Assessment of alcohol intake: Retrospective measures versus a smartphone application
Introduction
In research focusing on the causes and consequences of problem drinking, accurate assessment of alcohol use – both in terms of volume and pattern – is vital. Currently, such research relies primarily on data collected using well-validated and reliable retrospective self-report measures administered in the laboratory. Retrospective measures can be broadly categorised into two main types: quantity-frequency surveys and retrospective diaries. In their simplest form, quantity-frequency surveys ask participants to specify how much they usually drink and how often (Utpala-Kumar & Deane, 2010). This permits average daily consumption to be determined rapidly and efficiently, but does not allow for any investigation of pattern of intake (Del Boca & Darkes, 2003). Moreover, such surveys have been found to under-estimate total consumption by almost 30% when compared to prospective daily assessments (Heeb & Gmel, 2005) and up to 50% when compared to per capita sales of alcohol (Stockwell et al., 2004, Stockwell et al., 2008). This discrepancy is attributed, at least in part, to difficulties participants have conceptualizing their usual consumption, especially if their intake is highly variable across drinking occasions, a tendency to overlook occasional high intake sessions, and recall bias (Del Boca and Darkes, 2003, Stockwell et al., 2004). Participants are consequently thought to report modal rather than average consumption in response to quantity-frequency surveys (Utpala-Kumar & Deane, 2010).
The accuracy and detail of data obtained from quantity-frequency surveys can be enhanced somewhat by the inclusion of additional components. Beverage-specific questions, an explicit reference period, and including items that assess high intake behaviour have all been found to generate increased estimates of consumption and more comprehensive information about pattern of intake (Dawson, 2003, Del Boca and Darkes, 2003, Rehm, 1998). A number of measures adopt this type of approach. The Alcohol Use Questionnaire (AUQ), for instance, combines beverage-specific quantity-frequency questions about weekly intake over a six-month period with items that assess drinking behaviour and subjective effects (Mehrabian & Russell, 1978). Weekly intake, rate of consumption, and a composite binge score index can be derived from this measure (Townshend & Duka, 2002). Although this provides some insight into total intake and pattern of consumption, it nonetheless has still been found to under-estimate total intake when compared to prospective daily assessments (Townshend & Duka, 2002). Furthermore, binge behaviour is frequently determined based on tertile or median splits of the sample (Townshend and Duka, 2005, Townshend et al., 2014). Results are therefore inextricably bound to the sample studied (Bø, Billieux, & Landrø, 2016).
Retrospective diary methods of collecting alcohol consumption information ask participants to recall day-to-day intake over a preceding designated time period (Del Boca & Darkes, 2003). Alcohol Timeline Followback (TLFB; Sobell & Sobel, 1992) is a commonly employed, well-validated example of this technique. Such diaries furnish researchers with considerable information pertaining to volume consumed and pattern of intake (Kypri, Langley, & Stephenson, 2005). In a review of more than thirty papers, however, Feunekes, van't Veer, van Staveren, and Kok (1999) found these diaries significantly under-estimate consumption levels – by about 20% – when compared to quantity-frequency surveys and prospective assessments. A study comparing repeated 7-day TLFB with 30-day TLFB revealed how more frequent assessments identified higher volume of intake, greater frequency of binge episodes and fewer abstinent days, with the absolute value of volumetric discrepancies between the two measures increasing as a function of length of recall (Hoeppner, Stout, Jackson, & Barnett, 2010). Longer assessment time periods yield less precise data (Ekholm, 2004); however, collecting data over narrow time frames distorts alcohol consumption information because drinking behaviour has been found to vary considerably from week to week (Del Boca et al., 2004, Goldman et al., 2011).
Real-time assessment of alcohol intake and pattern of consumption potentially overcome disadvantages associated with retrospective measures. Such assessment enables drinking to be recorded repeatedly, in close proximity to the event, often in the natural environment, and in the absence of the researcher (Trull & Ebner-Priemer, 2013). To date, studies comparing daily intake recorded via hand-held electronic devices and interactive voice response systems with that captured using retrospective methods have yielded varied results. In some cases, real-time assessments have revealed significantly higher consumption (Searles, Helzer, Rose, & Badger, 2002) or greater variability of intake (Carney, Tennen, Affleck, Del Boca, & Kranzler, 1998), while in others no significant differences have been identified (Bernhardt et al., 2009). Moreover, such devices pose a significant cost to researchers, potentially limiting the scope of the research (Kuntsche and Labhart, 2013, Trull and Ebner-Priemer, 2013). They additionally place considerable burden on participants, possibly diminishing compliance (Kuntsche and Labhart, 2013, Shiffman et al., 2008). With the advent of smartphones and application technologies (apps), real-time assessment limitations can be reduced. Researchers can take advantage of high smartphone ownership (Poushter, 2016), for instance, by using apps on participants' own devices to collect data. Apps enable considerable information about drink type, size, and ethanol content to be quickly logged, reducing the burden on participants in terms of time required to record information. Their capacity to compute rate of consumption and standard drink calculations reduce the potential for inaccurate reporting.
Few studies to date have validated app assessment of alcohol intake and pattern of consumption with retrospective measures. Monk, Heim, Qureshi, and Price (2015) found healthy participants (n = 51) recorded greater consumption when using an app for 7 days as compared to when such information was gathered retrospectively using researcher-generated beverage-specific surveys. Similarly, alcohol intake recorded daily via an app over a six-week period was shown to be significantly higher than when reported using TLFB; indeed, discrepancies between the measures increased over time (Dulin, Alvarado, Fitterling, & Gonzalez, 2017). Participants (n = 25) in this study, however, were diagnosed as alcohol dependent and were simultaneously undertaking treatment modules deployed by the app (Dulin et al., 2017). In both studies, sample size was relatively modest and retrospective data were collected after real-time recording, potentially enhancing participants' recall of drinking information and thus underestimating differences between real-time and retrospective data. This later point is a limitation often cited by researchers in this area (Carney et al., 1998, Perrine et al., 1995, Toll et al., 2006). Moreover, neither study examined differences in pattern of alcohol intake across measures.
The aim of this study was to examine differences between alcohol consumption information captured using an app for 21 days with data recorded via traditional, well-validated retrospective measures – namely, 21-day TLFB and the AUQ – in a large healthy sample. We chose a 21-day time frame in order to capture drinking variability; this appears to be particularly important when assessing binge patterns of alcohol intake (Courtney & Polich, 2009). We hypothesised that, relative to retrospective reports, indices related to alcohol intake and pattern of consumption would be greater when recorded via CNLab-A. Further, we expected app data to better accord with Australian Bureau of Statistics (ABS) apparent per capita alcohol consumption.
Section snippets
Participants
The present study consists of data from 671 participants (M age 23.12, SD = 7.24, range: 16–56, 70% female) that form a subset of an ongoing project – entitled CheckMyControl – investigating the relationship between alcohol use and various social and cognitive factors in the healthy population (see Fig. 1). Participants were recruited via adverts posted in and around the University of Melbourne, researcher networks, and social media posts. The University of Melbourne Human Ethics Committee
Sample demographics
At the time of testing, 39.9% of the sample was aged under 20, 46.2% was aged 20 to 29, and 13.9% was aged 30 or over. Participants had completed, on average, 14.80 years of education (SD = 2.31). Their mean AUDIT and AUQ binge scores were 7.97 (SD = 5.49) and 20.09 (SD = 19.74) respectively. Alcohol and other drug use characteristics are presented in Table 1.
App compliance and adherence
On average, participants used CNLab-A 20.27 (SD = 1.88) days out of 21. As data submission was either event- or notification-contingent, there was
Discussion
This study investigated differences between alcohol intake and pattern of consumption information collected via a smartphone app with data gathered using traditional, well-validated, retrospective measures in a large healthy sample. We hypothesised that, relative to both a 21-day TLFB and the AUQ, indices related to quantity of intake and pattern of consumption would be greater when recorded via the 21-day CNLab-A app. Results revealed participants recorded drinking on a significantly greater
Role of funding source
This research was supported by an Australian National Health and Medical Research Council grant (1050766), and an Australian Research Council fellowship (FT110100088). The funding bodies had no role in the study design, collection, analysis or interpretation of data, writing the manuscript, or in the decision to submit the paper for publication.
Contributors
AP and RH designed the study. JP and AP developed the smartphone application. LB and RS programmed electronic protocols and built the database. AP collected data and conducted the statistical analysis. Both AP and RH contributed to the writing and editing of this article. All authors have read and approved the final manuscript.
Conflict of interest
No conflict declared.
Acknowledgements
The authors wish to thank Jiajie Li and Jemie Effendy who assisted with upgrades to the smartphone application.
References (44)
- et al.
Measuring alcohol consumption: A comparison of graduated frequency, quantity frequency, and weekly recall diary methods in a general population survey
Addictive Behaviors
(2005) - et al.
How good is fine-grained timeline follow-back data? Comparing 30-day TLFB and repeated 7-day TLFB alcohol consumption reports on the person and daily level
Addictive Behaviors
(2010) - et al.
Current forms of inhibitory training produce no greater reduction in drinking than simple assessment: A preliminary study
Drug and Alcohol Dependence
(2017) - et al.
Correspondence between Interactive Voice Response (IVR) and Timeline Followback (TLFB) reports of drinking behavior
Addictive Behaviors
(2006) 4307.0.55.001 – Apparent consumption of alcohol, Australia, 2013–14
- et al.
AUDIT: The alcohol use disorders identification test: Guidelines for use in primary care
(2001) - et al.
Alcohol assessment among college students using wireless mobile technology
Journal of Studies on Alcohol and Drugs
(2009) - et al.
Binge drinking is characterized by decisions favoring positive and discounting negative consequences
Addiction Research & Theory
(2016) - et al.
Levels and patterns of alcohol consumption using timeline follow-back, daily diaries and real time “electronic interviews”
Journal of Studies on Alcohol
(1998) Statistical power analysis for the behavioral sciences
(1988)
Binge drinking in young adults: Data, definitions, and determinants
Psychological Bulletin
Methodological issues in measuring alcohol use
Alcohol Research and Health
The validity of self-reports of alcohol consumption: State of the science and challenges for research
Addiction
Up close and personal: Temporal variability in the drinking of individual college students during their first year
Journal of Consulting and Clinical Psychology
Comparisons of alcohol consumption by timeline follow back vs. smartphone-based daily interviews
Addiction Research & Theory
Influence of the recall period on self-reported alcohol intake
European Journal of Clinical Nutrition
Alcohol intake assessment: The sober facts
American Journal of Epidemiology
How many versus how much: 52 weeks of alcohol consumption in emerging adults
Psychology of Addictive Behaviors
Understanding standard drinks and drinking guidelines
Drug and Alcohol Review
Using personal cell phones for ecological momentary assessment: An overview of current developments
European Psychologist
Episode-centred analysis of drinking to intoxication in university students
Alcohol and Alcoholism
Ethanol consumption: How should we measure it? Achieving consilience between human and animal phenotypes
Addiction Biology
Cited by (26)
Momentary versus retrospective reports of alcohol or cannabis use, sexual activity, and their co-occurrence
2021, Addictive BehaviorsCitation Excerpt :Potentially explaining this, we asked participants to retrospectively report number of days rather than the more specific assessment (i.e., number of drinks per day) used in other studies (Monk et al., 2015; Patrick & Lee, 2010). An alternative explanation is that recall biases were systematically underestimated because participants were asked to retrospectively report their substance use behavior for the same period that they provided momentary data (Poulton, Pan, Bruns, Sinnott, & Hester, 2018); future work could address this concern by using the timeline follow-back procedure (TLFB; Sobell & Sobell, 1992). We also found that alcohol- and cannabis-involved sexual activity tended to be overreported on the retrospective survey.
Understanding the effectiveness of advertisements about the long-term harms of alcohol and low-risk drinking guidelines: A mediation analysis
2021, Social Science and MedicineCitation Excerpt :Another limitation is our reliance on self-report measures of alcohol consumption. In future studies, more accurate measurement of consumption using, for example, ecological momentary assessment methods, could provide greater sensitivity for detecting even small changes in consumption patterns (Poulton et al., 2018). In analyses testing for interactions by risky drinking status, we defined low-risk and high-risk drinkers using past week consumption instead of average past 12-month consumption.
Ubiquitous computing for person-environment research: Opportunities, considerations, and future directions
2021, Measuring and Modeling Persons and SituationsRetrospective and Real-Time Measures of the Quantity of E-cigarette Use: An Ecological Momentary Assessment Study
2023, Nicotine and Tobacco ResearchImpact of personalized alcohol intake and cognitive feedback on alcohol use behavior in hazardous drinkers: A quasi-randomized trial
2023, Alcohol: Clinical and Experimental Research