Secondary Logo

Journal Logo

Epidemiology

Development of a Mathematical Model to Estimate the Cost-Effectiveness of HRSA's Ryan White HIV/AIDS Program

Goyal, Ravi PhDa; Hu, Cindy PhDa; Klein, Pamela W. PhDb; Hotchkiss, John MSa; Morris, Eric MSa; Mandsager, Paul MSPHb; Cohen, Stacy M. MPHb; Luca, Dara PhDa; Gao, Jessica BAa; Jones, Andrew MAc; Addison, West BAc; O'Brien-Strain, Margaret PhDc; Cheever, Laura W. MD, ScMb; Gilman, Boyd PhDa

Author Information
JAIDS Journal of Acquired Immune Deficiency Syndromes: February 1, 2021 - Volume 86 - Issue 2 - p 164-173
doi: 10.1097/QAI.0000000000002546

Abstract

INTRODUCTION

In the United States, the Health Resources and Services Administration's Ryan White HIV/AIDS Program (RWHAP) funds a combination of core medical and support services to help people with HIV live long and healthy lives by filling any gaps in care and treatment (see Figure 1, Supplemental Digital Content, https://links.lww.com/QAI/B560).1–3 Enacted in 1990 as the Ryan White Comprehensive AIDS Resources Emergency Act and codified most recently under Title XXVI of the Public Health Service Act (P.L. 115‒408 Title 42, US Code, December 31, 2018), the RWHAP is the only federally funded program focused on providing care and treatment for HIV and is the third largest source of federal funding for HIV care in the United States, after Medicaid and Medicare.4,5 The RWHAP supports more than half of more than 1 million people diagnosed with HIV nationwide.6,7 The RWHAP has achieved notable successes in linking people with HIV to and retaining them in care, prescribing antiretroviral therapy (ART), and achieving viral suppression.8,9 However, the long-term public health and cost impact of the RWHAP has not yet been quantified.

Mathematical models have been useful for assessing the cost-effectiveness of a range of HIV-related interventions such as preexposure prophylaxis,10,11 ART,12 routine HIV screening,13 treatment adherence counseling,14 and, more recently, HIV care coordination.15–17 These models traditionally assess the cost-effectiveness of a single intervention. However, the RWHAP funds a comprehensive and complex system of care consisting of a wide range of services, based on need as determined by local, city, and state jurisdictions. As a result, developing a model to assess the cost-effectiveness of the RWHAP faces 4 unique challenges: (1) determining which services to include in the model, (2) estimating the impact of those services on care retention and viral suppression, (3) quantifying the need for and receipt of such services, and (4) measuring the cost of those services.

In this article, we present a new and innovative mathematical model for estimating the cost-effectiveness of the RWHAP. Here, we focus on the basic structure of the model, including the definition of HIV care and treatment services and their impact on retention in care and viral suppression, the proportion of unmet need for these services, the transmission of HIV through sexual contact and injection drug use (IDU), and the transition of agents through the stages of care based on receipt of needed services. We also present the results of the model validation. In a companion article in the same issue of this journal, we define the counterfactual and present the results of the cost-effectiveness and sensitivity analyses.18

METHODS

Model Overview

We constructed an open-cohort, stochastic, agent-based model parameterized to represent the characteristics of the HIV epidemic in the United States in 2016. The model was written in the R programming language. The model simulates an agent's progression through 7 stages: (1) HIV negative, (2) HIV positive but status unknown, (3) diagnosed and status known but not receiving care, (4) in care and treatment but not virally suppressed, (5) virally suppressed, (6) left care and treatment, and (7) death (Figure 1). In addition, the model simulations interactions among agents through sexual contact and sharing of needles. Each month, an individual either remains in the current stage or transitions to another stage. The interactions a person without HIV has among people with HIV affect the rate the person will transition from stage 1 to stage 2, whereas the services a person with HIV receives affect the rate at which a person transitions among stages 3–7. The receipt of ancillary services, conditional on needing them, affects the rate of transitions between stages 4, 5, and 6 (ie, care retention and viral suppression), as illustrated by the purple arrows in Figure 1.

F1
FIGURE 1.:
Structure of model, including stages of care and possible transition pathways between stages. Purple arrows indicate transitions affected by receipt of needed ancillary services, including MCM, mental health and substance abuse services, and other non-OAHS medical care and support services. People can transition from any care stage (stages 2–6) to death (stage 7).

HIV Care and Treatment Services

For the purposes of the model, we divided the HIV care and treatment services funded under the RWHAP into 5 mutually exclusive categories: (1) outpatient ambulatory health services (OAHS), (2) ART, (3) medical case management (MCM), (4) mental health and/or substance abuse services (MH/SA), and (5) other outpatient medical care and support services. The first category, OAHS, is the RWHAP term for diagnostic or therapeutic services provided by a licensed health care provider in an outpatient medical setting. The last category includes all other RWHAP-funded non-OAHS medical and support services, such as home-based and community-based health services, non-MCM, health education and risk reduction counseling, medical transportation, legal assistance, and housing services, among others (HIV/AIDS Bureau 2018). Throughout this article, we refer to third, fourth, and fifth categories as ancillary services.

Input Modules

The model consists of 5 modules. The first module initializes the population being simulated. It reflects the HIV population in the United States in terms of its demographic and risk group characteristics, service needs and receipt, care stage, and health state. It also specifies the demographic and risk group characteristics of the HIV-negative population with a nonnegligible risk of acquiring the disease. The second module simulates HIV transmission through sexual contact or sharing of contaminated needles to people who are HIV negative. The third module modifies a person's health state over time. The fourth module governs monthly transitions through the 7 stages based in part on the services needed and received. The fifth module assigns each person a monthly cost and quality-of-life value, both of which vary by health state and care stage.

Initialization Module

The initial model population reflects the distribution of people with HIV in the United States with respect to a set of person-level attributes, including (1) RWHAP eligibility, (2) demographic characteristics, (3) risk group, (4) need for and receipt of services, (5) care stage, and (6) health state. Using the 2016 RWHAP Services Report (RSR) and 2015 AIDS Drug Assistance Program Data Report, we also estimated the proportion of people with HIV served by the RWHAP, as well as the percentage of those who were in care and treatment but not virally suppressed (stage 4) and those in care and treatment and virally suppressed (stage 5).19–21 We used 2016 HIV surveillance data from the Centers for Disease Control and Prevention (CDC) to assign people with HIV who seek care outside of the RWHAP for the different stages.14Figure 2 provides a high-level breakdown of people with HIV based on whether they are served by the RWHAP and by stage. The nodes that are parameterized in the model, denoted by P, represent the 8 groups characterized in the initialization module (The nonparameterized nodes are denoted by N.). Material Section 1 and Tables 1-13, Supplemental Digital Content, https://links.lww.com/QAI/B560 provide more details on the initialization process.

F2
FIGURE 2.:
Estimated number of people with HIV in the United States, by receipt of services through the RWHAP and by stage of care. Source: CDC. “Estimated HIV incidence and prevalence in the United States 2010–2015.” HIV Surveillance Supplemental Report, vol. 23, Number 1, March 2018. Available at https://www.cdc.gov/hiv/pdf/library/reports/surveillance/cdc-hiv-surveillance-supplemental-report-vol-23-1.pdf. Accessed July 10, 2018. Centers for disease control and prevention. “Monitoring selected national HIV prevention and care objectives by using HIV surveillance data—United States and 6 dependent areas, 2015.” HIV Surveillance Supplemental Report, vol. 22, Number 2, July 2017. Available at https://www.cdc.gov/hiv/pdf/library/reports/surveillance/cdc-hiv-surveillance-supplemental-report-vol-22-2.pdf. Accessed July 10, 2018, Health Resources and Services Administration. Ryan White HIV/AIDS Program Annual Client-Level Data Report, Ryan White HIV/AIDS program services report (RSR). 2016; https://hab.hrsa.gov/sites/default/files/hab/data/datareports/RWHAP-annual-client-level-data-report-2016.pdf. Figure includes 6 boxes or nodes that are not included in the model structure (boxes N1–N6, indicated by black text font) and 8 nodes (boxes P1–P8, indicated by blue text font) that are parameterized in the model. The model nodes (P) represent the 8 care stages included in the model.

The RSR and CDC surveillance data were used to estimate the demographic and risk group characteristics of individuals served by the RWHAP and those with HIV who receive care outside the RWHAP within each care stage. The risk group indicates the mechanism by which HIV can be transmitted to uninfected individuals. Everyone is assigned to 1 of the 6 viral load copies per milliliter categories (0–200, 201–400, 400–3,499, 3500–9,999, 10,000–49,999, and 50,000 or higher) and a CD4 count. Both quantities are determined by the duration since infection and care stage. In the model, an individual's CD4 count determines the quality of life and health care costs, whereas the viral load category determines the probability of HIV transmission. People with a viral load of 200 copies per milliliter or lower are considered virally suppressed.

The need for and receipt of services are based on an individual's demographic (age, gender, and race/ethnicity) and risk group [men who have sex with men (MSM) only, IDU only, MSM and IDU, and heterosexual contact] characteristics and whether or not they are served by the RWHAP. We assume all people with HIV need and receive OAHS and ART, as per national HIV treatment guidelines.22 However, the need for ancillary services varies by demographic and risk group. Based on a study of unmet need among RWHAP-funded and non-RWHAP–funded facilities, individuals served by the RWHAP are assumed to have a higher probability of receiving needed ancillary services than those who obtain care outside the RWHAP.8 As discussed below, not receiving a needed ancillary service modifies the probability of remaining in care and becoming virally suppressed.23

Finally, the model includes a representation of the population in the United States at nonnegligible risk of acquiring HIV. We defined this population to include MSM and IDU, as well as youth and adults who are neither MSM nor IDU but have an elevated risk of acquiring HIV through heterosexual contact. This population represents approximately 2.4%of the total US population older than 13 years and is approximately 6.1 times the number of people with HIV in the United States.19,24,25 The demographic and risk group distributions for the at-risk population were based on the characteristics of the newly infected population in 2015 as reported in HIV incidence reports from the CDC.19–20

HIV Transmission Module

As mentioned above, there are 2 types of interactions among the simulated population that enable HIV transmission: (1) sexual contact and (2) IDU. Both of these interactions are represented as networks in the model. The model includes 2 sexual contact networks: one for MSM and another for non-MSM transmissions. Both sexual networks are dynamic, that is, relationships form and dissolve over time. The rate of sexual transmission depends upon the number of sexual partners per month, number of discordant partners, sexual partner formation and dissolution rates, type of sexual contact, and transmission probability between partners. Because the transmission probabilities were only known for non-MSM contacts, we had to calibrate the probabilities for MSM contacts by multiplying the non-MSM probabilities by a fixed factor to match the HIV incidence rates produced by the CDC; the multiplier was identified based on a systematic search of plausible values. Key parameters for the sexual networks are presented in Table 129–32 (See Material, Section S.3, Table 14, Supplemental Digital Content, https://links.lww.com/QAI/B560).

TABLE 1. - Summary of Key Input Parameters Used in Model for Assessing the Cost-Effectiveness of the RWHAP
Variable Equation/Value Reference
Initialization module
 Percentage of people with HIV in United States Varies by demographic/risk group, 0.04–16.1 Table 1, Supplemental Digital Content, http://links.lww.com/QAI/B560
 Percentage of people with HIV who are undiagnosed Varies by demographic/risk group, 0–17.6 Table 2, Supplemental Digital Content, http://links.lww.com/QAI/B560
 Percentage of RWHAP clients who are in care and treatment and virally suppressed Varies by demographic/risk group, 0–16.0 Table 3, Supplemental Digital Content, http://links.lww.com/QAI/B560
 Percentage of RWHAP clients who are in care and treatment but not virally suppressed Varies by demographic/risk group, 0–21.3 Table 4, Supplemental Digital Content, http://links.lww.com/QAI/B560
 Percentage of RWHAP clients who left care Varies by demographic/risk group, 0–15.0 Table 5, Supplemental Digital Content, http://links.lww.com/QAI/B560
 Percentage of people diagnosed with HIV who do not receive RWHAP services, are in care and treatment, and are virally suppressed Varies by demographic/risk group, 0–14.8 Table 6, Supplemental Digital Content, http://links.lww.com/QAI/B560
 Percentage of people diagnosed with HIV who do not receive RWHAP services and are in care and treatment but are not virally suppressed Varies by demographic/risk group, 0–12.8 Table 7, Supplemental Digital Content, http://links.lww.com/QAI/B560
 Percentage of people diagnosed with HIV who do not receive RWHAP services and left care and those who were never in care Varies by demographic/risk group, 0–14.2 Table 8, Supplemental Digital Content, http://links.lww.com/QAI/B560
 Percentage of population at an elevated risk for becoming HIV infected Varies by demographic/risk group, 0–19.5 Table 13, Supplemental Digital Content, http://links.lww.com/QAI/B560
 Percentage of people with HIV who need a given ancillary service Varies by demographic/risk group, 11.9–100 Table 10, Supplemental Digital Content, http://links.lww.com/QAI/B560
 Percentage of people with HIV who receive a given ancillary service Varies by demographic/risk group, 76.9–93.5 Table 11, Supplemental Digital Content, http://links.lww.com/QAI/B560
Transition module
 HIV Negative (stage 1) to HIV positive (stage 2) Intrinsic to the model HIV transmission module
 Monthly probability of transitioning from HIV positive but undiagnosed (stage 2) to diagnosed (stage 3) Varies by gender and risk group, 0.012–0.028 Table 15, Supplemental Digital Content, http://links.lww.com/QAI/B560
 Monthly probability of transitioning from diagnosed (stage 3) to care and treatment (stage 4) 32
 Monthly probability of transitioning from care and treatment (stage 4) to viral suppression (stage 5) (1 − P)6 = 1 − 0.835, P = 0.259 33
 Monthly probability of transitioning from care and treatment (stage 4) to left care (stage 6) 0.075/12 = 0.00625 34
 Monthly probability of transitioning from viral suppression (stage 5) to left care (stage 6) 0.075/12 = 0.00625 34
 Monthly probability of transitioning from virally suppression (stage 5) to care and treatment (stage 4) 0.078/12 = 0.0065 35
 Monthly probability of transitioning from left care (stage 6) to care and treatment (stage 4) 0.169/12 = 0.0141 34
 Stages 1–6 to death (stage 7), death rate per 100 person years Varies by age and CD4 count, 0.22–91.2 Table 16, Supplemental Digital Content, http://links.lww.com/QAI/B560
 Increase in probability of transitioning from care and treatment (stage 4) or viral suppression (stage 5) to left care (stage 6) because of unmet need Varies by service category, 67.2%–518.1% Table 19, Supplemental Digital Content, http://links.lww.com/QAI/B560
 Decrease in probability of transitioning from care and treatment (stage 4) to viral suppression (stage 5) because of unmet need Varies by service category, 47.8%–99.7% Table 20, Supplemental Digital Content, http://links.lww.com/QAI/B560
Health state module
 CD4 cell count at the time of infection, cells/μL Mean: 830 12
SD: 29
 Rate of CD4 decline if untreated, cells/μL 4.183 per month 29
 Distribution of viral load within each disease phase Varies by viral load category and by disease phase Table 9, Supplemental Digital Content, http://links.lww.com/QAI/B560
Outcome module
 Monthly cost of OAHS $119.6* 37,39
 Monthly cost of prescription medications $1374.2 37,39
 Monthly cost of MCM $233.8 15,38
 Monthly cost of mental health and substance use services $384.7 52
 Monthly cost of other non-OAHS medical and support services $98.2 15,38
Monthly cost of hospital inpatient services, by CD4 stratum 37
 <50 cells/μL $2,112
 51–200 cells/μL $872
 201–350 cells/μL $380
 351–500 cells/μL $238
 501+ cells/μL $186
Monthly cost of hospital emergency department services, by CD4 stratum 37
 <50 cells/μL $115
 51–200 cells/μL $52
 201–350 cells/μL $30
 351–500 cells/μL $22
 501+ cells/μL $16
Quality-of-life weights by CD4 stratum
 <100 cells/μL 0.603 40
 100–199 cells/μL 0.603 40
 200–349 cells/μL 0.719 40
 350–499 cells/μL 0.836 40
 500+ cells/μL 0.836 40
 Discounting rate for cost and QALMs 3% 41,42
Transmission module
 Discordant multiplier for the IDU network 0.04 Model calibration
 No. of individuals with whom IDU share needles per month 5.9 53
 Transmission probability of contaminated syringe 0.0063 24
 No. of injections per month 3 23
 Percentage of contaminated syringes 84% 23
 Mean number of sexual partners per month 0.8 8
 Mean length of sexual partnership Youth: 29 mo 8
Adults: 128 mo
 Discordant multiplier for the MSM network 0.6 Model calibration
HIV monthly transmission probabilities through MSM by viral load category 25–28
 More than 50,000 copies/mL 0.0151
 10,000–50,000 copies/mL 0.0135
 3500–9999 copies mL 0.0070
 400–3499 copies/mL 0.0034
 200–399 copies/mL 0.0003
 Fewer than 200 copies/mL 0
 Discordant multiplier for the heterosexual network 0.5 Model calibration
HIV monthly transmission probabilities through heterosexual contact by viral load category 25–28
 More than 50,000 copies/mL 0.0075
 10,000–50,000 copies/mL 0.0068
 3500–9999 copies mL 0.0035
 400–3499 copies/mL 0.0017
 200–399 copies/mL 0.0001
 Fewer than 200 copies/mL 0
*All costs are shown in 2016 US dollars, after being inflated by the medical care component of the consumer price index.
Inpatient costs are from 7 HIVRN sites with relevant data. Emergency department costs are from 3 HIVRN sites with relevant data. Costs are shown in 2016 US dollars, after being inflated by the medical care component of the consumer price index.

For IDU transmissions, a needle-sharing network is modeled, where HIV can be transmitted through IDU by sharing needles with someone with HIV. However, individuals who participate in needle sharing can also transmit through sexual contact at rates based on the MSM risk and viral load categories. The rate of IDU transmission is based on the probability of having a discordant needle-sharing partner, number of partners, number of injections, probability of syringe contamination, and transmission probability of contaminated syringe. Key parameters of the needle-sharing network are reported in Table 1.26–28 Because of limited data on the evolution of needle-sharing networks over time, we assume that a person's IDU partners are the same for the full simulation. In addition, the behavior of a person who uses injection drugs, such as abstaining from use, does not change over the simulation.

Health State Module

An individual's health state is defined by CD4 count and viral load category. CD4 affects medical costs, quality of life, and probability of death, whereas viral load affects the probability of the virus being transmitted to an HIV-negative sexual partner. During each month of the simulation, these health state metrics change based on the person's care stage. If the person is HIV positive and not virally suppressed, their CD4 cell count declines at a constant rate per month (Table 1).33 If the individual achieves viral suppression, their CD4 count rebounds to their preinfection value.

An individual is in the acute phase for the first 2-month postinfection before transitioning to the chronic phase. All people with HIV were assumed to have more than 50,000 copies per milliliter during the acute phase. During the chronic phase, each person is assigned to 1 of the 6 viral load categories (or set points). The individual's category does not change except if they reside in the viral suppression stage, in which case they are assumed to have fewer than 200 copies per milliliter. However, if individuals leave viral suppression, their viral load returns to their assigned category. We estimated the distribution of viral load categories assigned during the chronic phase based on data from the NA-ACCORD (see Table 9, Supplemental Digital Content, https://links.lww.com/QAI/B560).34

Transitions Module

Individuals stochastically transition between stages in monthly increments. The monthly probability of an undiagnosed person with HIV transitioning from stage 2 to stage 3 (ie, receiving an HIV diagnosis) is conditional on gender and risk factor. This probability is estimated from CDC surveillance data (see Table 15, Supplemental Digital Content, https://links.lww.com/QAI/B560).35 The monthly probability of transitioning from diagnosis (stage 3) to care and treatment (stage 4) is based on estimates by Medland et al.36 This probability is exogenously determined and identical for all individuals. Transitioning to stage 4 indicates receipt of both OAHS and ART; without these 2 services, an individual has a zero probability of transitioning to care and treatment.

Once an individual enters care and treatment (stage 4), they can transition to viral suppression (stage 5), leave care (stage 6), or remain in care and treatment but not virally suppressed (stage 4). These transition probabilities are conditional on the ancillary services the individual needs and receives (see purple arrows in Figure 1). For individuals with all of their needs addressed, their month transition probability to stage 5 is based on an 83.5% probability of achieving viral suppression within 6 months of initiating ART,37 whereas their transition probability to leave care (stage 6) is based on the loss to follow-up rate reported by Schackman et al.38 The model assumes that the rate of loss to follow-up is the same for people who are in care and treatment but not virally suppressed (stage 4) and those who are virally suppressed (stage 5). For individuals without all their needs addressed, these probabilities are modulated based on the research by Messeri et al23, as discussed below.

From stage 5 (viral suppression), an individual can transition back to care and treatment (stage 4), leave care (stage 6), or remain at stage 5. As with transitions for stage 4, we start with transition probabilities conditional on individuals having all their needs addressed. The probability of transitioning from stage 5 to stage 4 is based on a viral rebound rate of 7.8 per 100 person-years,39 whereas the transition from stage 5 to stage 6 is the same as the loss to follow-up rate for the transition from stage 4 to stage 6. As with transitions from stage 4, the probabilities from stage 5 for individuals with unmet needs are modulated.

At each stage, an individual can transition to death (stage 7) based on age- and CD4-adjusted mortality rates from the Concerted Action on Seroconversion to AIDS and Death in Europe (CASCADE) study (see Material, Table 16, Supplemental Digital Content, https://links.lww.com/QAI/B560).40

To modulate transition probabilities for individuals with unmet needs, we obtained estimates of the impact of each ancillary service on retention from Messeri et al23 (see Material, Section S.4, Table 19, Supplemental Digital Content, https://links.lww.com/QAI/B560). However, there is limited research on the marginal impact of each ancillary service category on achieving and maintaining viral suppression. To estimate the value of these marginal impacts, we assumed that each ancillary service has the same relative impact on adherence because it has on retention. Given this assumption, we selected—based on a systematic search of plausible values—the impacts of the services such that the model projected a viral suppression rate among people with HIV in care and treatment of 89% (Material, Section S.4, Table 20, Supplemental Digital Content, https://links.lww.com/QAI/B560). The impact of the ancillary service categories was selected after calibration of the HIV transmission probability for MSM contacts—the only other parameter calibrated; this approach produced results within our validation requirements.

Outcome Module

The total cost of HIV-related medical and support services and the number of quality-adjusted life months (QALMs) for each person with HIV is calculated in monthly increments. The cost inputs are from the perspective of the US health care system; they reflect the cost of delivering care and treatment to everyone with HIV, rather than the amount paid by an individual or payer. Total costs vary based on the individual's care stage and health state. Monthly costs were based on average annual cost estimates in the literature for 5 of the 7 service categories included in the cost calculations (OAHS, ART, MH/SA, hospital inpatient, and emergency department).41 These annual costs were updated to 2016 dollars using the medical care component of the consumer price index and then converted to monthly inputs for the model (Table 1). The monthly cost of the remaining 2 service categories (MCM and other non-OAHS medical and support services) was estimated from 2016 RSR and 2016 RWHAP allocations reports.20,42 The QALMs, which vary by CD4 count, are based on findings reported by Tengs and Lin (Table 1).43,44 Based on the recommendation of the Panel on Cost Effectiveness in Health and Medicine, we discounted costs and QALMs annually by 3%.45,46

Model Validation

We based our validation approach on the recommendations outlined by a task force appointed by the International Society for Pharmacoeconomics and Outcomes Research and the Society for Medical Decision Making, using both cross-validation (comparison results with estimates from similar models) and external validation (comparison with real-world values).47 To evaluate the model's validity, we derived 4 key outcomes from the model and compared them with external benchmarks. The outcomes were (1) HIV incidence rate, (2) mortality rate for people with HIV, (3) life expectancy of people with HIV, and (4) lifetime costs of HIV care and treatment. For HIV incidence and mortality rates, we compared our results with rates reported by the CDC,19,48 as well as with published estimates from similar models.49,50 For life expectancy and lifetime costs of HIV care, we compared our results with published estimates from other HIV microsimulation models.12,16,38 We report estimates from years 5 and 10 to allow enough time for the model to generate stable results while still aligning with currently available near-term estimates. We followed convention by defining a valid result as one in which model predictions were within 15% of the external benchmark.48

RESULTS

The estimated HIV incidence rates for the MSM-only risk group at years 5, 10, and 25 and for the IDU-only risk group at year 50 are within 15% of the external benchmarks for the corresponding risk group and thus met our validity criterion (Table 2). The HIV incidence rates for the MSM + IDU group are more than 15% higher than the benchmark. However, the benchmark for this group is lower than the benchmark for the MSM-only group. In contrast, our model assumes that on average, individuals in the MSM + IDU group have a higher risk of acquiring HIV than the MSM-only and IDU-only groups because they have multiple infection modalities (sexual contact and needle sharing). As a result, the fact that the model results are higher than the benchmark for this group are to be expected. The HIV incidence rate for the heterosexual risk group is less than half the external benchmark, but the values for this group are small and the benchmark based on HIV surveillance is likely imprecisely estimated.

TABLE 2. - Model Validation Versus External Benchmarks: Predicted HIV Incidence and Mortality Rates per 100,000 Person Years
Model Prediction Benchmark
Year 5 Year 10
HIV incidence by risk group, per 100,000 person years*
 MSM only 500§ 502§ 529
 IDU only 467 515 388
 MSM + IDU 643 635 476
 Heterosexual 1.1 1.2 3.3
Mortality among people with HIV by ART status, per 100,000 person years
 All diagnosed people with HIV 1503§ 1663§ 1650
 Diagnosed people with HIV and on ART 1423§ 1461§ 850–1630
Rates expressed per 100,000 person years.
Rodger, A.J, R. Lodwick, M. Schechter, S. Deeks, J. Amin, R. Gilson, R. Paredese, E. Bakowskaf, F.N. Engsigg, and A. Phillips, A. “Mortality in Well Controlled HIV in the Continuous ART Arms of the SMART and ESPRIT Trials Compared with the General Population.” AIDS 2013:27;973–979.
Antiretroviral Therapy Cohort Collaboration. “Prognosis of HIV-1-Infected Patients up to 5 Years After Initiation of HAART: Collaborative Analysis of Prospective Studies.” AIDS 2007:21;1185–1197.
The Antiretroviral Therapy Cohort Collaboration. “Life Expectancy of Individuals on Combination ART in High-Income Countries: A Collaborative Analysis of 14 Cohort Studies.” Lancet 2008:372;293–299.
*Centers for Disease Control and Prevention. “Estimated HIV Incidence and Prevalence in the United States, 2010–2015.” HIV Surveillance Supplemental Report 2018, vol. 23, no. 1, 2018.
Centers for Disease Control and Prevention. Monitoring selected national HIV prevention and care objectives by using HIV surveillance data—United States and 6 dependent areas, 2016. HIV Surveillance Supplemental Report 2018; 23 (No. 4).
Trickey, A, M.T. May, J. Vehreschild, N. Obel, M.J. Gill, H. Crane, C. Boesecke, H. Samji, S. Grabar, C. Cazanave, M. Cavassini, L. Shepherd, A. d'Arminio Monforte, C. Smit, M. Saag, F. Lampe, V. Hernando, M. Montero, R. Zangerle, A.C. Justice, T. Sterling, J. Miro, S. ingle, J.A.C. Sterne, and Antiretroviral Therapy Cohort Collaboration. “Cause-Specific Mortality in HIV-Positive Patients Who Survived 10 Years After Starting Antiretroviral Therapy.” PLoS One 2016;11:e0160460.
§Model results are within ±15% of benchmark. Model results are shown for years 5 and 10 of simulation.

The model results for mortality rates for all people diagnosed with HIV (regardless of treatment status) at years 5 and 10 are within 15% of the benchmark based on CDC surveillance data (Table 2).47 The mortality rates for people diagnosed with HIV who were on ART at years 5 and 10 also fall within the range of the estimates found in the literature.49–51 Therefore, we consider the model to produce externally valid mortality rates for people with HIV who have and have not initiated ART.

In keeping with previous studies, we estimated average life expectancy for people with HIV who were not on ART to validate our model's ability to capture the natural history of disease progression (Table 3).12,52 We estimated the mean life expectancy of individuals with a CD4 count of 50 cells per microliter and 200 cells per microliter to be 1.52 years and 3.78 years, respectively, both of which are within 15% of the external benchmark.52 The model predicted the average life expectancy of people with HIV with a CD4 count of 500 cells per microliter to be about 33% above the external benchmark. Advances in medical care may explain the increased life expectancy predicted by our model compared with the benchmarks.

TABLE 3. - Model Validation Versus External Benchmarks: Predicted Average Life Expectancy by CD4 Count
CD4 Model Prediction Benchmark*
50 cells/μL 1.52 years 1.39 yrs
200 cells/μL 3.78 years 3.33 yrs
500 cells/μL 9.36 yrs 7.05 yrs
*Freedberg, K.A., E. Losina, M.C. Weinstein, A.D. paltiel, C.J. Cohen, G.R. Seage, D.E. Craven, H. Zhang, A.D. Kimmel, and S.J. Goldie. “The Cost Effectiveness of Combination ART for HIV Disease.” New England Journal of Medicine,2001;344:824–831.
Model results are within ±15% of external benchmark.

Finally, we estimated lifetime treatment costs of people with HIV who had ever initiated treatment to be $362,385. This is within 15% of the lifetime cost of $341,266 estimated for individuals with ART using the Cost-Effectiveness of Preventing AIDS Complications model.38 It is also within the $335,100 lifetime costs for people in medical care coordination for HIV disease estimated by Flash et al.16 Thus, we consider that our model meets our external validity criterion for this outcome.

DISCUSSION

The microsimulation presented in this paper represents the first model to capture the complexities of the RWHAP care delivery system in the United States, including simulating HIV transition along the stages of the HIV care continuum. The model required development of innovative strategies to calculate the need for and receipt of ancillary services among people with HIV under alternative systems of care, determine their costs, and estimate their impact on care retention and viral suppression. Most of the model outputs we tested—those that are critical to producing reasonable outcomes for the cost-effectiveness analysis—satisfied the validation criterion of within 15% of their respective benchmarks.48 Although further refinements are possible (such as allowing ART and OAHS costs to vary within the care and treatment stages based on a person's health state), the results presented in this article suggest that the model captures the dynamics of the current HIV epidemic in the United States and serve as a useful tool for evaluating the cost-effectiveness of a comprehensive system of care for HIV.

Several data limitations should be noted. First, some of the inputs needed to parameterize the model could not be estimated from existing data or obtained from published sources. In these instances, we made simplifying assumptions or leveraged calibration techniques used in other similar models faced with the need to synthesize inputs from multiple sources, reconcile inconsistencies between sources, and develop new inputs.53 A second limitation was the lack of information about people with HIV who do not receive services under the RWHAP. This limits our ability to compare RWHAP clients with those who are either not receiving services or receiving services funded by other programs, such as Medicare, Medicaid, or the Veterans Administration. We made assumptions on the conditional probability distribution of these characteristics and imputed missing values when necessary from existing data. Third, the model currently measures unmet need only among those in care. If unmet need is greater among those not in care than those in care, the model will underestimate the true extent of unmet need among people with HIV. Fourth, by expanding the number of HIV providers, the presence of the RWHAP might create more opportunities for linking newly diagnosed people to care. If the RWHAP facilitates linking patients to care and treatment more quickly after diagnosis than would otherwise have occurred in the absence of the program, the results of the cost-effectiveness analysis will be underestimated.

We intend to use this model to evaluate the cost-effectiveness of the RWHAP relative to a non-RWHAP scenario and to estimate the potential impact of new initiatives and policies—such as the recently enacted national plan to end the HIV epidemic—on the cost-effectiveness of the program.54 We will also use the model to assess the distribution of the relative cost and health outcomes of the RWHAP across demographic subgroups and services.

ACKNOWLEDGMENTS

This article was made possible by the contributions of the Ryan White HIV/AIDS Program grant recipients and subrecipients that provided data to the Health Resources and Services Administration.

REFERENCES

1. Health Resources and Services Administration. Ryan White HIV/AIDS Program Service Categories Crosswalk with the HIV Care Continuum. 2015.
2. Centers for Disease Control and Prevention. Understanding the HIV Care Continuum. 2018. Available at: https://www.cdc.gov/hiv/pdf/library/factsheets/cdc-hiv-care-continuum.pdf. Accessed December 31, 2018.
3. HIV/AIDS Bureau. Ryan White HIV/AIDS Program Services: Eligible Individuals & Allowable Uses of Funds. Washington, DC: HAB; 2018.
4. Kaiser Family Foundation. U.S. Federal Funding for HIV/AIDS: Trends Over Time. 2019. Available at: https://www.kff.org/hivaids/fact-sheet/u-s-federal-funding-for-hivaids-trends-over-time/. Accessed December 31, 2019.
5. Health Resources and Services Administration. Ryan White HIV/AIDS Program Legislation. 2019. Available at: https://hab.hrsa.gov/about-ryan-white-hivaids-program/ryan-white-hivaids-program-legislation. Accessed December 31, 2019.
6. Health Resources and Services Administration. Ryan White HIV/AIDS Program Annual Client-Level Data Report. 2017. Available at: http://hab.hrsa.gov/data/data-reports. Accessed December 31, 2018.
7. Mandsager P, Marier A, Cohen S, et al. Reducing HIV-related health disparities in the health Resources and Services Administration's Ryan white HIV/AIDS program. Am J Public Health. 2018;108:S246–S250.
8. Weiser J, Beer L, Frazier EL, et al. Service delivery and patient outcomes in Ryan White HIV/AIDS Program–funded and–nonfunded health care facilities in the United States. JAMA Intern Med. 2015;175:1650–1659.
9. Bradley H, Viall AH, Wortley PM, et al. Ryan White HIV/AIDS program assistance and HIV treatment outcomes. Clin Infect Dis. 2015;62:90–98.
10. Gomez GB, Borquez A, Case KK, et al. The cost and impact of scaling up pre-exposure prophylaxis for HIV prevention: a systematic review of cost-effectiveness modelling studies. PLOS Med. 2013;10:e1001401.
11. Bernard CL, Brandeau ML, Humphreys K, et al. Cost-effectiveness of HIV preexposure prophylaxis for people who inject drugs in the United States. Ann Intern Med. 2016;165:10–19.
12. Freedberg KA, Losina E, Weinstein MC, et al. The cost effectiveness of combination antiretroviral therapy for HIV disease. N Engl J Med. 2001;344:824–831.
13. Walensky RP. Cost-effectiveness of HIV interventions: from cohort studies and clinical trials to policy. Top HIV Med. 2009;17:130–134.
14. Freedberg KA, Hirschhorn LR, Schackman BR, et al. Cost-effectiveness of an intervention to improve adherence to antiretroviral therapy in HIV-infected patients. J Acquir Immune Defic Syndr. 2006;43:S113–S118.
15. Krebs E, Dale LM. The impact of localized implementation: determining the cost-effectiveness of HIV prevention and care interventions across six United States cities. HIV Spec. 2020;12:20–27.
16. Flash MJ, Garland WH, Martey EB, et al. Cost-effectiveness of a Medical Care Coordination Program for People with HIV in Los Angeles County. Open Forum Infectious Diseases.
17. Stevens ER, Nucifora KA, Irvine MK, et al. Cost-effectiveness of HIV care coordination scale-up among persons at high risk for sub-optimal HIV care outcomes. PLoS One. 2019;14:e0215965.
18. Goyal R, Luca D, Klein P. Cost-Effectiveness of the Health Resources and Services Administration's Ryan White HIV/AIDS Program. J Acquir Immune Defic Syndr. 2020. doi: 10.1097/QAI.0000000000002547. epub ahead of print.
19. Centers for Disease Control and Prevention. Monitoring selected national HIV prevention and care objectives by using HIV surveillance data—United States and 6 dependent areas, 2016. HIV Surveill Supplemental Rep. 2018;23. Available at: http://www.cdc.gov/hiv/library/reports/hiv-surveillance.html. Accessed July 10, 2018.
20. Health Resources and Services Administration. Ryan White HIV/AIDS Program Annual Client-Level Data Report. 2016. Available at: http://hab.hrsa.gov/data/data-reports. Accessed December 31, 2017.
21. Health Resources and Service Administration. 2015 Ryan White HIV/AIDS Program ADAP Data Report (ADR). 2016. Available at: https://hab.hrsa.gov/program-grantsmanagement/ryan-white-hivaids-program-adap-data-report-adr. Accessed December 31, 2017.
22. U.S. Department of Health and Human Services, Panel on Antiretroviral Guidelines for Adults and Adolescents. Guidelines for the Use of Antiretroviral Agents in Adults and Adolescents with HIV. Washington, DC: DHHS; 2019. Available at http://www.aidsinfo.nih.gov/ContentFiles/AdultandAdolescentGL.pdf. Accessed December 31, 2019.
23. Messeri PA, Abramson D, Aidala AA, et al. The impact of ancillary HIV services on engagement in medical care in New York City. AIDS Care. 2002;14:15–29.
24. Lansky A, Finlayson T, Johnson C, et al. Estimating the number of persons who inject drugs in the United States by meta-analysis to calculate national rates of HIV and hepatitis C virus infections. PLoS One. 2014;9:e97596.
25. Purcell DW, Johnson CH, Lansky A, et al. Estimating the population size of men who have sex with men in the United States to obtain HIV and syphilis rates. Open AIDS J. 2012;6:98–107.
26. Morris M, Kurth AE, Hamilton DT, et al. Concurrent partnerships and HIV prevalence disparities by race: linking science and public health practice. Am J Public Health. 2009;99:1023–1031.
27. Kaplan EH, Heimer R. A model-based estimate of HIV infectivity via needle sharing. J Acquir Immune Defic Syndr. 1992;5:1116–1118.
28. Patel P, Borkowf CB, Brooks JT, et al. Estimating per-act HIV transmission risk: a systematic review. AIDS. 2014;28:1509.
29. Attia S, Egger M, Müller M, et al. Sexual transmission of HIV according to viral load and antiretroviral therapy: systematic review and meta-analysis. AIDS. 2009;23:1397–1404.
30. Li Z, Purcell DW, Sansom SL, et al. Vital signs: HIV transmission along the continuum of care - United States, 2016. MMWR Morb Mortal Wkly Rep. 2019;68:267–272.
31. Rodger AJ, Cambiano V, Bruun T, et al. Risk of HIV transmission through condomless sex in serodifferent gay couples with the HIV-positive partner taking suppressive antiretroviral therapy (PARTNER): final results of a multicentre, prospective, observational study. Lancet. 2019;393:2428–2438.
32. Gopalappa C, Farnham PG, Chen Y, et al. Progression and transmission of HIV/AIDS (path 2.0) A new, agent-based model to estimate HIV transmissions in the United States. Med Decis Making. 2017;37:224–233.
33. Rodríguez B, Sethi AK, Cheruvu VK, et al. Predictive value of plasma HIV RNA level on rate of CD4 T-cell decline in untreated HIV infection. JAMA. 2006;296:1498–1506.
34. Kitahata MM, Gange SJ, Abraham AG, et al. Effect of early versus deferred antiretroviral therapy for HIV on survival. N Engl J Med. 2009;360:1815–1826.
35. Centers for Disease Control and Prevention, Epidemiology Program Office. Morbidity and Mortality Weekly Report: MMWR. US Department of Health and Human Services, Public Health Service; 2017.
36. Medland NA, Chow EP, McMahon JH, et al. Time from HIV diagnosis to commencement of antiretroviral therapy as an indicator to supplement the HIV cascade: dramatic fall from 2011 to 2015. PLoS One. 2017;12:e0177634.
37. Sax PE, DeJesus E, Mills A, et al. Co-formulated elvitegravir, cobicistat, emtricitabine, and tenofovir versus co-formulated efavirenz, emtricitabine, and tenofovir for initial treatment of HIV-1 infection: a randomised, double-blind, phase 3 trial, analysis of results after 48 weeks. Lancet. 2012;379:2439–2448.
38. Schackman BR, Fleishman JA, Su AE, et al. The lifetime medical cost savings from preventing HIV in the United States. Med Care. 2015;53:293.
39. O'Connor J, Smith C, Lampe FC, et al. Durability of viral suppression with first-line antiretroviral therapy in patients with HIV in the UK: an observational cohort study. Lancet HIV. 2017;4:e295–302.
40. Dunn D, Woodburn P, Duong T, et al. Current CD4 cell count and the short-term risk of AIDS and death before the availability of effective antiretroviral therapy in HIV-infected children and adults. J Infect Dis. 2008;197:398–404.
41. Gebo KA, Fleishman JA, Conviser R, et al. Contemporary costs of HIV health care in the HAART era. AIDS. 2010;24:2705.
42. Health Resources and Services Administration. Ryan White HIV/AIDS Program (RWHAP) Recipient Allocation and Expenditure Reports. 2016. Available at: https://hab.hrsa.gov/program-grants-management/grant-recipient-allocation-and-expenditure-reports. Accessed December 31, 2018.
43. Farnham PG, Gopalappa C, Sansom SL, et al. Updates of lifetime costs of care and quality-of-life estimates for HIV-infected persons in the United States: late versus early diagnosis and entry into care. J Acquir Immune Defic Syndr. 2013;64:183–189.
44. Tengs TO, Lin TH. A meta-analysis of utility estimates for HIV/AIDS. Med Decis Making. 2002;22:475–481.
45. Russell LB, Gold MR, Siegel JE, et al. The role of cost-effectiveness analysis in health and medicine. JAMA. 1996;276:1172–1177.
46. Sanders GD, Neumann PJ, Basu A, et al. Recommendations for conduct, methodological practices, and reporting of cost-effectiveness analyses: second panel on cost-effectiveness in health and medicine. JAMA. 2016;316:1093–1103.
47. Linley L, Johnson AS, Song R, et al. Estimated HIV Incidence and Prevalence in the United States 2010–2015. 2018.
48. Goldie SJ, Yazdanpanah Y, Losina E, et al. Cost-effectiveness of HIV treatment in resource-poor settings—the case of Côte d'Ivoire. N Engl J Med. 2006;355:1141–1153.
49. Trickey A, May MT, Vehreschild J, et al. Cause-specific mortality in HIV-positive patients who survived ten years after starting antiretroviral therapy. PLoS One. 2016;11:e0160460.
50. Rodger AJ, Lodwick R, Schechter M, et al. Mortality in well controlled HIV in the continuous antiretroviral therapy arms of the SMART and ESPRIT trials compared with the general population. AIDS. 2013;27:973–979.
51. Antiretroviral Therapy Cohort Collaboration. Life expectancy of individuals on combination antiretroviral therapy in high-income countries: a collaborative analysis of 14 cohort studies. Lancet. 2008;372:293–299.
52. Ciaranello AL, Morris BL, Walensky RP, et al. Validation and calibration of a computer simulation model of pediatric HIV infection. PLoS One. 2013;8:e83389.
53. Rydzak CE, Cotich KL, Sax PE, et al. Assessing the performance of a computer-based policy model of HIV and AIDS. PLoS One. 2010;5:e12647.
54. U.S. Department of Health and Human Services. Ending the HIV Epidemic: A Plan for America. Washington, DC: DHHS; 2018.
Keywords:

HIV/AIDS; health care cost; HRSA; RWHAP; cost-effectiveness; mathematical modeling

Supplemental Digital Content

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.