C

RA

RGHA

a

Hb

c

d

e

f

IRtAt

0d

The American Journal of Medicine (2005) 118, 292–300

LINICAL RESEARCH STUDY

outine human immunodeficiency virus testing:n economic evaluation of current guidelines

ochelle P. Walensky, MD, MPH,a,b Milton C. Weinstein, PhD,d April D. Kimmel,a

eorge R. Seage III, ScD, MPH,c Elena Losina, PhD,e Paul E. Sax, MD,b

ong Zhang, SM,a Heather E. Smith,a Kenneth A. Freedberg, MD, MSc,a

. David Paltiel, PhDf

From the Divisions of Infectious Disease and General Medicine, Department of Medicine, Massachusetts Generalospital, and the Partners AIDS Research Center, Harvard Medical School, Boston, Massachusetts;

Division of Infectious Disease, Brigham and Women’s Hospital, Boston, Massachusetts;Department of Health Policy and Management, Center for Risk Analysis, andDepartment of Epidemiology, Harvard School of Public Health, Boston, Massachusetts;Department of Biostatistics, Boston University School of Public Health, Boston, Massachusetts; and

Yale School of Medicine, New Haven, Connecticut.

BACKGROUND: The Centers for Disease Control and Prevention guidelines recommend humanimmunodeficiency virus (HIV) counseling, testing, and referral for all patients in hospitals with an HIVprevalence of �1%. The 1% screening threshold has not been critically examined since HIV becameeffectively treatable in 1995. Our objective was to evaluate the clinical effect and cost-effectiveness ofcurrent guidelines and of alternate HIV prevalence thresholds.

METHODS: We performed a cost-effectiveness analysis using a computer simulation model of HIVscreening and disease as applied to inpatients in U.S. hospitals.

RESULTS: At an undiagnosed inpatient HIV prevalence of 1% and an overall participation rate of33%, HIV screening increased mean quality-adjusted life expectancy by 6.13 years per 1000 inpatients,with a cost-effectiveness ratio of $35 400 per quality-adjusted life-year (QALY) gained. Expansion ofscreening to settings with a prevalence as low as 0.1% increased the ratio to $64 500 per QALY gained.Increasing counseling and testing costs from $53 to $103 per person still yielded a cost-effectivenessratio below $100 000 per QALY gained at a prevalence of undiagnosed infection of 0.1%.

CONCLUSION: Routine inpatient HIV screening programs are not only cost-effective but wouldlikely remain so at a prevalence of undiagnosed HIV infection 10 times lower than recommendedthresholds. The current HIV counseling, testing, and referral guidelines should now be implementednationwide as a way of linking infected patients to life-sustaining care.© 2005 Elsevier Inc. All rights reserved.

KEYWORDS:HIV/AIDS;HIV EIA;Testing;Screening;Cost-effectiveness

This research was funded by the National Institute of Allergy andnfectious Diseases (K23AI01794, K24AI062476, K25AI50436,01AI42006, Center for AIDS Research P30AI42851), the National Insti-

ute of Mental Health (R01MH65869), the National Institute on Drugbuse (R01DA015612), and the Centers for Disease Control and Preven-

Requests for reprints should be addressed to Rochelle P. Walensky,MD, MPH, Division of General Medicine, Massachusetts General Hospi-tal, 50 Staniford Street, 9th Floor, Boston, Massachusetts 02114.

E-mail address: rwalensky@partners.org.

ion (S1396-20/21).

002-9343/$ -see front matter © 2005 Elsevier Inc. All rights reserved.oi:10.1016/j.amjmed.2004.07.055

(viwHgidrohwrs

pgrudtullss

M

S

rftrstparpwdedHosodhtc

ptpwC

S

abUaHopba

aiai

ttauatdtrcta

Hmior

D

AmsiiaiiP

293Walensky et al Routine HIV Testing

In 1993, the Centers for Disease Control and PreventionCDC) published guidelines for human immunodeficiencyirus (HIV) counseling, testing, and referral in an effort todentify the estimated 300 000 patients in the United Statesith undiagnosed HIV infection.1,2 Because undiagnosedIV prevalence tends to mirror HIV diagnosis rates, theseuidelines, updated in 2001, recommend routine counsel-ng, testing, and referral in hospitals with an HIV dischargeiagnosis rate of �1%.3,4 Although seroprevalence studieseveal that many hospitals exceed the 1% prevalence thresh-ld, voluntary HIV counseling, testing, and referral servicesistorically have been offered only in prenatal settings orhen patients present with an HIV-associated illness or

equest an HIV test, and not routinely in the inpatientetting.4–7

To date, only a very small number of inpatient facilitiesrovide HIV counseling, testing, and referral according touidelines.8 Other infrequent interventions have exploredoutine HIV testing in the outpatient, hospital-associated,rgent care, and emergency department settings,6,9–12 withemonstrated success identifying cases of HIV infec-ion.6–11 Insufficient resources are commonly cited for fail-re to adhere to counseling, testing, and referral guide-ines.13 This paper sought to quantify the life expectancyosses attributable to unidentified HIV infection and to as-ess the cost-effectiveness of routine inpatient HIV coun-eling, testing, and referral.

ethods

tudy overview

We constructed an inpatient screening model (hereaftereferred to as the “screening module”), building upon theoundation of a previously designed and published model ofhe natural history and treatment of HIV disease (hereaftereferred to as the “disease model”).14–16 The purpose of thecreening module is to simulate the detection of HIV infec-ion in a general inpatient target group, with a specifiedrevalence of undetected HIV infection, whose membersre offered routine voluntary HIV counseling, testing, andeferral. Whether HIV is detected or not, all HIV-infectedatients in the defined target group enter the disease model,hich tracks disease progression based on natural historyata until death.17–19 Only identified infected patients areligible for HIV-specific care in the model. Patients may beetected as infected by one of three mechanisms: routineIV screening via the program under examination; devel-pment of an opportunistic infection leading to the diagno-is of HIV; or later HIV testing outside the hospital, basedn current background HIV testing rates. Patients who areetected after discharge through the latter two mechanismsave the opportunity to receive HIV-related care at a laterime. This analysis evaluates the clinical outcomes and

ost-effectiveness of an inpatient screening program com- i

ared with no screening program under alternative assump-ions regarding undetected HIV prevalence, screening andost-test counseling costs, and program participation. Thisork has been approved by the Partners Human Researchommittee.

creening module

Hypothetical inpatients enter the screening module onet a time and are randomly assigned an HIV serostatusased on a user-specified prevalence of undetected HIV.ninfected patients are offered an HIV test, which they may

ccept or refuse. Patients who refuse testing accumulate noIV testing cost. Patients who accept the test accrue the costf both the test and pretest counseling time. Uninfectedatients who obtain a negative test result receive no healthenefit and accrue age-, race-, and sex-specific life expect-ncy and quality-adjusted life expectancy.

The model uses input prevalence data to specify whethern infected patient has acute (primary) or chronic HIVnfection. Patients with acute HIV infection who are testedre not identified, receiving false-negative HIV enzymemmunoassay results.

Infected patients presenting with an opportunistic infec-ion are presumed to be HIV diagnosed via evaluation ofheir presenting infection. Infected patients who do not haven opportunistic infection enter the disease model withntreated progression of their disease until they are detectednd become eligible for HIV-specific treatment. These pa-ients, when screened, are offered an HIV test and may alsoecline testing. Patients who are HIV infected, tested, andest positive will receive care only if they return for their testesults and keep their HIV care appointment (linkage toare). Patients identified as infected by a mechanism otherhan the screening program are presumed to enter HIV caret a later time.

To incorporate the various possible means by whichIV-infected patients who are offered testing may ulti-ately fail to receive care, we defined an “index of partic-

pation” as the product of the following probabilities: beingffered and accepting the HIV test, and returning for testesults and being linked to HIV care.20

isease model

The disease model—the Cost-effectiveness of PreventingIDS Complications model—is a state-transition simulationodel of HIV-infected patients. Mutually exclusive health

tates—acute HIV infection, chronic HIV infection, acute clin-cal events, and death—incorporate the patients’ relevant clin-cal details, including CD4 cell count, HIV RNA (viral load),nd cumulative history of opportunistic infections.14–16 Eachnfected patient enters the model one at a time and is trackedndividually until death to simulate the patient’s life course.atient statistics are accrued and can be studied individually or

n aggregate, and are summarized as the frequency and type of

op

awcipiiprWaqop

tumtr

po

clppr

1suRt

D

C

294 The American Journal of Medicine, Vol 118, No 3, March 2005

pportunistic infections, mean time on therapy, mean life ex-ectancy, quality-adjusted life expectancy, and lifetime cost.

The disease model can distinguish between identifiednd unidentified HIV-infected patients. While all patientsith HIV infection enter the disease model, they only be-

ome eligible for antiretroviral therapy and opportunisticnfection care once infection is detected. All undiagnosedatients follow natural history disease progression accord-ng to their HIV RNA titer and are at risk of opportunisticnfections based on their CD4 cell count. Once infectedatients are identified, they enter into care, which includesegular CD4 cell count and HIV RNA laboratory tests.

hen patients reach a CD4 cell count of 200 cells/mm3,ntiretroviral therapy is administered.21 Four unique se-uential antiretroviral regimens, with a diminishing efficacyf viral load suppression, are assumed to be available toatients once they are identified as infected (Table 1).

HIV-infected patients are at risk of opportunistic infec-ions based on their CD4 cell count.14,15,17 If a previouslynidentified patient develops an opportunistic infection, theodel presumes that an HIV diagnosis is then made and that

he patient is immediately referred to care. In care, patients

Table 1 Base case input data

Variable Base C

Total sampleHIV prevalence 1%CD4 count among those infected

Acute infection (1.7%) 534 ceChronic infection (98.3%) 320 ce

Index of participationTest offer/acceptance rate 37%Rate of return/linkage to care 88%

HIV-infected sampleAntiretroviral starting criterion CD4 �Antiretroviral efficacy at viral suppression

1st line 70% a2nd line 60% a3rd line 34% a4th line 22% a

Enzyme immunoassay test characteristicsSensitivity 99.6%Specificity 97.5%

Testing costsHIV test $3Counseling $50Linkage to care $25

Antiretroviral therapy-associated costs (per month)1st line $10152nd line $11223rd line $14114th line $1292CD4 cell count (per test)* $83HIV RNA (per test)* $110Genotypic antiretroviral resistance testing (per test) $400

HIV � human immunodeficiency virus.*Performed every 3 months.

eceive guideline-recommended opportunistic infection pro- U

hylaxis regimens at the appropriate CD4 cell count thresh-lds.44

HIV-infected patients may die from non–HIV-relatedauses, opportunistic infections, or other chronic HIV-re-ated causes, depending on CD4 cell counts and prior op-ortunistic infection history.17,18 The non–HIV-relatedrobability of death is based on background age-, sex-, andace-dependent death rates.45

To achieve stability in estimates, we simulated more than00 million patients in the screening module. Both thecreening module and the disease model are programmedsing C and compiled in C�� 6.0 software (Microsoft,edmond, Washington). Further disease model specifica-

ions are described in detail elsewhere.14–16

ata

A summary of model input data is found in Table 1.

ohort characteristics

Because we were examining a cohort of inpatients in the

Range Explored Reference

0.001%–100% (2,3)

3 400–600 cells/mm3 (22–25)3 50–500 cells/mm3 (20,26,27)

10%–100% (13,28–30)10%–100% (29,31–34)

ells/mm3 No therapy � CD4 � 350 cells/mm3 (21)

eeks 60%–95% (35)eeks 50%–80% (36)eeks 20%–50% (37)eeks 10%–40% (37)

95%–100% (38)95%–100% (38)

— (39)$25–$100 (39–41)$15–$100 (38)

$507–$1522 (42)$561–$1683 (42)$705–$1411 (42)$646–$1938 (42)

— (43)— (43)— (43)

ase

lls/mmlls/mm

200 c

t 48 wt 24 wt 12 wt 12 w

nited States, we used the 2000 Nationwide Inpatient Sam-

pa�

T

zfccsinnaaiwrfi

UH

ltrafGpr

I

hcw9rrp

trtp2icv

T

tRMUtdvr

A

iCrTanma

R

B

p6is1ptcAcaaeCmula

M

tp

295Walensky et al Routine HIV Testing

le to estimate the mean (� SD) age (55.8 � 11.9 years)nd sex distribution (39% male) of hospitalized patients18 years of age.46

est characteristics

We used published test characteristics of the HIV en-yme immunoassay with a sensitivity for chronic HIV in-ection of 99.6% and a specificity of 97.5%, and a per testost of $3, which includes kit, laboratory, and confirmationosts.38,39,41 Although the literature suggests a pretest coun-eling cost of $25,39–41 HIV counseling may be performedn many settings by higher-paid professionals (eg, registeredurses, social workers, or physicians). Recognizing that it isot always possible to provide counseling for $25, wedopted a conservatively high value of $50 as our base casessumption and explored values in the range of $25 to $100n sensitivity analyses. We further assumed that acute casesould not be identified by enzyme immunoassay.47,48 Test

esults that are positive by enzyme immunoassay are con-rmed with Western blot (100% specific).40,41

ndetected HIV prevalence: CD4 cell count andIV RNA

Blinded seroprevalence studies suggest that the preva-ence of undetected HIV disease is approximately equal tohe prevalence of diagnosed HIV disease.4 Using previouslyeported data on patients presenting for initial HIV care, wescertained a mean CD4 cell count of 320 � 260 cells/mm3

or patients with chronic undiagnosed HIV infection.20

iven the importance of this uncertain variable, we ex-lored how variations in these CD4 estimates influenced ouresults in sensitivity analyses.

ndex of participation

Rates of acceptance for voluntary HIV antibody testingave been documented at 3% to 100%; in 1990, rates spe-ific to hospitalized inpatients ranged from 11% to 91%,13

hile prenatal testing acceptance rates ranged from 28% to8%.28,30 For the base case, we used an offer/acceptanceate of 37%, regardless of serostatus, as reported by aoutine HIV counseling, testing, and referral demonstrationrogram.29

Rates of return for HIV test results for uninfected pa-ients (67%) were obtained from the CDC.31 An 88% rate ofeturn and linkage to care for infected patients came fromhe demonstration program;29 this figure is consistent withublished reports of nonreturn rates varying from 10% to7%.32–34 Therefore, our base case overall index of partic-pation was as follows: (offer/accept) � (return/linkage toare) � (0.37) � (0.88) � 0.33. We considered alternative

alues ranging from 0.01 to 1.00 in sensitivity analyses. p

reatment costs

Costs for antiretroviral therapy and opportunistic infec-ion treatment and prophylaxis were provided by the 2001ed Book.42 Laboratory test costs were taken from theedicare Fee Schedule.43 The AIDS Costs and Servicestilization Survey provided charges for treatment of oppor-

unistic infections and routine HIV- and acquired immuno-eficiency syndrome (AIDS)-related care that were con-erted to economic costs using a national cost-to-chargeatio for HIV/AIDS.49,50

nalytic framework

The analysis was conducted from the societal perspectiven accordance with the recommendations of the Panel onost-Effectiveness in Health and Medicine using a discount

ate of 3%.51 All costs are presented in 2001 U.S. dollars.he model yields results denominated in costs, life expect-ncy, and quality-adjusted life expectancy; cost-effective-ess ratios of the screening strategies are reported incre-entally to the no screening strategy in dollars per quality-

djusted life-year (QALY) gained.

esults

ase case

A routine inpatient HIV screening program increasedrojected, discounted life expectancy from 5602.56 to215.15 QALYs per 1000 HIV-infected patients, or approx-mately 7.35 quality-adjusted life-months per infected per-on (Table 2). At a 37% test acceptance rate, screening of000 uninfected patients on average cost $19,800, or ap-roximately $20 per uninfected person. By advancing theime of identification with screening, the mean CD4 cellount at detection was increased from 196 to 244 cells/mm3.t an undetected HIV prevalence of 1% and a pretest

ounseling and test cost of $53, quality-adjusted life expect-ncy per 1000 inpatients increased by 6.13 QALYs and costn additional $216 600 per 1000 persons, yielding a cost-ffectiveness ratio of $35 400 per QALY gained (Table 3).ost-effectiveness ratios for the screening program re-ained favorable: $64 500 per QALY gained, even at an

ndetected HIV prevalence of 0.1%. At very high preva-ences of HIV infection (10%), these ratios plateaued atpproximately $32 400 per QALY gained.

echanisms of detectionWithout a screening program, 53% of patients were iden-

ified after presenting with an opportunistic infection, com-ared with 35% with a screening program. The screening

rogram identified 32% of infected patients.

S

P

in$an$r

H

ttHcttghi

I

iclpHttl0m1

L

gasleHii

296 The American Journal of Medicine, Vol 118, No 3, March 2005

ensitivity analyses

rogram/testing costsWhen the cost of testing (counseling and test costs)

ncreased from $53 to $103 per patient, the cost-effective-ess ratio for an inpatient screening program increased to38 400 per QALY gained at a 1% undiagnosed HIV prev-lence and to $94 900 per QALY gained at a 0.1% undiag-osed HIV prevalence (Figure 1). At a lower testing cost of28 and an HIV prevalence of 0.1%, the cost-effectivenessatio for inpatient testing was $49 400 per QALY gained.

IV-infected CD4 cell countTo understand the differences in CD4 cell count at de-

ection with and without a screening program, we examinedhe mean CD4 cell count estimate of undetected chronicallyIV-infected inpatients (base case [� SD]: 320 � 260

ells/mm3). The difference in mean CD4 cell count at theime of detection increased as the mean count in the unde-ected HIV-infected inpatients increased (Figure 2). Overall,reater screening program benefits were realized (withigher cost-effectiveness ratios) when the undetected HIV-nfected patients had higher CD4 cell counts.

Table 2 Costs and life expectancy per 1000 persons by infect

HIV-infected person, not screenedHIV-infected person, screenedHIV-uninfected person, not screenedHIV-uninfected person, screened

HIV � human immunodeficiency virus.

Table 3 Life expectancy, costs, and cost-effectiveness ratiosinfection

HIV Prevalence

TotalCosts/1000Persons ($)

0.01%No screen 10 300Screen 32 100

0.1%No screen 103 100Screen 142 700

1% (base case)No screen 1 031 400Screen 1 248 000

10%No screen 10 313 900Screen 12 301 600

HIV � human immunodeficiency virus.

ndex of participationWe varied the probabilities of each component of the

ndex of participation (offer/accept and return/linkage toare) from 0% to 100% in 10% increments at HIV preva-ences of 0.1% and 1% (Table 4). Programs with increasingarticipation rates and higher prevalences of undiagnosedIV infection were more attractive, with lower cost-effec-

iveness ratios. Even in a program with poor rates of par-icipation, in which both test offer/acceptance and return/inkage to care were 20% (index of participation � 0.20 �.20 � 0.04), the program’s cost-effectiveness ratio re-ained below $50 000 per QALY at an HIV prevalence of

%.

ead- and length-time biasBy model design, lead-time bias of the screening pro-

ram was not a factor. We confirmed this in a series ofnalyses whereby neither the screened group nor the un-creened group received any HIV care; the quality-adjustedife expectancy for the two groups was identical. We alsoxamined length-time bias by examining the effect for eachIV RNA set point. At an HIV prevalence of 1%, if all

nfected patients had an HIV RNA set point of �500 cop-es/mL (long length-time bias), then the life expectancy for

d screening status

l Costs/1000ons ($)

Quality-Adjusted LifeExpectancy/1000Persons (years)

139 000 5602.56837 800 6215.15

0 17 120.0119 800 17 120.01

tine HIV screening at varying prevalences of undetected HIV

Quality-Adjusted LifeExpectancy/1000Persons (years)

Cost-effectiveness($/QALY)

17 118.8517 118.91 356 000

17 108.4917 109.10 64 500

17 004.8317 010.96 35 400

15 968.2616 029.52 32 400

ion an

TotaPers

103122

for rou

1QsQp(i5iyg

O

aCPscnispcAsnQatc$at

D

Tr

hiserlidt

rpHNpfgscpsvwr

eosct1o6vpcp

FtcwhTae

Fmlva

297Walensky et al Routine HIV Testing

000 screened HIV-infected patients would be 6870.40ALYs, compared with 6378.85 QALYs per 1000 un-

creened patients (cost-effectiveness ratio � $39 500 perALY gained). At the opposite extreme, if all infectedatients had an HIV RNA set point of �30 000 copies/mLshort length-time bias), the life expectancy per 1000 HIV-nfected patients would be 5758.23 QALYs (screen) versus084.33 QALYs (no screen). In this more rapidly progress-ng cohort, the screening program was more cost-effective,ielding a cost-effectiveness ratio of $33 400 per QALYained.

ther sensitivity analyses

In other sensitivity analyses, we explored the efficacynd the cost of antiretroviral therapy as well as alternativeD4 cell counts for the initiation of antiretroviral therapy.revalence thresholds for efficient HIV screening remainedtable. While cost-effectiveness ratios fluctuated with theost of antiretroviral therapy, overall policy conclusions didot change. For example, at an HIV prevalence of 1%,ncreasing antiretroviral costs by 50% increased the HIVcreening cost-effectiveness ratio from $35 400 to $45 800er QALY; decreasing antiretroviral costs by 50% de-reased the cost-effectiveness ratio to $24 900 per QALY.nalyses examining access to antiretroviral therapy demon-

trated that at an HIV prevalence of 1%, the cost-effective-ess of HIV screening programs exceeded $50 000 perALY only if fewer than 45% of patients had access to

ntiretroviral therapy. Even at a prevalence of 0.1%, accesso antiretroviral therapy needed to be less than 54% for theost-effectiveness ratio of HIV screening to exceed100 000 per QALY. Alternative linkage to care costs hadsmaller overall effect on cost-effectiveness ratios than did

esting costs.

iscussion

he 2001 CDC guidelines for HIV counseling, testing, and

$0

$50,000

$100,000

$150,000

$200,000

$250,000

$300,000

$350,000

$400,000

0.01 0.1 1

HIV Prevalence (%)

Cos

t-ef

fect

iven

ess

($/Q

ALY

)

$103

$53 (base case)

$28

Cost of counseling and testing

igure 1 Sensitivity analysis examining the cost of human im-unodeficiency virus (HIV) counseling and testing. At HIV preva-

ences of �1%, the lines reflecting cost-effectiveness ratios con-erge. Only at prevalences of �0.1% do the costs of counselingnd testing drive the cost-effectiveness ratio.

eferral recommend routine screening of all inpatients in a

ospitals with an HIV prevalence of �1%.3 This thresholds largely based on a single-blinded HIV seroprevalencetudy of 20 acute care U.S. hospitals conducted by Janssent al. in 1992.4 In that study, hospitals with AIDS diagnosisates of �1/1000 discharges correlated to HIV seropreva-ence rates of �1%. The authors estimated that a routinenpatient HIV screening program in similar hospitals wouldetect nearly a quarter of a million HIV-infected, asymp-omatic patients.

Since that landmark study, several important events haveeshaped the HIV epidemic: effective antiretroviral thera-y,21,52 a revised of definition of AIDS,53 and mandates forIV (rather than only AIDS) reporting in many states.54

either the Janssen study nor the current guidelines incor-orate the cost of a testing program in the recommendationsor routine testing.3,4 It is these costs, however, that areenerally cited as barriers to HIV counseling and testingervices.13 Although cost-effectiveness assessments haveonsidered HIV testing in other venues (eg, prenatal andremarital settings), these analyses have not addressedcreening as understood in the current guidelines.3,55,56 Pre-ious cost-effectiveness analyses in the inpatient settingere conducted prior to the advent of highly active antiret-

oviral therapy.57,58

Intentionally mirroring the CDC’s recommendations, wevaluated the clinical benefits, costs, and cost-effectivenessf a routine HIV screening program under alternative as-umptions about HIV prevalence, testing costs, CD4 cellount of the undiagnosed patients, and program participa-ion. We found that at an undiagnosed HIV prevalence of%, routine inpatient HIV screening programs would notnly increase mean HIV quality-adjusted life expectancy by.13 years per 1000 persons, but would also offer goodalue for money, with a cost-effectiveness ratio of $35 400er QALY gained.51 This program compares favorably withost-effectiveness estimates for routine, standard screeningrograms in other chronic diseases,59–61 including type 2

63

57

51

45

4034

2924

19

0

50

100

150

200

250

300

350

400

50 100 150 200 250 300 350 400 450 500

Mean CD4 among Unscreened, Undiagnosed HIV-Infected Inpatients (cells/mm3)

aMea

n C

D4

at D

etec

tion

(cel

ls/m

m3 )a

a

Routine Screening

No Routine Screening

igure 2 Sensitivity analysis demonstrating the effects of rou-ine screening at various potential values for the mean CD4 cellount in a group of hospitalized patients. This mean CD4 valueill likely vary based on the severity and duration of undiagnoseduman immunodeficiency virus (HIV) infection in such patients.he vertical distance between the screening program (triangles)nd no screening program (squares) represents the opportunity forarlier intervention, referral to HIV care, and antiretroviral ther-

py.

d(Qre

aidaeptdcamchsmfiato

wpewss(bs

ih

Istoe

aopaptaqattmtramasd

acrplwdc

lt

298 The American Journal of Medicine, Vol 118, No 3, March 2005

iabetes ($70 000 per QALY gained), hypertension$80 400 per QALY gained), and colon cancer ($57 700 perALY gained). Even at an HIV prevalence as low as 0.1%,

outine HIV screening programs would likely remain cost-ffective ($64 500 per QALY).

Central to the HIV screening issue is who will pay for thedded expenditure. Thus far, testing recommendations andnitiatives remain unfunded mandates. If studies continue toemonstrate the prevalence of undiagnosed HIV infection29

nd the cost-effectiveness of HIV screening from the soci-tal perspective, payers might be encouraged to cover thisractice, either separately or as part of the basis for paymento hospitals. Those who currently fund guideline-concor-ant care, such as private insurers, Medicare, and Medicaid,ould include routine, inpatient HIV testing as a reimburs-ble point of care.62 Barring coverage changes, hospitalsay elect to either absorb the cost or cut back on other less

ost-effective programs. Such a commitment would alsoave to recognize the personnel required to offer HIV coun-eling and testing to all inpatients. Regardless of who isost appropriate to pay for screening, this analysis quanti-es the costs of routine HIV testing and of achieving itsssociated life expectancy benefits and demonstrates thathe cost-effectiveness of HIV screening falls below that ofther funded chronic disease screening programs.59–61

Because data on test acceptance and linkage are limited,e conducted extensive sensitivity analysis with regard toarticipation behavior. Our results demonstrate that movingven a small number of HIV-infected patients into care,ith the associated increases in survival and costs, has the

ingle largest effect on the cost-effectiveness ratio of acreening program. Only at very low HIV prevalences�0.1%) do high program costs and low participation ratesegin to alter this HIV-care driven cost-effectiveness ratioubstantially.

Why is screening for HIV so cost-effective? HIV testings inexpensive, accurate, and identifies a disease for which

Table 4 Cost-effectiveness of routine inpatient HIV screening

Probability of Test Offer andAcceptance

Probability of Returnfor Results andLinkage to Care

Index of participation � 0.040.10 0.400.20 0.200.40 0.10

Index of participation � 0.400.40 1.000.50 0.800.80 0.501.00 0.40

Index of participation � 0.900.90 1.001.00 0.90

HIV � human immunodeficiency virus; QALY � quality-adjusted life

ighly effective (although expensive) treatment is available. n

nfected patients have years of life to gain from viral loaduppression and the prevention of opportunistic infec-ions.52 Therefore, even in low-prevalence settings, theverall cost of screening is effectively spread over years ofxtended life for those who are infected.

There are important limitations to this study. First, thenalysis did not account for HIV screening benefits in termsf secondary infections averted. If later infections may berevented by HIV counseling and case identification, thisnalysis would underestimate the full benefit of a screeningrogram.63–65 The additional benefit of simplified occupa-ional health protocols related to needle stick injuries waslso not considered. Second, although the model includeduality-of-life estimates for health states, the short-termnxiety and fear over the several days that the patient awaitsest results is difficult to capture adequately when looking atotal life years as the clinical endpoint. From a program-atic standpoint, the CDC guidelines do not discuss prac-

ical issues of implementation, such as hospital readmissionates, HIV incidence among patients who are readmitted,nd the optimal frequency of repeat testing.3 Intentionallyodeling our analysis after these guidelines, we also did not

ddress these issues. However, in other settings, universalcreening programs have resulted in short-term decreases inisease incidence.66,67

Given that reported efforts of routine HIV screeningmong inpatients and outpatients have all documented suc-ess at HIV case identification, this analysis suggests thatesources may be well utilized by initiating HIV screeningrograms in hospitals with a diagnosed seroprevalence of asow as 0.1%.6,8–11 These programs should be establishedith frequent assessments of their yield in terms of newiagnoses of HIV infection, cost, and cost-effectiveness toonfirm their continued value.

Although HIV counseling, testing, and referral guide-ines have been published in the United States since 1993,hey are seldom implemented; a resulting 300 000 patients

am at alternative levels of program participation

Cost-effectiveness ($/QALY)

HIV Prevalence � 0.1% HIV Prevalence � 1%

106 000 39 800171 700 46 400316 600 60 400

60 600 34 90067 100 35 60088 200 37 700

103 200 39 200

60 300 34 90063 600 35 200

progr

-year.

ationwide remain HIV infected and undiagnosed.1–3 Be-

ffvttptnts

R

1

1

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

2

3

3

3

3

3

3

3

3

299Walensky et al Routine HIV Testing

ore dismissing expanded HIV counseling, testing, and re-erral programs as unaffordable, we should consider thealue of such efforts. We have demonstrated that an inpa-ient routine HIV screening program is not only cost-effec-ive, but would likely remain so at an undiagnosed HIVrevalence 10 times lower than the currently recommendedhreshold. With a renewed CDC emphasis on HIV testing,ationwide action should focus on appropriate implementa-ion of these guidelines in an effort to offer effective, life-ustaining care to those who are infected.68

eferences

1. Fleming PL, Wortley PM, Karon JM, DeCock KM, Janssen RS.Tracking the HIV epidemic: current issues, future challenges. Am JPublic Health. 2000;90:1037–1041.

2. CDC. Recommendations for HIV testing services for inpatients andoutpatients in acute-care hospital settings. MMWR Recomm Rep. 1993;42:1–6.

3. CDC. Revised guidelines for HIV counseling, testing, and referral.MMWR Recomm Rep. 2001;50:1–57.

4. Janssen RS, St Louis ME, Satten GA, et al. HIV infection amongpatients in U.S. acute care hospitals. Strategies for the counseling andtesting of the hospital patients. The Hospital HIV Surveillance Group.N Engl J Med. 1992;327:445–452.

5. St Louis ME, Rauch KJ, Petersen LR, Anderson JE, Schable CA,Dondero TJ. Seroprevalence rates of human immunodeficiency virusinfection at sentinel hospitals in the United States. The Sentinel Hos-pital Surveillance Group. N Engl J Med. 1990;323:213–218.

6. Goggin MA, Davidson AJ, Cantril SV, O’Keefe LK, Douglas JM. Theextent of undiagnosed HIV infection among emergency departmentpatients: results of a blinded seroprevalence survey and a pilot HIVtesting program. J Emerg Med. 2000;19:13–19.

7. Rhee KJ, Albertson TE, Kizer KW, Burns MJ, Hughes MJ, AscherMS. A comparison of HIV-1, HBV, and HTLV-I/II seroprevalencerates of injured patients admitted through California emergency de-partments. Ann Emerg Med. 1992;21:397–401.

8. Walensky RP, Losina E, Steger-Craven KA, Freedberg KA. Identify-ing undiagnosed human immunodeficiency virus: the yield of routine,voluntary inpatient testing. Arch Intern Med. 2002;162:887–892.

9. Pincus JM, Crosby SS, Losina E, King ER, LaBelle C, Freedberg KA.Acute human immunodeficiency virus infection in patients presentingto an urban urgent care center. Clin Infect Dis. 2003;37:1699–1704.

0. del Rio C, Paredas CF, Duffus W, et al. Routinely recommended HIVtesting at an urban urgent-care clinic—Atlanta, Georgia, 2000. MMWRMorb Mortal Wkly Rep. 2001;50:538–541.

1. Kelen GD, Shahan JB, Quinn TC. Emergency department-based HIVscreening and counseling: experience with rapid and standard sero-logic testing. Ann Emerg Med. 1999;33:147–155.

2. Glick NR, Silva A, Zun L, Whitman S. HIV testing in a resource-poorurban emergency department. AIDS Educ Prev. 2004;16:126–136.

3. Irwin KL, Valdiserri RO, Holmberg SD. The acceptability of volun-tary HIV antibody testing in the United States: a decade of lessonslearned. AIDS. 1996;10:1707–1717.

4. Freedberg KA, Losina E, Weinstein MC, et al. The cost effectivenessof combination antiretroviral therapy for HIV disease. N Engl J Med.2001;344:824–831.

5. Freedberg KA, Scharfstein JA, Seage GR III, et al. The cost-effec-tiveness of preventing AIDS-related opportunistic infections. JAMA.1998;279:130–136.

6. Weinstein MC, Goldie SJ, Losina E, et al. Use of genotypic resistancetesting to guide HIV therapy: clinical impact and cost-effectiveness.

Ann Intern Med. 2001;134:440–450.

7. Multicenter AIDS Cohort Study (MACS) Public Dataset: Release PO4.Springfield, Virginia: National Technical Information Service; 1995.

8. Enger C, Graham N, Peng Y, et al. Survival from early, intermediate,and late stages of HIV infection. JAMA. 1996;275:1329–1334.

9. Mellors JW, Munoz A, Giorgi JV, et al. Plasma viral load and CD4�lymphocytes as prognostic markers of HIV-1 infection. Ann InternMed. 1997;126:946–954.

0. Samet JH, Freedberg KA, Savetsky JB, Sullivan LM, Stein MD.Understanding delay to medical care for HIV infection: the long-termnon-presenter. AIDS. 2001;15:77–85.

1. Guidelines for the Use of Antiretroviral Agents in HIV-Infected Adultsand Adolescents. November 10, 2003. The Panel on Clinical Practicesfor Treatment of HIV Infection convened by the Department of Healthand Human Services (DHHS) and the Henry J. Kaiser Family Foun-dation. Available at http://aidsinfo.nih.gov.

2. Carcelain G, Blanc C, Leibowitch J, et al. T cell changes after com-bined nucleoside analogue therapy in HIV primary infection. AIDS.1999;13:1077–1081.

3. Kaufmann GR, Zaunders JJ, Cunningham P, et al. Rapid restoration ofCD4 T cell subsets in subjects receiving antiretroviral therapy duringprimary HIV-1 infection. AIDS. 2000;14:2643–2651.

4. Malhotra U, Berrey MM, Huang Y, et al. Effect of combinationantiretroviral therapy on T-cell immunity in acute human immunode-ficiency virus type 1 infection. J Infect Dis. 2000;181:121–131.

5. Walensky RP, Goldie SJ, Sax PE, et al. Treatment for primary HIVinfection: projecting outcomes of immediate, interrupted, or delayedtherapy. J Acquir Immune Defic Syndr. 2002;31:27–37.

6. Dybul M, Bolan R, Condoluci D, et al. Evaluation of initial CD4� Tcell counts in individuals with newly diagnosed human immunodefi-ciency virus infection, by sex and race, in urban settings. J Infect Dis.2002;185:1818–1821.

7. Sackoff JE, Shin SS. Trends in immunologic and clinical status ofnewly diagnosed HIV-positive patients initiating care in the HAARTera. J Acquir Immune Defic Syndr. 2001;28:270–272.

8. Roome A, Hadler J, Birkhead G, et al. HIV testing among pregnantwomen—United States and Canada, 1998-2001. MMWR Morb MortalWkly Rep. 2002;51:1013–1016.

9. Walensky RP, Losina E, Skolnik PR, et al. Voluntary HIV testing aspart of routine medical care—Massachusetts, 2002. MMWR MorbMortal Wkly Rep. 2004;53:523–526.

0. Stringer EM, Stringer JS, Cliver SP, Goldenberg RL, Goepfert AR.Evaluation of a new testing policy for human immunodeficiency virusto improve screening rates. MMWR Morb Mortal Wkly Rep. 2001;50:1054–1058.

1. Update: HIV counseling and testing using rapid tests—United States,1995. MMWR Morb Mortal Wkly Rep. 1998;47:211–215.

2. CDC. HIV Counseling and Testing in Publicly Funded Sites: 1995Summary Report. Atlanta, Georgia: U.S. Department of Health andHuman Services 1997.

3. Moore M, Gerber AR, Campbell CH Jr, Dillon BA, West GR, Vald-iserri RO. A study of clients returning for counseling after HIV testing:implications for improving rates of return. Pub Health Rep. 1993;108:12–18.

4. Bell RA, Truax SR, Ruiz JD, Sun RK, Molitor F. Predictors of failureto return for HIV test result and counseling by test site type. AIDSEduc Prev. 1999;11:1–13.

5. Staszewski S, Morales-Ramirez J, Tashima KT, et al. Efavirenz pluszidovudine and lamivudine, efavirenz plus indinavir, and indinavirplus zidovudine and lamivudine in the treatment of HIV-1 infection inadults. Study 006 Team. N Engl J Med. 1999;341:1865–1873.

6. Hammer SM, Squires KE, Hughes MD, et al. A controlled trial of twonucleoside analogues plus indinavir in persons with human immuno-deficiency virus infection and CD4 cell counts of 200 per cubicmillimeter or less. AIDS Clinical Trials Group 320 Study Team.N Engl J Med. 1997;337:725–733.

7. Baxter JD, Mayers DL, Wentworth DN, et al. A randomized study ofantiretroviral management based on plasma genotypic antiretroviral

resistance testing in patients failing therapy. CPCRA 046 Study Team

3

3

4

4

4

4

4

4

4

4

4

4

5

5

5

5

55

5

5

5

5

6

6

6

6

6

6

6

6

6

300 The American Journal of Medicine, Vol 118, No 3, March 2005

for the Terry Beirn Community Programs for Clinical Research onAIDS. AIDS. 2000;14:F83–F93.

8. Mylonakis E. Laboratory testing for infection with the human immu-nodeficiency virus: established and novel approaches. Am J Med.2000;109:568–576.

9. Varghese B, Peterman TA, Holtgrave DR. Cost-effectiveness of coun-seling and testing and partner notification: a decision analysis. AIDS.1999;13:1745–1751.

0. Farnham PG, Gorsky RD, Holtgrave DR, Jones WK, Guinan ME.Counseling and testing for HIV prevention: costs, effects, and cost-effectiveness of more rapid screening tests. Public Health Rep. 1996;111:44–53.

1. Ekwueme DU, Pinkerton SD, Holtgrave DR, Branson BM. Cost com-parison of three HIV counseling and testing technologies. Am J PrevMed. 2003;25:112–121.

2. Drug Topics Red Book. Montvale, New Jersey: Medical Economics;2001.

3. Clinical Diagnostic Laboratory Fee Schedule (CLAB), 2001. HealthCare Financing Administration (HCFA) of the Department of Healthand Human Services (DHHS). Available at: www.cms.hhs.gov/providers/pufdownload/default.asp#labfree. Accessed September 22,2003.

4. Kaplan JE, Masur H, Holmes KK. Guidelines for preventing oppor-tunistic infections among HIV-infected persons—2002. Recommen-dations of the U.S. Public Health Service and the Infectious DiseasesSociety of America. MMWR Recomm Rep. 2002;51:1–52.

5. Statistical Abstract of the United States: 1998. 118th ed. Washington,D.C.: Bureau of the Census 1998.

6. Statistics for U.S. Hospital Stays, Diagnosis Related Groups (DRGs),Healthcare Cost and Utilization Project (HCUP) Nationwide InpatientSample (NIS). Agency for Healthcare Research and Quality. Rock-ville, Maryland. Available at: http://hcup.ahrq.gov/HCUPnet.asp. Ac-cessed December 23, 2003.

7. Daar ES, Little S, Pitt J, et al. Diagnosis of primary HIV-1 infection.Los Angeles County Primary HIV Infection Recruitment Network.Ann Intern Med. 2001;134:25–29.

8. Hecht FM, Busch MP, Rawal B, et al. Use of laboratory tests andclinical symptoms for identification of primary HIV infection. AIDS.2002;16:1119–1129.

9. AIDS Cost and Services Utilization Survey: Public Use Tapes 4 and 5.No. PB94189891. Springfield, Virginia: National Technical Informa-tion Service; 1994.

0. Finkler SA. The distinction between cost and charges. Ann Intern Med.1982;96:102–109.

1. Gold MR, Siegel JE, Russell LB, Weinstein MC. Cost Effectiveness inHealth and Medicine. New York, New York: Oxford University Press;

1996.

2. Palella FJ Jr, Delaney KM, Moorman AC, et al. Declining morbidityand mortality among patients with advanced human immunodeficiencyvirus infection. N Engl J Med. 1998;338:853–860.

3. 1993 Revised classification system for HIV infection and expandedsurveillance case definition for AIDS among adolescents and adults.MMWR Recomm Rep. 1992;41:961–962.

4. CDC:HIV/AIDS. Surv Rep. 1999;11:1–44.5. Immergluck LC, Cull WL, Schwartz A, Elstein AS. Cost-effectiveness

of universal compared with voluntary screening for human immuno-deficiency virus among pregnant women in Chicago. Pediatrics. 2000;105:E54.

6. McKay NL, Phillips KM. An economic evaluation of mandatorypremarital testing for HIV. Inquiry. 1991;28:236–248.

7. Owens DK, Nease RF Jr, Harris RA. Cost-effectiveness of HIVscreening in acute care settings. Arch Intern Med. 1996;156:394–404.

8. Lurie P, Avins AL, Phillips KA, Kahn JG, Lowe RA, Ciccarone D.The cost-effectiveness of voluntary counseling and testing of hospitalinpatients for HIV infection. JAMA. 1994;272:1832–1838.

9. The cost-effectiveness of screening for type 2 diabetes. CDC DiabetesCost-Effectiveness Study Group, Centers for Disease Control andPrevention JAMA. 1998;280:1757–1763.

0. Littenberg B, Garber AM, Sox HC Jr. Screening for hypertension. AnnIntern Med. 1990;112:192–202.

1. Frazier AL, Colditz GA, Fuchs CS, Kuntz KM. Cost-effectiveness ofscreening for colorectal cancer in the general population. JAMA. 2000;284:1954–1961.

2. U.S. Preventive Services Task Force: Guide to Clinical PreventiveServices. 2nd ed. Baltimore, Maryland: Williams and Wilkins, 1996.

3. The NIMH Multisite HIV Prevention Trial: reducing HIV sexual riskbehavior. The National Institute of Mental Health (NIMH) MultisiteHIV Prevention Trial Group. Science. 1998;280:1889–1894.

4. Kamb ML, Fishbein M, Douglas JM Jr, et al. Efficacy of risk-reductioncounseling to prevent human immunodeficiency virus and sexuallytransmitted diseases: a randomized controlled trial. Project RESPECTStudy Group. JAMA. 1998;280:1161–1167.

5. Gibson D. Effectiveness of brief counseling in reducing HIV riskbehavior in injecting drug users: final results of randomized trials ofcounseling with and without HIV testing. AIDS Behav. 1999;3:3–12.

6. Morrison AS. Screening in Chronic Disease. 2nd ed. New York, NewYork: Oxford University Press; 1992.

7. Cancer Facts and Figures, 2002. Vol. 2003. Atlanta, Georgia: Amer-ican Cancer Society, 2002.

8. Advancing HIV prevention: new strategies for a changing epidemic—United States, 2003. MMWR Morb Mortal Wkly Rep 2003;52:329–

332.