Randomized Trial of an Innovative Electronic Screening and Brief Intervention for Reducing Drinking among Women of Childbearing Age

Abstract

Objectives:

To evaluate the efficacy of an innovative, self-administered, electronic Screening and Brief Intervention (e-SBI) in English and Spanish, “DrinkWise”, for reducing drinking among non-pregnant women of childbearing age.

Methods:

A parallel design, Phase I trial included 185 non-pregnant women reporting risky drinking (8 or more drinks in a week or 3 or more drinks in a day) who were recruited from two publicly funded Nutritional Assistance for Women, Infants and Children (WIC) program sites in the United States from 2016 to 2017. Participants were 18 to 44 years in age, 75% of Hispanic ethnicity, 44% Spanish speakers, 30% had not completed high school, and 15% were currently breastfeeding. Participants were randomized to receive (intervention condition, n=99) or not receive (control condition, no intervention, n=86) DrinkWise and followed at 3 and 6 months.

Results:

Women receiving DrinkWise had greater reductions in the odds of self-reported weekly alcohol use (OR=0.22, SE=0.12, p<0.01) and heavy alcohol use (OR= 0.23, SE=0.14, p<0.05) at 6-month follow-up than controls, with no group differences at 3-month follow-up. Compared to heavy drinking controls, heavy drinkers receiving DrinkWise showed a trend (p=.06) for greater reductions in drink (pour) size from 3- to 6-month follow-up.

Conclusions:

DrinkWise may be efficacious in reducing drinking among low-income women of childbearing age and provides a low cost tool for increasing access to recommended SBI among childbearing-age women. Studies should continue to build DrinkWise’s evidence base.

INTRODUCTION

Fetal alcohol spectrum disorders (FASD), a leading preventable birth defect, continues to be a public health issue in the United States(Ethen et al. 2009). Recent estimates indicate that FASD prevalence among first graders ranges from 1.1% to 5.0% (May et al. 2018). Delayed pregnancy recognition and fear of stigma, related to disclosing alcohol use, hamper efforts to reduce drinking among pregnant women(Jacobson et al. 2002). Recommended Screening and Brief Intervention (SBI) remains underutilized, particularly due to lack of staff and time resources in public health settings, where it is needed the most (Hingson et al. 2012, Agley et al. 2018).

Heavy drinking increases risk for unintended pregnancy and prenatal drinking. In 2011, 45% of pregnancies in the U.S. were unintended, with rates of unintended pregnancy among women below the federal poverty level being up to three times the national average (Finer and Zolna 2016). National estimates indicate that 3.3 million women of age 15 to 44 years are at risk for alcohol exposed pregnancies because they are drinking, sexually active, and not using birth control; and that 3 in 4 women who want to get pregnant do not stop drinking alcohol when they stop using birth control (CDC Newsroom). These data highlight the urgent need for primary prevention with non-pregnant women of childbearing age.

Self-administered electronic SBI (e-SBI) helps overcome barriers to in-person SBI implementation, allows for standardized and consistent SBI delivery, and can be especially useful in resource-strained busy public health settings. Recent e-SBI programs show promise in reducing drinking among women of childbearing age(Tenkku et al. 2011) but involve multiple sessions that increase user burden. In contrast, single-session e-SBI programs show a lack of effects on drinking at 2-, 3-, and 6-month follow-up(Delrahim-Howlett et al. 2011, Montag et al. 2015, Ondersma et al. 2016). In the larger SBI literature, repeated e-SBI is reported to be important for efficacy(Sundström et al. 2017) but several repeated administrations over an extended period of time do not increase efficacy(Jonas et al. 2012). Studies with childbearing-age women have yet to study efficacy of repeated e-SBI.

There are other gaps in research on e-SBI for childbearing-age women. While Hispanics are projected to be one-fourth of the U.S. population by 2050(U.S. Census Bureau 2004), e-SBI programs are generally available only in English. Including Hispanic women in prevention efforts is important. Hispanic women have the highest birth rates in the U.S. (Sutton and Mathews 2006), with fertility rates of 71 versus 58 per thousand for non-Hispanic White women in 2016(Child Trends 2018). Recent national data show a convergence in alcohol use patterns of White and Hispanic women of childbearing age, including in heavy drinking (Kanny et al. 2013). Despite research that shows that heavier drinkers pour and consume larger-than-standard drinks(Kaskutas and Graves 2000, Kaskutas and Graves 2001), studies have also not considered alcohol drink size. Interventions that reduce drink sizes could attenuate fetal effects of drinking by potentially reducing overall consumption and heavy drinking among childbearing-age women.

We developed an innovative, self-administered, e-SBI in English and Spanish that assesses the drinker’s beverage-specific drink sizes (pour size for a drink) and provides feedback to increase awareness about drinking, including on discrepancies between the drinkers’ and standard alcohol drink sizes. We previously reported feasibility and acceptability of a version of this e-SBI developed for use with pregnant women(Nayak et al. 2014). In the present study, we examined efficacy of e-SBI for non-pregnant women “DrinkWise”, administered twice, and report findings from a Phase I randomized trial with risky drinkers followed at 3 and 6 months.

METHODS

Setting and sample.

Participants were recruited at two Special Supplemental Nutrition Program for Women, Infants and Children (WIC) sites in northern California. Federally-funded, WIC provides nutrition and breastfeeding education, nutritious foods, and health and social service referrals for low-income pregnant, post-partum, and breastfeeding women and children up to age five.

Inclusion/Exclusion Criteria:

Non-pregnant, childbearing age (18–44) women, who had not been pregnant in the past six months were eligible for the trial if they had at least 6^th grade education, were fluent in English or Spanish, and met risky drinking criteria for drinking in the past 3 months. As in recent studies of SBI with women of childbearing age (Delrahim-Howlett et al. 2011), risky drinking was defined as the consumption of 3 or more drinks in a day or 8 or more drinks in a week. Consumption of 8 or more drinks in a week has also been included as a risk indicator for birth defects and growth deficiencies in several prior studies (Floyd et al. 2007, O’Connor and Whaley 2007, Feldman et al. 2012). Participants who reported alcohol-related problems (that they or someone they know was concerned about their drinking) or major depression (screening positive on the PHQ-8(Kroenke et al. 2009)), while not excluded, were referred for further screening and referral.

The Intervention.

Like other SBI programs, DrinkWise includes the assessment of drinking, feedback on drinking, information on standard sizes (14g pure ethanol in the U.S.) for alcoholic beverages (e.g., 12 ounces of beer), alcohol-related harms, and benefits of reducing consumption. It also includes SBI components of a personalized plan for reducing consumption with goal setting, an analysis of high-risk situations for drinking, and suggested coping strategies.

The normative feedback component in other SBI is replaced in DrinkWise by two innovative components: a) drink size assessment (DSA), to assess beverage-specific drink sizes for each alcoholic beverage consumed by the person using the program, using a life-size poster display of drink containers (9 glasses) placed above the study computer (“vessels display”), as well as on-screen photographs of these and other containers (glasses, bottles, cans); and b) drink size feedback (DSF), to educate the person using the program about standard drink sizes, and display discrepancies between her drink sizes and standard drinks, highlighting that the drinker could be drinking more than she thinks she does. Screenshots that illustrate DSA and DSF implementation are provided in a supplemental file.

DrinkWise takes about 10 minutes to complete. DrinkWise was developed independently, and without prior knowledge of or any relationship to DrinkWise Australia(https://drinkwise.org.au), an independent, not-for-profit organization established in 2005 for a healthier and safer drinking culture in Australia.

Procedure.

Women were informed about the study by large, brightly colored, posters on the program site walls and by WIC nutritionists. The nutritionists showed potential participants the separate, designated study area, where a Research Assistant provided more detailed study information and began the study program using a unique ID for each participant.

Participants used the computerized program to complete informed consent, screening and, if eligible for the trial, to complete baseline and follow-up assessments. Trial participants were randomized to receive DrinkWise (intervention condition) or no intervention (control condition) using computerized, simple randomization allocation (0 or 1 for the control or intervention condition, respectively). Each study computer had a 24-inch touch screen monitor and, on the wall above it, a life-size vessels display poster for the drink size assessment.

Women used the computerized program independently, with Research Assistants (RAs) available on-site to answer questions and address unanticipated problems. The RAs checked in with participants 6 weeks after the baseline and 3-month follow-up assessments, using text messages, telephone, or regular mail to update contact information as needed and to provide reminders about the next follow-up appointments. Participants were paid $5.00 if screened only and, if eligible for the trial, $30.00 for each assessment completed (baseline, 3- , and 6-month follow-up). Follow-up assessments were scheduled to coincide with participants’ WIC follow-up appointments. Because the trial and brief intervention were for non-pregnant women only, participants who became pregnant at follow-up were excluded from follow-up assessments.

As per practice in clinical settings and a prior study(O’Connor and Whaley 2007), where pregnant women were screened at each visit and re-administered the intervention if reporting any alcohol use, participants in the intervention condition who reported any drinking at the 3-month follow-up were re-administered DrinkWise (see Figure 1). Those in the control condition who reported drinking at the 3-month follow-up completed the drink size assessment (DSA) only. At the 6-month follow-up; all respondents drinking were administered DSA. This allowed for a better assessment of drinking quantity for both groups at each follow-up.

Figure 1.

Recruitment and follow-up overview

Participants were told that the study program would “ask about different drinks, such as alcohol and sweetened drinks, and may give you information on healthy drinking”. Hence, control participants may not have been aware that they did not receive the DrinkWise intervention. Computerized screening is reported to be perceived as less judgmental and elicit greater disclosure than in-person screening for other sensitive behaviors, such as inter-personal violence (Chang et al. 2012). Hence, the outcomes assessed via computer are assumed to be less biased than those assessed by study staff. For participant privacy, study information was kept completely separate from the WIC program and staff. The study was approved by the California Health and Human Services Agency Committee for the Protection of Human Subjects (project number: 15–04-1970) and the Public Health Institute’s Institutional Review Board (IRB #I15–002).

Measures.

Outcomes.

Alcohol use patterns are reported to be more sensitive indicators of e-SBI efficacy than abstinence(Rooke et al. 2010, Guide to Community Preventive Services 2012). As in prior studies with women of childbearing age(O’Connor and Whaley 2007, Delrahim-Howlett et al. 2011), we assessed self-reported frequency of alcohol consumption (collapsed into weekly or more often versus not), usual number of drinks of any beverage in a day, and maximum number of drinks of any beverage in a day.

Risky alcohol use, as defined previously, was a primary outcome. Information from the usual and maximum number of drinks in a day was used to determine if participants met the 3 or more drinks per day criteria for risky drinking. Given the dose-response relationship between drinking and FASD, the relevance of heavy drinking for teratogenic effects(O’Connor and Whaley 2007), and the increased risk for unplanned pregnancy and prenatal drinking associated with heavy drinking, we also included self-reported weekly alcohol use (once a week or more often) and heavy alcohol use or drinking as primary outcomes. Heavy drinking was defined as maximum consumption of 5 or more drinks of any beverage in a day.

Finally, drink size, unique to our study assessment, was examined as an outcome in post-hoc analyses. The developmental nature and relatively small sample size in our Phase I trial precluded estimates of variability in drink size, and thereby, planned analyses focused on drink size. Drink size assessment is described next.

Drink size assessment (DSA) determined the drink container or vessel (e.g., glass or bottle) and the pour size that the participant usually drank, separately for each beverage type consumed (e.g. beer, wine, spirits). The assessment, previously employed by Kaskutas and colleagues in a trial with pregnant women at a large health maintenance organization(Witbrodt et al. 2008, Armstrong et al. 2009) administered by social workers, was aided by a vessel display of actual glasses and by photographs of other drink containers with letter markings on each vessel to indicate pour size. Pour size was used to estimate: a) drink size for the primary (most frequently consumed) beverage, calculated as the ratio of the volume of the respondent’s poured drink size to the volume of a standard drink for the specific beverage type (e.g., 1.5oz/45ml for a drink of spirits); and b) corrected maximum quantity (e.g., a reported maximum of 2 drinks with a drink size of 1.5 was corrected to 3 drinks). Thus, drink size information was used to correct the maximum number of drinks consumed for the primary beverage by considering participants’ drink sizes rather than standard drink sizes. See screenshots in the supplemental file for other examples.

DSA was administered (as part of Drinkwise), to intervention participants at baseline, and at 3-month follow-up to all those reporting drinking since their baseline assessment. DSA was administered (without drink size feedback) to drinking control participants at the 3-month follow-up, and all drinking participants at the 6-month follow-up. This enabled us to correct self-reported maximum consumption amounts for both conditions at the 3- and 6-month follow-up.

Alcohol assessment increases awareness of drinking and can reduce drinking in controls via “assessment reactivity”(Kypri et al. 2004, Kypri et al. 2007), occluding differences between study conditions(Chang et al. 2005). Hence, DSA was not administered to controls at baseline, despite its usefulness in assessing drinking quantity, to limit assessment reactivity at baseline to that inherent in the assessment of drinking outcomes.

Our trial also included secondary outcomes to examine possible adverse effects of the intervention, including any tobacco use, any drug use, sweetened beverage use, and depression. Results suggest a lack of adverse effects, with no increases in secondary outcomes, at follow-ups for the intervention versus control condition. Further detailed information on secondary outcomes is beyond the scope of the current paper, given its focus on efficacy of the intervention in reducing alcohol use outcomes, but available at clinicaltrials.gov

Analysis.

Generalized estimating equations (GEE)(Diggle et al. 2002) tested DrinkWise’s efficacy, controlling for demographic differences between study condition and study sites. GEE models estimate population average or marginal models, and are often used in longitudinal studies because they account for within-subject non-independence (clustering) of observations across multiple waves of data collection and use all available data in model estimation(Twisk and de Vente 2002).

Participants who reported abstinence at follow-up (n=26, 17.1%) or missed the 3-month follow-up (n=4) did not received DSA (controls) or the DrinkWise re-administration (intervention condition) which assess and intervene with current drinking. Therefore, we included a term in all analyses to control for Drinkwise re-administration/DSA at 3-month follow-up (1 for yes, 0 for no). Approximately 70% (n=126 of 181) of all eligible participants had a value on 1 on this term.

We did not use any type of imputation methods to account for attrition per Twisk and Vente (Twisk and de Vente 2002, p. 337), who reported that not imputing at all may be better than any of the imputation methods applied when GEE is used to analyze a longitudinal dataset with missing data. All analyses were conducted in Stata v15.1(StataCorp. 2017).

DrinkWise efficacy was examined by testing for differences between study condition (e-SBI versus control) in outcomes at each follow-up compared to the baseline, that is change in drinking following each DrinkWise administration, hence at 3 months and 6 months each versus baseline. Because we re-administered DrinkWise, consistent with findings in the literature for efficacy of repeated SBI, the 6-month follow-up was the primary time point in the analyses. Given the 100% prevalence of risky drinking in both study groups at baseline (trial inclusion criteria), risky alcohol use at follow-ups was not compared to baseline. Instead, between-group differences at 3 month were assessed and changes from 3- to 6-month follow up in risky drinking were compared between study condition.

Continuous alcohol use variables (e.g., drink size) were log transformed to adjust for their non-normal distribution. Analyses included categorical terms for time, condition, and time-by-condition interactions. Predicted probabilities of outcomes were estimated, separately by condition, for significant interactions. Effect sizes were estimated for significant outcomes using intra-class correlations for the specific outcome, assumed power of .80 to detect a protective effect, medium squared multiple correlation (r²=.13) of the covariate to other covariates, and α=.05(Cohen 1988).

RESULTS

Over a 12-month period (February 2016 through February 2017), 664 women began the screening program, 18 quit part-way; 20 screened out as ineligible. Of the 626 women screened on alcohol use, 187 met criteria for risky drinking and 185 were randomized to the control (n=86) and intervention condition (n=99), respectively and completed the baseline assessment (see Figure 1).

Twenty-five participants completed only the baseline assessment (13.51%); 146 (78.9%) completed all assessments. At each follow-up, four women reported being pregnant (total=8) and were not further assessed. Follow-up rates of 82.2% (152/185) and 78.9% (146/185) at 3 and 6 months, indicated acceptable study retention(Fewtrell et al. 2008), consistent with prior studies using resource intensive, in-person assessment that reported follow-up rates of at least 73% at 6 months(Floyd et al. 2007, O’Connor and Whaley 2007). Attrition did not differ by study condition. Participants lost to follow-up (n=25) did not differ from those completing follow-ups in baseline demographics or alcohol use.

Baseline Participant Characteristics.

Women completing screening did not differ from the larger population of non-pregnant women at each site on any demographic characteristics (see Table 1). Trial participants were predominantly Hispanic (75%), 44% were Spanish speakers, 30% had not completed high school, and 70% had given birth more than once, respectively. Forty percent of trial participants had a child of age 1 year or younger, 15% reported currently breastfeeding.

Table 1.

Participant Demographics by Study Condition

	Control (n=86) % (n)	Intervention (n=99) % (n)
Age
18-20	3.5% ( 3)	1.0% ( 1)
21-29	41.9% (36)	52.5% (52)
30-39	47.6% (41)	44.4% (44)
40-44	7.0% ( 6)	2.0% ( 2)
Race/ethnicity
White/Caucasian	15.1% ( 13)	18.2% (18)
Black	4.7% ( 4)	6.1% ( 6)
Hispanic	76.7% (66)	72.7% (72)
Other	3.5% ( 3)	3.0% ( 3)
Marital Status+
Married or Living with Partner	74.4% (64)	57.6% (56)
Single, Divorced or Widowed	25.6% (22)	42.4% (43)
Language*
English	47.7% (41)	62.6% (62)
Spanish	52.3% (45 )	37.4% (37)
Education
Some high school or less	29.1% (25)	30.3% (30)
Completed high school	39.5% (34)	32.3% (32)
Some college or more	31.4% (27)	36.4% (37)
Age of youngest child
6 to 11 months	17.4% (15)	9.1% ( 9)
1-3 years	67.5% (58)	67.7% (67)
4-5 years	15.1% (13)	23.2% (23)
Currently Breastfeeding*	22.1% (19)	9.1% ( 9)
	Mean (SD)	Mean (SD)
Number of pregnancies (gravida)	3.06 (1.68)	2.75 (1.47)
Number of births (parity)	2.43 (1.24)	2.20 (1.11)

^*p<.05,

⁺p<.10

Participants receiving DrinkWise were less likely to use the program in Spanish or to be breastfeeding (p<.05), and tended to be single, widowed, or divorced (p<.10). Participants receiving DrinkWise were more likely than controls to report drinking two or more alcoholic beverage types (37.4%, n=37 vs 22.1%, n=19, p<.05) but did not differ on any other alcohol use measure at baseline (not shown in tables). Overall, about 5% (n=10) of all participants reported that they themselves or someone else was concerned about their drinking.

Alcohol Use Outcomes.

Table 2 presents alcohol use at baseline, 3-month, and 6-month follow-up, separately for participants in the control versus intervention condition.

Table 2.

Alcohol Use over Time by Study Condition

Assessment Time point	Baseline (N=185)		3-month Follow-up (N=152)		6-month Follow-up (N=146)
Alcohol Use Measure	Control (N=86) % (n)	Intervention (N=99) % (n)	Control (N=72) % (n)	Intervention (N=80) % (n)	Control (N=68) % (n)	Intervention (N=78) % (n)
Abstinencea			22.2% (16)	12.5% (10)	27.9% (19)	24.4% (19)
Risky Alcohol Usea	100.0% (86)	100.0% (99)	66.7% (48)	75.0% (60)	60.3% (41)	57.7% (45)
Weekly Alcohol Use (Once a week or more)b	16.3% (14)	22.2% (22)	22.2% (16)	21.3% (17)	23.5% (16)	11.5% ( 9)
Heavy Alcohol Use (Maximum Quantity (any beverage) 5+)	18.6% (16)	19.2% (19)	25.0% (18)	17.5% (14)	23.5% (16)	9.0% ( 7)
Drink Size-related Measures		M (SD)	M (SD)	M (SD)	M (SD)	M (SD)
Primary Beverage Drink size (compared to standard)c		0.8 (0.9)	0.8 (0.9)	0.8 (0.7)	0.7 (0.7)	0.7 (0.6)
Drink-Size-Corrected Maximum Quantity (Primary Beverage)c		2.3 (3.4)	2.5 (3.3)	2.6 (5.2)	2.3 (2.5)	1.8 (1.8)
Heavy Drinkers only		M (SD), n=19	M (SD), n=13	M (SD), n=15	M (SD), n=10	M (SD), n=12
Primary Beverage Drink size (compared to standard)c		0.96 (1.45)	0.74 (1.04)	1.13 (1.21)	0.76 (0.51)	0.6 (0.36)
Drink-Size-Corrected Maximum Quantity (Primary Beverage)c,d		4.16 (2.01)	3.63 (4.37)	5.49 (10.43)	3.15 (2.92)	1.97 (1.42)

^aHad to be a risky drinker at baseline,

^bAsked at follow-up only to drinkers,

^cNot assessed in controls at baseline,

^dquantity in terms of standard size drinks

Note: 1. Valid percentages (of participants at each time point are presented. A total of 156 participants were re-assessed at the 3-month follow-up and but 4 each reported that they had become pregnant since the last interview and, at that point, became ineligible to continue in the study.

2. Drinking quantity refers to number of drinks in a day, Heavy drinkers reported a maximum quantity of 5 or more drinks in a day

Table 3 presents results from models testing differences between study conditions in drinking outcomes over time. Odds ratios (OR) and standard errors (SE) are presented for dichotomous outcomes (weekly drinking vs not) and coefficients and SE for continuous outcomes (e.g., drink size). Study attrition was relatively minimal and did not differ by study condition or by demographics and alcohol use assessed at baseline.

Table 3.

Tests of Differences between Study Condition in Changes in Drinking Over Time

Alcohol Use Measurea	Risky alcohol use		Weekly Drinking (Once a Week or More)		Heavy alcohol use (maximum, any beverage) 5+ in a day		Primary Beverage Drink Size (relative to a standard drink)		Drink-Size- Corrected Maximum Quantity (Primary Beverage)
	OR	SE	OR	SE	OR	SE	Coef	SE	Coef	SE
Time Effect
3m (vs Baseline)	-		1.53	0.60	1.76	0.68
6m (vs Baseline)	-		2.12+	0.87	2.14+	0.86			−0.05	0.17
Condition Effect
Intervention (vs Control at Baseline or 3mb, c)	1.48	0.57	1.26	0.54	0.73	0.30			−0.08	0.19
Time*Condition Effect
Baseline to 3m Intervention (vs Baseline to 3m Control)b	−		0.58	0.29	0.48	0.25
Baseline to 6m Intervention (vs Baseline to 6m Control)	−		0.22**	0.12	0.23*	0.14
3m to 6m Intervention (vs 3 m to 6m control)	0.85	0.36							−0.12	0.23
Heavy Drinkers d (Drink Size only)
Time*Condition Effect
3m to 6m Intervention (vs 3m to 6m Control)							−0.82+	0.44	−0.78	0.57

⁺<.10,

^*<.05,

^**<.01; 3m=3-month follow-up, 6m=6-month follow-up

^aOutcomes from planned and post-hoc analyses (on drink size) are shown;

^bDrink size data and corrected maximum amounts are not available for the control condition at baseline;

^cFor risky drinking, the condition effect is at 3 month only, given no differences at baseline due to risky drinking being trial inclusion criteria.

^dHeavy drinkers reported a maximum quantity of 5 or more drinks in a day

Note: GEE Models adjusted for demographic differences between the intervention and control conditions at baseline and between study site (e.g., race/ethnicity, language, marital status, educational attainment, and breastfeeding status), and controlled for whether the participant was still drinking at the 3-month follow-up and therefore received the drink size assessment/intervention at that follow-up.

Primary Outcomes.

Table 3 shows significant Time-by-Condition effects for two primary outcomes at the 6-month but not the 3-month follow-up. Compared to controls, women receiving the DrinkWise intervention had a greater reduction in the odds of self-reported weekly alcohol use (OR=0.22, p<0.01) and heavy alcohol use at 6 months compared to baseline (OR= 0.23, p<0.05). There were no significant differences in risky alcohol use by study condition at 3 months and changes in risky drinking from 3- to 6-month follow-up did not differ by study condition.

Figure 2 shows the predicted probabilities for significant outcomes, self-reported weekly alcohol use and heavy alcohol use, at baseline and follow-up, illustrating the contrast between study conditions over time. For both frequent and heavy drinking (ICC=0.75, 0.63 respectively), the detectable ORs at the population level, with a .80 power was estimated to be less than .10, indicating a protective effect of medium size for DrinkWise.

Figure 2.

Predicted Probabilities for Alcohol Use Outcomes with greater reduction in the DrinkWise (e-SBI) condition versus controls

Post-hoc findings.

Small average drink sizes (< 1, see Table 1) reported by participants indicated that a floor effect may have masked between study condition differences in changes in drink size over time (Type II error). Given that heavy drinkers consume larger than standard drinks, we conducted post-hoc analyses of changes in drink size among participants who reported self-reported heavy drinking (5 or more drinks of any beverage in a day) at baseline (last rows, Table 1 and 2). A trend was found for a larger reduction in drink size of the primary beverage, from 3 to 6 months (Coeff= −0.82, p=0.06), among heavy drinkers in the intervention condition than heavy drinking controls. Drinkers reporting 5 or more drinks in a day who received DrinkWise (n=12) reduced their primary beverage drink size by almost half (1.1 to 0.6 times a standard drink), while drink size remained unchanged for heavy drinking controls in the control condition. Drink size corrected maximum quantity for these drinkers reduced from 5 drinks at 3 months to under 2 drinks at 6 months while that for the controls remained at about 3 drinks. However, differences by study condition in changes in primary beverage maximum quantity failed to reach significance.

Given the trend in drink size reductions among those reporting 5 or more drinks in a day, we also conducted post-hoc analyses using a lower threshold (4 or more drinks in a day) for heavy drinking (not shown in table). This second heavy drinking variable was consistent with exceeding daily limits specified in the NIAAA guidelines for low-risk drinking, and with the lack of a known safe threshold for drinking during pregnancy and increasing risk for FASD risk with increasing consumption amounts (Feldman et al. 2012). A statistically significant difference was found between study condition in change in primary beverage drink size from 3-month to 6-month follow-up (Coeff= −0.54, SE=0.27, p=0.02) among those reporting 4 or more drinks in a day. Those reporting 4 or more drinks in a day who received DrinkWise (n=25) reduced their primary beverage drink size (Mean: 0.96 at 3 months to 0.59 at 6 months) more than controls reporting 4 or more drinks in a day, whose drink size remained unchanged (Mean=0.8; n=17). Differences in changes in drink-size corrected maximum quantity between groups were not statistically significant.

DISCUSSION

We report on the efficacy of an innovative, user-friendly, self-administered e-SBI in English and Spanish, “Drinkwise”, in reducing drinking among childbearing-age women. Compared to controls, women completing DrinkWise showed a greater reduction from baseline to 6-month follow-up in the likelihood of self-reported weekly and heavy alcohol use than controls. These larger reductions in drinking for intervention participants cannot be explained by assessment reactivity as controls did not reduce their drinking, despite receiving drink size assessment at the 3-month first follow-up. Birth defects and growth deficiencies associated with prenatal alcohol exposure are documented to be dose-related without evidence of a threshold(Feldman et al. 2012). Any reduction in frequency of drinking and of heavy drinking, thus, can reduce fetal effects, particularly among women prior to pregnancy recognition, including those unable to abstain after stopping birth control to conceive (nearly 75% (CDC Newsroom)) or during pregnancy (about 10 to 11% of pregnant women (Chasnoff et al. 2005, Tan et al. 2015)).

The lack of efficacy findings for risky alcohol use may be due to lack of changes in weekly consumption quantity. Weekly quantity has been shown to not be useful in predicting alcohol-related harms, while exceeding daily limits (e.g., 4 or more drinks for women) was a strong predictor of alcohol harms(Dawson et al. 2012). Hence, estimated medium effect sizes for reducing weekly and heavy alcohol use, along with the low cost advantage of e-SBI(Ondersma et al. 2016), substantiate DrinkWise’s promise as a tool to increase women’s access to SBI, especially in busy public health settings. Our findings argue for the continued evaluation of DrinkWise’s as a FASD prevention tool.

Post-hoc analyses that showed greater drink size reductions among heavy drinkers in the intervention versus control condition also suggest DrinkWise efficacy. Even though differences in drink-size-corrected maximum quantity failed to reach statistical significance, likely due to the small sample sizes for heavy drinkers (n=10 and 12), the mean maximum quantity for heavy drinkers receiving DrinkWise reduced from over 5 to less than 2 drinks. Consistent with observations that heavy drinkers may need drink size feedback the most(Kaskutas and Graves 2001), greater drink size reductions among heavy drinkers receiving DrinkWise suggest that these drinkers may especially benefit from DrinkWise.

Our study has several strengths. To our knowledge, this is the first study to report on efficacy of a bilingual e-SBI. An online search using Pubmed, Ebscohost, Google Scholar and the web did not yield any published reports of e-SBI in English and Spanish. Three-quarters of our study participants were Hispanic and one-third (37%) of all those receiving our e-SBI “DrinkWise” completed it in Spanish. Our study’s contribution in demonstrating e-SBI efficacy with low-income Hispanic women is particularly significant as low-income minority women are less likely to receive appropriate counseling for prenatal drinking(Abel 1995).

We also demonstrated successful implementation of our personnel-free e-SBI in a busy, public health setting and obtained higher follow-up rates than in recent e-SBI studies with childbearing-age women(Ondersma et al. 2016). Prior studies that reported no efficacy for e-SBI used in-person or telephonic follow-up assessments(Delrahim-Howlett et al. 2011), which could have resulted in greater under-reporting of drinking compared to in our study. We previously reported higher rates of drinking among pregnant women using e-SBI compared to that obtained by WIC staff(Nayak et al. 2014). Increased detection of drinking enabled by the use of electronic assessment can be particularly useful to reduce Type II errors in trials, as under-reporting of drinking, particularly by controls, can lead to null findings and the premature discarding of promising interventions.

SBI re-administration has previously not been used in studies with childbearing-age women. Lowered odds of reporting weekly and heavy drinking for the intervention versus control condition found at 6 months and not at 3 months, i.e., following the DrinkWise re-administration, suggest the need for re-administering e-SBI for efficacy. SBI re-administration is documented to maintain, or enhance reductions in the larger SBI literature(Jonas et al. 2012). Administering SBI twice may be the most feasible re-administration delivery format for resource-limited public health settings.

Another methodological innovation of our study was assessing drink size for both study groups at follow-ups to better measure drinking. While, overall, those receiving and not receiving DrinkWise did not differ in changes in drink size over time, heavy drinking participants, showed a trend for greater reduction in drink size when receiving DrinkWise versus usual care. Our study is among the first to show that e-SBI may be particularly efficacious for childbearing-age women who report heavy drinking.

Our study limitations include those inherent in a Phase I trial, such as relatively small sample sizes, limited number of study conditions, and the lack of a true control (placebo) group. Online activities on health behaviors other than drinking, such as nutrition, for the control group, would provide a time control to equalize study conditions on time spent on the computer. Our follow-up periods were also relatively short and the use of simple randomization, separately within each of the two study sites, may have resulted in a slightly larger intervention group (n=99 vs 86 for the control condition). Future larger trials should include longer follow-ups, block randomization methods to achieve study group balance, and a time control for those not receiving the intervention.

Drink sizes for study groups were not compared at baseline, given the lack of drink size assessment (DSA) for controls and the lack of baseline differences between study condition in drinking. Findings for drink size reduction for those receiving DrinkWise occurred only from 3 to 6 months and not from baseline to 3 months, a change that may be tied to receiving the drink size feedback twice for those who were still drinking at 3 months (roughly 70% of participants). Controls drinking at 3 months received only DSA and received it only once (not at baseline), and did not show any change in drink size at 6 months. A larger trial could examine efficacy of repeated drink size assessment and drink size feedback on changes in reported drink size using a multiple arm design (e.g., including groups with and without DrinkWise re-administration and drink size assessment). Using multiple groups was outside the scope of our Phase I trial.

A multi-arm trial that includes control groups with and without the drink size assessment, including at baseline, and an intervention group without drink size feedback would also help address a study limitation of drink size assessment and feedback possibly confounding reported changes in drinking. Multiple control and intervention groups can help examine whether drink size assessment has an assessment reactivity-related effect on reducing drinking at follow-up and if drink size feedback increases accuracy of reporting actual consumption at follow-up, such as more drinks, especially by those with larger-than-standard drink sizes. Larger trials should also include more detailed assessment of alcohol consumption to better assess FASD risk, by considering the strength of the beverage consumed (e.g., 3% in a light beer versus 7% in malt liquor) and binge consumption via quantity consumed in a 2-hour period in a day.

We had very low, albeit expected, pregnancy rates at follow-ups (under 5% at 6 months) and also did not assess contraception use, an important risk factor for alcohol-exposed pregnancies. Given that DrinkWise focuses on drinking among non-pregnant women, we were unable to examine drinking among participants who became pregnant at follow-up. A larger trial that assesses contraceptive use and includes a DrinkWise version for pregnant women can assess its true preventive impact on prenatal drinking. Larger trials should also oversample heavy drinkers to assess for possible moderating effects for heavy drinking on DrinkWise efficacy. Despite these limitations, inherent in a Phase I trial, our findings indicate efficacy for DrinkWise and document its potential for use in low-resource settings and with Spanish speakers.

CONCLUSIONS

“DrinkWise”, an electronic SBI that includes drink size assessment and feedback, can reduce weekly and heavy alcohol use in childbearing-age women, thereby reducing risk for poor fetal outcomes associated with maternal alcohol use. More research that builds the evidence base for this innovative bilingual e-SBI is indicated. Future trials should systematically examine effects of re-administering DrinkWise, including its innovative components of drink size assessment and feedback. These trials should use larger samples, a multi-arm design, longer term follow-up to study longer-term efficacy, and examine possible moderating effects for heavy drinkers, including those reporting binge drinking.