The Cost-Effectiveness of Human Immunodeficiency Virus Testing and Treatment Engagement Initiatives in British Columbia, Canada: 2011–2013

We used a dynamic human immunodeficiency virus (HIV) transmission model to determine the cost-effectiveness of HIV care interventions in British Columbia, Canada, in 2011–2013. HIV testing and treatment initiation interventions were cost-effective, while the treatment retention intervention was not.

Abstract

Background

Recognition of the secondary preventive benefits of antiretroviral therapy (ART) has mobilized global efforts to “seek, test, treat, and retain” people living with human immunodeficiency virus [HIV]/AIDS (PLHIV) in HIV care. We aimed to determine the cost-effectiveness of a set of HIV testing and treatment engagement interventions initiated in British Columbia, Canada, in 2011–2013.

Methods

Using a previously validated dynamic HIV transmission model, linked individual-level health administrative data for PLHIV, and aggregate-level HIV testing data, we estimated the cost-effectiveness of primary care testing (hospital, emergency department [ED], outpatient), ART initiation, and ART retention initiatives vs a counterfactual scenario that approximated the status quo. HIV incidence, mortality, costs (in 2015$CDN), quality-adjusted life years (QALYs), and incremental cost-effectiveness ratios were estimated. Analyses were executed over 5- to 25-year time horizons from a government-payer perspective.

Results

ED testing was the best value at $30216 per QALY gained and had the greatest impact on incidence and mortality among PLHIV, while ART initiation provided the greatest QALY gains. The ART retention initiative was not cost-effective. Delivered in combination at the observed scale and sustained throughout the study period, we estimated a 12.8% reduction in cumulative HIV incidence and a 4.7% reduction in deaths among PLHIV at $55258 per QALY gained. Results were most sensitive to uncertainty in the number of undiagnosed PLHIV.

Conclusions

HIV testing and ART initiation interventions were cost-effective, while the ART retention intervention was not. Developing strategies to reengage PLHIV lost to care is a priority moving forward.

Following recent scientific findings on the secondary preventive benefits of combination antiretroviral treatment (ART) in observational [1–4] and experimental studies [5, 6], as well as confirmed individual health benefits from early ART initiation [7], international efforts are now focused on identifying and implementing evidence-based interventions to maximize population-level diagnosis, ART access, and, ultimately, viral suppression to address the global human immunodeficiency virus (HIV)/AIDS pandemic [8]. Substantial, coordinated efforts will be required to identify, implement, and evaluate effective interventions to treat and prevent HIV in order to achieve the goals of the ambitious United Nations 90-90-90 target, which calls for 90% of those living with HIV to be diagnosed, 90% of those diagnosed to access ART, and 90% of those on ART to reach viral suppression by 2020 [8]. While HIV testing to identify and reduce the size of the undiagnosed HIV-positive population remains critical [9], engaging and retaining people living with human immunodeficiency virus (PLHIV) on ART poses perhaps the greatest challenge to achieving the UN 90-90-90 target [10]. A range of interventions comprising a combination implementation strategy [11], accounting for the needs of specific demographic groups, and carefully considering the most practical means and opportune settings to intervene will be necessary in diverse settings globally.

On this basis, the British Columbia Ministry of Health launched the Seek and Treat for Optimal Prevention of HIV/AIDS (STOP HIV/AIDS) pilot program in 2 regional health authorities (Vancouver Coastal and Northern) in British Columbia (BC) in 2010. The pilot provided increased funding for HIV testing and treatment [12] and targeted funds for public health intervention and entailed a commitment to ongoing monitoring and evaluation, aided by comprehensive linked administrative databases [13, 14], to ascertain the value of implemented interventions and guide future investment decisions.

To establish an effective HIV response with long-term sustainability, there is an urgent need for efficiency in allocating scarce resources for the variety of public health interventions available in HIV/AIDS care [11]. Interventions to improve the HIV care cascade at its various stages may vary substantially in their incremental value. Further, the impact of individual interventions may be enhanced or diminished when delivered in combination with others (ie, their effects may not simply be additive), and this impact can change over time [15]. Prioritizing interventions for implementation, scale-up, or disinvestment on the basis of the incremental health benefits they provide per dollar invested provides clear and objective criteria to inform resource allocation decisions.

Our objective was to determine the cost-effectiveness of a set of HIV testing and treatment engagement interventions initiated in BC, as part of the STOP HIV/AIDS pilot program, which ran from 2011 to 2013. We drew on individual-level data and published information on the scale and effect of a selection of interventions executed in BC as part of the STOP HIV/AIDS pilot and satisfied our objective using a validated dynamic HIV transmission model built using population-level linked health administrative data.

METHODS

Model Description

We adapted and extended an existing deterministic transmission model previously used to estimate the health benefits and costs of expanded HIV screening and treatment in the United States [16] and BC [17]. The adult population of BC aged 15–64 years were partitioned into compartments on the basis of HIV risk behavior (men who have sex with men [MSM], people who inject drugs [PWID], MSM-PWID, and heterosexuals), screening status, and HIV infection status. Among those HIV infected, individuals were further classified as infected, diagnosed, on ART, and off-therapy and partitioned according to CD4 cell count (CD4≥500 µL, 350–499 µL, 200–349 µL, <200 µL). Disease progression was differentiated among those on ART [18] and not on ART [19] and estimated as a function of CD4 count, stratified into the 4 categories noted above. Health state transitions occurred at monthly intervals (Table 1).

Table 1.

Input Parameters for the Dynamic Compartmental Human Immunodeficiency Virus Transmission Model

Description	Value	Source
Number of individuals in susceptible compartment i at time t (initial [1996]a value)	2.64M	[14, 20, 21, 22]
Number of individuals in HIV-infected (undiagnosed) compartment i at time t (initial value)	3694	[14, 22, 23]
Number of individuals in HIV-diagnosed compartment i at time t (initial value)	3844	[14, 22, 23]
Number of individuals in HIV treatment compartment i at time t (initial value)	0	[14, 22, 23]
Number of individuals in HIV off-treatment compartment i at time t (initial value)	0	[14, 22, 23]
Monthly entry rate of individuals into compartment i	Time-varying	[20]
Average duration uninfected individuals in compartment i remain “identified” after screening	12 months	Assumption
Monthly HIV screening rate for individuals in compartment i	Figure 1	[1]
High risk (PWID, MSM/PWID) (multiplier)	2.6	[24]
Total sufficient contact rate	Time-varying	[3–6, 16, 21, 25, 26, 27–56]
Mortality and maturation rate for individuals in compartment i
Monthly mortality rate: heterosexual (susceptible)	0.00027	[57]
Monthly mortality rate: MSM (susceptible)	0.00034	[57]
Monthly mortality rate: PWID (susceptible)	0.00238	[27]
Monthly mortality rate: MSM/PWID (susceptible)	0.00246	[27]
Monthly mortality rate: heterosexual (infected/diagnosed)
CD4: ≥500	0.00079	[58]
CD4: 350–499	0.00079	[58]
CD4: 200–349	0.00136	[59]
CD4: <200	0.00853	[18]
Monthly mortality rate: MSM (infected/diagnosed)
CD4: ≥500	0.00045	[28]
CD4: 350–499	0.00045	[58]
CD4: 200–349	0.00175	[59]
CD4: <200	0.00853	[18]
Monthly mortality rate: PWID, including MSM/PWID (infected/diagnosed)
CD4: ≥500	0.00250	[58]
CD4: 350–499	0.00250	[58]
CD4: 200–349	0.00357	[59]
CD4: <200	0.00856	[18]
Monthly mortality rate: on ART (×1000–1)	Time-varying	[14, 22, 23]
Non-PWID
CD4: ≥500	0.474 (1996); 0.399 (2010)
CD4: 350–499	0.608 (1996); 0.541 (2010)
CD4: 200–349	1.167 (1996); 1.043 (2010)
CD4: <200	8.259 (1996); 7.883 (2010)
PWID
CD4: ≥500	0.973 (1996); 0.833 (2010)
CD4: 350–499	1.181 (1996); 1.078 (2010)
CD4: 200–349	2.557 (1996); 2.354 (2010)
CD4: <200	8.288 (1996); 8.021 (2010)
Monthly maturation rate	0.00125	[20]
Symptom-based monthly case finding rate for infected individuals in compartment i
Low risk (CD4: 200–349)	0.00874	[60]
High risk (CD4: <200)	0.01842	[60]
HIV disease progression rate for individuals not on ART		[19, 60–62]
CD4: ≥500 to CD4: 350–499	0.02209
CD4: 350–499 to CD4: 200–349	0.02209
CD4: 200–349 to CD4 <200	0.02209
CD4: <200 to death	0.00250
HIV disease progression rate for individuals on ART in compartment i	Supplementary Materials, Figure A6	[14, 22, 23]
Rate of ART discontinuation	Figure 2, panels E and F	[14, 22, 23]
Probability of direct ART initiations immediately after diagnosis	Supplementary Materials, Figure A4	[14, 22, 23]
Rate of individuals from diagnosed compartment i initiating antiretroviral treatment	Figure 2, panels A and B	[14, 22, 23]
Rate of ART reinitiation among those who had discontinued treatment	Figure 2, Panels C and D	[14, 22, 23]
Indicator variable for eligibility for HIV treatment initiation	Time-varying	[63, 64, 65–71]
1996–1997	CD4 <500
1998–2001	CD4 <350
2002–2003	CD4 <200
2004–2007	CD4 <350
2008–2010	CD4 <500
Post-2011	All CD4
HIV transmission parameter
Needle-sharing parameters
Number of injections (monthly) at baseline	19.5	[21, 26]
Reduced injections due to opioid agonist treatment	0.75	[27, 72]
Probability of shared injection at baseline	0.21	[21, 28]
Probability of transmission per shared injection: CD4 >500	0.002	[27, 29, 30]
Probability of transmission per shared injection: CD4: 350–499	0.002
Probability of transmission per shared injection: CD4: 200–349	0.003
Probability of transmission per shared injection: CD4 <200	0.003
Reduced probability of transmission on ART	0.9	[4]
Homosexual sex parameters
No sexual partners (annual)	2.7	[16, 21, 31]
Condom use probability	0.5	[25, 32–36]
Probability of transmission: CD4: >500	0.04	[16, 37–40]
Probability of transmission: CD4: 350–499	0.04
Probability of transmission: CD4: 200–349	0.05
Probability of transmission: CD4: <200	0.1
Reduced probability of transmission on ART	0.96	[6]
Decreased number of sexual partners due to diagnosis	0.5	[41–44
Condom effectiveness	0.9	[45]
Heterosexual sex parameters
No sexual partners: MSM	0.08	[32]
No sexual partners: MSM/PWID	0.08	[46, 47]
No sexual partners: PWID	1.6	[47, 48]
No sexual partners: heterosexual	0.9	[32, 49–51]
Condom use probability: MSM	0.3	[34, 36, 46]
Condom use probability: MSM/PWID	0.3
Condom use probability: PWID	0.3	[35, 46]
Condom use probability: heterosexual	0.3	[51]
Probability of transmission: CD4: >500	0.025	[3, 37, 52–56]
Probability of transmission: CD4: 350–499	0.025
Probability of transmission: CD4: 200–349	0.035
Probability of transmission: CD4: <200	0.065
Decreased number of sexual partners due to diagnosis	0.5	[44]
Reduced probability of transmission on ART	0.96	[5]
Condom effectiveness	0.9	[45]
Costs (2015$CDN)
ART costsc: PWID	Time dependentb	[73]
CD4: >500	435 (1996); 1276 (2010)
CD4: 350–499	464 (1996); 1287 (2010)
CD4: 200–349	483 (1996); 1287 (2010)
CD4: <200	478 (1996); 1294 (2010)
ART costsc: non-PWID	Time dependentb	[73]
CD4: >500	494 (1996); 1295 (2010)
CD4: 350–499	523 (1996); 1306 (2010)
CD4: 200–349	542 (1996); 1305 (2010)
CD4: <200	537 (1996); 1313 (2010)
Non-ART medical costsd: PWID	Time dependentb	[74]
CD4: >500	788 (1996); 745 (2010)
CD4: 350–499	825 (1996); 784 (2010)
CD4: 200–349	883 (1996); 1022 (2010)
CD4: <200	1896 (1996); 1992 (2010)
Non-ART medical costsd: non-PWID	Time dependentb	[74]
CD4: >500	317 (1996); 309 (2010)
CD4: 350–499	362 (1996); 330 (2010)
CD4: 200–349	471 (1996); 471 (2010)
CD4: <200	1098 (1996); 763 (2010)
HIV enzyme-linked immunosorbent assay antibody test	13	[75]
Confirmatory Western blot test	21	[75]
Medical care costsd: HIV-negative, MSM	205	[76]
Medical care costsd: HIV-negative, PWID	518	[73, 76]
Medical care costsd: HIV-negative, heterosexual	217	[76]
Intervention costse		Appendix
Hospital-based care testing	32081 (IMP); 12750 (SUS)
Emergency department testing	42798 (IMP); 9892 (SUS)
Outpatient clinic testing	49215 (IMP); 14785 (SUS)
ART initiation	66922 (IMP); 32271 (SUS)
ART retention	230734 (IMP); 111263 (SUS)
Annual discount rate	0.03
Quality-adjusted life years
Susceptible	1.00	[77]
Infected: CD4: ≥500	0.91	[78–81]
Infected: CD4: 350–499	0.79	[78–81]
Infected: CD4: 200–349	0.79	[78–81]
Infected: CD4: <200	0.72	[78–81]
Diagnosed: CD4: ≥500	0.87	[78–81]
Diagnosed: CD4: 350–499	0.72	[78–81]
Diagnosed: CD4: 200–349	0.72	[78–81]
Diagnosed: CD4: <200	0.72	[78–81]
On ART: CD4>500	0.87	[78–81]
On ART: CD4: 350–499	0.83	[78–81]
On ART: CD4: 200–349	0.83	[78–81]
On ART: CD4: <200	0.82	[78–81]
PWID multiplier	0.90	[27, 82]

Abbreviations: ART, highly active antiretroviral treatment; CDN, Canadian dollars; HIV, human immunodeficiency virus; IMP, implementation phase; MSM, men who have sex with men; PWID, people who inject drugs; SUS, sustainment phase.

^a“Initial” values refer to 1996 values, the start of the model calibration period (1996–2010).

^bFigures presented for time-dependent parameters are 1996 and 2010 values.

^cIncludes ART medication costs and associated pharmacy dispensation costs.

^dIncludes non-ART medication costs and associated pharmacy dispensation costs, costs of physician billings for outpatient care, and hospitalization cost.

^eIntervention costs were differentiated between the implementation phase and the sustainment phase, with the latter excluding fixed costs of initiating the interventions.

HIV transmission occurred through homo- and heterosexual sex and needle sharing according to the assumption of proportional mixing [83, 84]. Table 1 provides the estimated baseline (1996) sexual and injection risk behaviors used to calibrate the model, with changes in these behaviors corresponding with proxies of injection and sexual risk behavior during the period [26, 27], as we have detailed previously [85].

The model was parameterized using comprehensive linked individual health administrative and registry data for the population of diagnosed PLHIV. Further details regarding the construction and composition of the HIV-positive cohort and available databases are provided elsewhere [13, 14]. The model was validated against population-level estimates of overall HIV prevalence, HIV incidence (overall and by risk group), the number of diagnosed PLHIV (overall and by ART receipt), the number of deaths among PLHIV by calendar year, and the size of the HIV-negative population during the period 1996–2010.

Interventions Assessed

We focused on the following 5 distinct interventions that were part of the STOP HIV/AIDS initiative: HIV testing in hospital, emergency departments (EDs), and outpatient clinic settings; ART initiation initiatives; and ART retention initiatives (Table 2) [86–88]. Each intervention was delivered at the Vancouver Coastal Health Authority (VCHA), home to 25% of BC’s population and 50% of BC’s population of PLHIV [13, 14]. We used observed aggregate-level testing rates and individual-level ART initiation and reinitiation rates in VCHA during the study period to estimate the effect of these interventions.

Table 2.

Description of Selected British Columbia Ministry of Health Seek and Treat for Optimal Prevention of Human Immunodeficiency Virus/AIDS Interventions Delivered in 2011–2013

Intervention	Description	Delivery
Hospital-based testing	Integrate the routine offering of HIV testing into clinical practice in hospitals	Simplified pretest guideline while maintaining verbal informed consent Included a new patient follow-up process allowing for post-test counseling, education, public health follow-up, and linkage to care to be delegated to Vancouver Coastal Health Authority, communicable disease control nurses
ED testing	Integrate the routine offering of HIV testing into clinical practice in EDs	Introduced similarly as hospital-based testing
Outpatient clinic testing	Increase the routine offering of HIV testing to adult patients who had not been tested in the last year or presented specific risk, clinical symptoms, or the diagnosis of another sexually transmitted disease	Simplified pretest guideline while maintaining verbal informed consent. Consisted of training, continuing medical education credits, approved by the College of Family Physicians of Canada, and ongoing in-practice support
ART initiation	Expanded support to help gain access to ART	Low-barrier patient-centric support to liaise with a continuum of health and social services was provided by a multidisciplinary team that included healthcare professionals, social workers, and trained peer navigators
ART retention	Expanded case management to maintain ART adherence and help ART reinitiation among patients who have discontinued ART	When required, one-on-one adherence monitoring and support was provided Following a referral for case management, outreach workers attempted to locate patients who had discontinued ART

The STOP initiative period was 2011–2013.

Abbreviations: ART, highly active antiretroviral treatment; ED, emergency department; HIV, human immunodeficiency virus.

In order to construct the counterfactual “status quo” level of testing, ART initiation, and retention, we estimated and extended trends in annual rates of testing and ART initiation from 2006 to 2010 (excluding prior years when initiation was constrained by clinical guidelines [63, 64]), as well as the probabilities of ART dropout, derived using multistate Markov models [89, 90] (Figures 1 and 2). We derived the costs for the implementation and sustainment phases of each respective intervention using financial data provided by VCHA [91, 92]. We assumed no public health intervention costs for the status quo scenario. Further detail on the calculations used to estimate the scale of delivery across HIV risk groups, effectiveness, and costs are provided in the Supplementary Materials.

Figure 1.

Historical trends and estimated British Columbia Ministry of Health Seek and Treat for Optimal Prevention of HIV/AIDS intervention effects for human immunodeficiency virus (HIV) test initiatives. Historical aggregate-level HIV testing rates (left-hand sides of panels A–C) were extrapolated to generate counterfactual HIV testing rates, approximating the status quo (ie, no additional public health investments to scale-up testing) for the intervention period (dashed lines). Information on actual testing rates in Vancouver Coastal Health Authority, adjusted using external data, were used to estimate testing rates for each setting and by HIV risk group during the intervention period. Abbreviations: ED, emergency department; HIV, human immunodeficiency virus; MSM, men who have sex with men; PWID, people who inject drugs.

Figure 2.

Historical trends and estimated British Columbia Ministry of Health Seek and Treat for Optimal Prevention of human immunodeficiency virus/AIDS intervention effects for highly active antiretroviral treatment (ART) initiation and retention initiatives. The observed monthly probability of ART initiation among diagnosed people who inject drugs (PWID; panel A) and non-PWID (panel B) for British Columbia from 1996–2010 was used to generate counterfactual ART initiation probabilities approximating the status quo (ie, no additional public health investments to support ART initiation) for the intervention period (dashed line). ART initiation probabilities in Vancouver Coastal Health Authority, observed in linked individual-level data, were used to determine the intervention effect. The same procedure was used to determine probabilities of ART reentry (panels C and D). To adequately capture changes in the probability of ART dropout within the dynamic simulation model, we extrapolated counterfactual CD4-based transition probabilities using multistate Markov models for PWID and non-PWID (panels E and F) and estimated observed dropout probabilities during the intervention period using similar procedures as described above. Abbreviations: ART, highly active antiretroviral treatment; PWID, people who inject drugs.

Analyses

The primary model-projected outcomes included HIV incidence, mortality, and quality-adjusted life-years (QALYs) gained, as well as the costs of medical care, including costs attributable to ART (drug and pharmacy dispensation costs), non-ART medical care (HIV and non-HIV diagnostic tests, hospital-based care, family practice care, non-ART prescriptions), and costs attributable to HIV testing and public health intervention. The costs of ART and non-ART medical care among PLHIV were updated from previously published studies [93, 94]. We applied QALY weights derived from the peer-reviewed literature for HIV-negative individuals and PLHIV in and out of treatment, adjusting for injection drug use [76–81].

We estimated incremental cost-effectiveness ratios (ICERs) for each component intervention compared to the counterfactual status quo. In order to capture the long-term individual benefits of ART and second-order transmission effects, we considered 5-, 10-, and 25-year time horizons following the intervention period (2011–2013), for a (maximum) study period of 2011–2038. We considered a third-party payer perspective, presented all costs in 2015$CDN, and discounted costs and QALYs at an annual rate of 3%. Cost-effectiveness analysis was conducted according to well-established methods and conformed to guidelines on cost-effectiveness analysis and dynamic transmission modeling [95–97].

Otherwise, we executed sensitivity analyses on the size of the undiagnosed population [98], the uncertainty in the (unobserved) distribution of testing across HIV risk groups for each setting (hospital, outpatient clinic, ED), the effectiveness of each intervention, and the costs of the ART initiation and retention interventions. Further details are provided in the Supplementary Materials.

RESULTS

Model Validation

Estimated prevalence in the model was within 5.7% of the independently estimated Public Health Agency of Canada (PHAC) prevalence during the model calibration period (1996–2010), while estimated HIV incidence was within 2.7% of the PHAC estimate [98]. Overall model-estimated numbers of diagnosed cases were within 1.9% of observed figures at the midpoint of each calendar year, and the estimated number of deaths among diagnosed PLHIV was within 9% of observed figures. Finally, the modeled figures for the total susceptible population (aged 15–64 years) were within 1% of observed figures (further details provided in the Supplementary Materials).

HIV Testing Interventions

Overall, we estimated that the specified interventions increased testing rates from 222 to 1814 per 100000 BC residents aged 15–65 years during the intervention period. We further estimated the hospital-based testing initiative would result in 131 fewer incident HIV cases and 42 fewer deaths among PLHIV if the initiative were maintained at its current level over a 25-year time horizon (2011–2038; Figure 3). This intervention would result in higher ART and testing costs and would carry a public health intervention cost of $3.98 million but would offset non-ART medical costs by $4.83 million in present value over the 25-year study period. Also, it would result in an increment of 295.88 QALYs, for an ICER of $34544 per QALY gained (Table 3).

Figure 3.

Epidemiological effects of the British Columbia Ministry of Health Seek and Treat for Optimal Prevention of HIV/AIDS initiative interventions evaluated (25-year time horizon). Estimated effects of each human immunodeficiency virus (HIV) care intervention on averted incident cases (panel A) and deaths among people living with human immunodeficiency virus (panel B), by HIV risk group, over the complete study time horizon. *Combined initiatives to prevent treatment dropout and enhance reengagement. **Reverting back to the counterfactual status quo-levels of HIV testing and treatment engagement for the remainder of the study time horizon. Abbreviations: ART, highly active antiretroviral treatment; ED, emergency department; HET, heterosexual; HIV, human immunodeficiency virus; MSM, men who have sex with men; PLHIV, people living with human immunodeficiency virus; PWID, people who inject drugs.

Table 3.

Benefits, Costs, and Incremental Cost-Effectiveness of British Columbia Ministry of Health Seek and Treat for Optimal Prevention of Human Immunodeficiency Virus/AIDS Interventions

Time Horizons Following the Intervention Period	ART Costs	Non-ART Costs	Human Immunodeficiency Virus Testing Costs	Interventiona Costs	Total Costs	QALYs	Incremental Cost- Effectiveness Ratiob
5-year time horizon	$CDN (M)	$CDN (M)	$CDN (M)	$CDN (M)	$CDN (B)	(Millions)
Status quo	$871.47	$601.30	$16.75	--	$67.15	24.96	--
	Δ cost (M)	Δ cost (M)	Δ cost (M)	Δ cost (M)	Δ cost (M)	Δ QALYs
Hospital-based testing	1.88	–0.32	0.77	1.89	4.31	7.04	$612051
Outpatient clinic testing	2.08	–0.32	1.40	2.63	5.87	7.26	$808749
ED testing	2.89	–0.47	1.42	2.11	6.09	10.40	$585985
ART initiation	6.82	–1.69	0.00	4.28	9.67	94.91	$101877
ART retentionc	5.90	–0.47	0.00	14.75	20.34	43.16	$471385
Combined interventions, sustained	19.55	–3.27	3.58	25.67	46.26	164.12	$281892
Combined interventions, 2011–2013 onlyd	10.67	–2.19	0.83	15.18	25.02	122.30	$204578
10-year time horizon	$CDN (M)	$CDN (M)	$CDN (M)	$CDN (M)	$CDN (B)	(Millions)
Status quo	$1429.18	$896.87	$26.08	--	$104.45	38.81	--
	Δ cost (M)	Δ cost (M)	Δ cost (M)	Δ cost (M)	Δ cost (M)	Δ QALYs
Hospital-based testing	4.31	–1.16	1.29	2.53	7.36	45.06	$163396
Outpatient clinic testing	4.92	–1.20	2.34	3.37	9.83	48.82	$201410
ED testing	6.74	–1.75	2.38	2.61	10.58	69.01	$153352
ART initiation	9.52	–2.60	0.00	5.90	13.50	204.29	$66102
ART retentionc	9.52	–0.92	0.00	20.35	29.38	92.42	$317911
Combined interventions, sustained	34.50	–7.47	5.99	34.77	70.23	457.78	$153414
Combined interventions, 2011–2013 onlyd	12.67	–3.48	0.83	15.18	26.45	253.30	$104427
25-year time horizon	$CDN (M)	$CDN (M)	$CDN (M)	$CDN (M)	$CDN (B)	(Millions)
Status quo	$2778.53	$1510.33	$48.22	--	$193.01	71.69	--
	Δ cost (M)	Δ cost (M)	Δ cost (M)	Δ cost (M)	Δ cost (M)	Δ QALYs
Hospital-based testing	6.26	–4.83	2.53	3.98	10.22	295.88	$34544
Outpatient clinic testing	7.75	–5.28	4.59	5.04	14.54	333.29	$43623
ED testing	9.79	–7.76	4.67	3.73	14.03	464.44	$30216
ART initiation	11.79	–4.10	0.00	9.55	19.60	586.52	$33423
ART retentionc	16.56	–2.82	0.00	32.93	48.51	304.02	$159551
Combined interventions, sustained	49.74	–23.28	11.74	55.23	105.34	1906.30	$55258
Combined interventions, 2011–2013 onlyd	9.76	–6.20	0.83	15.18	23.17	637.42	$36356

Abbreviations: ART, combination antiretroviral therapy; B, billions; CDN, Canadian dollars; ED, emergency department; M, millions; QALY, quality-adjusted life-year.

Δ incremental costs/QALYs.

^aCosts of public health intervention.

^bIncremental cost-effectiveness ratio (ICER) of the intervention vs the counterfactual status quo: ICER = (Cost_intervention – Costs_{status quo})/(QALY_intervetnion – QALY_{status quo}).

^cART retention includes the initiatives targeting preventing ART dropouts and enhancing reengagement among treatment-discontinued people living with human immunodeficiency virus.

^dReverting back to the counterfactual status quo-levels of human immunodeficiency virus testing and treatment engagement for the remainder of the study time horizon.

In contrast, we estimated the outpatient clinic testing initiative will result in a reduction of 145 incident cases and 48 deaths among PLHIV for an ICER of $43623 per QALY gained if sustained through 2038. Finally, given the enhanced ability to reach populations at higher risk of HIV seroconversion (PWID and MSM), we estimated the ED testing initiative would have the largest impact on provincial HIV incidence, resulting in 214 cases averted and 67 deaths among PLHIV for an ICER of $30216 per QALY gained if sustained at the current scale throughout the study period.

ART Engagement Initiatives

We estimated a 10.6% increase in ART initiation for PWID and 7.9% increase for non-PWID from 2011 to 2013 vs the status quo counterfactual. The effects of the ART retention initiative were limited, including a 0.9%/4.2% increase in the probability of reengaging PWID/non-PWID on ART and a 4.7%/5.6% reduction in the probability of dropout among PWID/non-PWID (overall, across CD4 strata).

If sustained at the current scale throughout a 25-year time horizon, we estimated the ART initiation intervention would avert 75 incident cases and 46 deaths, resulting in an increment cost of $19.60 million in present value, with an additional $11.79 million attributable to ART costs and $9.55 million in public health intervention costs. However, this would be offset by $4.10 million in savings in non-ART medical costs. For an aggregate gain of 586.52 QALYs, we estimated an incremental cost-effectiveness of $33423 per QALY gained. Otherwise, we anticipated the ART retention initiative will avert 82 incident cases and 30 deaths at a cost of $159551 per QALY gained.

Combined Effect of Selected STOP HIV/AIDS Interventions

We considered the effects of the combination of HIV testing and ART engagement initiatives if sustained throughout the study period and if discontinued (reverting back to pre-STOP levels) after the 2011–2013 intervention period. If sustained, we anticipate 595 incident HIV cases averted and 217 deaths averted among PLHIV, at a cost of $55258 per QALY gained. If the intervention was to be discontinued after the 3-year pilot period, the intervention would only avert 111 incident cases and 50 deaths among PLHIV, at a cost of $36356 per QALY gained.

Incremental cost-effectiveness ratios for the majority of the interventions at shorter time horizons (5, 10 years) were outside the cost-effective range. However, the ART initiation initiative remained around $100000 per QALY even at a 5-year time horizon, demonstrating the more immediate value of engaging PLHIV in ART and the longer-term benefits that result from earlier detection of HIV and averted HIV incidence resulting from HIV testing interventions.

Sensitivity Analyses

The results of our sensitivity analyses are presented in Figures 4–6. Focusing on the sustained implementation of the combined interventions, use of the upper and lower bounds of PHAC-estimated HIV prevalence for the province of BC in 2010 resulted in substantial differences in the estimated number of incident HIV cases averted (595 at baseline; 441 and 1011 at the lower and upper bounds) and deaths averted (217; 160–385), with the baseline ICER of $55258, ranging from $68401 to $39112 (Figure 4). Attributing a greater proportion of HIV tests toward MSM and PWID improved (decreased) ICERs for the HIV testing initiatives, while attributing a greater proportion of tests toward heterosexuals increased ICERs (Figure 5); however, ICERs remained within a cost-effective range in all instances. Removing implementation costs for each intervention resulted in small changes in ICERs, most notably decreasing the ICER for the ART retention initiative to $145403 per QALY. In 2-way sensitivity analysis, the ART retention initiative never fell below $50000, even after simultaneously doubling effectiveness and decreasing costs by 90%. Finally, improving the effectiveness of each intervention jointly increased the number of incident cases and deaths averted and decreased the ICER, while decreasing effectiveness by 50% increased the ICER to $81837 per QALY gained (Figure 6). Detailed results are presented in the Supplementary Materials.

Figure 4.

Results of sensitivity analyses on human immunodeficiency virus (HIV) prevalence. The size of population of people living with HIV in 2010 was set to the lower or upper bounds of provincial HIV prevalence estimates in 2010 from the public health agency of Canada. All results were for the combined intervention, sustained over the complete study time horizon. Abbreviations: ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life-years.

Figure 5.

Sensitivity analysis on proportion of risk groups in human immunodeficiency virus (HIV) testing in hospital, outpatient clinic, and emergency department (ED) settings. The baseline distribution of HIV testing in each setting was as follows: hospital-based: 92.7% heterosexual, 3.3% men who have sex with men (MSM), 4.0% people who inject drugs (PWID); outpatient clinic: 96.6% heterosexual, 2.1% MSM, 1.3% PWID; ED: 94.7% heterosexual, 3.2% MSM, 2.0% PWID. The proportions of heterosexuals tested in each scenario were adjusted accordingly. All results were for the combined intervention, sustained over the complete study time horizon. Abbreviations: HIV, human immunodeficiency virus; ICER, incremental cost-effectiveness ratio; MSM, men who have sex with men; PWID, people who inject drugs; QALY, quality-adjusted life-years.

Figure 6.

Sensitivity analysis on the effectiveness of the combined British Columbia Ministry of Health Seek and Treat for Optimal Prevention of human immunodeficiency virus/AIDS pilot interventions. All results were for the combined intervention, sustained over the complete study time horizon. Abbreviations: ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life-years.

DISCUSSION

We estimated the cost-effectiveness of a selected set of HIV testing and treatment engagement initiatives implemented as part of BC’s STOP HIV/AIDS pilot from 2011 to 2013. Using World Health Organization criteria for assessing cost-effectiveness [99] with a 25-year time horizon, the ICERs we found to represent 59%–312% of BC’s Gross Domestic Product (GDP) per capita ($51135), suggesting, by international standards [100], that all but the ART retention initiative can be classified as “very cost-effective” (<1× GDP per capita: hospital-based, outpatient clinic, and ED testing; ART initiation and combined STOP pilot) or “cost effective” (<3× GDP per capita: combined STOP pilot). Delivered in combination at the current scale and sustained throughout our study period, we estimated the interventions could reduce cumulative HIV incidence by 13% and deaths among PLHIV by 5% at an incremental cost of $105 million in present value, or $3.8 million per year over the 25-year study time horizon.

Individually, ED testing provided the best value for the money and the greatest impact on preventing HIV incidence and mortality among PLHIV, while the ART initiation strategy provided the greatest gain in QALYs. Despite large increases in testing, we did not observe “saturation” in testing due to the limited scale of these interventions and the projected growth in the BC population [20]. Further, while we could only attribute tests across risk groups with indirect estimates, results were robust in sensitivity analyses, varying the proportion of tests across risk groups. These points underline the value of HIV testing, which is likely to be even higher in settings with larger populations of undiagnosed PLHIV.

The ART retention initiative was the least successful, notably leading to a very modest 0.9% increase in ART reinitiation rates among treatment-discontinued PWID and was only considered (marginally) cost-effective by excluding startup costs or drastically increasing effectiveness. While we could only estimate the effectiveness of this intervention coarsely—a subset of Vancouver Coastal Health clients not identified in our databases was targeted—the intervention’s impact was nonetheless not detectable at the Health Authority level. This may be partially attributable to an increasingly selective client base that the HIV care system continues to struggle to retain in ART, despite a high level of public health programming and fully covered medication costs. This context is important in considering the generalizability of our results to other settings. Nonetheless, a recent systematic review of interventions to improve ART adherence demonstrated that while some interventions were effective, their estimated effects were modest and waned over time [101]. Refined interventions and, possibly, new medication technologies [102, 103] will be critical to the success of HIV treatment and prevention strategies in BC and elsewhere moving forward.

We note that we have not provided an exhaustive account and evaluation of all interventions executed as part of the STOP HIV/AIDS pilot project but rather a selection of those with sufficient reported information to provide reasonable estimates of scale of delivery and effectiveness. In particular, several smaller-scale targeted HIV testing campaigns were executed, including a highly effective small-scale bathhouse testing campaign and a peer-driven testing campaign that had a 0.2% HIV-positivity rate and that reconnected 324 individuals (6.8%) into care [104, 105].

Given the demonstrated value of the assessed interventions, the feasibility of scale-up and sustainment of these and other interventions across the province must be considered carefully. Our findings pertain to interventions executed by VCHA, whose underlying epidemic has a distinct profile from that of other health authorities. The STOP HIV/AIDS initiative is now an ongoing provincial program, and each of the 4 other health authorities have implemented similar programs according to the gaps in their regional care cascades [106]. It is not necessarily the case that these interventions, at their observed scale and effectiveness, would have a comparable impact in other regions within or outside BC, which may also have different underlying levels of existing services.

This analysis featured several limitations worthy of further discussion. First, the interventions assessed were delivered on a nonrandomized basis, and information on the distribution of tests among the key HIV risk groups had to be estimated. We emphasize that our counterfactual status quo scenario—a characterization of what testing and treatment engagement rates would have been in the absence of the STOP pilot—were estimates based on historical trends in these rates over time. Despite these limitations, a flexible modeling framework that is amenable to sensitivity analysis and can capture long-term outcomes provides perhaps the best opportunity to assess such interventions within a causal framework [107].

Second, while the majority of our model was built with linked individual-level data for the population of PLHIV in BC, the size of the undiagnosed population and its distribution across risk groups was drawn from a national modeling effort from the Public Health Agency of Canada [98], which did not include much of the population-level data available to the STOP HIV/AIDS study team. Ascertaining accurate estimates of the size of the undiagnosed population of PLHIV and the distribution of this population across HIV risk groups is of paramount importance in establishing the value of HIV testing campaigns, in particular, in absolute terms and relative to other interventions.

Otherwise, we have previously outlined limitations due to infectivity being modeled indirectly through CD4-based stages of disease progression and no explicit account of higher infectivity in the 6 months following seroconversion. Our model was nonetheless able to produce risk group–specific incidence estimates and reproduce key aspects of the HIV epidemic at the population level in BC with a high degree of precision [17].

We have demonstrated the cost-effectiveness and estimated the long-term epidemiological impact of 5 key components of a coordinated combination HIV prevention strategy executed in BC. Our results demonstrate the substantial value these programs have added despite limited-scale implementation and underline the need to expand and sustain public health intervention efforts to curb the HIV epidemic in BC.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. We acknowledge Michelle Olding for assistance with manuscript preparation, as well as all British Columbia Ministry of Health and Vancouver Coastal Health Decision support staff who were involved in data access and procurement, including Miranda Compton Vancouver Coastal Health Authority (VCHA), Theodora Consolacion (BC-CDC), Monika Lindegger, (BC-CDC), Al Cassidy (BC Ministry of Health), and Joleen Wright and Karen Luers (VCHA). We also acknowledge Ciro Panessa, Nancy South, and Mark Gilbert for their contributions to the STOP/HIV AIDS study group. Bohdan Nosyk is a Michael Smith Foundation for Health Research Scholar.

The STOP HIV/AIDS Study Group is comprised of the following: Rolando Barrios, VCHA; Patty Daly, VCHA; Reka Gustafson, VCHA; Perry RW Kendall, British Columbia (BC) Ministry of Health; Gina McGowan, British Columbia Ministry of Health; Irene Day, BC Centre for Excellence in HIV/AIDS; Kate Heath, BC Centre for Excellence in HIV/AIDS; Robert S Hogg, BC Centre for Excellence in HIV/AIDS; Julio SG Montaner, BC Centre for Excellence in HIV/AIDS; and Bohdan Nosyk, BC Centre for Excellence in HIV/AIDS.

Disclaimer. The funders had no direct role in the conduct of the analysis or the decision to submit the manuscript for publication.

Financial support. This study was funded by the BC Ministry of Health–funded “Seek and Treat for Optimal Prevention of HIV & AIDS” pilot project, a grant from the National Institutes of Health/National Institute on Drug Abuse (R01-DA-041747), and by Genome Canada (142HIV).

Potential conflicts of interest. J. S. G. M. has received limited unrestricted funding, paid to his institution, from Abbvie, Bristol-Myers Squibb, Gilead Sciences, Janssen, Merck, the MAC AIDS FUND, and ViiV Healthcare. All other authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.