Characteristics of Women with Alcohol Use Disorders Who Benefit from Intensive Motivational Interviewing

Abstract

Background:

Women with alcohol disorders have more severe problems related to their drinking than men. They have higher mortality from alcohol-related accidents and enter treatment with more serious medical, psychiatric, and social consequences.

Objective:

This study assessed the effects of Intensive Motivational Interviewing (IMI), a new, 9-session counseling intervention for women with drinking problems.

Methods:

A randomized clinical trial was conducted with 215 women. Most were white (83%), college educated (61%), and older (mean age 51). Half received IMI and half a standard single session of MI (SMI) along with an attention control (nutritional education).

Results:

Generalized estimating equations models showed women who were heavy drinkers at baseline in the IMI condition reduced heavy drinking more than those in the SMI condition at 2-, 6-, and 12-month follow-up. Analyses of disaggregated subgroups showed IMI was most effective for women with low psychiatric severity, more severe physical and impulse control consequences associated with drinking, and higher motivation. However, formal 3-way interaction models (condition by moderator by time) showed significant effects primarily at 2 months.

Conclusions:

Improvements associated with IMI were limited to heavy drinking and varied among subgroups of women. Studies of women with more diverse characteristics are needed.

Reviews of alcohol studies targeting women raise a number of specific concerns that need attention. Relative to men, women with alcohol disorders present more serious health, mental health, and social consequences related to their drinking.^1,2 However, despite more serious consequences, women are less likely than men to receive help³ and interventions for women with alcohol use disorders have been understudied relative to men.¹

Studies show women perceive integrated programs that address a variety of issues, including mental health and parenting, as the most helpful.⁴ However, most programs have not sufficiently considered how to adapt approaches to be more gender responsive. A limited number of studies provide some support for gender-specific programs that are specifically designed for women, while other studies show mixed gender programs are equally effective.^1,5 Whether programs are mixed or gender specific, there is evidence indicating those that adapt services to be responsive to the needs of women, such as addressing psychiatric symptoms, trauma, pregnancy, parenting, and relationship issues have better outcomes.^2,6

Although research shows that a number of treatment interventions impact alcohol problems among women and men, the effect sizes found are typically modest.⁷ In addition, some longitudinal studies show interventions affect outcomes such a retention, but not use of alcohol or drugs.⁸ Other studies show improvements decrease over time, especially for individuals with more serious alcohol problems.⁹ There is a clear need for interventions that have larger and more sustained impacts. However, interventions designed to maximize and sustain improvement may differ by gender.

Development of Intensive Motivational Interviewing

This study reports findings on an intervention designed to effect significant, sustained reductions in drinking among women with alcohol use disorders, intensive motivational interviewing (IMI). IMI builds on all of the standard supportive and directive interventions emphasized in motivational interviewing (MI), including reflections, open questions, summary statements, affirmation of strengths, developing discrepancies, and change planning,¹⁰ However, IMI expanded the number of sessions to nine, rather than the standard one or two sessions used in MI. Descriptions of the development of IMI can be found in several previous publications.^{10, 11, 12}

Initially, IMI began as a treatment of methamphetamine (MA) dependence.¹³ The rationale for developing a more intensive, 9-session dose of MI that went beyond the standard single MI session included the view that a higher number of sessions delivered over a longer period of time might be more effective for persons with more severe drug and other co-occurring problems.¹⁴ Nine sessions were chosen because that was viewed as being significantly more substantive than the standard single session of MI yet brief enough to keep a focus on MI interventions and avoid becoming a more generic form of therapy.

In addition to the larger number of sessions, there are operational differences between standard MI and IMI. For example, an innovative aspect of IMI was that it was designed to expand implementation of the change plan over multiple sessions. In standard MI interventions, the change plan is typically developed during the final session. Thus, it is a static process with no opportunity to monitor how the change plan proceeds or adjust over time. In contrast, the change plan in IMI is continually evolving over time. IMI emphasizes modifications of targeted goals and plans for achieving them as well as new goals being developed as stated goals are met.

After the development of an IMI manual for MA dependence, a modified version of IMI was developed to address the needs of persons on probation or parole living in sober living recovery homes (SLHs).¹⁵ Motivational interviewing case management (MICM) combined MI-based techniques with standard case management to assist new residents in several respects: 1) adapt to the new living situation, 2) access needed services in the community, 3) comply with terms of probation or parole, and 4) reduce the risk of HIV infection and transmission to others. Instead of 9 weekly sessions, MICM includes three weekly sessions followed by monthly sessions up to one year. The monthly sessions are designed to be flexible and responsive to individual needs; often they are shorter duration and conducted by phone.

Structure of IMI Manuals

Although there have been several variations of IMI manuals addressing different populations in different clinical contexts,¹⁰ all of the manuals retain all the standard supportive and directive interventions of MI-based approaches. The overall focus for the first three sessions is taken from the MI manual intervention developed by Obert and Farentinos¹⁶ for NIDA Clinical Trials Network (CTN) studies. The focus for first session is “problem identification.” The second session focuses on ambivalence, which includes the pros and cons of making a change in substance use. The primary goal is to meet the client where they are at in terms of their drinking and establish a therapeutic alliance. During subsequent sessions, the therapist guides the interaction to expand on reasons to make changes. This represents an effort to increase “change talk,” which has been found to be associated with improved outcomes.¹⁷ In the third session, the client and therapist collaborate on developing the change plan, which specifies changes the client wants to make and strategies for achieving them. Sessions four to eight focus on implementation of the change plan, which might include discussion of achievements, setbacks, revision of goals, revision of strategies for achieving goals and development of new goals. During session 9, the therapist and client focus on termination by reviewing what has transpired over the course of therapy, gains the client has been made, and ongoing plans for change.

Research on IMI

Formal testing of the first version of IMI began with initial pilot testing of the manual used to treat persons with methamphetamine (MA) dependence.¹³ Based on encouraging findings showing significant decreases in MA use after treatment, a randomized clinical trial of 217 MA dependent persons was conducted. At 6 months, those who received IMI reduced psychiatric symptoms more than a comparison group consisting of a single session of MI that was enhanced with a nutrition education groups to achieve time equivalence between the two study conditions.¹⁸ However, MA outcomes did not differ between the treatment and comparison group. An unexpected finding was that among women who had co-occurring alcohol problems (N= 87 ) those who received IMI (N=40) made more significant reductions in their drinking over the 6-month follow-up than women in the comparison condition. Men did not show similar improvements on drinking measures.¹⁹ The largest improvement in drinking outcomes for women were at the most distal follow-up time point, 6 months.

There has been limited research on the types of individuals that respond best to IMI interventions. One exception is the study of MICM, which assessed 330 probationers and parolees entering 49 SLHs. Residents in 22 houses (n = 149 individuals) were randomly assigned to receive a “Motivational Interviewing Case Management” (MICM) intervention and residents in the other 27 houses (n = 181 individuals) received SLH residency as usual. At 6-and 12-month follow-up, both study conditions showed significant improvement relative to baseline on substance abuse, criminal justice, HIV risk, and employment outcomes. For persons who attended at least one MICM session, there were better criminal justice outcomes compared to the SLH as usual group.¹⁵ Latent class analysis was used to assess whether high and low recovery capital interacted with the intervention to effect outcome.²⁰ The class of high recovery capital, characterized by lower psychiatric severity, lower trauma, and higher motivation, received significantly more benefit from MICM than the low recovery group. A number of outcomes were better, including measures of criminal justice, drug and psychiatric problems.

Modification of IMI for Women with Alcohol Disorders

Based on the favorable findings for alcohol reduction among women during the MA study, funding was obtained to conduct a larger randomized study of IMI focused on alcohol problems and tailor the intervention to maximize the ways it responded to the needs of women. The main changes in the manual included directing therapists to assess whether a variety of issues identified by Erol and Karpyak¹ and Greenfield et al² impacted drinking: 1) relationships, 2) parenting, 3) sexual and physical abuse, 4) self-esteem, 5) obstacles to treatment entry, and 6) co-occurring health and mental health concerns. Consistent with psychotherapy approaches for women in the general psychotherapy literature,²¹ therapists elicited discussion about reciprocal associations between drinking problems and the destructive social relationships These modifications directing therapists to incorporate discussion of issues relevant to women constitute additional ways that IMI differs from standard MI. Currently, there are no MI manuals designed to specifically address the needs of women.

Outcomes for Women Receiving IMI

Outcomes for women with alcohol use disorders who received IMI have been reported in previous papers assessing 2- and 6-month follow-up. Using a randomized trial of 215 women who were interviewed at baseline and 2-month follow-up, Polcin, et al¹¹ reported women in both the IMI and comparison conditions made significant reductions in percent days drinking (PDD), percent heavy drinking days (PHDD), and alcohol severity on the Addiction Severity Index. Among women who were heavy drinkers, defined as those drinking 14 or more days to the point of intoxication over the past 30 days at baseline, those assigned to IMI showed larger reductions in PDD and PHDD than women in the comparison condition. When the same sample of women were followed up at 6 months, there were significant between group effects showing larger reductions in PHDD among women in the IMI condition.¹² One limitation of the study was the lack of diversity by race and economic status. The paper noted most of the sample was white and middle or upper income. Therefore, it was stressed that additional studies were needed on women who were low income and members of minority groups,

Although there are numerous studies demonstrating the efficacy for standard MI interventions,^7,8 the effect sizes are often relatively small when MI is compared to other interventions and the duration of the effects are often limited. The early papers on IMI for women^11,12 represent a significant step forward in the MI literature because they demonstrate advantages of IMI over standard MI in terms of reduction of heavy drinking among women. However, the papers noted two significant weaknesses. First was the need to assess outcomes over longer time periods to ascertain how long improvements lasted. Second was the need to assess whether subgroups of women responded differentially to IMI. The current paper was designed to respond to these limitations by reporting heavy drinking outcomes at 12 months and the effects of a variety of potential moderators.

Purpose

There were two primary goals for the current paper. The first was to assess whether the 2- and 6-month outcomes showing larger reductions in PHDD for heavy drinking women in the IMI condition persisted at 12 months. Based on the 2- and 6-month findings, we hypothesized that larger reductions in PHDD would be found for heavy drinking women in the IMI condition at 12 months. The second goal was to examine whether subgroups of heavy drinking women benefited differentially from receiving IMI. We hypothesized that psychiatric severity and motivation would interact with IMI to influence outcome. Specifically, we expected the differences between IMI and the comparison condition would be strongest for women who were low on psychiatric severity and high on motivation. These hypotheses were based on the latent class analyses conducted on MICM.²⁰ Although MICM represents a modification of IMI and was used on a mixed gender sample of persons in SLHs, the latent class analysis has been the only analysis describing how the effect of an IMI-based intervention differed by participant characteristics. Findings showed that a class of study participants who were more highly motivated and had lower psychiatric severity and lower histories of trauma were able to benefit more from MICM. We therefore used those findings as a rationale for our hypothesized moderators.

In addition to motivation and psychiatric severity, we explored how severity of problems associated with drinking moderated the effect of IMI. For these analyses we used scales from the Drinker Inventory of Consequences, which assesses measures of drinking consequences in a number or areas, including physical, social, intrapersonal, impulse control, and interpersonal.²² The rationale for examining these scales is that discussing drinking related consequences is a central component of IMI. We reasoned that higher levels of problems at baseline would present more opportunities for therapists and clients to discuss reasons to make changes in drinking and strategies for their successful implementation.

Method

Sample

Data collection took place at an outpatient treatment program in Northern California. Women were primarily recruited through radio and newspaper advertising. Others contacted the clinic seeking services for alcohol problems. All participants were age 18 or older, able to speak English, capable of giving informed consent, and met DSM-V criteria for current alcohol use disorder. After an initial phone screen assessing these criteria, participants were screened in-person at the program where the study was conducted. Participants were assessed for detoxification symptoms and mental health disorders and were referred to outside services when it was clinically indicated, but they were not necessarily excluded from the study. However, women with serious, persistent medical or mental health problems were excluded and referred to more intensive services. Women with low to moderate severity of co-occurring drug problems (<6 DSM criteria) were included but those with high severity were excluded from study participation.

Procedures

Figure 1 indicates recruitment and retention data. Of the 384 women screened for the study, 69 were determined to be ineligible, 78 refused participation or did not show up for the baseline interview, and 22 were deemed ineligible at the baseline interview. After signing the informed consent and providing contact information for follow-up interviews, 215 participants completed the baseline assessment and were randomly assigned to receive IMI (n=108) or SMI (n=107) as developed by Martino et al (2006). The SMI condition serving as the control group consisted of a typical one-session MI intervention. Randomization included a block design to ensure equivalence across the study conditions of low and medium/high severity of alcohol problems as determined by DSM-V criteria. All participants were re-interviewed at 2-, 6-, and 12-month follow-up. Participants were paid $30 for their time participating in the study at baseline and $50 at the follow-up interviews.

Figure 1. Randomization and follow-up.

The IMI and the SMI comparison interventions were delivered by three female master’s degree therapists. Therapists provided sessions to women in both study conditions. Two of the three therapists had previous experience with IMI manuals. All received training on the specific aspects of IMI for women. To ensure competence and adherence to the IMI and SMI interventions, therapists audiotaped all sessions.

The Yale Adherence and Competence Scale II for Motivational Interviewing (YACS)²³ was used as a measure of competency and fidelity for both treatments IMI and SMI. A sample of 146 tapes (16% of all sessions) for IMI and SMI were randomly assigned to raters to be assessed using the YACS. Both raters were Ph.D. psychologists and had previous experience coding MI sessions. One functioned as the primary rater during our 5-year study of IMI for MA dependence (Polcin et al, 2014). In addition to assessing standard items, trained raters examined compliance with modifications made to ensure that issues related to women were addressed. These were coded dichotomously as yes (issue was brought up by the client or therapist) or no (issue was not brought up by client or therapists. All session ratings were well above the minimum level for competence defined by the YACS for frequency and skillfulness of MI-based interventions. Inter-rater agreement on 9 tapes rated by two coders was 100% for meeting the minimum standard for competence. During the two months of active treatment, all women in both conditions also attended weekly outpatient group treatment using a Craving Intervention Management (CIM)²⁴ model that was largely based on cognitive behavioral therapy principles. Study procedures were approved by the Institutional Review Board at the Public Health Institute.

Measures

Baseline Only

Demographics include ethnicity, age, education, and marital status.

DSM V Checklist for Drug and Alcohol Disorders was used to determine inclusion criteria of current alcohol use disorder and exclusion of high severity of drug problems. Items were based on the previous version of the instrument,^25,26 the DSM IV, but items were updated to reflect new criteria in the DSM V.

Primary Outcome

Timeline Follow-Back²⁷ (TLFB, heavy drinking days) was used to record the participant’s self-report of percent heavy drinking days (PHDD) (4+drinks) during the past 60 days.

Moderator Variables

All moderator variables tested were results from baseline analyses.

Addiction Severity Index (ASI) Alcohol Scale²⁸ is a standardized, structured interview that was used to assess heavy drinking at baseline. Psychometric properties for the scale include test-retest reliability of 0.84 and internal consistency alphas averaging 0.86.²⁹ One item was used to identify heavy drinking at baseline, which asked days drank to intoxication over the past 30 days. Heavy drinking was defined as 14 or more days of drinking to intoxication in the past month. This measure was dichotomous, coded as heavy drinking versus not.

Addiction Severity Index (ASI) Psychiatric Scale²⁸ was used to assess alcohol problem severity. In a review of the psychometric properties of the ASI Mäkelä²⁹ found the psychiatric severity scale to have strong internal consistency, test-retest reliability, and validity. Scores range from 0 to 1.00. For some analyses, such as analyses of subgroups, this measure was dichotomized at the mean.

Drinker Inventory of Consequences is a 50-item instrument that measures drinking related consequences in 5 areas: physical, social, intrapersonal, impulse control, and interpersonal. The measure was developed for Project MATCH and has excellent psychometric properties for the total scale as well as the five subscales.²² For some analyses, such as analyses of subgroups, these scales were dichotomized at the mean.

University of Rhode Island Change Assessment Scale (URICA) is a self-report measure that is used to assess an individual’s readiness to change when entering addiction treatment. Responses are used to group individuals into stages of change: precontemplation, contemplation, action, and maintenance. Willoughby and Edens³⁰ found the scale to have good reliability and validity when categorized into clusters representing ready to change versus not ready. For our analyses, the scale was dichotomized into categories we termed as pre-action (precontemplation and contemplation stages) and action (action and maintenance stages).

Analysis

Descriptive analyses were used to depict means and percentages for baseline characteristics and primary outcomes (PHDD) at each data collection timepoint (baseline, 2 months, 6 months, and 12 months). Generalized estimating equation (GEE) models with robust standard errors to adjust for clustering were used to assess within and between study condition differences over time as well potential moderator effects. Models testing within effects for PHDD included terms for condition and time (categorical with baseline as referent) and controlled for age, ethnicity, marital status, and education. Models testing between effects were identical to the within effects models but included an interaction term for condition by time. Before testing for moderator effects using three-way interaction models, we first examined between effects for subgroups based on the dichotomized baseline moderator variables described above. These models provided preliminary information on moderators that were then tested in three-way interaction models. In the three-way models, the effect of IMI on PHDD included terms for time and moderator variables.

Results

Demographic Characteristics

The mean age of study participants at baseline was 50.9 (sd=11.3). Participants were mostly white (83.3%), married (53.5%) and college educated (61.4%). Nearly all the women (94.4%) reported residence in a stable living situation, such as renting an apartment or owning a house. Over half (50.2%) reported an annual income over $65,000; 25% reported an income over $150,000. The overall follow-up rate was 87% at 2 months, 86% at 6 months and 85% at 12 months. There were no significant baseline differences on measures of drinking between women who were and were not followed up 2, 6, and 12 months.

Engagement in Treatment

Engagement in the treatment protocols was excellent. Overall, 79.4% of the 108 women randomized to IMI completed 7 or more sessions. Women in the IMI condition attended an average of 7.4 (sd=2.3) individual sessions and an average of 5.9 (sd=2.5) of the 9 CIM group sessions. Among the 107 women assigned to SMI, all attended the single session of MI and an average of 5.1(sd=2.7) CIM groups sessions. However, their attendance at the nutrition group, added to achieve time equivalence between the study conditions, was low, at an average of 2.8(sd=2.7) of the 8 nutrition sessions offered.

Descriptive Analysis

Table 1 shows the means for heavy drinking for both study conditions across all time points. PHDD at baseline did not differ between the two study conditions. The pattern for both study conditions showed a significant decrease in heavy drinking between baseline and two months and continuation of that improvement at 6- and 12- months. The table shows that reductions were larger for women in the IMI condition. At baseline the PHDD was 0.49(sd=0.33) which was reduced to 0.22(sd=0.31) at 12 months. The reduction was smaller for the comparison condition, 0.52(sd=0.32) at baseline and 0.31(sd=0.34) at 12 months.

Table 1.

Heavy Drinking (TLFB PHDD) by condition and time.

	Baseline	2-month	6-month	12-month
Condition	Mean(sd)	Mean(sd)	Mean(sd)	Mean(sd)
SMI	0.517 (0.323)	0.284 (0.319)	0.303 (0.342)	0.305 (0.343)
IMI	0.491 (0.334)	0.196 (0.251)	0.241 (0.324)	0.216 (0.305)

The SMI Condition includes all participants assigned to the single session of MI.

The IMI Condition includes all participants assigned to intensive MI.

Total Sample (N=215): SMI (n=108) and IMI (n=107).

Table 2 shows the mean and median values for the moderator variables: the six scales of the Drinkers Inventory of Consequences (DrINC), the URICA readiness to change, and the ASI Psychiatric severity scale. Independent samples t-tests indicated no significant differences (p<.050) between SMI and IMI groups on any of the moderator variables.

Table 2.

Baseline descriptive statistics of moderator variables by condition^.

Condition	Moderator	Mean(sd)
SMI
	Total	43.861 (21.290)
	Social ResponsibilityA	5.315 (4.096)
	ImpulsivityA	5.491 (3.802)
	IntrapersonalA	14.704 (5.754)
	InterpersonalA	8.444 (6.382)
	PhysicalA	9.907 (4.751)
	Readiness to ChangeB	0.280 (0.450)
	ASI Psychiatric Scale	0.259 (0.212)
IMI
	TotalA	40.748 (18.463)
	Social ResponsibilityA	4.897 (3.900)
	ImpulsivityA	5.495 (3.710)
	IntrapersonalA	13.981 (5.105)
	InterpersonalA	7.168 (5.482)
	PhysicalA	9.206 (3.880)
	Readiness to ChangeB	0.270 (0.447)
	ASI Psychiatric Scale	0.222 (0.189)

The SMI Condition includes all participants assigned to the single session of MI. The IMI Condition includes all participants assigned to intensive MI.

Total Sample (N=215): SMI (n=108) and IMI (n=107).

^ADrinker Inventory of Consequence Scales.

^BUniversity of Rhode Island Change Assessment Scale (URICA).

Independent samples t-tests revealed no significant differences between SMI and IMI groups on all moderator variables, all p’s > 0.117 for Total Sample.

Generalizes Estimating Equation Models

Results from our GEE models (Table 3) confirmed several hypotheses. Within condition findings at 12 months for the entire sample showed significant, sustained reductions in PHDD relative to baseline (β = −0.247, SE= 0.026, p<.01). When we compared between group improvement at 12 months for the entire sample, we found no differences between the two study conditions (β = −0.063, SE=0.051). However, the last column in the table shows a statistical trend at 12 months for women who were heavy drinkers at baseline. The heavy drinkers who were assigned to IMI trended toward better PHDD outcomes relative to the comparison group (β = −0.105, SE=0.062, p=.09).

Table 3.

Results of GEE models testing within, between, and subgroup effects for PHDD.

	a ITT Sample Models (N=215)			b Subgroup Models
	Within Effects	Between Effects		cURICA Readiness	dASI Psychiatric	cDrINC Physical	cDrINC Impulse	eHeavy Drinking
	Coef. (SE)	Coef. (SE)		Coef. (SE)	Coef. (SE)	Coef. (SE)	Coef. (SE)	Coef. (SE)
Study Condition (SMI referent)
IMI	−0.031 (0.115)	−0.003 (0.115)		0.291 (0.200)	−0.028 (0.061)	−0.040 (0.104)	−0.280 (0.087)*	0.108 (0.137)
Time (baseline referent)
2-months	−0.268 (0.023)*	−0.239 (0.032)*		−0.124 (0.031)*	−0.185 (0.048)*	−0.226 (0.038)*	−0.172 (0.044)*	−0.266 (0.038)*
6-months	−0.234 (0.023)*	−0.215 (0.033)*		−0.188 (0.043)*	−0.243 (0.053)*	−0.226 (0.040)*	−0.165 (0.047)*	−0.219 (0.039)*
12-months	−0.247 (0.026)*	−0.216 (0.038)*		−0.121 (0.047)*	−0.146 (0.060)**	−0.229 (0.046)*	−0.149 (0.053)*	−0.265 (0.043)*
Condition × Time Interaction
IMI × 2-months	--	−0.057 (0.045)		−0.230 (0.068)*	−0.144 (0.065)**	−0.153 (0.059)*	−0.130 (0.063)**	−0.119 (0.056)**
IMI × 6-months	--	−0.039 (0.046)		−0.126 (0.084)	−0.032 (0.067)	−0.110 (0.065)***	−0.109 (0.066)***	−0.121 (0.057)**
IMI × 12-months	--	−0.063 (0.051)		−0.200 (0.087)**	−0.147 (0.076)**	−0.152 (0.070)**	−0.137 (0.073)***	−0.105 (0.062)***
Intercept	0.564	0.548		0.456	0.666	0.523	0.460	0.676

Notes:

^aITT=intent to treat sample. Coefficients and standard errors have been adjusted for age, ethnicity, education, and marital status.

^bSubgroup samples based on dichotomized variables.

^cURICA dichotomized as action versus pre-action, DrINC physical and DrINC impulse control dichotomized above the median,

^dASI psychiatric dichotomized below median.

^eHeavy Drinking dichotomized as 14+ days heavy drinking (to intoxication) the past 30 days versus not.

^*p ≤ 0.01

^**p ≤ 0.05

^***p < 0.10

Table 3 also displays subgroup effects at 2-, 6-, and 12-month follow-ups for other characteristics. These analyses dichotomized moderator variables and then tested their effects for each subgroup. Consistent with our hypotheses, four characteristics were associated with better outcome: low psychiatric severity, more severe physical and impulse control consequences associated with drinking, and higher readiness for change. For women with these characteristics, IMI had a significant effect on PHDD relative to the control group. The coefficients in Table 3 show the effects were strongest and most consistent for moderators at the 2- and 12-month timepoints.

Comparable GEE models tested the effect of IMI for the corresponding subgroups: above the median on psychiatric severity, below the median on physical and impulse control consequences, and membership in the pre-change category for readiness for change. These, as hypothesized, resulted in no significant condition by time interactions (results not displayed). Although the physical consequences and impulse control scales on the DrINC were both moderators in these subgroups models, it is noteworthy that three other scale on the DrINC were not: Interpersonal, Intrapersonal, and Social Responsibilities.

Table 4 shows results of 3-way interaction models for the four moderator variables that were found to be associated with effective IMI in the subgroup analyses. Also shown in Table 4 is the interaction effect for IMI with heavy drinking (14+ days drinking to intoxication over the last 30 days at baseline). Consistent with the subgroup models, we found significant moderator effects at 2 months for higher readiness for change (β = −0.237, SE= 0.087, p<.01), lower psychiatric severity (β = −0.178, SE= 0.089, p<.05), and higher severity of physical problems on the DrINC (β = −0.186, SE= 0.092, p<.05). However, unlike the subgroup analysis, the impulse control scale on the DrINC did not show a moderator effect. Moderator effects at other time points were limited. There was a significant moderating effect for heavy drinking at 6 months (β = −0.222, SE= 0.080, p<.01)and a trend for moderation for readiness for change at 12 months (β = −0.183, SE= 0.106, p<.10).

Table 4.

Three-way interaction models by moderator.

	URICA Readiness	ASI Psych	DrINC Physical	DrINC Impulse	Heavy Drinking
	Coef. (SE)	Coef. (SE)	Coef. (SE)	Coef. (SE)	Coef. (SE)
Study Condition (SMI referent)
IMI	−0.085 (0.114)	−0.031 (0.121)	−0.124 (0.154)	0.069 (0.126)	0.000 (0.107)
Moderator (high referent)	−0.197 (0.067)*	−0.040 (0.062)	0.018 (0.066)	0.124 (0.063)**	0.298 (0.057)*
Time (baseline referent)
2-months	−0.279 (0.041)*	−0.290 (0.043)*	−0.254 (0.058)*	−0.301 (0.047)*	−0.123 (0.053)**
6-months	−0.228 (0.041)*	−0.187 (0.039)*	−0.200 (0.057)*	−0.263 (0.045)*	−0.203 (0.042)*
12-months	−0.248 (0.047)*	−0.282 (0.044)*	−0.195 (0.065)*	−0.283 (0.053)*	−0.009 (0.063)
Study Condition × Time Interaction
IMI × 2-month	0.008 (0.055)	0.033 (0.061)	0.035 (0.071)	0.012 (0.064)	−0.017 (0.062)
IMI × 6-month	−0.002 (0.053)	−0.043 (0.063)	0.021 (0.068)	0.032 (0.062)	0.100 (0.055)***
IMI × 12-month	−0.014 (0.061)	0.019 (0.068)	0.007 (0.077)	0.013 (0.070)	−0.109 (0.075)
Condition × Moderator Interaction
IMI × Moderator	0.258 (0.095)*	0.056 (0.090)	0.220 (0.089)*	−0.027 (0.089)	0.035 (0.079)
Moderator × Time Interaction
Moderator × 2-month	0.151 (0.051)*	0.107 (0.064)	0.029 (0.069)	0.128 (0.064)**	−0.144 (0.065)**
Moderator × 6-month	0.033 (0.059)	−0.054 (0.066)	−0.025 (0.069)	0.098 (0.065)	−0.017 (0.057)
Moderator × 12-month	0.121 (0.067)***	0.139 (0.075)***	−0.033 (0.080)	0.133 (0.074)***	−0.256 (0.076)*
Condition × Moderator × Time Interaction
IMI × Moderator × 2-month	−0.237 (0.087)*	−0.178 (0.089)**	−0.186 (0.092)**	−0.142 (0.090)	−0.103 (0.084)
IMI × Moderator × 6-month	−0.119 (0.099)	0.010 (0.092)	−0.129 (0.094)	−0.140 (0.091)	−0.222 (0.080)*
IMI × Moderator × 12-month	−0.183 (0.106)***	−0.167 (0.102)	−0.157 (0.104)	−0.149 (0.101)	0.002 (0.097)
Intercept	0.613	0.566	0.525	0.361	0.346

Notes: ITT=intent to treat sample. Coefficients and standard errors have been adjusted for age, ethnicity, education, marital status, and MI session exposure.

^*p ≤ 0.01

^**p ≤ 0.05

^***p ≤ 0.10

Discussion

There exists a vast literature on the effectiveness of MI, particularly for alcohol, but also other problems as well, including drugs, HIV risk, smoking, and mental health problems.¹⁷ The intervention described in the current paper is different from standard MI in terms of structure (e.g., number of sessions) and process (e.g., implementation of the change plan over time).

Previous research has shown IMI has some advantages over standard MI in terms of reducing heavy drinking among women.^11,12 Findings from the current study show outcomes at more distant time points (12 months) and the effects of a variety of moderators.

The 12-month findings from the current analyses of IMI for women with alcohol use disorders are consistent with earlier analyses at 2- and 6-month follow-up. Women who received either a single session of MI combined with the CIM group treatment or the new, 9-session IMI intervention combined with CIM showed significant reductions in PHDD at 12-months. These findings are consistent with earlier analyses at 2 and 6 months.^11,12 Also consistent with our earlier analyses, results using the full sample at 12-months did not show differenced in drinking outcome between the two study conditions. However, the 12-month analysis that examined PHDD among the subgroup of women who were heavy drinkers at baseline showed those receiving IMI trended toward larger reductions in PHDD. This finding replicated results at 2- and 6-month follow-up although the findings were stronger at these earlier time points. Overall, IMI appears to be beneficial for women who are heavy drinkers. However, women who are not heavy drinkers do not appear to derive greater benefit from the more intensive IMI intervention. The findings suggest they do as well with a single session.

The significance of the finding for heavy drinking women is bolstered by the fact that the field is increasingly focused on ways to sustain the improvements made during treatment. Typically, improvements made during treatment decrease over time, especially for individuals with more serious alcohol problems.⁹ One approach has been to implement ongoing recovery monitoring and case management after completion of treatment.³¹ However, results from the current study suggest that IMI is an example of an intervention that results in sustained improvement on its own, without investment in ongoing care after treatment ends.

It is important to note that despite large reductions in heavy drinking for women who were heavy drinkers, the level of drinking at follow-up continued to be concerning. At 6-month follow-up, women who received IMI were on average drinking heavily 24% of the days over the past 60. At 12 months, the average percent of heavy drinking over the past 60 days for women in the IMI group was 22%; only 12% (n=11) reported complete abstinence over the past 60 days. For some women, IMI might be working as a harm reduction intervention, where heavy drinking and drinking related harms are reduced and but not entirely eliminated.

Anecdotal reports from study therapists suggested that some women receiving IMI began treatment with a goal of reduced drinking and continued with that goal throughout treatment. However, a limited number shifted to a goal of abstinence at some point during their treatment. While a few of these women achieved abstinence, it was more common for them to continue some level of use with fewer drinking related consequences. There is a need for additional research to identify the characteristics of women who can sustain a goal of reduced drinking over longer time periods and those who retune to baseline drinking if they cannot achieve abstinence.

Subgroup Analyses

In addition to confirmation of our hypothesis that the effects of IMI on PHDD would continue at 12-month follow-up for women who were heavy drinkers at baseline, several of our hypotheses about subgroups were confirmed. Potential moderating variables, which were dichotomized at baseline and treatment effects were assessed for these subgroups. Women most likely to benefit included women who had higher motivation, more serious physical and impulse control problems relate to their drinking, and lower psychiatric severity at baseline. For the two DrINC scales (physical and impulse), effects were evident at all time points. For motivation and psychiatric severity, effects were seen at 2- and 12-month follow-up.

The finding that motivation and psychiatric severity at baseline were associated with better outcome replicates previous studies, including assessments of IMI-based interventions. In a review of studies on predictors of alcohol outcome, Adamson, Sellman, and Frampton³² found motivation and psychopathology to be strong predictors. These findings are intuitively appealing from a clinical perspective. Persons who are not motivated may be less likely to attend sessions, engage with the therapist, and follow through with the change plan. Persons with high levels of psychiatric severity may require more specialized services for co-occurring disorders.³³ The finding that higher psychiatric severity was associated with worse outcome is consistent with a recent study of an MI-based intervention, Motivational Interviewing Case Management.²⁰ That study used latent class analysis to show that lower levels of intrapersonal recovery capital among study participants, such as more severe psychiatric problems and less motivation for change, did not benefit from the intervention.

It was unclear why two scales on the Drinker Inventory of Consequences (physical and impulse control) interacted with IMI while three others did not (intrapersonal, interpersonal, and social responsibility. However, several factors may have been influential. Anecdotal reports from study therapists suggest that physical concerns were highly prevalent for the women. They were concerned about various medical conditions caused by drinking, but also other physical issues, such as weight gain and energy level.

Some of the women talked about improving their health through reduced drinking, but they also discussed additional health changes, such as increasing their exercise and eating healthier. This may have given therapist more leverage in terms of eliciting discussion about reasons to make changes in in drinking in response to physical concerns. In addition, the sample was relatively older, with the mean age being nearly 60. Increasing concerns about physical health issues that emerge with increased age and that might be exacerbated by drinking may also have played a role in increasing a focus on physical problems.

Most of the items on the impulse control scale address specific behaviors that might be experienced as embarrassing by this population of women. For example, items ask about arrests for driving under the influence, other arrests related to drinking, physical fights, destruction of property, injuries to others, and taking foolish risks. Reactions to these events from friends and family may result in embarrassment and regret that motivate change and enable the therapist to engage in discussion about impulsive consequences. Stigma may play a role as well, particularly among this cohort of older drinkers. While younger women cohorts tend to view drinking in ways that are similar to their male counter parts, that is not the case for older women, who are often more judgmental toward heavy drinking women.³⁴ Concern about negative judgements about impulsive behaviors from peers, friends, and family may have been a factor motivating change.

Three-Way Moderator Effects

Moderator effects in our three-way interaction models (condition by moderator by time) were less robust than the effects of moderators found in our subgroup analyses. Moderator effects in the three-way analyses were primarily seen at the 2-month time point. These included, motivation, DrINC (Physical), and psychiatric severity. The three-way moderator analysis of heavy drinking showed PHDD was significantly affected by IMI at 6 months, but not at other time points. The only significant moderator at 12 months was a trend (p<.10) for motivation. Nevertheless, the other three moderators were showing movement toward significance (all were p values <.15) and with a higher level of statistical power may have reached significance.

The three-way interactions showing that moderating effects were strongest at 2 months is conceptually appealing. One factor is the moderator is proximal to the outcome timepoint. One can easily understand how factors assessed at treatment entry (i.e., baseline) might influence the course of treatment and then treatment outcome two months later. However, it is also understandable that these effects might wane a year later. Among other factors, the status of the moderators many have changed substantially over a one-year period and measurement of those factors at timepoints more proximal to the 12-month follow-up might be more informative. Nevertheless, the importance of the four mediators found to effect PHDD at 2 months is evident in the directional trend toward association at 12 months, although none of the four moderators reached statistical significance at that timepoint.

Limitations and Need for Further Research

It is important to note several limitations to the study and issues requiring more research. First, although we found the effect of IMI on PHDD among heavy drinkers can be seen in at 2 months and trends toward continuing improvement up to 12 months, further research needs to address how long the beneficial effects last beyond the 12-month time period we assessed. Although our N of 183 at 12 months provided sufficient power to test hypotheses, a larger N would be able to more accurately detect smaller between group differences. Second, there is a need to assess why in our three-way analyses the effect of IMI on heavy drinking was strongest at the 6-month time point. Although we speculate about why some subgroup characteristics may have had better outcomes, more work is needed to understand the mechanisms of how different subgroups derive benefit from IMI. Specifically, mediation analyses would be helpful in this regard. Finally, there are limitations related to sample characteristics. We studied women who were older and represented a higher economic status relative to the overall population of women drinkers. In addition, most of the women in the study were white. More studies are needed to assess the effects of IMI on a broader range of women, particularly those who are minorities, lower income, younger, and suffering from serious co-occurring disorders such as mental illness.