Estimated Lifetime HIV–Related Medical Costs in the United States

Abstract

Background 

Lifetime cost estimates are a useful tool in measuring the economic burden of HIV in the United States. Previous estimation methods need to be updated, given improving antiretroviral therapy regimens and updated costs.

Methods 

We used an updated version of the agent-based model progression and transmission of HIV (PATH) 3.0 to reflect current regimens and costs. We simulated a cohort of those infected in 2015 until the last person had died to track the lifetime costs for treatment of HIV, including HIV health care utilization costs (inpatient, outpatient, opportunistic infection prophylaxis, non-HIV medication, and emergency department), opportunistic infection treatment costs, and testing costs. We assumed a median per-person diagnosis delay of 3 years and a 3% base monthly probability of dropout from care for a base-case scenario. Additionally, we modeled a most favorable scenario (median diagnosis delay of 1 year and 1% base dropout rate) and a least favorable scenario (median diagnosis delay of 5 years and 5% base dropout rate).

Results 

We estimated an average lifetime HIV-related medical cost for a person with HIV of $420,285 (2019 US$) discounted (3%) and $1,079,999 undiscounted for a median 3-year diagnosis delay and 3% base dropout rate. Our discounted cost estimate was $490,045 in our most favorable scenario and $326,411 in our least favorable scenario.

Conclusions 

Lifetime per-person HIV-related medical costs depend on the time from infection to diagnosis and the likelihood of dropping out of care. Our results, which are similar to previous studies, reflect updated antiretroviral therapy regimens and costs for HIV treatment.

In Brief

Using updated recommendations, costs, and regimens for antiretroviral therapy in our agent-based model, we estimated the discounted lifetime HIV–related medical cost for persons with HIV to be $420,285.

Lifetime cost estimates for persons with HIV (PWH) are important for informing economic evaluations of HIV prevention interventions.^1–3 These estimates are often used in measuring the economic burden of sexually transmitted diseases in the United States.^1,3,4 Although previous lifetime cost estimates can be updated yearly for inflation, recent improvements in antiretroviral therapy (ART) drug regimens have resulted in fewer persons dropping out of care because of virologic failure.⁵ This calls for an update to previous lifetime cost estimation methods and estimates.⁶ Additionally, the Guidelines for the Use of Antiretroviral Agents in Adults and Adolescents routinely update the average wholesale price (AWP) of ART and available generic components.⁷

Lifetime cost estimates incorporate costs of ART regimens for PWH when in care and can, therefore, be impacted by how long after HIV infection a person starts care and treatment, as well as how well they adhere to their regimens. A person’s CD4 count can have an impact on the amount of treatment needed. For example, having a lower CD4 count could lead to a higher probability of acquiring an opportunistic infection (OI).⁸ Dropping out of care or experiencing a gap in care can also affect estimated lifetime cost.

We used an agent-based simulation model, a computational model that tracks each individual's behaviors, to measure the estimated per-person HIV-related lifetime medical costs for a cohort of PWH infected in 2015 and followed until the last person had died. We chose 2015 to be consistent with previous studies in which the model⁹ was used and to reflect the most current data published by Centers for Disease Control and Prevention at the time (see Appendix, https://links.lww.com/OLQ/A606). Because the amount of time spent in care impacts the lifetime cost estimate, we varied the probability of dropping out of care and the time delay from infection to diagnosis to obtain a most favorable and a least favorable scenario.

METHODS

For our analysis, we used an updated version of PATH 3.0, Progression and Transmission of HIV, an agent-based model created in NetLogo,¹⁰ which has been used in previous analyses of HIV prevention interventions and costs.^6,9,11,12 The model was initialized with a population of 10,000 persons, representing a cross section of PWH in the United States in 2006. The initial population was weighted to represent PWH by age, sex, transmission risk group (men who have sex with men (MSM), persons who inject drugs (PWID)-male, PWID-female, PWID-MSM, heterosexual male, heterosexual female), and stage of the HIV care continuum (acutely infected, but not aware of infection; nonacutely infected and unaware; aware, but not in care; in care, but not virally suppressed; in care and virally suppressed).

The time step for the simulation was 1 month. For each time step of the simulation, PATH tracked the viral load and CD4 count of each individual in the model, considering disease progression and care-continuum stage. To determine if an event such as death, dropping out of care, or acquiring an OI occurs, rates and risks were converted to monthly probabilities, a random number between 0 and 1 was drawn, and if the random number was less than the probability, the event occurred. The death of each individual was determined by a calibrated monthly probability to match the number of deaths among PWH (regardless of cause) per year.¹³ During the acute phase of infection, the time step was weekly rather than monthly.

During the analytic period, which began in 2016, we modeled movement along the care continuum directly by adjusting the time from infection to diagnosis and the probability of dropping out of care. Before the analytic period, we matched the proportion of persons who were diagnosed with HIV infection, linked to care, prescribed ART, and achieved a suppressed viral load in the model to annual HIV surveillance data from 2006 to 2015 (see Appendix, https://links.lww.com/OLQ/A606).^14–21

For persons infected in 2015, we used a median diagnosis delay of 3 years, consistent with the study by Dailey et al.²² This resulted from calibrating the model to match the reported increase in the percentage of diagnosed infections among PWH from 2010 (82.8%) to 2015 (85.3%).²² We, therefore, assigned each person infected in 2015 a delay from a uniform distribution of 23 to 54 months, which accounts for the 2% to 3% of persons who die before diagnosis. We assumed that all PWH infected after 2015 began ART treatment when they were diagnosed.

For the probability of dropping out of care, we assumed that the probability was inversely proportional to the time spent on ART. For example, if k was the initial monthly dropout probability and y is the number of years a person has been on ART, then the monthly dropout probability would be:

$f\left(y\right)=\frac{k}{1+\frac{y}{2}}$

We assumed that 45% of those who dropped out re-entered care after 1 year, and the other 55% reentered care after 2 years.²³ Before the analytic period, we calibrated the model to match the decline in incidence from 2010 to 2015 according to HIV surveillance data.²⁴

Recent literature⁵ suggests that fewer patients change ART regimens because of virologic failure than a decade ago and that patients now usually change regimens for reasons related to cost, simplicity, or drug interactions. We updated regimens that began after 2015 as summarized in Figure 1. Using a uniform distribution, we assigned 1 of 4 regimens chosen from the current ART guidelines,⁷ the most prescribed regimens,⁵ and subject matter experts. We assumed that persons took the regimen for one (if they started ART with a CD4 count > 200 cells/μL) to 2 months (if they started ART with CD4 count ≤ 200 cells/μL) before they became virally suppressed. Some people switched their regimen at a probability of 41%.⁵ We assigned those who changed their regimen a random length of time to stay with the first regimen from a truncated normal distribution with a mean of 48 months⁵ and lower and upper bounds of 12 and 72 months. We assumed that switching to a different regimen affected the cost incurred, but not the viral suppression status. Our model also allowed some persons to have a short gap (less than a year in length) at some point in their regimen. According to Byrd et al,²⁵ 8% to 30% of people in their study experienced a gap. Using a median value of 19%, we changed this to a monthly probability of each person taking a gap that month, 1 – e^{−0.19 * (1/12)} = 0.016. Because 69% of those who experienced a gap never filled their prescriptions during the gap,²⁶ we assumed they were not virally suppressed and did not accrue costs (except for non-HIV medications). The duration of the gap was selected from a truncated normal distribution with a mean of 3 months, a standard deviation of 3 months, and lower and upper bounds of 1 and 12 months.

Figure 1: Flowchart for updated regimen program. A person with diagnosed HIV enters care and starts antiretroviral therapy (ART). If they modify, they will change regimens after an assigned duration with median of 48 months. Persons stick with their first or second regimens until death. They may also experience a gap in care (with median time of 3 months but less than a year) or drop out of care (for 1–2 years or over 2 years). Those who have not started ART yet or have dropped out are labeled as aware/not in care and have a higher viral load. Those who are in a gap are labeled as on ART, but not VLS. They have a viral load slightly higher than being completely VLS. Those who are on ART are VLS. VLS, virally suppressed.

For the cost analysis, we defined our cohort as those in the model who were infected in the year 2015, and we verified that the percentages of the cohort by age and transmission risk group matched incidence in 2015 (Table 1). We tracked the cohort until everyone in the cohort had died. During the simulation, we estimated ART costs for the use of the four regimens described above, and we calculated testing cost, including an HIV test, HIV-RNA test for viral load detection, CD4 test, and a genotype test for whenever a person started a new regimen (Table 2). If a person acquired an OI, which depended on the CD4 count and the monthly probability of acquiring an OI, we tracked the OI treatment cost. We also kept track of their viral load (Table 2), which was used in calculating their CD4 count. For those not in care, we kept track of their non-HIV medication costs (costs associated with HIV infection, but not related to ART or OI prophylaxis) (Table 2), which were based on CD4 count. Then, for those in care, HIV health care utilization costs included inpatient care, outpatient care, OI prophylaxis, non-HIV medication costs, and emergency department costs (Table 2) based on CD4 count. The non-HIV medication costs and HIV health care utilization costs were randomly selected from truncated normal distributions with means and standard deviations outlined in Table 2. Finally, we recorded the total lifetime cost, discounted at 3%, and undiscounted, for the cohort living with HIV. All costs are reported in 2019 US dollars, and those reported in different years were adjusted using the medical care component of the Consumer Price Index. Future costs were discounted at a rate of 3%.

TABLE 1 - Comparison of 2015 Cohort by Proportions of Age and Transmission Risk Group to the 2015 HICSB Incidence Data

		PATH Newly Infected in 2015		HICSB Incidence Data 2015 (14)
		(Total Cohort: 1006)		(Total Cohort: 38,500)
		Count	Percent	Count	Percent
Age at infection, y
13–24		247	25	10,200	27
25–34		359	36	13,700	36
35–44		185	18	6800	18
45–54		135	13	5000	13
≥55		80	8	2800	7
Transmission risk group
HET female		163	16	5800	15
HET male		68	7	2800	7
MSM		690	67	26,100	68
HET-PWID female		14	1	1000	3
HET-PWID male		29	3	1300	3
MSM-PWID		42	4	1400	4

PATH, Progression and transmission of HIV; MSM, men who have sex with men; PWID, persons who inject drugs; HICSB, HIV Incidence and Case Surveillance Branch; HET, heterosexual.

TABLE 2 - Input Parameters for PATH Important to Cost Analysis

	Input Value					Source
Epidemiologic variables
CD4 cell count when infected	750–900					32s *
HIV viral load set point	4.0–5.0					33s, 34s
Monthly probability of acquiring an OI	0.0004–0.003					8
Suppressed HIV viral load level	1–2.7					35s
Nonsuppressed HIV VL on ART	3.1–4.5					36s
Max number of regimens	4					Subject matter expert
Cost variables	Per Month or Event (2019 US$)
ART regimen cost
I. BIC/TAF/FTC	2368					37s
II. DTG/ABC/3TC	2763
III. EFV/TDF/FTC	1881
IV. EVG/COBI/FTC/TDF	2205
OI treatment cost
Pneumocystis pneumonia	967					31s
Mycobacterium avium complex	386
Toxoplasmosis	2271
Cytomegalovirus infection	601
Fungal infection (Candidiasis)	658
Toxoplasmosis	441
Cryptococcosis	441
HIV testing cost per new diagnosis	3599					6
CD4 test cost (each quarter)	215					27
HIV-1 RNA test (each quarter)	537					27
HIV genotype test cost (applied at beginning of new regimen)	326					27
Health care utilization cost	Mean cost (2019 US$) by CD4 count category per month (standard deviation)
	≤50	51–200	201–350	351–500	>500
Non-HIV medication	250 (22)	242 (13)	224 (10)	228 (9)	254 (9)	27
OI prophylaxis	150 (11)	79 (5)	17 (2)	6 (1)	4 (1)
Inpatient utilization	3055 (398)	1189 (130)	540 (54)	305 (49)	132 (19)
Outpatient utilization	85 (6)	78 (3)	68 (2)	61 (2)	57 (1)
ED utilization	123 (42)	56 (5)	32 (2)	24 (2)	17 (1)
Other variables
Discount rate	3%					28

*References 31s–37s are listed in the supplementary file.

All costs are reported in 2019 U.S. dollars. Viral load (VL), emergency department (ED).

PATH, Progression and Transmission of HIV; OI, opportunistic infection; ART, antiretroviral therapy; BIC/TAF/FTC, Bictegravir/Tenofovir Alafenamide/Emtricitabine (Biktarvy); DTG/ABC/3TC, Dolutegravir/Abacavir/Lamivudine (Triumeq); EFV/TDF/FTC, Efavirenz/Tenofovir Disoproxil Fumarate/Emtricitabine (Atripla); EVG/COBI/FTC/TDF, Elvitegravir/Cobicistat/Tenofovir Disoproxil Fumarate/Emtricitabine (Stribild).

We also examined different scenarios by varying time from infection to diagnosis and dropout rate. For our base case, we assumed a 3% base monthly dropout rate and a median delay to diagnosis of 3 years. We defined a least favorable scenario that included a 5% base dropout rate and a median delay to diagnosis of 5 years. Additionally, we defined a most favorable scenario that included a 1% base dropout rate and a median delay to diagnosis of 1 year.

Antiretroviral drug prices vary substantially by payers who pay or reimburse for the drug costs. The final prices often include mandatory or voluntary discounts, rebates, and negotiated reimbursement rates.⁷ Because there is no consensus on the most accurate prices for pharmaceuticals in the United States, the Second Panel on Cost-Effectiveness in Health and Medicine recommended that the Federal Supply Schedule (FSS) prices be used to reflect the societal marginal costs of drugs for health care costs and for cost-effectiveness analyses.²⁸ We used FSS prices for the antiretroviral drug costs (Table 2), which are estimated to be 53% of the AWP list prices.²⁹ We varied the base-case drug prices by ±25% in a sensitivity analysis to account for a wide variation in drug costs, including lower prices of some generic ART regimens.³⁰ Recent literature showed that switching from brand name to an available generic formulation of an ART regimen could reduce the costs by as much as 25%,³⁰ the lower bound in our sensitivity analysis. On the other hand, the average price paid to the drug manufacturer by other federal, nonfederal, or commercial payers could be much higher than the FSS cost.²⁹ We assumed the upper bound to be 25% higher than the FSS cost in our sensitivity analysis.

Additionally, we explored the effect of gap length on lifetime cost estimates by extending the mean gap length from its base case value of 3 to 9 months. We also decreased the probability of experiencing a gap to zero, and we increased it to 30% to explore those effects on lifetime costs. Finally, we varied the percentage of those who re-entered care 1 year (but less than 2 years) after dropping out by ± 30% from the base 45%.

RESULTS

We estimated the average lifetime HIV-related medical cost for a person with HIV to be $420,285 (2019 US$) discounted and $1,079,999 undiscounted (Table 3). For this base case estimate, we assumed a median delay from infection to diagnosis of 3 years and an initial 3% monthly probability of dropping out of care.

TABLE 3 - Results from Base Case Scenario, Least Favorable Scenario, and Most Favorable Scenario in 2019 US$*

	Base Case		Least Favorable Scenario		Most Favorable Scenario
	Mean	95% CI	Mean	95% CI	Mean	95% CI
Disc. lifetime costs	420,285	416,082–424,487	326,411	322,457–330,365	490,045	485,001–495,088
Undiscounted lifetime costs	1,079,999	1,066,896–1,093,103	873,402	860,823–885,982	1,205,458	1,190,661–1,220,255
Disc. health care utilization costs	64,037	63,381–64,694	50,193	49,560–50,825	75,420	74,600–76,239
Disc. drug regimen costs	328,026	325,128–330,925	252,524	249,763–255,284	417,701	414,646–420,757
Disc. OI treatment costs	360	354–366	369	363–375	372	366–379
Onset of AIDS, y	19.17	18.90–19.46	17.58	17.33–17.83	17.63	17.41–17.96
Additional life expectancy by age at infection (years)
<25	48.09	47.68–48.50	44.95	44.51–45.40	50.04	49.66–50.43
25–34	40.21	39.87–40.55	37.86	37.51–38.22	39.64	39.30–39.98
35–44	34.21	33.96–34.47	31.94	31.66–32.21	33.39	33.13–33.65
45–54	27.59	27.41–27.77	26.27	26.07–26.47	26.76	26.57–26.95
≥55	13.90	13.81–13.99	13.59	13.50–13.69	13.73	13.64–13.83
Overall	37.31	36.94–37.68	35.07	34.69–35.45	37.30	36.92–37.67
Duration on ART regimens by age at infection (years)
<25	39.62	39.23–40.01	31.85	31.34–32.24	47.20	46.82–47.59
25–34	31.97	31.64–32.31	25.29	24.97–25.61	35.36	34.98–35.74
35–44	26.09	25.82–26.35	19.90	19.64–20.16	27.88	27.55–28.22
45–54	20.24	20.04–20.44	15.29	15.10–15.48	21.28	21.01–21.54
≥55	8.02	7.93–8.11	5.26	5.19–5.34	9.25	9.12–9.38
Overall	29.34	28.98–29.70	23.01	22.67–23.35	32.97	32.56–33.38
Average age at death by age at infection, y
<25	66.67	66.26–67.07	63.44	63.04–63.88	68.47	68.09–68.86
25–34	69.68	69.30–70.02	67.48	67.13–67.83	69.11	68.78–69.45
35–44	73.78	73.53–74.03	71.41	71.14–71.69	72.92	72.67–76.15
45–54	77.04	76.86–77.23	75.85	75.64–76.05	76.34	76.15–76.53
≥55	83.90	83.81–83.99	83.59	83.50–86.69	83.73	83.64–83.83
Overall	71.79	71.46–72.12	69.59	69.23–69.94	71.77	71.45–72.09

* The base case scenario reflects a 3% base monthly dropout rate and 3-year median diagnosis delay. The least favorable scenario reflects a 5% base dropout rate and a 5-year diagnosis delay. The most favorable scenario reflects a 1% base dropout rate and a 1-year median diagnosis delay. Results are averages of the cohort for their total lifetime costs, except for regimen and OI treatment costs. These two costs are averaged over the persons in the cohort who actually accumulated costs in these categories. For onset of AIDS, this was averaged over the persons in the cohort who developed AIDS.

OI, opportunistic infection; ART, antiretroviral therapy; AIDS, acquired immune deficiency syndrome; CI, confidence interval; Disc., discounted.

With a median diagnosis delay of 5 years and a 5% base dropout rate in our least favorable scenario, the discounted lifetime cost estimate was $326,411 and the undiscounted estimate was $873,402. The most favorable scenario (median diagnosis delay of 1 year and a 1% base dropout rate) yielded a discounted estimate of $490,045 and an undiscounted estimate of $1,205,458.

The drug regimen cost was about 78% of the discounted lifetime costs for the base case and 77% of the discounted lifetime costs for the least favorable scenario. For the most favorable scenario, the drug regimen cost was 85% of the discounted lifetime costs.

We estimated an average additional life expectancy in the base case of 37.31 years, which was similar to the most favorable scenario's 37.30 years (Table 3). Average additional life expectancy in the least favorable scenario was 2 years shorter. The average age at death for the base case was 71.79 years, which is similar to the most favorable scenario's 71.77 years. However, the overall average time spent on ART (including those who never started ART) was 29.34 years for the base case scenario and 32.97 years for the most favorable scenario. Those infected between the ages of 13 to 24 years lived longer in the most favorable scenario, had a longer additional life expectancy, and spent almost 8 more years on ART (Table 3). Additionally, in the worse-case scenario, those infected between the ages of 13 to 24 years had a shorter additional life expectancy on average and spent eight less years on ART. As the age groups at infection increased, the difference between the base case and the most favorable scenario for both average age of death and average time on ART generally decreased.

In the sensitivity analysis, we varied the antiretroviral drug prices by ±25% from the base case and estimated discounted lifetime costs to range from $341,963 to $493,995 (Table 4). Changing the gap length and the probability of having a gap in treatment had little effect on lifetime cost estimates (results not shown). Changing the percentage of those who re-entered care 1 year (but less than two) after dropping out by ±30% from the base 45% only affected the costs by approximately 6% (results not shown).

TABLE 4 - Results of Sensitivity Analysis on Increasing and Decreasing Antiretroviral Therapy Prices by 25% (2019 US$)

	Discounted Lifetime Costs	Percent Difference	Undiscounted Lifetime Costs	Percent Difference	Discounted Drug Regimen Costs	Percent Difference
Base case	420,285	—	1,079,999	—	328,026	—
25% Less ART costs	341,963	−18.6%	877,387	−18.8%	245,115	−25.3%
25% More ART costs	493,995	17.5%	1,270,713	17.7%	410,606	25.2%

DISCUSSION

Because there are fewer PWH changing regimens due to virologic failure than in our previous HIV lifetime cost analysis,⁶ we updated our PATH model to reflect current ART regimens and prices. We developed a base case scenario that included a median 3-year delay from infection to diagnosis and a 3% base monthly probability of dropout from care. A proportion of the cohort could change their regimen due to reasons beyond that of virologic failure. In addition to dropping out of care, some persons experienced a gap of less than a year in their drug regimen. In our simulations, using a cohort of those infected in 2015, which we ran until all those in the cohort had died, we estimated an average discounted lifetime treatment cost of $420,285.

Our base case cost estimate was lower than the average discounted lifetime cost estimate of $501,000 (2019 US$) reported by Farnham et al⁶ in 2013 for persons who entered care at CD4 counts of 500 cells/μL. However, it must be noted that those authors assumed a best-case scenario where once persons entered care, they never dropped out and did not incur gaps in ART treatment. Our most favorable scenario gave an average discounted lifetime cost of $490,045, which is roughly equivalent to the Farnham et al estimates.

When comparing to Schackman et al's³ 2015 analysis, we observed similar results to our least favorable scenario. Those authors estimated a lifetime cost of $391,800 (2019 US$) for PWH who were infected at age 35 years. They assumed that PWH enter care 5.1 years after infection (similar to our least favorable scenario), but they had a two-year shorter additional life-expectancy of 29.5 years. For our least favorable scenario, those infected at age 35 to 44 years had an average additional life-expectancy of 31.9 years (Table 3), and for those infected at age 35 years (the youngest age in the group), it could be higher still. Schackman et al’s optimal care scenario, in which PWH entered care at CD4 count of 500/μL and stayed in care, was close to our most favorable scenario with their estimate of $435,200 ($522,240; 2019 US$). Though their estimate was higher, it is worth noting that our estimate assumed a 3% base dropout rate and the possibility of having a gap in care where PWH do not accumulate drug costs.

Because our most favorable scenario implied that persons spend more time on ART than in the other 2 scenarios, the average discounted regimen cost was higher with the most favorable scenario and lower with the least favorable scenario. Our discounted OI treatment costs did not vary substantially among scenarios because PWH will accrue OI treatment cost whether or not they are in care.

The additional life-expectancies for our base case scenario and our most favorable scenario were similar. This could be explained by the fact that fewer persons were suffering from virologic failure, optimizing whatever time they did spend on ART. Although our age at infection was based on the distribution of ages in 2015 incidence data¹⁴ and Farnham et al's paper used an average age at infection of 35 years, we observed an additional life-expectancy just three quarters of a year shy of that estimated by Farnham et al.⁶ This similarity resulted from the fact that 54% of the population in our analysis was infected between ages 25 and 44 years. When breaking down the age at death by age at infection, there was a larger gap for age at death between the most favorable scenario and base case for those infected between ages 13 and 24 years than for older age groups. Therefore, the younger the age at which a person is infected, the more they will benefit from being on ART longer.

Antiretroviral drug prices remain high in the United States and increase every year 4 to 6 times faster than the inflation rate.⁴ Although some generic components of the branded drug regimens are available in the market, their prices tend to remain high, ranging from 46% to 95% of the AWP of the branded regimens.^7,30 Martin and Shackman³⁰ showed that switching from brand-name to an available generic formulation of an ART regimen would yield a 25% reduction in the wholesale acquisition cost or the FSS cost. In a sensitivity analysis, we varied the antiretroviral drug cost by ±25% of the costs used in the base case, based on FSS prices, and we estimated that the average discounted lifetime costs of HIV care ranged from $341,963 to $493,995. These results represent potential variation in the lifetime costs based on other branded or generic drug prices and payers' perspectives, including the Medicaid drug rebate program under 340B or other federal and commercial payers' perspectives.

Our study had several limitations. While there are now many more drug regimens on the market, we limited our analysis to four regimens based on subject matter expert opinions and for simplicity. We also only included the costs of co-morbidities that were included in the data sources used in our analysis (30, 31s). Future studies could include costs associated with additional co-morbidities, chronic infections, and other non-HIV conditions, which would increase the lifetime cost estimates. We also did not include HIV prevention costs or lost productivity costs. Additionally, the PATH model uses parameters taken and calibrated from many different sources. These values should be updated as new surveillance data are released. Next, our parameters related to cost, disease progression, and death were not stratified by race/ethnicity. This could affect the average of total lifetime costs. Finally, we did not investigate the number of new infections created by the cohort, which would be an additional cost to society.

In conclusion, we estimated the average per-person discounted lifetime HIV-related medical costs for PWH to be $420,285. Our analysis incorporated updated drug regimen programs that reflect the fact that PWH are not switching regimens due to virologic failure as frequently. We explored various levels of adhering to drug regimens through dropping out of care and taking gaps in care, and our base case reflected a base monthly dropout probability of 3% and median delay from infection to diagnosis of 3 years.