An interactive voice response program to reduce drinking relapse: A feasibility study

2.1. Participants 

Thirty-three men and twenty-seven women, 20–61 years old (M = 41.93, SD = 9.2 years), treated for alcohol dependence at the Herrington Recovery Center, a residential treatment facility of the Rogers Memorial Hospital (Oconomowoc, WI), were asked to participate in the study upon discharge. Three patients were African American and 57 were Caucasian. The mean duration of their residential care was 57.9 days (SD = 21.3 days). The obtained sample was representative of patients receiving care at the Herrington Recovery Center. Study procedures and materials were reviewed and approved by the Western Institutional Review Board (Olympia, WA). The informed consent document signed by each participant explained that the study coordinator would contact all study participants at 30, 90, and 180 days after discharge for a personal interview. They were also informed that they would be asked to provide blood samples for biochemical analysis and that a collateral informant whom they designated would be interviewed by telephone. In compliance with the Office for Human Research Protections regulations, study participants were also permitted to withdraw their consent without consequence or further follow-up at any time. Upon completing data collection at each interval, patients were paid $75 for the 30-day follow-up, $125 for the 90-day follow-up, and $200 for the 180-day follow-up. Informed consent also explained that some participants would be randomly assigned to call a computer-automated telephone system (IVR) on a daily basis for the duration of the study and that no financial or clinical consequence would result from reporting drinking relapses or failing to comply with the daily reporting procedures. No additional participant compensation was associated with providing the daily IVR reports.

2.2. Procedures 

Volunteers to participate in the study were randomly assigned to one of three treatment groups: (1) prompted use of the IVR system; (2) ad libitum use of the IVR system; or (3) no use of the IVR system (control group). The content and functionality of the IVR system are described further subsequently. The study coordinator/counselor was instructed to make a personal telephone call to Group 1 participants any time they failed to make a daily call to the IVR system for 2 consecutive days. The purpose of the phone call was to inquire about the reasons for noncompliance with the reporting procedure and to encourage system use. If Group 1 participants did not begin using the system thereafter, the coordinator/counselor continued calling them daily for at least 10 days. After 10 consecutive days of prompting noncompliant participants to use the system without success, the coordinator/counselor continued to call the patients at least twice each week until system use began or they withdrew consent for study participation. Study participants in Group 2 had access to the same daily IVR reporting system but were not contacted or prompted to use it if they did not make daily calls to the system. Group 3 participants were not given access to the reporting system.

At treatment discharge and 30, 90, and 180 days later, the study coordinator met face to face with each study participant to obtain TLFB reports of drinking events and quantities (at treatment discharge, the referent interval was 90 days prior to intake) using a calendar and other memory aides to enhance the retrospective recall of daily drinking estimates (Sobell & Sobell, 1992). The following paper/pencil self-report instruments were also completed by each participant at each assessment: (1) Work and Social Adjustment Scale (WSAS), a 5-item measure of functional impairment regarding work, home management, social leisure, private leisure, and personal relationships (Marks, 1986, Mundt et al., 2002); (2) Obsessive–Compulsive Drinking Scale (OCDS), a 14-item scale measuring preoccupation with drinking and impulsive drinking behavior (Anton et al., 1995, Anton et al., 1996, Bohn et al., 1996); (3) SF-36, a widely used 36-item medical outcome survey instrument that produces both mental component scores (MCSs) and physical component scores (PCSs; QualityMetric, 1998); and (4) Drinker Inventory of Consequences (DrInC), a 50-item questionnaire of lifetime and recent consequences of alcohol consumption developed for Project MATCH (NIAAA, 1995).

In addition to the participant interviews, when possible, the study coordinator obtained blood samples for biochemical analyses (Carbohydrate-Deficient Transferrin by BioRad Laboratories, Hercules, CA; Early Detection of Alcohol Consumption Test by Alcohol Detection Services, Brookfield, WI; and Whole Blood-Associated Acetaldehyde by LabOne, Lenexa, KS). The study coordinator also contacted a collateral informant (e.g., spouse, parent, close friend, or sibling) by telephone and conducted the Form 90-ACS collateral interview on drinking (NIAAA, 1996).

Study participants randomized to either of the IVR intervention groups were given instructions regarding system use and asked to begin calling the system on a daily basis. The IVR system was composed of multiple interactive modules (see Fig. 1) that incorporated individualized feedback based on current and previously provided responses. System access was provided through a toll-free telephone number; entry into the IVR modules was granted upon registration of a valid ID number and a personally selected four-digit pass code. The IVR system branched to either the study participant daily reporting modules or the study coordinator/counselor menu options based on the user ID.

View Large Image Download to PowerPoint

Fig. 1. The functional structure and call flow through IVR modules in the pilot relapse reduction program.

2.3. Participant interface 

The drinking status of study participants was assessed first by asking them whether they had or had not consumed alcohol since their prior call (the date of their last call was identified or the prior 24 hours was referenced for the first call). Callers indicating that they had consumed alcohol were prompted to indicate the number of beer, wine, and liquor drinks (definitions of a standard drink were provided) and to confirm the responses. Callers indicating that they had not consumed alcohol also confirmed their responses and provided ratings to four questions that were adapted from the Penn Alcohol Craving Scale (Flannery, Volpicelli, & Pettinati, 1999) concerning the ease or difficulty of abstaining from drinking and the severity of drinking urges. Participants who indicated the occurrence of a drinking episode after a prior IVR report of abstinence were warned of a potential abstinence violation effect (Marlatt & Gordon, 1985) that could contribute to a more severe relapse. A report of abstinence following a prior call reporting a drinking episode was acknowledged and praised. Individual ratings of coping difficulty were monitored from call to call, and individualized messages were provided based on the direction of change between calls and the overall magnitudes of their ratings.

Participants reporting a drinking episode to the IVR system were sent to the Drinking Reasons module. In this module, they indicated whether each of 16 potential situations drawn from the Reasons for Drinking Questionnaire (RDQ; Zywiak, Connors, Maisto, & Westerberg, 1996) contributed to the initiation of their reported drinking episode. Positive endorsements of specific items on the RDQ were categorized as being related to negative emotions, social pressure/positive emotions, or urges/cravings. A reported drinking episode associated with 2 or more RDQ items within a specific category prompted tailored feedback about the identified factors, and brief advice was provided for coping with such situations in the future.

Participants reporting abstinence since their prior call were branched to the Resolve Not to Drink module. This module assessed their perceived likelihood of drinking in each of 12 potential high-risk drinking situations drawn from the Drinking Refusal Self-Efficacy Questionnaire (Young & Knight, 1989). Each situation assessed was related to emotional relief, opportunity, or social pressure/positive emotions. Endorsement of 2 or more situations within any of the domains resulted in feedback specific to the identified situations, and advice was offered for dealing with such situations should they occur subsequently.

After completion of either the Drinking Reasons module or the Resolve Not to Drink module, depending on the reported drinking status, all callers were given an opportunity to hear cognitive–behavior advice adapted from the Project MATCH CBT Manual (NIAAA, 1999). They were not required to listen to this module if they did not wish to.

All callers completed a seven-question relapse risk questionnaire adapted from multiple instruments that assessed personal efforts to avoid drinking, perceptions of personal stress and depression, use of social support systems to avoid drinking, and social pressures and high-risk situations such as planning to attend a party or celebration where alcohol would be served. Participants indicating an elevated risk of relapse were given feedback about the heightened risk of relapse and encouraged to seek help if needed.

At the end of the call, study participants were informed of the existence of any voice mail message that had been left for them by the study coordinator/counselor. They were allowed to listen to the message immediately or to save it for retrieval at a later time. They were also given an opportunity to record a message for the study coordinator/counselor and could request a personal phone call from the coordinator/counselor.

2.4. Coordinator–counselor interface 

Each morning, the study coordinator/counselor received an automatic facsimile report listing the following: (1) IDs of all study participants who had not called the IVR reporting system for 2 or more consecutive days; (2) IDs of all those who had left voice mail messages; and (3) IDs of those who had requested a personal phone call. The IVR system recognized the study coordinator based on her ID number. Upon entry to the coordinator/counselor module, she could select actions from the following options: (1) request a new daily summary report facsimile of the IDs of noncompliant participants, IDs of participants with messages, and IDs of participants requesting a call back; (2) send and retrieve voice mail messages with specific study participant ID numbers; and (3) request a facsimile report summarizing data from all the calls received from a specific participant during the prior 30 days. This report indicated the date of each call received, the reported drinking status, and a summary of the data entered (e.g., drinks consumed and contributing reasons, ratings of coping difficulty and situational confidence, and overall relapse risk).

3.1. Group equivalence at baseline 

No significant difference was evident between the randomized groups regarding sex, age, and length of stay in residential treatment. Alcohol consumption patterns from TLFB data referencing the 90 days prior to entering treatment were equivalent across the three groups, with overall means of 57.9 drinking days (SD = 24.2 days), 52.3 heavy drinking days (SD = 24.4 days; criterion of four drinks per day for women and five drinks per day for men), and 660 total drinks consumed (SD = 596). The groups were also similar with respect to the following measures: OCDS obsession (M = 7.7, SD = 4.2) and compulsion (M = 11.5, SD = 4.7) scores; SF-36 MCS (M = 37.3, SD = 12.5) and PCS (M = 52.3, SD = 9.2); total lifetime (M = 32.7, SD = 7.6) and recent (M = 60.8, SD = 25.3) consequences of drinking as measured by the DrInC; and functional impairment (M = 19.4, SD = 9.9) as measured by the WSAS. WSAS scores higher than 20 are indicative of moderately severe or worse psychopathology. Scores between 10 and 20 reflect significant functional impairment associated with mild-to-moderate clinical symptoms, whereas those lower than 10 are typically associated with subclinical populations (Mundt, Marks, et al., 2002).

3.2. Follow-up at 30 days 

During the first month after discharge, 4 subjects in Group 1 (prompted system use), 2 subjects in Group 2 (ad libitum system use), and 1 subject in Group 3 (control) withdrew their consent for study participation and dropped out of the study before the first follow-up interview. These dropout data (and those at the 90-day follow-up as discussed subsequently) suggest that the intensity/intrusiveness of daily reporting was aversive to a considerable number of the study participants. An average of 20.2 (SD = 7.4, range = 2–30) daily calls were received from the 34 participants in Groups 1 and 2 remaining in the study through the first follow-up interview.

At the 30-day follow-up interview, self-reported abstinence (TLFB data) was reported by 13 of the 16 Group 1 participants, 12 of the 18 Group 2 participants, and 13 of the 19 Group 3 participants. Of all the participants, 2 of the Group 1 participants, 5 of the Group 2 participants, and 7 of the Group 3 participants who self-reported abstinence had one or more positive biomarkers for drinking and/or collateral reports suggesting that they may have been drinking. Of the 15 participants self-reporting a relapse (n = 3, 6, and 6 for each group, respectively), the control subjects reported more drinking days and more heavy drinking days but fewer total drinks consumed than the patients from the other two groups. Mean outcome scores from the OCDS subscales, the SF-36 MCS, DrInC, and WSAS suggested that Group 1 participants reported somewhat less impairment than the participants from the two groups; however, these differences were not statistically significant and should not be overinterpreted.

3.3. Follow-up at 90 days 

Between the 1-month and 3-month follow-up periods, 2 additional subjects in Group 1 and 1 subject in Group 2 withdrew their study consent and were not interviewed at the second follow-up. An average of 36.7 (SD = 18.1, range = 0–60) daily calls were received from the 31 participants in Groups 1 and 2 remaining in the study through the second follow-up interview.

At the 90-day follow-up interview, self-reported abstinence (TLFB data) was reported by 8 of the 14 Group 1 participants, 13 of the 17 Group 2 participants, and 13 of the 19 Group 3 participants. Of all the participants, 2 of the Group 1 participants, 8 of the Group 2 participants, and 8 of the Group 3 participants self-reporting abstinence had one or more positive biomarkers for drinking and/or collateral reports suggesting that they may have been drinking. Of the 16 participants self-reporting a relapse between the first and second follow-ups (n = 6, 4, and 6 for each group for each group, respectively), the control subjects reported more drinking days, more heavy drinking days, and more total drinks consumed than the patients from the other two groups. Mean outcome scores from the OCDS subscales and DrInC indicated that Group 2 participants reported less impairment than the participants from the other two groups, with Group 3 participants indicating less impairment on the WSAS. Again, such differences are not statistically significant and must be interpreted cautiously.

3.4. Follow-up at 180 days 

Between the 3-month and 6-month follow-up periods, one subject in Group 2 and one subject in Group 3 withdrew their study consent and were not interviewed at the end of the study. An average of 51.4 (SD = 22.6, range = 10–89) daily calls were received from Groups 1 and 2 participants who completed the study.

At the final follow-up interview, self-reported abstinence (TLFB data) was reported by 7 of the 14 Group 1 participants, 14 of the 16 Group 2 participants, and 13 of the 18 Group 3 participants. Of all the participants, 4 Group 1 participants, 6 Group 2 participants, and 10 Group 3 participants self-reporting abstinence between the second and third follow-up intervals had one or more positive biomarkers for drinking and/or collateral reports suggesting that they may have been drinking. Of the 14 participants self-reporting a relapse between the 90- and 180-day follow-ups (n = 7, 2, and 5 for each group, respectively), the control subjects reported more drinking days, more heavy drinking days, and more total drinks consumed than the patients from the other two groups. Mean outcome scores from the OCDS subscales, the SF-36 MCS, DrInC, and WSAS suggest that Group 2 participants reported less impairments than the participants from the other two groups, but these differences are not statistically significant.

3.5. IVR system use by Groups 1 and 2 

The mean duration of the initial call (presenting full system instructions and definitions) was 9 minutes 21 seconds. Call duration decreased rapidly during the first 2 weeks of use as participants became familiar with the system and made use of keyword identifiers to provide their responses quickly (Mundt et al., 1997). The mean duration of the 379 calls received from Day 15 to Day 30 was 5 minutes and 2 seconds. During the 60-day follow-up period from Day 31 to Day 90, the mean call length was 4 minutes 40 seconds (n = 1,183 calls); during the final follow-up period, the 1,508 calls received took an average of 4 minutes 35 seconds to complete. There was no significant difference in the number of daily calls received from each group for any of the follow-up intervals. In fact, in all three follow-up intervals, the mean number of calls received from Group 2 participants was greater than that from Group 1 participants—with fewer participants withdrawing consent. These data are not statistically significant but suggest that the additional personal attention afforded to noncompliant Group 1 participants was not effective in increasing reporting behavior and may have been detrimental to study participation.

Forty participants were initially randomized to use the IVR reporting system. A total of 10 participants subsequently withdrew their consent and dropped out of the study, 6 from Group 1 and 4 from Group 2. Twenty participants opted to hear the CBT advice at least once after completing the drinking assessments. Additional features incorporated into the IVR system were voice mail messaging and requesting a follow-up call from the study coordinator. Of the participants, 30 chose to leave a message for the study coordinator at least once (M = 7.1, range 1–48 times) and 12 requested personal follow-up calls (M = 1.1, range = 1–21 times).

Of the 6,090 potential daily calls that could have been received from consenting participants within each follow-up interval, 3,464 were received (56.9%). Of the 30 participants randomized to Group 1 or 2 who completed the study, calling compliance ranged from 8.3% to 99.4%. In terms of compliance, 3 participants were less than 25% compliant, 7 had compliance rates between 25% and 50%, 13 called the system on 50–75% of the days, and 7 provided daily calls on more than 80% of the possible days.

3.6. IVR system use related to relapse events 

From the 30 Groups 1 and 2 participants who completed the study, 3,290 contemporaneous IVR daily reports were received. Of these 30 participants, 17 reported no relapse event at any of the follow-up intervals using the TLFB. Although some of these abstinence self-reports may be suspect, based on the collateral report and biochemical data, there is minimal reason to question the self-reports from participants acknowledging relapse events.

Contemporaneous use of the IVR system by the 13 participants who retrospectively reported relapse events via TLFB was examined. These participants reported a total of 276 drinking days via TLFB, whereas only 103 drinking days were reported via IVR. These data appear contrary to those from prior studies that found higher estimates of drinking using contemporaneous reporting methods relative to TLFB (e.g., Searles et al., 2002, Searles et al., 1995). In this study, we found that contemporaneous IVR calls were not made on 53.6% of the TLFB days when a drinking event was reported but missed for only 42.4% of TLFB abstinent days, suggesting that IVR calls on drinking days were more likely to be skipped than were calls on non-drinking days. Upon further examination, we found that 66.0% of the 103 contemporaneously reported IVR drinking days were preceded by 1 or more noncompliant reporting days. In addition, the mean number of drinks reported via IVR following a noncompliant reporting day (M = 10.4, SD = 13.9) is significantly greater than the number of drinks reported (M = 4.3, SD = 3.4) if an IVR call was received the prior day, t(101) = 2.53, p = .013. The number of previously missed reporting days is significantly correlated with the self-reported severity of the relapse (ρ = .485, p < .001).