|
指导留学生论文：Using Daily Interactive Voice Response Technology to Measure Drinking and Related Behaviors in a Pharmacotherapy Study
Henry R. Kranzler, Khamis Abu-Hasaballah, Howard Tennen, Richard Feinn, and Kevin Young
Background: Interactive voice response technology (IVR) allows investigators to collect daily measuresof drinking, medication adherence, mood, and other treatment-relevant variables that may change day today during a clinical trial. Despite these advantages, no published studies have used IVR in alcoholpharmacotherapy trials.
Methods: Subjects provided daily data via IVR during the 12-week treatment period. Seven subjectscompleted the trial.
Results: We found a high level of participant adherence to the IVR protocol, higher levels ofdrinking reported by IVR than by a commonly used recall method, and distinct within-day associationsbetween daily mood and alcohol consumption: these could not be obtained through traditional assessmentmethods.
Conclusions: IVR seems to be feasible for the collection of daily indicators of treatment outcomes andprocesses in pharmacotherapy studies among problem drinkers.
Key Words: Interactive Voice Response, Outcomes Assessment, Problem Drinkers, Pharmacotherapy,
Alcoholism.
AVARIETY OF METHODS have been developed tomeasure drinking behavior in alcohol-treatment trials.
http://www.ukassignment.org/liuxueshenglunwen/The most widely used approach is the timeline follow-backmethod (TLFB; Sobell and Sobell, 1992), in which aninterviewer uses a calendar of events identified by therespondent to facilitate the respondent’s retrospective estimatesof alcohol consumption on a day-to-day basis. Althoughaggregate levels of alcohol consumption obtainedwith the TLFB compare favorably with reports obtaineddaily, the TLFB cannot accurately identify day-to-day variationin drinking behavior (Carney et al., 1998). Dailyreports represent a novel and uniquely informative approachto data collection in alcohol-treatment studies, becausethey provide an opportunity to study the moderatingeffect of treatment on relations between drinking and subjectivestates such as mood (Collins et al., 1998; Kranzler etal., 2004). Furthermore, daily variation in mood or desireto
drink may help to elucidate the mechanism of treatmenteffects on drinking behavior (Kranzler et al., 2004; Tennen
et al., 2000).
Interactive voice response technology (IVR) is a relativelynew and increasingly popular technology that uses
the telephone to administer survey questions. The respondentanswers each question by pressing the keys on the
telephone keypad, and responses are entered automaticallyin a database. IVR has a number of advantages over traditional
face-to-face interviews, including interview consistency,access to difficult-to-reach populations such as recoveringdrug users and alcoholics (Alemi et al., 1994),mmediate data availability and accessibility, and convenienceboth for respondents and investigators.#p#分页标题#e#
In view of these advantages, IVR has been used in a widevariety of studies ranging from symptom and alcohol consumptionmonitoring to psychological assessment andtreatment (Corkrey and Parkinson, 2002). Bardone et al.(2000), Helzer et al. (2002), and Searles et al. (2000, 2002)have demonstrated the feasibility, reliability, and validity ofdaily IVR drinking reports in community and studentsamples.
This pilot study examined the feasibility of daily IVRto evaluate both outcomes and treatment processes in anopen-label study of daily versus targeted naltrexone toreduce drinking among problem drinkers. To our knowledge,this represents the first use of IVR for evaluatinoutcomes in an alcohol-treatment study.
From the Lowell P. Weicker, Jr. General Clinical Research Center (HRK,KA-H), the Department of Community Medicine and Healthcare (HT), andthe Alcohol Research Center, Department of Psychiatry (HRK, HT, RF, KY),University of Connecticut Health Center, Farmington, Connecticut.Received for publication December 24, 2003; accepted April 20, 2004.
Supported by NIH Grants P50 AA03510, K24 AA13736, and M01
RR06192 (University of Connecticut General Clinical Research Center).
Reprint requests: Henry R. Kranzler, MD, Department of Psychiatry, Universityof Connecticut Health Center, 263 Farmington Ave., Farmington, CT
06030-2103; Fax: 860-679-1316; E-mail: [email protected].
Copyright © 2004 by the Research Society on Alcoholism.
DOI: 10.1097/01.ALC.0000130806.12066.9C
0145-6008/04/2807-1060$03.00/0
ALCOHOLISM: CLINICAL AND EXPERIMENTAL RESEARCH
Vol. 28, No. 7
July 2004
1060 Alcohol Clin Exp Res, Vol 28, No 7, 2004: pp 1060–1064
METHODS
Overview
Nine heavy-drinking subjects were recruited through newspaper advertisements.After a telephone screening interview, subjects were interviewedin person by a research nurse, gave informed consent to participate,
and then underwent a clinical laboratory evaluation and physicalexamination. All subjects who were interviewed met study criteria and,consequently, were assigned to receive naltrexone, an orally availableopioid antagonist approved for the treatment of alcohol dependence. Thedose of naltrexone used was either daily (i.e., one 50-mg tablet) ortargeted [i.e., one 50-mg tablet to be used 3–5 days per week in anticipationof (and to prevent) a heavy-drinking episode]. Heavy-drinking episodes
ere defined as four or more drinks in a day for women and five ormore drinks in a day for men. By using the Inventory of Drinking Situations(Annis et al., 1987) as a guide, subjects provided a list of the five
riskiest settings in terms of the likelihood of their drinking in excess of thatamount. Subjects also received brief counseling every other week duringthe 12-week treatment period. Retrospective drinking assessments wereconducted at the time of study enrollment, every two weeks at eachtreatment visit, and at the end of treatment. Subjects were asked to calla#p#分页标题#e#
toll-free study telephone number to complete an IVR interview each dayof the 84-day study.
Subjects
Inclusion criteria were age 18 to 60 years, stable residence, ability toread English, average weekly intake of 18 standard drinks for women
and 24 standard drinks for men, and a desire to reduce their drinking.Women of childbearing potential were included in the study only if they
were practicing reliable birth control and had a negative serum pregnancytest. Subjects were excluded if they showed clinical evidence of physicaldependence on alcohol, were deemed to require more intensive clinical
care than that provided in the study, met current criteria for a DSM-IVdrug use disorder (other than nicotine), met lifetime criteria for opioiddependence, had used psychoactive drugs in the preceding month, or had
a serious psychiatric or medical illness.Five of the nine subjects enrolled in this pilot study were men, and all
were white. Mean age was 48.4 years (SD, 8.3 years), and mean educationwas 17.4 years (SD, 3.2 years). Seven individuals (78%) met DSM-IVcriteria for current alcohol dependence. Comorbid psychiatric disorders
were identified in four individuals, three of whom met criteria for alifetime (but not current) diagnosis of major depressive disorder. Twoindividuals met criteria for a lifetime (but not current) diagnosis of
cannabinoid use disorder (one of the individuals with major depressivedisorder had cannabinoid dependence; another individual met cannabinoidabuse criteria). Three subjects were receiving antidepressant therapy
at the time of their study participation. Subjects reported drinking on93.2% of days (SD, 5.2%), of which 50.2% of these drinking days wereheavy-drinking days (four or more drinks per day for women and five or
more drinks per day for men). Four of the subjects received the targetedtreatment, whereas the remaining five received the daily treatment.
Assessments
A demographic form and the Structured Clinical Interview for DSM-IV
(First et al., 1996) were administered to determine study eligibility. TheTLFB was administered to assess the subjects’ accuracy in reportingdrinking and medication adherence both at 2-week intervals and also onceat the completion of the study to cover the entire 12-week treatmentperiod.
IVR Technology
The IVR system used SmartQuest and DialQuest (Telesage, Inc.,Seattle, WA). Each question was answered numerically by using thetelephone keypad. The system allowed participants to respond to questionsbefore the question was completed, thus creating a shorter interviewas the participant became familiar with the system.
This IVR system offers the best combination of verifiable daily electronicdata entry, easy participant access to a data-entry portal, andautomated transfer to a Microsoft Access database (Redmond, WA), yetwithout requiring that participants be comfortable in a personal computeror handheld computer environment. This system also eliminates the inevitable#p#分页标题#e#
data-transcription and coding errors associated with paper-andpencilassessment. The IVR system was configured to process up to sixcalls simultaneously.
During their baseline visit, subjects met with a research assistant for abrief (10–15 min) IVR training session, at which time they were providedwith a toll-free number that directed them to six available lines. Each
subject was provided with an anonymous study identification number (toensure the confidentiality of all study assessments) and a personalizedalternate number (i.e., based on birth date) to use in the event that he orshe forgot the study number.
Subjects were asked to call the system from a touch-tone phone (which
could be a cell phone) between the hours of 5:00 and 9:00 PM. This timeframe offered sufficient flexibility to allow the individual to respond on thebasis of the day’s experience before the time of day when individuals
typically drink most heavily. We selected a 4-hr reporting window toprovide consistent timing of reports across days and to promote adherenceto the IVR regimen. Data were not obtained on days that subjects failedto call in during the 4-hr window. This feature reflects our interest in thetemporal relations among events and the overall high levels of response
that we anticipated would occur on the basis of our prior experience withdaily paper booklets (Kranzler et al., 2003).
Each subject was provided with a wallet-sized interview guide with keyterms for each interview question in the order in which it was presented bythe system. The system was programmed so that a subject who wished tochange a response after it was entered could do so by pressing the star (*)key to re-enter the response.
Subjects who did not call the toll-free number by 8:00 PM were calledby the system at a preferred phone number and reminded to complete theinterview. The reminder prompt was brief and provided no details of thestudy or its purpose. When the final IVR interview question was answered,the subject’s data for that day were stored in a data file that was automatically
associated with her or his identification number.
The IVR interview assessed mood, desire for alcohol, confidence toresist drinking, medication use and side effects, and alcohol consumptionsince the last day’s call. The main focus of this report is on medication
adherence and alcohol consumption—two key measures in the conduct ofmedication trials to treat alcohol dependence. We also contrast dailymood reports with mood assessment based on aggregate (between-person)reports.
Medication Adherence. Using IVR, subjects reported whether they
consumed a naltrexone tablet since the previous call. In TLFB interviewsconducted by a research nurse biweekly, subjects recalled the days on
which they took medication.
Daily Drinking. Each day, subjects recorded their alcohol consumptionfor the previous night (i.e., after the last IVR survey) and for that day.
They reported the number and quantity (in standard drinks) of each ofthree categories of alcoholic beverages—beer, wine, and liquor—separatelyfor “last night” (the time elapsed between the preceding day’s calland the end of the day) and “today” (the time elapsed since the beginningof the day and the time of the call).Daily Mood. Subjects were asked to rate their mood for that day byusing nine mood descriptors (active, angry, bored, enthusiastic, happy,nervous, relaxed, sad, and tranquil) derived from the mood circumplex(Larsen and Diener, 1992) and our previous research. Each mood wasrated on a five-point scale (0 “not at all” to 4 “extremely”). Themoods were combined into two groups: (1) positive mood, consisting ofthe items active, happy, enthusiastic, tranquil, and relaxed ( 0.90 based#p#分页标题#e#
on aggregated responses), and (2) negative mood, consisting of the itemssad, angry, nervous, and bored ( 0.86 based on aggregated responses).
INTERACTIVE VOICE RESPONSE IN HEAVY DRINKERS 1061
Consistent with the circumplex model, positive and negative moods were
not significantly related ( 0.48; p 0.19; n 9).
Statistical Analysis
Cohen’s and the Pearson coefficient (r) or Spearman coefficient ()
were used to measure concordance and association among measures,
respectively. Multilevel modeling with hierarchical linear modeling (Raudenbush
et al., 2001) was used to determine change over time in the
duration of phone calls and the number of daily drinks.
RESULTS
指导留学生论文IVR Adherence
Seven subjects (78%) completed the 12-week treatment
trial; one subject was lost to follow-up, and one discontinued
treatment because he found the medication not to be
therapeutic. There was substantial variation in the proportion
of daily IVR reports subjects provided, ranging from
12 to 84, with a total of 562 valid responses (mean, 74.3%
of the maximum possible responses; SD, 29.0%). When we
included only IVR reports from the first call to the day of
the last call (to account for the two subjects who discontinued
treatment prematurely), the average response rate was
83.0% (SD, 20.7%).
The IVR system made 178 reminder calls to participants
who had not made an IVR report by 8:00 PM. Subjects
received an average of 19.8 calls (SD, 12.5). Reminder calls
resulted in 54 calls to the system (i.e., a 30.3% rate of
response to the reminder calls, which represents 9.3% of all
calls received).
The average duration of a phone call was 3.9 min (SD,
0.98 min). Eighty-seven percent of the calls were less than
5 min long, and as subjects became familiar with the IVR
protocol over the first 10 weeks of the study, the duration
of each call decreased [F(1,467) 420.78; p 0.001], with
each successive call taking 0.02 min less than the previous
call. After the 10th week, the length of the calls remained
stable, taking on average 2.81 min.
Medication Adherence
When interviewed at 2-week intervals by using the TLFB,
subjects reported taking medication on 434 days, compared
with 423 days reported via IVR. The overall concordance
rate between TLFB and IVR was 78.2% ( 0.69; p
0.001).
Alcohol Consumption: IVR Versus TLFB
Drinking Days. On the basis of IVR, subjects reported
drinking on 479 days, compared with 470 days by using the
biweekly TLFB. The correlation between IVR and TLFB
for total number of drinking days by subject was 0.95 (n
9; p 0.001). The number of drinking days did not differ
by method at either the within-person level (t554 1.52; p#p#分页标题#e#
0.13) or the between-persons level (t8 0.83; p 0.43).
The two methods agreed for 457 drinking days (i.e., 95%)
and 63 abstinent days (i.e., 83%), resulting in excellent
concordance ( 0.75; p 0.001).
Drinks per Day. The correlation between the number of
daily drinks recorded with IVR and the biweekly TLFB was
r 0.69 (p 0.001), ranging from r0.12 to r 0.90 for
individual subjects. The correlation between IVR and the
12-week TLFB was r 0.52 (p 0.001), ranging from r
0.10 to r 0.60.
When examined between persons, there was no difference
in the number of drinks per day reported with daily
IVR and biweekly TLFB (IVR: mean, 3.15; SD, 1.22; biweekly
TLFB: mean, 2.75; SD, 1.40; t8 1.25; p 0.25).
However, subjects reported consuming more drinks on the
daily IVR than when they reported their alcohol use on the
12-week TLFB (IVR: mean, 3.22; SD, 1.32; 12-week TLFB:
mean, 2.44; SD, 0.81; t6 3.11; p 0.021).
By using multilevel modeling, with days nested within
subject, the mean number of drinks within persons was
higher for IVR (mean, 3.21; SD, 2.29) compared with both
the biweekly TLFB (mean, 2.54; SD, 11.69; t8 2.69; p
0.028) and the 12-week TLFB (mean, 2.50; SD, 1.40; t6
3.29; p 0.017).*
Within-Person Versus Between-Person Associations of Daily
Drinking and Mood
Daily reports allow a comparison of within-person and
between-person associations among study variables. We
compared within- and between-person associations of the
number of drinks consumed with positive and negative
mood. Participants reported consuming fewer drinks on
days during which they experienced a more positive mood
(r0.13; p 0.003), but the number of drinks on a given
day was not related to that day’s negative mood (r 0.05;
p 0.22). When the observations were aggregated across
subjects (n 9), producing between-persons drinking data
comparable to what is reported in most clinical trials, there
was no association between positive mood and the number
of drinks per day (0.02; p 0.97) or between negative
mood and the number of drinks per day ( 0.42; p
0.26).
DISCUSSION
This pilot study, conducted in preparation for a placebocontrolled
trial investigating the effects of targeted naltrexone
in problem drinkers, demonstrated the feasibility of
using daily automated telephone interviews to evaluate
outcomes in alcohol pharmacotherapy studies. IVR was
well received by study participants, who demonstrated a
* The correlation using within-subjects data violates the assumption of
independent observations and may result in an erroneous conclusion regarding
the relationship between the two variables of interest, whereas using the
aggregate between-subjects data reduces the sample size and removes the#p#分页标题#e#
variation that occurs within individuals. This is the main impetus for the
popularity of multilevel modeling, in which both the within-person and
between-person variation are modeled.
1062 KRANZLER ET AL.
high degree of adherence with the call requirements. The
use of reminder calls demonstrably enhanced adherence,
an important element in the application of IVR technology
to clinical trials. The duration of the calls was generally less
than 5 min, with a significant reduction over time through
the first 10 weeks of the trial.
Consistent with prior reports, daily IVR seems to provide
a more accurate estimate of drinking behavior than does a
commonly used retrospective recall method, whether the
recall method is applied over a 2-week or a 12-week interval.
When compared across subjects, we found convergence
on both the number of drinking days and mean daily consumption
between daily IVR and TLFB, which is in agreement
with findings reported by Bardone et al. (2000). However,
when compared by using within-person observations
and multilevel modeling, there was evidence that subjects
underreported alcohol use on the TLFB. On average, subjects
reported drinking a mean of 0.58 more drinks per day
when this was assessed with daily IVR compared with the
biweekly TLFB and 0.78 drinks more drinks per day compared
with the 12-week TLFB. These differences are consistent
with the findings of Searles et al. (1995, 2000), who
found that traditional quantity/frequency indicators of alcohol
use yielded lower levels of alcohol use compared with
daily IVR. In concluding that subjects underreported their
alcohol use on the TLFB, we assume that close-to-real-time
drinking reports (i.e., daily IVR) are more accurate than
recalled drinking and that recall methods introduce memory
decay and memory bias (Tennen and Affleck, 1996).
The extreme variation in the correlation for individual
subjects between daily IVR and TLFB also indicates that
for some individuals, TLFB-derived drinking reports bear
little resemblance to drinking reports captured close to
their real-time occurrence through daily IVR.
There was substantial variation in the proportion of daily
IVR reports, with the number of daily reports ranging from
12 to 84. That data were missing from some individuals is
potentially problematic, but no more so than in studies inwhich the TLFB is used retrospectively to measure drinking
指导留学生论文behavior. Another potential limitation of this study is thatsome individuals may have been intoxicated at the time
they provided daily reports. Using blood alcohol concentrationand collateral reports, Perrine et al. (1995) demonstrated
the reliability of daily IVR reports. Furthermore,we found no evidence of outlying reporting days during#p#分页标题#e#
which it took more time than usual to complete the dailyprotocol, which would suggest intoxication in the respondent.
Because subjects had a relatively low level of alcoholdependenceseverity, research on the feasibility of using
IVR in more severely dependent subjects is needed. In viewof the small sample size and the subjects’ low level of
dependence severity, caution should be exercised in generalizingthe findings to the universe of alcohol-dependentsubjects participating in treatment trials.
Despite these limitations, as demonstrated in this study,
IVR has potential utility in the conduct of alcoholtreatmenttrials. A common problem in alcohol-treatmentstudies is subjects’ failure to return to the treatment siteand to comply with efforts to collect TLFB information via
the telephone. An advantage of daily data collection via
IVR is that multilevel modeling, which is made possible by
daily reports, is robust with regard to missing data and that
it weights each participant’s data on the basis of the amountof data supplied by that individual.
A distinct advantage of obtaining daily drinking reports
in clinical trials is that they allow investigators to examine
within-person associations as well as between-persons associationsamong study variables. Indeed, we found that
although participants consumed fewer drinks on days duringwhich they experienced a more positive mood, therewas no association between positive mood and drinks perday when this was evaluated with traditional betweenperson
(aggregated) data. Many theories of alcohol usedescribe inherently within-person processes. For example,the self-medication hypothesis asserts that problem drinkersuse alcohol to alleviate negative affective states (Swendsenet al., 2000). Alcohol pharmacotherapy studies canbegin to determine whether treatment affects this withinpersonprocess and other hypothesized drinking dynamics
by incorporating daily reports as outcome indicators (Kranzleret al., 2004). Our findings indicate that daily IVR is afeasible and efficient way to measure outcomes and daily
drinking processes in clinical trials and that the range of its
application in such trials warrants further study.
ACKNOWLEDGMENT
We thank the staff of the Clinical Research and EvaluationUnit of the University of Connecticut Alcohol Research Center(particularly Lynn McLaughlin, RN, and Kristen Tremblay, BA)for their assistance in the conduct of this study.
REFERENCES
Alemi F, Stephens R, Parran T, Llorens S, Bhatt P, Ghadiri A, Eisenstein
E (1994) Automated monitoring of outcomes: application to treatmentof drug abuse. Med Decis Making 14:180–187.Annis HM, Graham JM, Davis CS (1987) Inventory of Drinking Situations
(IDS) User’s Guide. Alcoholism and Drug Addiction Research Foundation,
Toronto.
Bardone AM, Krahn DD, Goodman BM, Searles JS (2000) Using interactive
voice response technology and timeline follow-back methodologyin studying binge eating and drinking behavior: different answers todifferent forms of the same question? Addict Behav 25:1–11.#p#分页标题#e#
Carney MA, Tennen H, Affleck G, Del Boca FK, Kranzler HR (1998)Levels and patterns of alcohol consumption using timelinefollow-back,
daily diaries and real-time electronic interviews. J Stud Alcohol 59:447–
454.
Collins RL, Morsheimer ET, Shiffman S, Paty JA, Gnys M, PapandonatosGD (1998) Ecological momentary assessment in a behavioral drinking
moderation training program. Exp Clin Psychopharmacol 6:306–315.
Corkrey R, Parkinson L (2002) Interactive voice response: review of
studies 1989–2000. Behav Res Methods Instrum Comput 34:342–353.
First MB, Spitzer RL, Gibbon M, Williams JBW (1996) Structured Clinical
Interview for DSM-IV Axis I Disorders—Patient Edition (SCID-I/P, Version
2.0). Biometrics Research Department, New York State Psychiatric
Institute, New York.
INTERACTIVE VOICE RESPONSE IN HEAVY DRINKERS 1063
Helzer JE, Badger GJ, Rose GL, Mongeon JA, Searles JS (2002) Decline
in alcohol consumption during two years of daily reporting. J Stud
Alcohol 63:551–558.
Kranzler HR, Armeli S, Feinn R, Tennen H (2004) Targeted naltrexonetreatment moderates the relations between mood and drinking behavioramong problem drinkers. J Consult Clin Psychol, 72:317–327.
Kranzler HR, Armeli S, Tennen H, Blomqvist O, Oncken C, Petry N,Feinn R (2003) Targeted naltrexone for early problem drinkers. J Clin
Psychopharmacol 23:294–304.
Larsen RJ, Diener E (1992) Promises and problems with the circumplexmodel of emotion, in Review of Personality and Social Psychology (Clark
MS ed), pp 25–59. Sage, Newbury Park, CA.Perrine MW, Mundt JC, Searles JS, Lester L (1995) Validation of dailyself-reported alcohol consumption using interactive voice response
(IVR). J Stud Alcohol 56:487–490.Raudenbush SW, Bryk AS, Cheong YF, Congdon R (2001) HLM 5:Hierarchical Linear and Nonlinear Modeling. Scientific Software International,
Lincolnwood, IL.
Searles JS, Helzer JE, Rose GL, Badger GJ (2002) Concurrent and
retrospective reports of alcohol consumption across 30, 90 and 366 days:interactive voice response compared with the timeline follow back.
J Stud Alcohol 63:352–362.
Searles JS, Helzer JE, Walter DE (2000) Comparison of drinking patternsmeasured by daily reports and timeline follow back. Psychol Addict
Behav 14:277–286.
Searles JS, Perrine MW, Mundt JC, Helzer JE (1995) Self-report of
drinking using touch-tone telephone: extending the limits of reliabledaily contact. J Stud Alcohol 56:375–382.
Sobell L, Sobell M (1992) Timeline follow-back: a technique for assessingself-reported alcohol consumption, in Measuring Alcohol Consumption
(Litten R, Allen J eds). Human Press, Clifton, NJ.Swendsen J, Tennen H, Carney MA, Affleck G, Willard A, Hromi A
(2000) Mood and alcohol consumption: an experience sampling test of
the self-medication hypothesis. J Abnorm Psychol 109:198–204.#p#分页标题#e#
Tennen H, Affleck G (1996) Daily processes in coping with chronic pain:
methods and analytic strategies, in Handbook of Coping (Zeidner M,
Endler NS eds), pp 151–180. Wiley, New York.
指导留学生论文Tennen H, Affleck G, Armeli S, Carney MA (2000) A daily process
approach to coping: linking theory, research and practice. Am Psychol
55:626–636.
1064 KRANZLER ET AL.
|
