OUP user menu

Gender differences in the efficacy of brief interventions with a stepped care approach in general practice patients with alcohol-related disorders

Susa Reinhardt, Gallus Bischof, Janina Grothues, Ulrich John, Christian Meyer, Hans-jürgen Rumpf
DOI: http://dx.doi.org/10.1093/alcalc/agn004 334-340 First published online: 9 February 2008

Abstract

Aim: To analyse gender differences in the efficacy of stepped care brief interventions for general practice patients with alcohol problems. Methods: Data are part of “Stepped Interventions for Problem Drinkers,” in which 10,803 patients from 85 general practitioners were screened using alcohol related questionnaires; 408 patients were randomized (32% were female) to a control (booklet only) or two different intervention groups: stepped care (feedback, manual, and up to three counselling sessions depending on the success of the previous intervention) and fixed care (four sessions). Response rate for the 12 months follow-up was 91.7%. Results: Regression analysis revealed a significant effect size only in women (P = 0.039). After excluding alcohol dependents and binge drinkers, an effect size (R2) of 0.031 (P = 0.050) in women and an effect size (R2) of 0.069 (P = 0.057) in men was obtained. Among the patients in stepped care who, by the first assessment point, had reduced drinking to within safe-drinking limits, there was a tendency for females to have achieved this more often than males (40% vs. 24%; P = 0.089). Conclusions: In a heterogeneous sample, the intervention was only effective for women. Women tended to profit more from the first, less intensive intervention than men. When analysis was limited to those reporting “at risk” average daily consumption and “alcohol abuse,” the gender differences in efficacy appeared to be less, but the study was not sufficiently powered to affirm that.

Introduction

Providing brief interventions in medical settings is a promising approach in secondary prevention of problematic drinking behaviour (Babor and Higgins-Biddle, 2000; Bertholet et al., 2005). General practices have been shown to be an ideal setting for brief interventions due to high prevalence rates of alcohol use disorders (Hill et al., 1998). Furthermore, a study in Germany found at least one contact per year to a general practitioner (GP) for 80% of alcohol dependent and 67.4% for alcohol abusing individuals (Rumpf et al., 2000). Randomized controlled trials have shown a significant reduction in patient's alcohol consumption when interventions in GPs had taken place (Wallace et al., 1988; WHO Brief Intervention Study Group, 1996; Fleming et al., 1997; Poikolainen, 1999).

There is no consistent conclusion to draw on the topic of gender differences and meta-analyses of brief interventions: Earlier meta-analyses emphasized gender differences (Bien et al., 1993; Kahan et al., 1995; Wilk et al., 1997), whereas the more recent ones showed the equality of outcomes among men and women (Whitlock et al., 2004; Ballesteros et al., 2004; Bertholet et al., 2005). Earlier meta-analyses sometimes found a trend for a better efficacy of brief interventions in women (Wilk et al., 1997), others revealed no consistent results in women at all (Chang, 2002), and even others showed a better and stronger efficacy of brief interventions in men (Kahan et al., 1995). The meta-analysis by Poikolainen (1999) found efficacy of brief interventions only in women attributed to a lack of homogeneity in men and in both genders combined. This stands in contrast to an important meta-analysis on the topic of gender differences which found no difference between genders (Ballesteros et al., 2004). The authors included seven studies and used as outcome measures the quantity of weekly consumption and the frequency of drinkers who reported consumption below hazardous levels. Ballesteros et al. (2004) found similar standard effect sizes in men (d = −0.25) and women (d = −0.26) for the reduction of alcohol consumption and similar odds ratios in men (2.32) and women (2.31) for the frequency of individuals drinking below the harmful level. The meta-analysis by Bertholet and colleagues also showed an effectiveness of brief interventions for both men and women in reducing alcohol consumption at 6 and 12 months (Bertholet et al., 2005). More recent is the review by Kaner and colleagues (2007), where sub-group analysis confirmed the benefit of brief intervention in men, but not in women. Often a null finding in women was described-–that both control and intervention groups tend to reduce alcohol consumption in trials of brief interventions (Chang, 2002).

As an important new development in order to effectively treat patients with alcohol-related disorders, stepped care approaches are applied. Following Borsari et al. (2007, p. 131) stepped care is a “dynamic, performance-based procedure in which individuals not responding to an initial level of treatment that is the least intensive are then provided a more intensive treatment” and “within this framework, different levels of interventions are linked together.” Especially in times of limited resources, “it makes sense to provide all the time, expertise, and individual attention a patient needs, but not more” (Haaga, 2000, p. 547).

When this study was planned, only one study examined a stepped care approach for reducing alcohol consumption (Breslin et al., 1998). The authors evaluated a standard alcohol treatment for heavy drinking adults and provided a supplemental treatment for non-responders, who continued to drink above the recommended levels. But they did not find group differences between the non-responders with additional treatment and the ones without, arguing that the additional treatment was not intensive enough.

The main question of the following study of counselling for patients with problem drinking in general practice was to examine whether time can be saved using stepped care for conventional brief interventions, and to compare the outcome after these interventions to the outcome of a control group. Our results, which reveal that stepped care for individuals with at-risk drinking and/or alcohol use disorders is time saving and can have a similar effect to traditional “brief interventions” are published elsewhere (Bischof et al., in press-a).

The aim of this paper is to examine whether the patient's gender affects the efficacy of brief interventions in patients of general practitioners with problematic alcohol consumption.

Method

Recruitment of general practices and procedure

In our study “Stepped Interventions for Problem drinkers (SIP),” 81 general practices in the northern German city of Luebeck and its surrounding areas as well as four practices in the city of Kiel took part. Of the 241 practices in Luebeck, 82 had to be excluded because of various reasons and 78 were not willing to participate, resulting in a response rate of 50.9% among the physicians (in the city of Kiel the response rate was 31.8%). Data gathering ranged between 2 and 4 weeks per practice and was conducted by study nurses, medical students and research staff between January 2002 and March 2003.

Within the waiting rooms, patients were asked to fill out a screening questionnaire on alcohol consumption embedded in items on smoking, mental health and socio-demography. Participants were deemed positive when they scored five points or more in the Alcohol Use Disorder Identification Test (AUDIT, Allen et al., 1997) and/or two or more points in the Luebeck Alcohol dependence and abuse Screening Test (LAST, Rumpf et al., 1997). Patients fulfilling the inclusion criteria were asked to further participate and sign informed consent. Then, a questionnaire was sent via mail and after a few days a pre-intervention assessment to diagnose alcohol dependence and abuse according to the Diagnostic and Statistical Manual of Mental Disorders (DSM IV, American Psychiatric Association, 1995) was administered by phone.

Assessment

The AUDIT consists of ten items on recent alcohol use, alcohol-dependence symptoms, and alcohol-related problems. Recent studies report a Cronbach's alpha ranging between 0.75 and 0.97 (Reinert and Allen, 2007). Validity was tested against diagnoses of harmful use, dependence as well as at-risk drinking and was found to be generally good (Babor et al., 2001; Reinert and Allen, 2007). The recommended cut-off point is eight (Babor et al., 2001). However, in two German samples, five points turned out to be the best cut-off point to maximize sensitivity (Rumpf et al., 2002; Dybek et al., 2006). The screening tool LAST consists of seven items for detecting alcohol abuse and dependence. Two positive responses are the cut-off; internal consistency ranges between 0.69 and 0.81 (Cronbach's alpha); validity data show sensitivity ranging between 0.63 and 0.87 and a specificity of 0.88 to 0.93 (Rumpf et al., 1997).

For diagnosis according to DSM IV, the alcohol section of the Munich Composite International Diagnostic Interview (M-CIDI) (Wittchen et al., 1995) was used, which is the German version of the Composite International Diagnostic Interview (CIDI) (Robins et al., 1988). The M-CIDI is a fully structured and computerized interview and reports whether the diagnostic criteria are fulfilled. The inter-rater reliability, test-retest reliability and validity have been demonstrated to be good (Lachner et al., 1988; Andrews and Peters, 1998).

In addition to the M-CIDI, other questions concerning alcohol consumption and the core constructs of the transtheoretical model of behaviour change (TTM, Prochaska et al., 1992) were assessed, partly via telephone and partly by post.

The questionnaire MHI-5 (Berwick et al., 1991) consists of five items to assess mental health as the short form of the Mental Health Inventory (Veith and Ware, 1983; Davies-Avery et al., 1988). MHI-5 is appropriate to detect anxious and depressive disturbances within the last month (German translation Rumpf et al., 2001b) with a higher sum of the Likert-scaled and transformed items indicating a higher psychological pressure. The internal consistency showed a good performance with a Cronbach's alpha of 0.74, a sensitivity of 0.83 and a specificity of 0.78 (Rumpf et al., 2001b).

The stages of change were assessed using a German translation of the Readiness to Change Questionnaire—RCQ (Rollnick et al., 1992; Heather et al., 1993; Hannöver et al., 2002, 2003). Stages of change were allocated using the quick method (Heather et al., 1993). Here the Cronbach's alpha is reported for three subscales: precontemplation (0.67), contemplation (0.80), and action (0.83). Data suggest good predictive validity for the instrument (Hannöver et al., 2003).

Inclusion and exclusion criteria

Patients were eligible if they sent back the questionnaire and were either diagnosed with alcohol dependence or alcohol abuse according to the DSM-IV criteria or fulfilled criteria of at-risk or binge drinking only, descending in a hierarchical order, choosing always the most severe diagnosis. At-risk drinking was defined as an average alcohol consumption of more than 20 g alcohol per day for women and more than 30 g per day for men according to the common criteria of the British Medical Association (1995); binge-drinking was specified as more than 60 g alcohol for women and more than 80 g for men two or more times a month (adapted from a WHO study, Babor and Grant, 1992). At the time of planning this study, no general criteria were available. Some studies had used lower criteria, but BMA criteria are quite common.

Exclusion criteria were no alcohol consumption within the last 4 weeks, already in treatment for alcohol problems within the last year, severe or terminal illness, severe drug dependence, no telephone and not understanding or speaking German sufficiently.

Sample

The recruitment procedure is depicted in Fig. 1.

Figure 1

Recruitment of study participants.

In total, 13,033 consecutive patients aged 18 to 64 years were contacted during the practices and 10,803 patients were screened. Of these patients, 2,239 (20.7%) screened positive and 1,410 of those agreed to further participate (response rate 63%). In 79.2% (n = 1,119) of the cases, interviews could be conducted (response rate of screening positives 50%). Of those, 59.3% (n = 664) were not eligible (not fulfilling criteria of alcohol dependence, alcohol abuse, at-risk or binge-drinking). Of 408 final study participants, 124 were diagnosed as alcohol dependants and 59 as alcohol abusers. The inclusion criteria of at-risk drinking fulfilled 112, for binge-drinking 113 patients. Of the study participants, 32% were female. In order to include one female patient in the study, 47 of the women willing to take part had to be screened. For one male study participant, 17 screenings of willing men were necessary.

Randomization and intervention

After the diagnostic procedure, participants were randomly allocated to a control group or to one of the two intervention groups. The unit of randomization was the individual patient. Randomization was carried out prior to the telephone interview using sealed cards drawn by study staff from one container. This anomaly was necessary for organisational reasons—it did not lead to unbalanced groups, because envelopes with cards of ineligible patients were sealed again and put back in the container.

Participants in the control group (n = 139) received a booklet on health behaviour. Participants in both intervention groups received computerized feedback and a stage-tailored manual after the diagnostic procedure. The feedback consisted of individualized information tailored to the client's motivation to change and included normative feedback on drinking and drinking-related risks as well as core constructs of the TTM (Prochaska et al., 1992): stages and processes of change, decisional balance and self-efficacy. The computerized feedback alone was associated with outcomes no different from those of a traditional counselling session combined with the feedback (Bischof et al., in press-b). The feedback was supplemented by a stage-tailored manual. Counselling sessions for patients in the intervention groups were conducted via phone and based on Motivational Interviewing (MI, Miller and Rollnick, 2002) and Behaviour Change Counselling (BCC, Rollnick et al., 1999). The main content of each intervention was the enhancement of motivation to reduce the alcohol consumption or become abstinent, according to the individual's stage of change. Participants in fixed care (n = 131), received four counselling sessions up to 30 minutes each, directly after the diagnostic procedure and 1, 3, and 6 months later. Patients in stepped care (n = 138) received the same feedback and the manual as the first, less intensive, intervention. At the second contact after one month, the success of the first intervention was checked. If the patients reached safe drinking limits and were confident to maintain the reduction (at least six points on a 10-point Likert scale), they received no further intervention. If the first intervention was not successful, patients were offered a counselling session. The same procedure was conducted at the third contact after 3 months and the fourth contact after six months. The counselling sessions for the stepped care group lasted up to 40 minutes each, in order to equal the intervention time of the fixed care group.

Training of counsellors and supervisor

All interventions were conducted by three psychologists with expertise in clinical treatment and research. They completed a 4-week training in MI (principles, basic techniques and practical exercises). This workshop and later supervisions were conducted by the project manager and member of the MI community (HJR). All counselling sessions were audio taped and a random sample was coded for MI consistency by the other researchers. For supervision all four collaborators met on a weekly basis. The use of the manual was checked and possible discrepancies were corrected. Quality control included the rating of MI consistency and advanced training over the whole period.

Follow-up

Follow-up was scheduled after 12 months. A blinded personal interview was conducted by research staff who had no contact with the patient prior to the outcome assessment. There was a mean interval between baseline assessment and follow-up interview of 401 days (SD = 56.7). In the case of non-accessibility via telephone, participants were personally contacted at their homes. Of the baseline sample described above (n = 408), 3 (0.7%) had died. Of the remaining 405 participants, 27 (6.6%) were not attainable, 4 individuals (1%) withdrew their further participation, and 374 individuals (91.7%) were re-interviewed.

Statistical analyses

Data were analysed using SPSS for Windows, version 14.0. Outcome was analysed on the basis of intention to treat, assuming that the patient who could not be reached for the follow-up assessment did not change in outcome variables. The primary outcome measure was reduction of alcohol consumption (grams of pure alcohol). In order to test time-saving of the stepped care approach, the total amount of counselling in minutes was used. Groups were compared using t-test, Mann–Whitney U-test (used for comparing expended time for counselling between the intervention groups) and Kruskal–Wallis H-test. In addition, we conducted ordinary least squares (OLS) regression to compare the intervention groups concerning drinking outcomes, and group differences for expended time when efficacy of the intervention was held constant. Finally, we conducted an OLS regression to compare intervention and control group concerning drinking outcome. Because the distribution of our outcome variables were highly skewed, we conducted these analyses after transforming time expended for counselling and alcohol consumption using the logarithmic scale, which made the variables nearly normally distributed. After examining the residuals of these regressions for normality and checking for heteroscedasticity, we retransformed the regression coefficients for the categorical variables using the formula (eb − 1) × 100. This transformation is appropriate when the errors are normal and homoscedastic and should be interpreted as the percentage difference in drinking outcome between that variable and the comparison group (Manning, 1998). In addition, for analysing a gender specific intervention effect an interaction term of gender and intervention was calculated for the regression model.

Results

The first analysis was to compare male and female study participants. There was a significant difference in age, showing a younger age for female patients (t = 2.48, df = 402, P = 0.014). As expected male participants had a higher AUDIT sum (U = 15398,50; P = 0.016) and showed less negative mental health and therefore lower points in MHI-5 (U = 14113,50; P = 0.007). Education showed a slightly higher education for women (U = 9274,00; P = 0.061). Regarding group randomization, distribution of diagnoses and distribution of TTM stages analyses showed no gender differences. Participants in stepped care and fixed care did not differ significantly in any demographic variables and alcohol risk terms; more details are described elsewhere (Bischof et al., in press-a).

For the purpose of examining the efficacy of the intervention in relation to gender, the intervention groups (stepped care and fixed care) were collapsed yielding a higher statistical power.

Regression analysis for reduction of the drinking amount revealed no intervention effect comparing control and intervention groups with an effect size of R2 = 0.006 (P = 0.124, see Table 1). Splitting the sample by gender, analyses showed a significant effect size (R2 = 0.029) only in women (P = 0.039); men showed an effect size of R2 = 0.001 (P = 0.564). Based on this analysis, the intervention effect for women was 35.5% with a small to medium effect size and for men a reduction of 9.6% with no relevant effect size.

View this table:
Table 1

Prediction of alcohol consumption at follow-up; intervention status (0 = CG; 1 = IG) entered at step 2

PredictorR2dfFR2-changeTP-value
Overall baseline consumption0.0551/40623.69
   Intervention0.0612/40513.07.006−1.540.124
Men0.0391/27611.20
   Intervention0.0402/275 5.76.001−0.580.564
Women0.1121/12816.17
   Intervention0.1422/12710.48.029−2.090.039
  • CG = Control group. IG = Intervention groups

In order to check an interaction between gender and intervention, an interaction term was included in the regression analysis, but showed no significant result.

A calculation of number needed to treat (NNT)—using drinking reduction as an outcome—revealed that 10 women had to be treated to get an improvement in one female patient, which would not have occurred with the booklet only. The corresponding number for men was 17.

For the next analysis, patients with a diagnosis of alcohol dependence and patients fulfilling the criteria of binge-drinking were excluded. In this analysis men showed an R2-change of 0.031 with a tendency of P = 0.057 and women showed a change of R2 = 0.069 with a tendency of P = 0.050. Male at-risk drinkers and abusers within the intervention groups reduced 39.3% and female 36.1%, indicating again small to medium effect sizes (see Table 2). Within this subsample the gender–intervention interaction term was also included in the analysis, but again showed no significant contribution to the regression model.

View this table:
Table 2

Prediction of alcohol consumption at follow-up, intervention status (0 = CG; 1 = IG) entered at step 2 of the regression analysis; at risk drinkers and patients with alcohol abuse only

PredictorR2dfFR2-changeTP-value
Overall baseline consumption0.0031/1690.468
   Intervention0.0402/1683.470.037−2.540.012
Men0.0011/1150.090
   Intervention0.0322/1141.890.031−1.920.057
Women0.0581/523.19
   Intervention0.1272/513.700.069−2.010.050
  • CG = Control group. IG = Intervention groups.

Response to step 1 of the stepped care intervention occurred more often in females than males. At the second contact, constituting the first efficacy check, 40% of the female study participants in stepped care had responded to the first intervention (i.e. reached safe-drinking limits) in comparison to 24.4% of the male participants (P = 0.089). There was no difference at any other point in time (see Table 3).

View this table:
Table 3

Stepped care patients who reached safe drinking levels: Comparison between genders

T2 n (%)T3 n (%)T4 n (%)
Male20 (24.4)13 (21.3)11 (25.0)
Female14 (40.0)4 (22.2)4 (26.7)
P = 0.089P = 0.934P = 0.898
  • T2 = contact one month after intake. T3 = contact three months after intake. T4 = contact six months after intake.

In order to check whether differences in drinking behaviour help explain responses to step 1 (reduction consumption to safe limits), an analysis of the risk status at baseline for the reduction drinking was calculated with no significant difference (U = 1151,00; P = 0.118). Mean ranks showed a lower chance of dropping out for patients with a more severe risk status.

Participants in stepped care received roughly half the amount (M = 40.01; SD = 41.24) of intervention in minutes compared to fixed care participants (M = 80.31; SD = 40.30), showing a significant difference (P < 0.001). Less counselling time was consumed by women in stepped care than by men (P = 0.026). Also, the percentage reduction of drinking in women of the stepped care group was greater than in men (P = 0.050). In control and fixed care group, men and women showed no differences in the amount of drinking reduction.

Discussion

Whether the patient's gender affects the efficacy of brief interventions for alcohol problems is unclear in the literature (Poikolainen, 1999; Ballesteros et al., 2004; Kaner et al., 2007). This study provides new data on this point. In addition, this is the first study to analyse gender differences in a stepped care approach of brief interventions in problem drinking GP patients. In the heterogeneous sample, consisting of alcohol dependents, alcohol abusers, at-risk drinkers and binge drinkers, the efficacy of the intervention showed significant results and a small to medium effect size only for women. Since recent meta-analyses show that brief interventions are not effective in alcohol dependent individuals, we excluded these patients as well as binge drinkers from our sample in the next step of data analysis. An analysis of this sub-sample (at-risk drinkers and alcohol abusers) showed results in line with earlier studies, with brief intervention associated with tendencies and small to medium effect sizes in men and women.

A gender–intervention interaction term showed no additional contribution to the regression model, for the whole group and within the subsample. This is a major problem with the interpretation of the results and suggests that other variables are more important for efficacy than gender. The usual explanation for findings that men do better with brief intervention (Kaner et al., 2007) is that women show more assessment reactivity and therefore female control groups show improvements which mask the effects of intervention (Fleming et al., 1997). In the presented study gender does not seem to be the most important factor. Efficacy in brief intervention might be a question of severity in alcohol-related problems or influenced by other, yet to explore, variables.

Analysing the responses to the stepped care procedure showed that women tended to profit more than men from the first step, which in our study was the computerized feedback and the stage-tailored manual. In comparison to 40% of the female study participants, only 24% of the male participants left the programme after the first intervention, because they reduced drinking below recommended limits and revealed sufficient self-efficacy to maintain this reduction. This non-significant tendency for more women than men to benefit from the first step should be explored in further research. (Perhaps the small sample size of this sub-group contributed to the failure to show significance in that analysis.) Responders to the first stage did not differ from non-responders in risk status at baseline. Those closer to a cut-off point for at-risk drinking might be more likely to reach this point after the first intervention, since less behaviour change is required, than those with heavier drinking, where much more behaviour change is required. Again the small sample size might be a barrier in showing evidence for this consideration. Mean ranks in the analysis showed the expected order.

Another finding was that women received only about half the amount of intervention in stepped care condition. Following the higher rate of leaving the programme for women due to intervention success after the first step, the difference in time expended for counselling is explainable.

There are some limitations to this study. In particular, the original trial was not powered sufficiently for this subgroup analysis. Thus the lack of significant differences between control and intervention groups may be due to type II errors. In addition, a low cut-off in AUDIT for positive screening (5 points) produced an extraordinary high rate of patients screening positive, but not fulfilling the inclusion criteria of the study. This might reduce the representativeness of the study, but on the other hand strengthens the sensitivity of the screening instrument and therefore satisfies a high reachability.

However, the presented results could be first hints on explaining the heterogeneous results concerning gender differences in earlier studies and meta-analyses. Describing and explaining the content of brief interventions in detail is necessary to analyse gender specific responses.

Because treatment resources are limited, research needs to look for the most appropriate but cost effective way for providing brief interventions for male and female problematic alcohol drinkers. Future studies should focus on efficacies of brief interventions in different risk statuses (efficacy as a function of severity) and the processes within the intervention that leads to change and to what extent and in which ways men and women differ in these processes.

References

View Abstract