratio in the fourth quarter of 2017, United States. Ann Epidemiol. 2018;28(12):841–849. doi:10.1016/j.annepidem.2018.06.005

18.

Centers for Disease Control and Prevention. Updated guidelines for antiretroviral postexposure prophylaxis after sexual, injection drug use, or other nonoccupational exposure to HIV—United States. 2016. https://www.cdc.gov/hiv/pdf/programresources/cdc-hiv-npep-guidelines.pdf. Accessed December 19, 2019.

19.

Bien

Patel

Blackstock

Felsen

. Reaching key populations: PrEP uptake in an urban health care system in the Bronx, New York. AIDS Behav. 2017;21(5):1309–1314. doi:10.1007/s10461-016-1663-8

20.

Felsen

Bellin

Cunningham

Zingman

. Development of an electronic medical record-based algorithm to identify patients with unknown HIV status. AIDS Care. 2014;26(10):1318–1325. doi:10.1080/09540121.2014.911813

21.

Hanna

Felsen

Ginsberg

, et al. Increased antiretroviral therapy use and virologic suppression in the Bronx in the context of multiple HIV prevention strategies. AIDS Res Hum Retroviruses. 2016;32(10-11):955–963. doi:10.1089/AID.2015.0345

22.

SAS [computer software]. Version 9.4. Cary, NC: SAS Institute, Inc; 2014.

23.

Centers for Disease Control and Prevention. Division of HIV/AIDS Prevention strategic plan 2017-2020. https://www.cdc.gov/hiv/pdf/dhap/cdc-hiv-dhap-external-strategic-plan.pdf. Published 2016. Accessed December 30, 2018.

24.

Goldstein

Streed

Jr Cahill

. Being PrEPared—preexposure prophylaxis and HIV disparities. N Engl J Med. 2018;379(14):1293–1295. doi:10.1056/NEJMp1804306

25.

Jain

Krakower

Mayer

. The transition from postexposure prophylaxis to preexposure prophylaxis: an emerging opportunity for biobehavioral HIV prevention. Clin Infect Dis. 2015;60(suppl 3):S200–S204. doi:10.1093/cid/civ094

26.

Molina

Capitant

Spire

, et al. On-demand preexposure prophylaxis in men at high risk for HIV-1 infection. N Engl J Med. 2015;373(23):2237–2246. doi:10.1056/NEJMoa1506273

27.

Molina

Charreau

Spire

, et al. Efficacy, safety, and effect on sexual behaviour of on-demand pre-exposure prophylaxis for HIV in men who have sex with men: an observational cohort study. Lancet HIV. 2017;4(9):e402–e410. doi:10.1016/S2352-3018(17)30089-9

28.

Gilead Sciences. Safety and efficacy of emtricitabine and tenofovir alafenamide (F/TAF) fixed-dose combination once daily for pre-exposure prophylaxis in men and transgender women who have sex with men and are at risk of HIV-1 infection (DISCOVER). https://clinicaltrials.gov/ct2/show/NCT02842086?term=f%2Ftaf. Published 2016. Accessed December 30, 2018.

Evaluation of Algorithms Used for PrEP Surveillance Using a Reference Population From New York City,July 2016–June 2018

Abstract

Objective:

Methods:

Results:

Conclusions:

Keywords

Program Description

Algorithm Descriptions

Purpose of the Evaluation

Methods

Setting

Patient Sampling

Medical Record Review

Exclusion Criteria

Statistical Analysis

Results

Lessons Learned

Limitations

Footnotes

Authors’ Note

Acknowledgments

Declaration of Conflicting Interests

Funding

ORCID iD

References