Cost-effectiveness of Electronic- and Clinician-Delivered Screening, Brief Intervention, and Referral to Treatment for Women in Reproductive Health Centers

Abstract

Aims:

To determine the cost-effectiveness of electronic- and clinician-delivered SBIRT (Screening, Brief Intervention and Referral to Treatment) for reducing primary substance use among women treated in reproductive health centers.

Design:

Cost-effectiveness analysis based on a randomized controlled trial.

Setting:

New Haven, Connecticut, USA.

Participants:

A convenience sample of 439 women seeking routine care in reproductive health centers who used cigarettes, risky amounts of alcohol, illicit drugs, or misused prescription medication.

Interventions:

Participants were randomized to enhanced usual care (EUC, N=151), electronic-delivered SBIRT (e-SBIRT, N=143), or clinician-delivered SBIRT (SBIRT, N=145).

Measurements:

The primary outcome was days of primary substance abstinence during the 6-month follow-up period. To account for the possibility that patients might substitute a different drug for their primary substance during the 6-month follow-up period, we also considered the number of days of abstinence from all substances. Incremental cost-effectiveness ratios and cost-effectiveness acceptability curves determined the relative cost-effectiveness of the three conditions from both the clinic and patient perspectives.

Findings:

From a healthcare provider perspective, e-SBIRT is likely (with probability greater than 0.5) to be cost-effective for any willingness-to-pay value for an additional day of primary-substance abstinence and an additional day of all-substance abstinence. From a patient perspective, EUC is most likely to be the cost-effective intervention when the willingness to pay for an additional day of abstinence (both primary-substance and all-substance) is less than $0.18, and e-SBIRT is most likely to be the cost-effective intervention when the willingness to pay for an additional day of abstinence (both primary-substance and all-substance) is greater than $0.18.

Conclusions:

e-SBIRT could be a cost-effective approach, from both healthcare provider and patient perspectives, for use in reproductive health centers to help women reduce substance misuse.

1. INTRODUCTION

Despite high rates of substance misuse both in the United States and internationally [1,2], most individuals who misuse substances neither receive nor feel they need substance use treatment [2–4]. Screening, Brief Intervention, and Referral to Treatment (SBIRT) seeks to identify those with substance misuse via proactive universal screening and engage them in changing substance misuse. Its brevity makes it acceptable regardless of interest in treatment; systematic reviews find it efficacious for reducing unhealthy alcohol [4,5] and tobacco use [6–8], although several studies find it ineffective for drug use [9–11]. The American Congress of Obstetricians and Gynecologists recommends SBIRT in reproductive health settings [12].

The cost-effectiveness of SBIRT is less clear. Although a number of studies conducted full economic evaluations of SBIRT for hazardous drinking [13], systematic reviews of this work suggest relatively weak evidence for the cost-effectiveness of SBIRT [14,15]. To our knowledge, only three studies conducted full economic evaluations of SBIRT for substances other than alcohol, and these results are mixed [16–18]. There is a notable gap in cost-effectiveness assessments when SBIRT targets a spectrum of substances rather than only one [19]. Additionally, prior economic evaluations did not include electronically-delivered SBIRT, a delivery modality that offers cost-reducing advantages such as less clinician training and supervision, less time commitment from clinicians, and reliable delivery of treatment [20,21]. Using data from a recently completed 3-group randomized trial, the present analysis reports on the cost-effectiveness of electronic- and clinician-delivered SBIRT relative to enhanced usual care among medical treatment-seeking women with a broad range of risky substance use.

2. METHODS

Methods and results of the trial are described in the main report [22]. The design is summarized below, followed by the methods used for the economic evaluation. Outcome and resource utilization data for these analyses are taken from the trial and combined with cost data obtained prospectively from study personnel and the healthcare providers used in the trial. Data collection occurred between September 5, 2011, and January 28, 2015. The trial was approved by Yale Institutional Review Board, included a certificate of confidentiality from the National Institute on Drug Abuse, and is registered at ClinicalTrials.gov, .

2.1. Trial Design

Women visiting two urban, academic reproductive health clinics located in New Haven, CT were voluntarily screened for cigarette smoking or misuse of alcohol, illicit drugs, or prescription medication. Screening took place during regularly scheduled reproductive health visits and was overseen by research assistants who used an audio-enabled, computer-assisted self-interview tool. Women provided consent before screening. Eligible participants were women at least 18 years old who scored positive on the Alcohol, Smoking, and Substance Involvement Screening Test (ASSIST) [23].¹ According to ASSIST guidelines, for all substances other than alcohol, the cutoff for inclusion in the trial was a score of >= 4; for alcohol, we used the standard cutoff of >= 11 for nonpregnant women but used >= 6 for pregnant women. Exclusions were (1) current or imminent incarceration or hospitalization; (2) inability to provide contact information for 2 persons; (3) participation in substance use treatment or self-help programs in the past 3 months; (4) or inability to interview and consent in English.

Research staff approached 2,421 women for screening, 439 of whom screened positive and were randomly assigned to the three conditions: enhanced usual care (EUC, 151), computer-delivered SBIRT (e-SBIRT, 143), or clinician-delivered SBIRT (SBIRT, 145). There were no differences between conditions on any baseline demographic or substance use variables [22]. Eighteen percent of participants were pregnant. Self-reported primary substances (i.e., the substance the participant identified as most problematic) were cigarettes (57%), alcohol (12%), cannabis (20%) and other drugs (11%). Participants reported using their primary substance an average of 23.7 days (SD = 7.9) during the 4 weeks preceding the baseline assessment.

All participants received a handout listing local treatment resources, including inpatient, residential, and outpatient treatment programs (along with brief program descriptions, payment accepted, hours of operation, addresses and phone numbers), self-help and internet recovery resources, and tips for quitting on one’s own. Both SBIRTs were based on motivational interviewing (MI) and focused on supporting the importance of, and a woman’s confidence in, cutting down or quitting substances [24]. The e-SBIRT featured an interactive, three-dimensional, mobile, talking, animated narrator that delivers a single, tailored MI-based brief intervention via a tablet computer (the same tablet that was used for screening) [20]. The SBIRT was a single, MI-based intervention and followed the same sequence of MI components provided in e-SBIRT. Clinicians delivering SBIRT (two study nurses, three social workers, and one obstetrician-gynecologist) received a 15-hour SBIRT workshop training followed by practice cases with feedback and coaching and monthly group supervision during the trial, all provided by the senior author (SM) per current standards [24]. As in prior studies [25–28], providers were required to meet a performance criterion prior to seeing randomized participants. Afterward, a sample (50%) of audio recorded SBIRT sessions were rated for MI fidelity [28].

EUC involved a brief interaction with patients about their substance use and review of the handout listing local treatment resources.² This interaction took 2 minutes, whereas the mean (SD) duration of the SBIRT interventions was 21.06 (6.9) minutes for e-SBIRT and 22.03 (5.8) minutes for SBIRT (including time to review the handout). Patients reported strong satisfaction (on average, “considerably” to “extremely” satisfied) with both SBIRT interventions [29]; no clinician satisfaction data were collected in the trial.

Assessments were completed at baseline and 1-, 3- and 6-month post randomization and were kept brief (30 minutes) to minimize pre-intervention effects [30,31]. Self-reported days of substance use (covering the period since the last survey was completed) were collected via the Timeline Followback (TLFB) method [32,33] and corroborated with urine toxicology tests. Retention rates exceeded 84% at all follow-up points and were comparable among groups. Of the 439 participants in the main trial, 57 (13.0%) were missing the 6-month assessment (EUC, 18; e-SBIRT, 18; SBIRT, 21). Missing 6-month assessment data were imputed following Briggs et al. [34] in which standard regression analysis is used to provide estimates of the missing data conditional on complete variables in the analysis. Specifically, predictive equations were obtained from regressing number of days of abstinence on a set of demographic characteristics (age, race, years of education, employment status, pregnancy status, primary drug, ASSIST score) and, when available, number of days of abstinence reported at the most recent (i.e., 1-month or 3-month) follow-up assessment. Predictive equations from the regressions had R² values ranging from .28 (patients with no follow-up assessment) to .56 (patients with 3-month assessment). Eleven (2.5%) participants were excluded from analyses due to malfunctioning timestamp data from the interventions.

2.2. Economic Evaluation

To extrapolate the implementation cost of the interventions from the trial to a more realistic real-world scenario, the economic evaluation assumes each intervention is integrated into the standard operation procedures of a single healthcare provider so that all incoming patients would self-screen in the reception area via a tablet computer provided by the receptionist (as opposed to a comparatively small number of incoming patients being approached for study participation by research assistants). This assumption has the following two advantages.

First, it allows us to more accurately model the efficiencies inherent in e-SBIRT. In the trial, participants in all three conditions returned their tablet to the research assistant upon completing the screen, whereupon the research assistant assessed the ASSIST score and then, as appropriate, routed participants to the brief intervention. If each intervention were integrated into the standard operating procedures of the healthcare provider, e-SBIRT patients who screened positive could move directly from the screen to the intervention using the same tablet, thereby saving the receptionist the step of interpreting the ASSIST score and routing the patient to the intervention. During the trial, research assistants estimated the mean duration of screen service support time (including time to explain the screen to patients, interpret the ASSIST score, and prepare the tablet for the next patient) to be 4 minutes for each intervention. The base scenario of the economic evaluation assumes the same 4 minutes for EUC and SBIRT, and 3 minutes for e-SBIRT (because the receptionist would not need to interpret the ASSIST score, etc.).

Second, by assuming all incoming patients would be screened, the fixed costs of implementing each intervention can be allocated over a more reasonable number of patients than occurred in the trial. Assuming a full-scale implementation of each intervention, however, requires decisions about several cost-related parameters, including annual number of unique patient visits, number of tablets required, and the useful life of the tablets and training. As explained below, the base scenario assumed 6,500 unique patient visits annually, 2/3/2 tablets would be required to support EUC/e-SBIRT/SBIRT, and the useful life of the tablets and training would be 5 years.

The two trial healthcare providers averaged a combined total of 6,500 unique patient visits annually (6,000 in one clinic and 500 in the other); the base scenario assumes the same total for a single clinic. The trial healthcare providers operated 2,444 hours per year (47 hrs/week * 52 weeks/year). A single site with the same operating hours and 6,500 unique patient visits annually would experience an average arrival rate of 2.66 patients per hour. In the trial, EUC and SBIRT averaged 11.2 minutes of tablet time per patient visit,³ or 5.36 patients per hour per tablet, while e-SBIRT averaged 13.7 minutes of tablet time per patient visit,⁴ or 4.38 patients per hour per tablet. Assuming a uniform distribution of patient arrivals, a single tablet would be utilized 50% of the time (2.66/5.36) in EUC and SBIRT, and 61% of the time (2.66/4.38) in e-SBIRT. The base scenario assumes 2 tablets for EUC and SBIRT (to handle peak demand of 10.72 patients per hour, or 4.0 times the average hourly demand) and 3 tablets for e-SBIRT (to handle peak demand of 13.14 patients per hour, or 4.9 times the average hourly demand). All scenarios assume the purchase of “rugged” tablets at $1,675 apiece (more below), so no breakage is assumed.

In 2017, the average turnover rate of US nurses working in the specialty of “women’s health” was 13.3% (i.e., the average duration in the same job was 7.5 years) [35]. The average expected life span (replacement cycle length) of tablets in the United States, from 2017 to 2022, is 5.12 years [36], and the economic evaluation assumes the healthcare provider will use a “rugged” tablet that is either IP65 or MIL-STD810G tested for durability [37]. Thus, the base scenario conservatively assumes 5 years as the useful life of the tablets and training.

To determine how the results would likely change had the trial been implemented under alternative realistic conditions, we conducted sensitivity analyses in which we considered two alternative scenarios – one favorable to the SBIRTs and one unfavorable to the SBIRTs – that made different assumptions about twelve key cost-related parameters (including the three described above) listed in Table 1. Unless specified otherwise, the base scenario of the economic evaluation is based on the actual costs (i.e., resource utilizations multiplied by unit costs) as occurred in the trial.

Table 1 –

Key parameter assumptions – by scenario

	Base Scenario	Favorable Scenario	Unfavorable Scenario
Annual # of unique patient visitsa	6,500	7,150	5,850
# tablets needed in EUC/e-SBIRT/SBIRTb	2/3/2	2/3/2	4/6/4
Useful life of fixed start-up activities and materials (yrs)c	5	7	3
% of patients who screen positived	18.13	25	10
e-SBIRT license fee ($/positive screen)e	1.99	0.99	2.99
Clinician trainingf	As in trial	1 training session for all clinicians	As in trial
Clinic overhead rate (%)g	30	25	35
Wages of study personnelh	As in trial	−10%	+10%
Screen service support time in EUC/e-SBIRT/SBIRT (mins)i	4/3/4	3/2/3	5/4.5/5
Intervention service support time in EUC/e-SBIRT/SBIRT (mins)j	0/0/8	0/0/7	0/3/9
Discount rate (%)k	3	2	5
Value of patient time ($/hr)l	9.60	7.25	11.50

^aSection 2.2 provides the rationale for the base scenario. The favorable/unfavorable scenarios assume +/− 10% of the base (i.e., 7,150/5,850 unique patient visits annually).

^bSection 2.2 provides the rationale for the base scenario. The unfavorable scenario assumes twice as many tablets (4/6/4) are necessary (perhaps due to unanticipated breakage).

^cSection 2.2 provides the rationale for the base scenario.

^dDuring the study, 18.13% of patients screened positive, and the base scenario assumes the same. The favorable/unfavorable scenarios assume 25%/10% of patients screen positive to reflect potential variation in patient substance use should the interventions be implemented elsewhere.

^eMany websites sell mobile behavioral healthcare applications ranging in price from free to $2.99 [46,47]. Although e-SBIRT was free in the trial, to reflect a real-world implementation of e-SBIRT, the base scenario assumes a hypothetical license fee of $1.99 per e-SBIRT patient (charged only for patients who screen positive), while the favorable/unfavorable scenarios assume $0.99/$2.99.

^fIn the trial, multiple workshops were conducted at both sites to train a total of 8 providers (2 of whom did not reach certification standards). The base and unfavorable scenarios assume the same. With a single clinic, however, fewer workshops might be required and so the favorable scenario assumes a single workshop would be used to train all providers at the same time.

^gThe overhead rate at the two trial clinics was 30% and the base scenario assumes the same. The favorable/unfavorable scenarios assume 25%/35% to reflect potential variation in clinic setting and size.

^hWages of study personnel in the base scenario are same as in the trial. The favorable/unfavorable scenarios assume +/− 10% to reflect potential geographic variation in labor costs should the interventions be implemented elsewhere.

ⁱSection 2.2 provides the rationale for the base scenario. The unfavorable scenario assumes smaller efficiency gains from streamlining the transition between the screen and e-SBIRT intervention.

^jDuring the trial, study personnel estimated the average duration of intervention service support time for SBIRT (including time for the clinician to prepare prior to the brief intervention and to take notes following the intervention) to be 8 minutes. There was no intervention service support time for EUC and e-SBIRT during the trial. The base scenario assumes the same. The unfavorable scenario assumes 3 minutes of intervention service support time for e-SBIRT.

^kThe discount rate is used to estimate the equivalent annual cost of the start-up (one-time) fixed costs [38]. Following recommendations by the Second Panel on Cost-Effectiveness in Health and Medicine [45], the base scenario assumes a 3% discount rate, while the favorable/unfavorable scenarios assume 2%/5%.

^l$9.60 was the minimum wage in the state of CT (where the trial clinics are located) in 2016. $7.25 was the federal minimum wage in 2016. $11.50 was the highest minimum wage across all US states in 2016.

2.2.1. Effectiveness measures

Because both SBIRTs were designed to improve primary substance abstinence among women treated in reproductive health centers, the main outcome in this study was the number of days of primary substance abstinence during the 6-month follow-up period. However, to account for the possibility that patients might substitute a different drug for their primary substance during the 6-month follow-up period, we also considered the number of days of abstinence from all dangerous substances (cigarettes, alcohol, marijuana, cocaine, opiates, benzodiazepines, methamphetamine, and other (typically self-reported as ecstasy or PCP)).

2.2.2. Cost analysis

Cost data were collected prospectively during the trial. A research assistant administered a survey to the Chief Financial Officer at each reproductive health clinic to obtain healthcare provider-specific cost data (e.g., clinician wages, overhead and fringe rates, cost of space). Data to estimate fixed costs were obtained from study records (e.g., costs of training and supervision). Data to estimate variable costs were obtained from computer timestamps (screen times, e-SBIRT intervention times), recorder timestamps (SBIRT intervention times), surveys of study personnel (e.g., service support times for screening), and study records (e.g., wages of research assistants). All labor costs included fringe benefits and overhead. Research-specific costs (e.g., incentive payments for study participation) were excluded from the analysis. Costs were estimated from both the healthcare provider and patient perspectives ⁵ and adjusted to 2016 US dollars using the Consumer Price Index.

Because most of the implementation costs are easily expressed on an annual recurring basis, the start-up (one-time) fixed costs for training and hardware were converted to an annual equivalent cost following Drummond et al. [38]. Following Barbosa et al. [39], annual costs were then normalized by the expected number of patients who screened positive each year and presented in terms of the average cost per positive screen. Transportation costs are excluded from the analysis (from the patient perspective) because screening took place during regularly scheduled reproductive health visits.

2.2.3. Cost-effectiveness analysis

The relative cost-effectiveness of the three conditions was determined using incremental cost-effectiveness ratios (ICERs) and cost-effectiveness acceptability curves (CEACs). The ICER measures the additional cost per unit of outcome gained [38, 40–41], and is defined in this study as the incremental cost of using a given intervention, compared to the next least costly intervention, to produce an additional day of abstinence during the 6-month follow-up period. Because ICERs are point estimates, CEACs are presented to illustrate the statistical uncertainty in our study due to our sample. CEACs show the probability that each intervention would be cost-effective, given the observed data, under different assumptions about the value of an additional day of abstinence [41,42]. Costs and effects for each condition were bootstrapped with 2,000 replicates to produce the CEACs

3. RESULTS

3.1. Cost analysis

Table 2 shows the respective base, favorable, and unfavorable scenario “average cost per positive screen” from the clinic perspective for each of the conditions, in total and disaggregated by variable cost vs. fixed cost. A detailed breakdown of the variable and fixed costs in the base scenario is presented in Tables 3 and 4, respectively. Table 5 shows the respective base, favorable, and unfavorable scenario “average cost per positive screen” from the patient perspective, in total and disaggregated by screen cost vs. intervention cost.

Table 2 –

Average total cost per positive screen – by scenario (clinic perspective)^a

	EUC (n = 151)	e-SBIRT (n = 141)	SBIRT (n = 136)
(a) Base scenario
Average variable cost per positive screen ($)	19.29	18.06	48.79
Average fixed cost per positive screen ($)	0.86	1.15	7.70
Average total cost per positive screen ($)	20.15	19.21	56.49
(b) Favorable scenario
Average variable cost per positive screen ($)	11.12	10.60	35.86
Average fixed cost per positive screen ($)	0.39	0.53	3.48
Average total cost per positive screen ($)	11.51	11.13	39.34
(c) Unfavorable scenario
Average variable cost per positive screen ($)	43.23	46.38	78.01
Average fixed cost per positive screen ($)	4.99	6.94	25.51
Average total cost per positive screen ($)	48.22	53.32	103.52

^aAssumptions underlying each scenario are described in Table 1.

Table 3 –

Detailed average variable cost per positive screen – base scenario (clinic perspective)

	EUC ($) (n = 151)	e-SBIRT ($) (n = 141)	SBIRT ($) (n = 136)
Screen
Receptionist timea	2.42	1.82	2.42
Spaceb	0.05	0.05	0.05
Materials (ear covers)c	0.48	0.48	0.48
Average Screen Cost per patient	2.95	2.35	2.95
Average Screen Cost per positive screend	16.26	12.96	16.26
Brief Intervention & Referral to Treatment (BIRT)
Interventionist timee	2.04	2.04	30.60
Spacef	0.07	0.15	1.01
Materials (brochure)g	0.92	0.92	0.92
License feeh	0.00	1.99	0.00
Average BIRT Cost per positive screen	3.03	5.10	32.53
Average variable cost per positive screen	19.29	18.06	48.79

^avalue of receptionist time * screen service support time, where value of receptionist time = $36.36/hr (i.e., $18.90/hr * 1.48 (fringe rate) * 1.30 (overhead rate)) and screen service support time = 4/3/4 minutes for EUC/e-SBIRT/SBIRT.

^btime for screen tasks * sq ft of space for screen tasks * cost of space, where time for screen tasks = 11.2/10.2/11.2 minutes for EUC/e-SBIRT/SBIRT, space = 15 sq ft, and cost of space = $.0184 per sq ft per hr.

^cTwo ear covers for tablet headphones (for privacy) @ $0.24.

^dAverage screen cost per positive screen equals the average screen cost per patient divided by the percentage of all screened patients who trigger an intervention. For example, the average screen cost per positive screen in the EUC condition in the base scenario = $2.95 ÷ 0.1813 = $16.26.

^eEUC and e-SBIRT: weighted mean value of provider time * time to deliver brochure, where weighted mean value of provider time = $61.15/hr (i.e., $37.33/hr * 1.26 (fringe rate) * 1.30 (overhead rate)) and time to deliver brochure = 2 minutes. SBIRT: $2.04 + weighted mean value of provider time * (intervention service support time + intervention time), where intervention service support time = 8 minutes and intervention time = 20.03 minutes.

^fEUC: 2 minutes to deliver brochure * 110 square feet (brochure delivered in private room) * $.0184 per sq ft per hr. e-SBIRT: (19.06 minutes for brief intervention * 15 square feet (BI done in reception area with tablet computer and headphones) + 2 minutes to deliver brochure * 110 square feet (brochure delivered in private room)) * $.0184. SBIRT: (20.03 minutes for brief intervention + 8 minutes of intervention service support time + 2 minutes to deliver brochure) * 110 square feet (all done in private room with/by clinician)) * $.0184.

^gBrochures cost $.92 each at copy center.

^hHypothetical license fee for patients who screen positive and receive intervention via tablet.

Table 4 –

Detailed fixed costs – base scenario (clinic perspective)

	EUC ($) (n=151)	e-SBIRT ($) (n=141)	SBIRT ($) (n=136)
ONE-TIME (START-UP) FIXED COSTS
Train receptionists (2) to process the screen
Trainer timea	103	69	103
Receptionist timeb	164	109	164
Spacec	2	1	2
Receptionist training subtotal	269	179	269
Train providers (8) to deliver brochure
Trainer timed	34	34	34
Provider timee	363	363	363
Spacef	1	1	1
Provider brochure training subtotal	398	398	398
Brochure
Time to collect information and design/layoutg	759	759	759
Tabletsh	3350	5025	3350
Storage locker for tablets	32	32	32
Train providers (8) to deliver SBIRT
Workshops
Trainer timei			5614
Provider timej			5864
Spacek			88
Materialsl			984
Workshops subtotal			12,550
Provider (8) time reviewing SBIRT training materialsm			5555
Total One-Time (Start-Up) Fixed Cost	4,808	6,393	22,912
Equivalent Annual Costn	1,019	1,355	4,857
ANNUAL (ON-GOING) FIXED COSTS
Supervision
Supervisor time rating tapeso			1974
Supervisor time giving feedbackp			1112
Interventionist time receiving feedbackq			1124
Spacer			9
Supervision subtotal			4,219
Total Annual (On-Going) Fixed Cost	0	0	4,219
AVERAGE FIXED COST PER POSITIVE SCREENs	0.86	1.15	7.70

^avalue of trainer time * time to train receptionists to deliver/process tablets for screen, where value of trainer time = $45.67/hr (i.e., $27.02/hr * 1.30 (fringe rate) * 1.30 (overhead rate)) and time to train receptionists = 2.25/1.50/2.25 hrs for EUC/e-SBIRT/SBIRT. Receptionist training time for e-SBIRT is less than EUC/SBIRT because no need for e-SBIRT receptionist to learn how to interpret ASSIST score. Assumes receptionists are trained together.

^bvalue of receptionist time * time to train receptionist to deliver/process tablets for screen * number of receptionists, where value of receptionist time = $36.36/hr (i.e., $18.90/hr * 1.48 (fringe rate) * 1.30 (overhead rate)), time to train receptionist = 2.25/1.50/2.25 hrs for EUC/e-SBIRT/SBIRT, and number of receptionists = 2.

^ctime to train receptionists * sq ft of space where training occurs * cost of space, where time to train receptionists = 2.25/1.5/2.25 hrs for EUC/e-SBIRT/SBIRT, space = 110 sq ft, and cost of space = $.0086 per sq ft per hr.

^dvalue of trainer time * time to train providers to deliver brochure, where value of trainer time = $45.67/hr (as above) and time to train providers = 0.75 hrs.

^emean value of provider time * time to train provider to deliver brochure * number of providers, where mean value of provider time = $60.54/hr (i.e., $36.96/hr * 1.26 (fringe rate) * 1.30 (overhead rate)), time to train provider = 0.75 hrs, and number of providers = 8. Two providers in the study did not complete certification, so only six providers provided brochures to patients.

^ftime to train providers to deliver brochure * square feet of space where training occurs * cost of space, where time to train providers = 0.75 hrs, space = 110 square feet, and cost of space = $.0086 per sq ft per hr.

^gvalue of designer time * time to design brochure, where value of designer time = $63.22/hr (i.e., $37.41/hr * 1.30 (fringe rate) * 1.30 (overhead rate)) and time to design brochure = 12 hrs.

^hcost of tablet * number of tablets, where cost of tablet = $1,675 and number of tablets = 2/3/2 for EUC/e-SBIRT/SBIRT. $1,675 is the mean cost of rugged tablets (tested for IP65 or MIL-STD810G durability) as reviewed in Computer World [37].

ⁱvalue of trainer time * time to deliver workshops, where value of trainer time = $112.80/hr (i.e., $66.75/hr * 1.30 (fringe rate) * 1.30 (overhead rate)) and time to deliver workshops to all 8 providers = 49.8 hrs.

^jweighted mean value of provider time * mean time providers spend in training * number of providers, where weighted mean value of provider time = $64.80/hr, mean time providers spend in training = 11.3125 hrs, and number of providers = 8. Two providers in the study did not complete certification, so only six providers delivered SBIRT.

^ktime to deliver workshops * sq ft of space where workshops occur * cost of space, where time to deliver workshops = 49.8 hrs, space = 205 sq ft, and cost of space = $.0086 per sq ft per hr.

^lnumber of providers trained * (cost of training manual + cost of digital tape recorder) + cost of training DVDs + cost of DVD player, where number of providers trained = 8, cost of training manual = $10.81, cost of digital tape recorder = $75.67, cost of training DVDs = $97.28, and cost of DVD player = $194.57.

^mweighted mean value of provider time * mean time providers spend reviewing SBIRT training materials outside of workshop * number of providers, where weighted mean value of provider time = $59.10/hr, mean time providers spend reviewing SBIRT training materials outside workshop = 11.75 hrs, and number of providers = 8.

ⁿone-time (start-up) fixed cost * annuity factor, where annuity factor depends on the assumed discount rate and useful life of the start-up activities. In the base scenario, the annuity factor is 4.7171 (assuming 3% discount rate, 5 years of useful life, and start-up costs paid at the beginning of the program).

^ovalue of supervisor time * annual number of hrs supervisor rates tapes, where value of supervisor time = $112.80/hr (as above) and annual number of hrs supervisor rates tapes = 17.5 hrs.

^pvalue of supervisor time * annual number of hrs supervisor gives feedback, where value of supervisor time = $112.80/hr (as above) and annual number of hrs supervisor gives feedback = 9.8 hrs. In trial, supervisor frequently provided feedback to multiple providers at same time.

^qweighted mean value of provider time * mean time providers spend receiving feedback annually * number of providers, where weighted mean value of provider time = $59.52/hr, mean time providers spend receiving feedback annually = 2.36 hrs, and number of providers = 8.

^rtime to supervise providers annually * sq ft of space where supervision occurs * cost of space, where time to supervise providers annually = 9.8 hrs, space = 110 sq ft, and cost of space = $.0086 per sq ft per hr.

^sThe average fixed cost per positive screen is equal to the annual (on-going) fixed cost plus the equivalent annual (start-up) fixed cost, all divided by the expected number of annual positive screens. For example, the average fixed cost per positive screen in the SBIRT condition in the base scenario = ($4,219 + $4,857)/(6,500*.1813) = $7.70.

Table 5 –

Average total cost per positive screen – by scenario (patient perspective)^a

	EUC (n = 151)	e-SBIRT (n = 141)	SBIRT (n = 136)
(a) Base scenario
Average screen cost per positive screen ($)b	1.64	1.61	1.61
Average intervention cost per positive screen ($)c	1.92	4.97	5.12
Average total cost per positive screen ($)	3.56	6.58	6.73
(b) Favorable scenario
Average screen cost per positive screen ($)	1.18	1.15	1.15
Average intervention cost per positive screen ($)	1.45	3.75	3.87
Average total cost per positive screen ($)	2.63	4.90	5.02
(c) Unfavorable scenario
Average screen cost per positive screen ($)	2.06	2.02	2.02
Average intervention cost per positive screen ($)	2.30	5.95	6.14
Average total cost per positive screen ($)	4.36	7.97	8.16

^aAssumptions underlying each scenario are described in Table 1.

^bvalue of patient time * (time to receive tablet and instructions from receptionist + screen time + time to return tablet), where value of patient time = $9.60/hr, time to receive tablet and instructions = 2 minutes, screen time = 7.72/7.55/7.55 minutes for EUC/e-SBIRT/SBIRT, and time to return tablet = 0.5 minutes.

^cEUC: $9.60/hr * (2 minutes to receive brochure + assumed 10 minutes to read brochure). e-SBIRT: $9.60/hr * (19.06 minutes to receive brief intervention via tablet + 2 minutes to receive brochure + assumed 10 minutes to read brochure). SBIRT: $9.60/hr * (20.03 minutes to receive brief intervention from clinician + 2 minutes to receive brochure + assumed 10 minutes to read brochure).

3.2. Effectiveness

Table 6 shows the average patient outcomes from the trial. As seen in Table 6, although the average number of all-substance abstinent days during the 6-month follow-up period was less than the average number of primary-substance abstinent days across conditions (as expected), the differences between e-SBIRT and the other conditions were similar (i.e., ~16.5 additional abstinent days for e-SBIRT vs. EUC, and no significant difference between the two SBIRTs). The trial found no evidence of heterogeneity in effect sizes between pregnant and non-pregnant women, although the trial was not powered to detect such differences [22].

3.3. Cost-effectiveness

3.3.1. Clinic perspective

Figure 1 presents the CEACs in the base scenario for each of the conditions from the clinic perspective. Each CEAC shows the probability that a given condition is cost-effective given the observed data. Note that the CEACs are a function of the threshold willingness-to-pay of the decision maker for an additional day of abstinence. As Figure 1 shows, e-SBIRT is likely (with probability > 0.5) to be cost-effective for any willingness-to-pay value for an additional day of abstinence (both primary-substance and all-substance). This result is not surprising inasmuch as the results in Tables 2 and 6 show that e-SBIRT is approximately as inexpensive as EUC and approximately as effective as SBIRT.

Figure 1:

Clinic perspective, base scenario. Cost-effectiveness acceptability curves for (a) days of primary-substance abstinence, and (b) days of all-substance abstinence. For a given threshold value, the probability of being cost-effective is equivalent to the proportion of the 2,000 bootstrapped replicates for which each intervention (EUC, e-SBIRT, SBIRT) had the highest net benefit [42].

3.3.2. Patient perspective

Figure 2 presents the CEACs in the base scenario for each of the conditions from the patient perspective. As Figure 2 and Tables 5 & 6 show, EUC is most likely to be the cost-effective intervention when the patient’s willingness to pay for an additional day of abstinence (both primary-substance and all-substance) is less than $0.18 (i.e., ($6.58 - $3.56)/(16.66 days or 16.35 days)), and e-SBIRT is most likely to be the cost-effective intervention when the patient’s willingness to pay for an additional day of abstinence is greater than $0.18.

Figure 2:

Patient perspective, base scenario. Cost-effectiveness acceptability curves for (a) days of primary-substance abstinence, and (b) days of all-substance abstinence. For a given threshold value, the probability of being cost-effective is equivalent to the proportion of the 2,000 bootstrapped replicates for which each intervention (EUC, e-SBIRT, SBIRT) had the highest net benefit [42].

Table 6 –

Average patient outcomes – by condition^a

	EUC (n = 151)	e-SBIRT (n = 141)	SBIRT (n = 136)
Days of Primary-Substance Abstinenceb	45.26 (55.59)	61.92 (64.10)	61.74 (63.43)
Days of All-Substance Abstinencec	28.01 (44.99)	44.36 (58.44)	39.17 (55.52)

^aValues are means with standard deviations in parentheses.

^bDifference between e-SBIRT and EUC was significant: 16.66 days (p-value = .018). Difference between SBIRT and EUC was significant: 16.48 days (p-value = .020). Difference between e-SBIRT and SBIRT was non-significant: 0.18 days (p-value = .982).

^cDifference between e-SBIRT and EUC was significant: 16.35 days (p-value = .008). Difference between SBIRT and EUC was non-significant: 11.16 days (p-value = .061). Difference between e-SBIRT and SBIRT was non-significant: 5.19 days (p-value = .450).

3.3.3. Sensitivity Analyses:

The remaining portions of Tables 2 and 5 show the average cost per positive screen in the favorable and unfavorable scenarios from the healthcare provider and patient perspectives, respectively. The results of the sensitivity analyses support the base scenario results described above.

From the healthcare provider perspective, as Table 2 shows, e-SBIRT is less expensive than SBIRT in all three scenarios, while e-SBIRT is approximately as inexpensive as EUC in all three scenarios. In the unfavorable scenario, compared to EUC, the incremental cost of using e-SBIRT to obtain an additional day of abstinence (both primary-substance and all-substance) is $0.36 (i.e., ($53.32 - $48.22)/(16.66 days or 16.35 days)).

From the patient perspective, compared to EUC, the incremental cost of using e-SBIRT to obtain an additional day of abstinence (both primary-substance and all-substance) ranges from $0.18 in the base scenario to $0.14 in the favorable scenario to $0.22 in the unfavorable scenario.

4. DISCUSSION

This study found that e-SBIRT – compared to EUC and SBIRT – is likely to be a good value, from both the healthcare provider and patient perspectives, for improving abstinence among women seeking medical care in reproductive health clinics. In short, e-SBIRT is approximately as inexpensive as EUC and approximately as effective as clinician-delivered SBIRT. This finding is robust to sensitivity analyses in which key parameters were varied widely to reflect alternative possible implementation scenarios.

The results illustrate the possibilities and limitations of using technology to deliver SBIRT. For example, although the screen was administered using a computer tablet, average per patient screen costs were similar to those found in other studies [43,44] because healthcare provider employees were factored in to provide the tablet, explain the screen to incoming patients, receive the tablet and prep it for the next patient, etc. Requisite staff time may decline as technology continues to grow in stability, ease of use, and familiarity. Using a computer to deliver SBIRT reduced the cost of clinician-delivered SBIRT by approximately one-half (unfavorable scenario) to two-thirds (base scenario), with similar patient outcomes.

The present study has several strengths. First, it is based on a randomized controlled trial that (a) enrolled a large group of women using a variety of substances, (b) used similarly formatted brief MI-based interventions to isolate potential differences in SBIRT delivery method, and (c) obtained high follow-up rates. Second, all cost data were collected prospectively alongside the trial and include both variable costs and fixed costs. Third, findings are supported by sensitivity analyses on key cost-related parameters. Fourth, detailed table notes containing unit costs, resource utilizations, and the assumptions underlying the economic evaluation are provided to enable both practitioners and researchers to model alternative implementation scenarios.

The study also has limitations. First, service support times in the base scenario were estimated using surveys of study personnel who performed the relevant tasks, as opposed to using time-motion methods [44]. However, the estimated mean service support times used in the base scenario of the present study are similar to those reported in the literature.⁶ Moreover, the findings are robust to sensitivity analyses that varied the mean service support times. Second, the trial was conducted within urban academic healthcare settings, thereby limiting the findings’ generalizability. For example, implementing e-SBIRT in alternative settings may necessitate modification costs which, presumably, would be covered by the license fee. Third, the trial did not include a “usual care” condition or a no treatment control, so the present study is able to shed light only on the cost-effectiveness of the SBIRTs compared to each other and to EUC. However, even if we make the conservative assumption that a zero cost “usual care” condition would result in the same outcome as EUC, e-SBIRT would still be a good value: $1.15 (($19.21 - $0)/(16.66 days)) and $0.39 (($6.58 - $0)/(16.66 days)) per additional day of primary substance abstinence from the healthcare provider and patient perspectives, respectively. Fourth, the economic analysis relied on imputed missing effectiveness data for 13.0% of participants. Listwise deletion (i.e., excluding participants with missing data) produced ICERs and CEACs that are slightly more favorable to both SBIRTs (results not shown). Fifth, the absence of QALYs as a patient outcome measure limits the comparability of our findings. Finally, although the Second Panel on Cost-Effectiveness in Health and Medicine [45] recommends adopting a healthcare sector and a societal perspective alongside any other relevant perspective, the data requirements necessitated by both of these perspectives are beyond the scope of this study.

In conclusion, this study adds to the growing literature on the value of SBIRT programs by providing the first cost-effectiveness analysis of SBIRT delivered to women in reproductive health settings as well as the first cost-effectiveness analysis of SBIRT delivered by a computer versus a clinician.