Assessment of an Electronic and Clinician-Delivered Brief Intervention on Cigarette, Alcohol and Illicit Drug Use Among Women in a Reproductive Healthcare Clinic

Abstract

Background:

Women are at highest risk for development of a substance use disorder during their reproductive years. We recently evaluated the efficacy of an electronic screening, brief intervention and referral to treatment (e-SBIRT) and a clinician-delivered SBIRT (SBIRT) compared with enhanced usual care (EUC) for reducing overall substance use among women recruited from reproductive health clinics. The present study assessed the impact of the SBIRT interventions within three primary substance subgroups: cigarettes, illicit drugs, and alcohol.

Methods:

This is a secondary analysis from a 3-group randomized trial comparing e-SBIRT and SBIRT to EUC. For the present study, participants (N=439) were grouped according to their primary substance: cigarettes, alcohol, or illicit drugs. Differences in days per month of primary substance use over time between treatment groups were examined using generalized estimating equations, modelling linear as well as quadratic effects of time.

Results:

Cigarettes were the most frequently reported primary substance (n=251), followed by illicit drugs (n=137) and alcohol (n=51). For primary cigarette use the interaction between the linear effect of time and treatment was significant for SBIRT (β (SE) = −0.067 (0.029), p=0.020), but not e-SBIRT, suggesting greater reductions in cigarette use over the first 3 months following treatment with SBIRT compared to EUC. However, the significant interaction of SBIRT with time-squared (β (SE) = 0.009 (0.004), p=0.049) showed that reductions in cigarette use attenuated over time, such that after month 3, monthly reductions in cigarette use were similar between groups. Results followed a similar pattern for primary illicit drug use among the e-SBIRT group in which the interaction of e-SBIRT treatment with linear time (β (SE) = −0.181 (0.085), p=0.033) and quadratic time (β (SE) = 0.028 (0.012), p=0.018) were statistically significant suggesting greater reductions in illicit drug use with e-SBIRT versus EUC, which attenuated with time. Neither SBIRT nor e-SBIRT was associated with a significant reduction in days of alcohol use per month, as compared to EUC.

Conclusions:

Reproductive-age women appear to respond differently to electronic- and clinician-delivered interventions, depending on their primary substance. SBIRT reduced use of cigarettes, and e-SBIRT reduced illicit drug use. Although neither intervention reduced primary alcohol use, the sample size was small (n = 51), suggesting a need for further testing in a larger sample.

1. Introduction

Women are at highest risk for development of a substance use disorder during their reproductive years (18–44) (Compton, Thomas, Stinson, & Grant, 2007). There are important general medical adverse effects for reproductive age women who misuse substances. For example, the prevalence of HIV among women who inject drugs is significantly higher compared to men (Des Jarlais, Feelemyer, Modi, Arasteh, & Hagan, 2012). In addition, substance use in pregnancy is a major risk factor for adverse pregnancy outcomes and increased infant morbidity and mortality (Cnattingius, 2004; Forray & Foster, 2015).

Most women of child-bearing age receive the majority of their medical care in reproductive health settings (Scholle & Kelleher, 2003; Stovall, Loveless, Walden, Karjane, & Cohen, 2007). Treatment interventions provided in these settings are thus likely to reach a high proportion of women who are suffering from or at risk of developing a substance use disorder, and could potentially leverage the tendency of women who are pregnant or planning to conceive to adopt positive health behaviors (Williams, Zapata, D’Angelo, Harrison, & Morrow, 2012). Screening, Brief Intervention, and Referral to Treatment (SBIRT) interventions within general medical settings for adult patients are effective for reducing unhealthy alcohol (Bertholet, Daeppen, Wietlisbach, Fleming, & Burnand, 2005; Whitlock, et al., 2004) and tobacco use (Hettema & Hendricks, 2010; Lai, Cahill, Qin, & Tang, 2010; Stead, Bergson, & Lancaster, 2008). However, the evidence for SBIRT interventions for illicit substance use in general medical settings is mixed, with some studies finding significant decreases in illicit drug use (Bernstein, et al., 2005) while others find no effect (Roy-Byrne, Bumgardner, Krupski, & et al., 2014; R. Saitz, Palfai, Cheng, & et al., 2014). Further, SBIRT interventions have proven difficult to implement, raising calls for consideration of technology-delivered methods (Marsch, Carroll, & Kiluk, 2014).

To this end we recently evaluated whether SBIRT delivered electronically or by clinician to women in reproductive healthcare settings is superior to enhanced usual care for any substance misuse and treatment-seeking for substance use disorders (Martino, et al., 2017). We found that, overall, SBIRT delivered via therapist or computer was significantly more effective than enhanced usual care in reducing days of primary substance use, though not for increasing rates of post-SBIRT treatment or self-help utilization. In that study, we operationalized “primary substance” as the substance participants self-identified as most problematic (with an additional requirement of having used that substance at least once in the past 28 days). The predominant substance in that analysis was tobacco in the form of cigarettes, which may have driven the results. In this study, we further assessed the impact of the SBIRT interventions by exploring outcomes in three primary substance subgroups (cigarettes, alcohol and illicit drugs) over the six months following the interventions. Given that multiple substance use is common, we also examined other secondary substance use, regardless of its primacy for the three subgroups. We were interested in evaluating the potential differential effects of SBIRT versus usual care on substance use among different substance subgroups.

2. Methods

2.1. Design

This is a secondary analysis from the Screening to Augment Referral to Treatment (Project START), a 3-group randomized trial that tested the efficacy of an electronically-delivered SBIRT (e-SBIRT) and a clinician-delivered SBIRT (SBIRT) compared with enhanced usual care (EUC) (Martino, et al., 2017). The target population was women in two reproductive healthcare settings screened for use of cigarettes, alcohol, illicit drugs or misuse of prescription medication. Following screening (see below for eligibility criteria) and baseline assessment, eligible women were randomized to and received one of the three treatment modalities on site. Randomization was via a computerized urn randomization program (Stout, Wirtz, Carbonari, & Del Boca, 1994) that balanced allocation across conditions on the primary substance used and on pregnancy status. In-person assessments were completed 1-, 3- and 6-month post randomization. The study was approved by an institutional review board and included a certificate of confidentiality from the National Institute on Drug Abuse, which protects identifiable research information from forced disclosure. In addition, in the state of Connecticut, substance use in pregnancy is not subject to mandatory reporting or testing for suspected drug use and is not considered child abuse. All data collected for research purposes, including urine toxicology results, remained confined to the research database and were not shared with providers in the clinic. Data collection occurred between April 20, 2011 and January 28, 2015.

2.2. Participants

We screened pregnant and non-pregnant women who presented for outpatient reproductive healthcare at one of two hospital-based clinics. Women provided verbal and written consent to be screened with the World Health Organization Alcohol, Smoking, and Substance Involvement Screening Test (ASSIST) (Humeniuk, et al., 2008) to determine preliminary eligibility. Women were eligible for randomization if they were at least 18 years old, reported use of a primary substance in the month prior to randomization, and had an ASSIST score of 4 or greater for any substance except alcohol, for which the standard cutoff of ≥ 11 was used for non-pregnant women and a revised standard of ≥ 6 was used for pregnant women. Women were ineligible for randomization if they participated in substance use treatment or self-help programs in the 3 months prior to screening, could not speak English, or would be unavailable for the follow-up assessments. Women who met eligibility criteria were then formally asked to participate in the study and provide written consent for participation in the treatment trial.

2.3. Assessments

The baseline assessment was conducted via computer and collected demographic information, the ASSIST, primary substance (nicotine/alcohol/marijuana/other illicit drug or prescription narcotic), substance use diagnoses from the Mini-International Neuropsychiatric Inventory 5.0.0 Clinician-Rated(Sheehan, et al., 1998) and Fagerstrom Test for Nicotine Dependence (FTND).(Heatherton, Kozlowski, Frecker, & Fagerstrom, 1991) At the 1-, 3- and 6-month follow-up visits, the FTND and post-SBIRT substance use treatment utilization were collected via computer. The Timeline Followback (L. Sobell & Sobell, 1992; L. C. Sobell, Brown, Leo, & Sobell, 1996) recall method was used to collect data on any substances used during the 28 days prior to baseline, and all days between follow-up visits; this was used to generate days of use per month (i.e. 28 days) for each substance used. Urine samples were collected at each assessment and analyzed by Redwood Toxicology Labs (Santa Rosa, CA) for drug (amphetamines, barbiturates, benzodiazepines, buprenorphine, cocaine, methadone, methaqualone, opiates, oxycodone, phencyclidine, propoxyphene, cannabinoids), alcohol (EtG/EtS) and nicotine (cotinine). The false negative rates for the e-SBIRT, SBIRT and EUC were low (0.050, 0.061, and 0.057), respectively suggesting similarity between groups in accuracy of self-reported substance use (Martino, et al., 2017).

2.4. Treatment Conditions

The e-SBIRT, previously described (Ondersma, Chase, Svikis, & Schuster, 2005), featured an interactive, three-dimensional, animated narrator to deliver components of a single, 20-minute MI-based brief intervention. The main components of the intervention are summarized in detail elsewhere (Ondersma, et al., 2005), and included: 1) feedback regarding the negative consequences of drug use, their self-reported readiness to change, and their drug use as compared to that of all adult women; (2) pros and cons of drug use and related change, chosen from a list of options - the positive and negative aspects of drug use; and (3) a summary and query regarding their interest in change, followed by optional goal-setting regarding drug use.

The SBIRT was a single, 20-minute MI-based intervention that followed the same sequence of MI components provided in e-SBIRT described above. SBIRT was delivered by trained study nurses (n=2), social workers (n=3), and one obstetrician-gynecologist with good fidelity as assessed via independent raters (Martino, et al., 2017). According to established standards, two-way random effects, average-measures intra-class correlation coefficients (ICC) for the fundamental and advanced MI strategy adherence and competence scores showed good to very good inter-rater reliability (adherence ICC: fundamental = .77 and advanced = .85; competence ICC: fundamental = .78 and advanced = .71). Raters indicated that MI-inconsistent items almost never occurred in the sessions, thereby severely restricting rating variance and disallowing meaningful ICC calculation. In both SBIRT arms, the intervention focused on the participant’s stated primary substance only. EUC involved a brief interaction (approximately 2 min) with patients about their substance use and review of a handout listing local treatment resources. The handout “enhanced” usual care in that it was not usually offered to non-study-enrolled women in the clinic.

2.5. Measures

2.5.1. Primary Substance Subgroups.

Based upon efficacy studies in the literature, we segregated participants into three primary drug subgroups for our analysis: cigarettes, illicit drugs, and alcohol. From the randomized participants, we grouped the subset of women who identified any illicit drug (including cannabis, cocaine, heroin, non-prescription use of benzodiazepines or opioids, methamphetamines, or other illicit drugs) as their primary substance into the illicit substance drug group. Similarly, the nicotine cigarette and alcohol groups were made up of those who self-identified those substances as primary. Participants were included in only one primary substance group. Primary substance was determined during screening by asking the following question: “Of all of the substances you have used in the past 28 days, which one would you say is the biggest problem for you? Which has caused you the most problems, or which have you used the most?”

2.5.2. Secondary Substance Subgroups.

We also evaluated other substance use based on levels of cigarette, alcohol or illicit drug use, regardless of whether or not the substance was identified as the participant’s primary drug. Secondary cigarette users were defined as non-pregnant women smoking one cigarette per day on average in the month prior to baseline assessment (Dierker, et al., 2008) and pregnant women reporting any smoking in the month prior to baseline assessment. Secondary alcohol users were defined according to the National Institute of Alcohol Abuse and Alcoholism (NIAAA) definition for non-pregnant women of more than three drinks on any day or more than seven drinks in a week in the month prior to baseline assessment (National Institute on Alcohol Abuse and Alcoholism, 2016). Pregnant women were classified as secondary alcohol users if they reported any alcohol consumption in the month prior to baseline. All women, regardless of pregnancy status, reporting any use of an illicit drug (i.e., marijuana, cocaine, heroin, non-prescription use of opioids or benzodiazepines, methamphetamines, or other illicit drugs) during the month prior to baseline were included in the secondary illicit drug use subgroup. Women could be included in more than one secondary substance use subgroup if they met criteria for more than one substance.

2.6. Data Analysis

The characteristics of participants were summarized using descriptive statistics. Differences between treatment groups in primary substance use (collapsing across type) were presented in the main paper (Martino, et al., 2017). For the present study, differences in days per month of tobacco products, illicit drugs, or alcohol (whichever was primary for that participant) over time between treatment groups were examined using generalized estimating equations (GEE) to account for within-subject correlation in the outcome variable (days per month of substance use), that was measured longitudinally, by specifying an auto-regressive correlation structure. Because our outcome was an overdispersed (i.e., variance > mean) count variable, we specified a negative binomial distribution with a log-link function. β coefficients were estimated using quasi-likelihood estimation methods, and sandwich standard errors were estimated. Explanatory variables included treatment group (0=EUC, 1=e-SBIRT, 2=SBIRT), time modeled as a continuous variable for months 0–6, the interaction between treatment and time, and pregnancy status (pregnant=0, non-pregnant=1). We included both the linear and quadratic effect of time to test for a curvilinear relationship between substance use and time since the effects of substance use treatments often attenuate with time. The interaction between treatment and time was the primary effect of interest. We also examined the three-way interaction between treatment, time and pregnancy status to see if differences in days per month of substance use over time between treatment groups differed between pregnant and non-pregnant women. We tested six different GEE models, one for each of the three primary and secondary substance subgroups. In each of these models, the days per month of substance use outcome was specific to the subgroup. For instance, days per month of alcohol use was the outcome for the subgroup of participants whose primary drug was alcohol and the subgroup of participants who were secondary users of alcohol. Cohen’s d effect sizes were calculated for each visit time point (baseline, and months 1,3,6). The effect size for change from baseline was calculated by subtracting the baseline Cohen’s d from the later time points’ Cohen’s d. Statistical significance was set at p < 0.05; since this is a secondary and exploratory analysis we did not adjust for multiple comparisons. All analyses utilized the intention-to-treat approach and were conducted using SAS 9.4 (Cary, NC).

3. Results

2,421 women were screened and 439 were randomized to one of the three conditions (e-SBIRT, SBIRT or EUC), for details on the participant flow in the trial please refer to the previously published flow diagram (Martino, et al., 2017). The subject characteristics by primary substance are presented in Table 1. Cigarettes were the most frequently reported primary substance (n=251), followed by illicit drugs (n=137) and alcohol (n=51). The study population was predominantly African-American, non-pregnant women in their mid-thirties, with a high school education or greater, not currently employed and not living with a spouse or partner. Women in the illicit drug subgroup identified the following as their primary drug: cannabis (n=90), cocaine (n=38), opioids (n=4), phencyclidine (n=3) and benzodiazepines/sedatives (n=2). Figure 1 shows a bar graph of each of the primary substance use categories and the proportion of concomitant substance use with each primary substance.

Table 1.

Baseline Characteristics of Randomized Participants by Primary Substance Subgroups

Characteristic	Cigarettes (N = 251)	Alcohol (N=51)	Illicit Drugsa (N = 137)
Age in years, mean (SD)	34 (11)	36 (12)	35 (11)
Race/Ethnicity, N (%)
African-American, non-Hispanic	151 (60)	41 (80)	101 (74)
Caucasian-American, non-Hispanic	41 (16)	4 (8)	13 (9)
Hispanic-American	44 (18)	5 (10)	16 (12)
Other Race/Ethnicityb	15 (6)	1 (2)	7 (5)
Marital Status, N (%)
Married / living with partner	120 (48)	21 (41)	56 (41)
Other living situation	131 (52)	30 (59)	81 (59)
Pregnant, N (%)	58 (23)	2 (4)	20 (15)
Education, N (%)c
Some high school or less	78 (31)	18 (35)	50 (37)
High school graduate	99 (40)	24 (47)	58 (43)
Beyond high school	73 (29)	9 (18)	28 (21)
Employment status, N (%)
Full time	43 (17)	8 (16)	15 (11)
Part time	52 (21)	12 (24)	18 (13)
Not working	156 (62)	31 (61)	104 (76)
Days/month using primary substance, mean (SD)	27 (4)	18 (9)	20 (10)
ASSIST primary substance score, mean (SD)d	21 (7)	25 (9)	23 (10)
Secondary Substance Subgroups, N (%)e
Nicotinef	247 (98)	24 (47)	93 (68)
Alcoholg	57 (23)	48 (94)	60 (44)
Illicit Drugsh	78 (31)	21 (41)	136 (99)
Treatment group, N (%)
Enhanced Usual Care	90 (36)	13 (25)	48 (35)
e-SBIRT	80 (32)	23 (45)	40 (29)
SBIRT	81 (32)	15 (29)	49 (36)

Abbreviations: SD, standard deviation; e-SBIRT, electronically-delivered Screening, Brief Intervention, and Referral to Treatment; SBIRT, clinician-delivered Screening, Brief Intervention, and Referral to Treatment; EUC, Enhanced Usual Care; ASSIST, Alcohol, Smoking, and Substance Involvement Screening Test;

^aIncludes cannabis (n=90), cocaine (n=38), opiates (n=4), phencyclidine (n=3), benzodiazepines/sedatives (n=2);

^bIncludes multiracial (n=16), Native American (n=5), Pacific Islander (n=1), and Romanian (n=1);

^cN=2 women missing education data;

^dPossible range of scores is 0 to 39 with higher scores indicating riskier use;

^eSome participants met at-risk criteria for more than one substance so column totals are greater than group sample size;

^f≥1 cigarette/day on average in past 30 days;

^g>3 drinks/day or >7 drinks/week in past 28 days;

^hany illicit drug use in past 28 days

Figure 1.

Bar graphs depicting primary substance use along with co-use of other substances. Primary substance is identified on the x-axis.

Eighty women who met criteria for use of a primary substance were pregnant. Of the 699 eligible screened patients, the proportion of women who consented to participate was similar between pregnant and non-pregnant women, 74.65% and 74.95%, respectively (p = 0.994), as was the proportion that were randomized to a treatment group, 59.15% and 63.96%, respectively (p = 0.290). On average, pregnant women were younger than non-pregnant women, and African-American women were less likely to be pregnant compared to the other racial/ethnic groups. No differences were found in marital status, education, employment status, baseline days/month of primary substance use, or treatment group assignment between pregnant and non-pregnant women. Pregnant women were more likely than non-pregnant women to identify nicotine as their primary drug, and less likely to identify alcohol as their primary drug. The number of pregnant women reporting secondary substance use was the following: cigarette (n=69), alcohol (n=14) and illicit drugs (n=36). Pregnant women were less likely to be in the alcohol secondary substance group compared to non-pregnant women.

The number of women meeting criteria for secondary substance use for each of the subgroups is shown in Table 1. Among women in the secondary cigarette use subgroup (n=364), 68% reported cigarettes, 26% reported illicit drugs, and 7% reported alcohol as their primary substance. For women in the secondary alcohol use subgroup (n=165), 29% reported alcohol, 36% reported illicit drugs, and 35% reported cigarettes as their primary substance. In the secondary illicit drug use subgroup (n=235), 58% of women reported illicit drugs, 33% reported cigarettes, and 9% reported alcohol as their primary substance.

Results from the GEE models are presented in Table 2 and are depicted visually in Figure 2. Table 3 shows modeled means for the number of days per month of use by treatment group for each primary substance and each secondary substance use subgroup. Days per month of substance use were lower on average among pregnant versus non-pregnant women, and this difference was significant in all models except among the illicit drug as primary drug subgroup. No significant three-way interactions between treatment, time, and pregnancy status were found in any model, so results are presented without this three-way interaction term. Findings by substance for both the primary and secondary substance subgroup analyses are presented below.

Table 2.

Results of generalized estimating equation models predicting days per month of substance use among primary and secondary substance subgroups

	Days / Month of Cigarette Use			Days / Month of Alcohol Use			Days / Month of Illicit Drug Use
	Primary N=251	Secondary N=364		Primary N=51	Secondary N=165		Primary N=137	Secondary N=235
Parameter	β (Std Error)	β (Std Error)		β (Std Error)	β (Std Error)		β (Std Error)	β (Std Error)
Intercept	3.324*** (0.017)	3.329*** (0.012)		2.749*** (0.129)	2.615*** (0.094)		3.081*** (0.069)	2.971*** (0.068)
Treatment group
e-SBIRT	−0.049 (0.029)	−0.021 (0.019)		0.350* (0.165)	0.216 (0.150)		0.003 (0.100)	−0.142 (0.108)
SBIRT	0.011 (0.020)	−0.001 (0.018)		0.037 (0.203)	−0.059 (0.146)		−0.117 (0.100)	0.002 (0.094)
EUC	Reference	Reference		Reference	Reference		Reference	Reference
Month (Linear effect of time)	−0.033** (0.012)	−0.056*** (0.013)		−0.328 (0.195)	−0.070 (0.088)		−0.099* (0.045)	−0.131** (0.044)
Month*Month (Quadratic effect of time)	0.001 (0.002)	0.004* (0.002)		0.041 (0.026)	−0.003 (0.014)		0.005 (0.007)	0.012± (0.007)
Treatment × (Linear) Time Interaction
Month * e-SBIRT	−0.052 (0.027)	−0.025 (0.024)		0.090 (0.211)	−0.238* (0.117)		−0.181* (0.085)	−0.053 (0.076)
Month * SBIRT	−0.067* (0.029)	−0.032 (0.024)		−0.099 (0.278)	−0.310* (0.130)		−0.077 (0.078)	−0.046 (0.067)
Month * EUC	Reference	Reference		Reference	Reference		Reference	Reference
Treatment × (Quadratic) Time Interaction
Month*Month * e-SBIRT	0.005 (0.004)	0.002 (0.003)		−0.014 (0.028)	0.035* (0.018)		0.028* (0.012)	0.009 (0.011)
Month*Month * SBIRT	0.009* (0.004)	0.004 (0.004)		0.011 (0.036)	0.048* (0.019)		0.014 (0.011)	0.007 (0.010)
Month*Month * EUC	Reference	Reference		Reference	Reference		Reference	Reference
Pregnant (vs. not pregnant)	−0.118* (0.055)	−0.145** (0.052)		−0.828*** (0.230)	−1.299*** (0.372)		−0.239 (0.173)	−0.482** (0.171)

Abbreviations: e-SBIRT, electronically-delivered Screening, Brief Intervention and Referral to Treatment; SBIRT, clinician-delivered Screening, Brief Intervention, and Referral to Treatment; EUC, Enhanced Usual Care;

^*p < 0.05;

^**p < 0.01;

^***p < 0.001

Figure 2.

Change in substance use over time by primary and secondary substance use subgroups. Graphs depict modeled means from the generalized estimating equation model for the mean number of days per month of cigarette (A) and (B), alcohol (C) and (D), and illicit drug (E) and (F) use for the primary substance and secondary substance subgroups, respectively. Model was adjusted for treatment group, linear and quadratic effects of time, treatment × (linear and quadratic) time interactions, and pregnancy status. Standard error bars are shown. Abbreviations: e-SBIRT, electronically-delivered Screening, Brief Intervention and Referral to Treatment; SBIRT, clinician-delivered Screening, Brief Intervention, and Referral to Treatment; EUC, Enhanced Usual Care.

Table 3.

Modeled Means by Subgroups

		Days/Month of Cigarette Use		Days/Month of Alcohol Use		Days/Month of Illicit Drug Use
	Month	Mean	(SE)	Mean	(SE)	Mean	(SE)
Primary Drug e-SBIRT	0	25.7	(0.6)	21.5	(2.3)	21.1	(1.5)
	1	23.8	(0.8)	17.4	(1.9)	16.5	(1.6)
	3	21.0	(1.2)	13.3	(2.1)	12.3	(1.9)
	6	19.0	(1.5)	13.4	(2.4)	12.9	(1.9)
SBIRT	0	27.3	(0.3)	15.7	(2.4)	18.7	(1.4)
	1	25.0	(0.7)	10.8	(2.4)	16.0	(1.3)
	3	22.0	(1.1)	6.9	(2.8)	13.2	(1.7)
	6	21.0	(1.3)	7.8	(2.8)	13.3	(1.9)
EUC	0	27.0	(0.4)	15.1	(2.0)	21.0	(1.4)
	1	26.2	(0.5)	11.4	(1.4)	19.2	(1.4)
	3	24.6	(0.7)	8.2	(2.3)	16.4	(1.6)
	6	22.7	(1.1)	9.2	(2.4)	14.2	(1.9)
Secondary Use e-SBIRT	0	26.6	(0.4)	15.2	(1.7)	15.7	(1.3)
	1	24.7	(0.6)	11.5	(1.3)	13.3	(1.2)
	3	22.0	(0.9)	8.0	(1.2)	10.8	(1.4)
	6	20.3	(1.2)	7.5	(1.2)	10.7	(1.4)
SBIRT	0	27.1	(0.3)	11.5	(1.2)	18.2	(1.2)
	1	25.0	(0.5)	8.3	(1.1)	15.5	(1.1)
	3	22.4	(0.9)	5.6	(1.1)	12.7	(1.3)
	6	21.5	(1.0)	6.0	(1.3)	12.4	(1.5)
EUC	0	27.2	(0.3)	12.2	(1.2)	18.1	(1.2)
	1	25.8	(0.4)	11.4	(1.5)	16.1	(1.2)
	3	23.8	(0.7)	9.7	(1.8)	13.6	(1.3)
	6	22.5	(1.0)	7.3	(1.2)	12.5	(1.4)

Abbreviations: SE, standard error; e-SBIRT, electronically-delivered Screening, Brief Intervention, and Referral to Treatment; SBIRT, clinician-delivered Screening, Brief Intervention, and Referral to Treatment; EUC, Enhanced Usual Care

Note: Means were modeled using generalized estimating equation models predicting days per month of substance use among primary substance and at-risk use subgroups. Models were adjusted for treatment group, linear and quadratic effects of time, treatment × (linear and quadratic) time interactions, and pregnancy status.

3.1. Cigarette Use

In the primary substance subgroup analysis, the GEE model showed that the change in number of days per month of cigarette use differed between treatment groups (see Figure 2, panel A). For the SBIRT group, the treatment group × linear time interaction term was negative and significant, and the treatment group × quadratic time interaction term was positive and significant, suggesting greater initial declines in days per month of cigarette use - but attenuation of these effects over time - compared to the EUC group (Table 2). The effect sizes for the change in days per month of cigarette use between baseline and months 1, 3 and 6, ranged from small to medium for SBIRT (Cohen’s d = 0.31, 0.40, and 0.24, respectively). Results followed a similar pattern for the e-SBIRT group, but the treatment by time interactions did not reach statistical significance. In the secondary substance analysis (Figure 2, panel B), no significant differences were detected between treatment groups in days of cigarette use per month.

3.2. Alcohol Use

There were no significant differences in the change in number of days of alcohol use per month between treatment groups in the primary substance analysis (Figure 2, panel C). In the secondary alcohol use subgroup (Figure 2, panel D), both e-SBIRT and SBIRT, compared to EUC, showed significantly greater initial declines in days per month of alcohol use, followed by the effects leveling off (Table 2). Effect sizes for the change in days per month of alcohol use between baseline and months 1, 3 and 6, compared to EUC, were small for SBIRT (Cohen’s d = 0.24, 0.31, and 0.07, respectively) and ranged from small to medium for e-SBIRT (Cohen’s d = 0.25, 0.41, and 0.24, respectively).

3.3. Illicit Drug Use

In the illicit drug subgroup of the primary substance analysis (Figure 2, panel E), e-SBIRT, compared to EUC, showed significant greater initial declines in days per month of drug use, followed by an attenuation of effects (Table 2). The effects sizes for the change in number of days per month of illicit drug use between baseline and months 1, 3, and 6 were small for e-SBIRT (Cohen’s d = 0.28, 0.37, and 0.11, respectively). SBIRT followed the same pattern of results, but none of the interaction terms reached statistical significance (Table 2). No significant differences between treatment groups were found in the secondary illicit drug subgroup (Figure 2, panel F; Table 2).

4. Discussion

This study examined the impact of SBIRT delivered electronically or by a clinician in reproductive health clinics on subgroups of primary and secondary users of cigarette, alcohol and illicit drugs. Effect sizes were not large, but were significant and notable considering they stemmed from a one-time, brief intervention. When compared to EUC, the SBIRT intervention significantly reduced the number of days of cigarette use per month among women who identified cigarettes as their most problematic substance. In contrast, effects on secondary cigarette use were not demonstrated. For women reporting alcohol as their primary substance there were no significant differences in use between treatment groups; however, for women with secondary alcohol use, both e-SBIRT and SBIRT, compared to EUC, showed significantly greater declines in days per month of alcohol use. For primary illicit substance use we found that e-SBIRT, compared to EUC, showed significantly greater decreases in days per month of drug use. Like cigarette use, when illicit drug use was examined according to secondary use, neither SBIRT modality demonstrated a significant intervention effect.

The finding that SBIRT is efficacious for primary cigarette use is consistent with the literature in this area (Aveyard, Begh, Parsons, & West, 2012; Ferreira-Borges, 2005; Fiore, 2008; Yilmaz, Karacan, Yoney, & Yilmaz, 2006). However, the lack of e-SBIRT effects on primary cigarette use is surprising given prior results in this area (Ondersma, et al., 2012; Strecher, Shiffman, & West, 2005). The lack of impact e-SBIRT and SBIRT on secondary cigarette use is not as surprising, since the motivational interview focused on enhancing change in the primary substance identified by the participant and thus differences for primary substances are likely to be more robust.

We did not show a significant effect for either SBIRT or e-SBIRT on primary alcohol use in this analysis. However, effect sizes for SBIRT and e-SBIRT (respectively, d = .24 and .25 at 1 month, and .31 and .41 at 3 months) on secondary alcohol use were similar to those for positive effects with tobacco and illicit drugs, suggesting that evaluation in a larger sample may be appropriate (n = 51 for primary alcohol use in this sample). Among the larger sample of women with secondary alcohol use (n = 165), SBIRT and e-SBIRT each showed significant effects. Notably, however, most women who reported using alcohol identified a different substance as their primary concern; only 29% of women with secondary alcohol use also identified alcohol as their most problematic substance. It is possible that reducing use of their primary substance also led to reductions in drinking. In addition, the efficacy of SBIRT is better established for hazardous alcohol use (Jonas, Garbutt, Amick, & et al., 2012; Madras, et al., 2009; O’Donnell, et al., 2014) than for heavy use or dependence (Richard Saitz, 2010); most of the women in our sample were comprised of risky rather than severe alcohol users.

Women who endorsed illicit drug use responded better to e-SBIRT, possibly due to the greater anonymity and freedom from stigma provided by a computer delivered intervention (Beatty, Chase, & Ondersma, 2014; Durant, Carey, & Schroder, 2002; Grekin, et al., 2010; Ostrea, Brady, Gause, Raymundo, & Stevens, 1992). Thus, it is worth further exploring the efficacy of technology-based interventions for more stigmatized substances, as well as for substance use in pregnancy, which also carries significant stigma.

This study had several strengths. It screened and enrolled a large group of women with a variety of substance use histories and included both pregnant and non-pregnant women. It also had high follow-up rates and biochemical corroboration of primary substance use. In terms of the SBIRT delivery, it used similarly structured brief MI-based interventions and independent verification of MI fidelity in the SBIRT condition. The study was conducted in urban academic hospital-based clinics that serve predominantly non-Hispanic African-Americans, and while this may limit the generalizability of our results, it is also a strength. Low income minority women are less likely to receive substance use treatment (Buser, 2009; Murphy, Mahoney, Hyland, Higbee, & Cummings, 2005), face more barriers to treatment delivery (e.g. lack of time, childcare, transportation, cost, etc.) (Merzel, English, & Moon-Howard, 2010), and experience a disproportionate burden of substance use-related diseases (Haiman, et al., 2006; Jemal, et al., 2008; US Dept of Health and Human Services, 1998). Therefore, the utility of our brief interventions among low-income minority women, which can be easily delivered in reproductive healthcare settings and overcome some of the treatment barriers experienced by this population, is a significant strength.

The main limitation of our study is that, as a secondary analysis, it was not designed to detect differences in use between substance groups; all analyses were thus exploratory, and findings should be interpreted cautiously. Another limitation is that we are unable to test or control for any potential assessment reactivity, since we did not include a screen-only group that did not receive any assessment or follow-up. However, several SBIRT studies have taken into account the lack of an unassessed control group by including a screen-only control group in addition to a control group that undergoes the same assessments as the brief intervention, and found no differences in outcomes between control groups with or without (screen-only) assessments (Cherpitel, et al., 2010; D’Onofrio, et al., 2012; Daeppen, et al., 2007). Furthermore, the actual participation in a randomized control trial, with the longer and repeated assessments, stipulation of long-term monitoring and formal consent process is likely to attenuate effect sizes. As such, the lack of a true usual care condition actually makes our results more conservative and likely an underestimate of the true effects of our interventions. Another limitation is that we are unable to comment on the characteristics of the women that declined to be screened, and whether women with more severe substance use declined to participate. Finally, despite the confidentiality of the toxicology results and the lack of legal repercussions for reporting substance use, some women, particularly those with more severe substance use or those reluctant to engage in treatment at baseline, may not be willing to disclose their substance use and therefore not benefit from our interventions.

5. Conclusion

Although exploratory, these results demonstrate the potential utility of delivering SBIRT interventions to address primary tobacco and illicit drug use in reproductive health settings with women who are not presenting for substance use treatment. There may be an interaction between SBIRT delivery modality (in person vs. electronic) and primary substance use among reproductive-age women when compared to enhanced usual care. Further research is needed to better understand how to increase the effect of SBIRT and the potential moderating factors that may be contributing to these differences in response to mode of delivery.

Highlights.

• SBIRT delivered electronically or via clinician in reproductive clinics was examined

• The impact of SBIRT on cigarettes, alcohol, and illicit drugs was evaluated

• Clinician-delivered SBIRT reduced use of cigarettes

• Electronically-delivered SBIRT reduced illicit drug use