A Systematic Approach to Standardizing Drinking Outcomes From Timeline Followback Data

Abstract

Objective:

The timeline followback (TLFB) interview is the gold standard for the quantitative assessment of alcohol use. However, self-reported “drinks” can vary in alcohol content. If this variability is not accounted for, it can compromise the reliability and validity of TLFB data. To improve the precision of the TLFB data, we developed a detailed standard operating procedure (SOP) to calculate standard drinks more accurately from participant reports.

Method:

For the new SOP, the volume and alcohol content by volume (ABV) of distinct types of alcoholic beverages were determined based on product websites and other reliable sources. Recipes for specific cocktails were constructed based on recipes from bartending education websites. One standard drink was defined as 0.6 oz (14 g) of absolute alcohol. Standard drink totals were contrasted for the new SOP approach and the standard procedure, which generally assumed that one self-reported drink was equivalent to one standard drink.

Results:

Relative to the standard TLFB procedure, higher numbers of standard drinks were reported after implementing the TLFB SOP.

Conclusions:

Variability in procedures for conversion of self-reported alcohol consumption to standard drinks can confound the interpretation of TLFB data. The use and reporting of a detailed SOP can significantly reduce the potential for such inconsistencies. Detailed and consistent procedures for calculation of standard drinks can enhance the quality of TLFB drinking data.

Introduction

The timeline followback (TLFB) method ¹ is considered the gold standard measure of substance use. ² In the alcohol use disorder research setting, alcohol use is often quantified using the TLFB. ¹ TLFB is an assessment in which clinicians or study personnel provide patients with calendar-based memory cues to help them construct chronological reports of their recent alcohol (and/or other substance) use. Data obtained from these interviews can be used to quantify various aspects of drinking behavior, such as drinking per day percentage of abstinent days, percentage of heavy drinking days, and drinks per drinking day.

Because alcoholic beverages contain highly variable amounts of pure alcohol, each participant’s self-reported alcohol consumption (eg, “two margaritas”) must be converted to “standard drinks,” to allow meaningful evaluation of alcohol consumption. In North America, the pure alcohol content of a standard drink is defined as 14 g (0.6 fluid ounces) of ethanol. By this measure, using commonly referenced volumes and alcohol content (ie, percent alcohol by volume, ABV), multiplying ABV times volume shows that each of the following represent 0.6 oz of ethanol or one “standard drink”: a single 5-oz glass of wine (assuming 12% alcohol by volume (ABV), a 12 oz. bottle of beer (assuming 5% ABV), or a single mixed drink (assuming 1.5 oz of liquor, at 40% ABV). ³ It is important to note that we are using the standard ounces and ABV, but want to acknowledge that there are many different proofs, ABVs and sizes of alcoholic beverages. This SOP is not all encompassing but rather a list of the most common drinks, ABVs and sizes of drinks. The research team relied on the participants to recall the ounces and ABV, and if no information was provided would go with the standard or default values in the tables below.

These conversions rely on a number of assumptions, which present opportunities for inaccuracies and inconsistencies if additional information is not obtained. For example, if a participant reports drinking “a beer,” then without further information, this might be interpreted as one standard drink, as noted above. However, a 16-oz triple India Pale Ale may contain up to 1.8 oz of ethanol—that is, 3 standard drinks. Similarly, high variability in the content and volume of mixed drinks introduces the potential for inaccuracies and poor interrater reliability when making the conversion to standard drinks. Study staff who are conducting the TLFB assessments must be trained to obtain as much information as possible from the participant about the type and volume of beverage, and staff responsible for converting that information to standard drinks must receive clear and detailed instructions for calculating the alcohol content of several types of alcoholic beverages.

Our research team at NYU became aware of the need to refine the procedures for calculating standard drinks while resolving some significant discrepancies in independent double entry of TLFB data in a recently completed clinical trial for alcohol use disorder (Bogenschutz et al ⁴ ; NCT#02061293). In the process of quality assurance monitoring, we developed a comprehensive Standard Operating Procedure (SOP) incorporating the best information available to estimate the alcohol content of all types of alcoholic beverages. In this brief report, we outline the detailed SOP, explain how it is used, and provide examples of how it affects the calculation of drinking outcomes.

Method

We first compiled a list of alcoholic beverages that had been documented on the TLFB in our recently completed trial. Since our participants reported a wide range of volumes, frequently using colloquialisms (eg, a “nip” of brandy), a comprehensive list of drink sizes was also generated, for later calculation purposes. The beverages were entered into an Excel spreadsheet that separated them by beverage type (ie, beer, wine/champagne, liquor, cocktails/mixed drinks). Our priority was to generate standard values for the ABV of each of the beverages our participants reported.

ABV

For all drinks, the SOP took beverage types/brands into consideration when such information was available. When specific brands of beer, wine or liquors were reported, we use the published ABV of that brand. In the absence of specific information about brand of beer, we used 5% ABV that has been used historically. ¹ For wine, although 12% ABV has historically been used ¹ our review of popular brands indicated that using a slightly higher ABV was more accurate. According to Martinez et al ⁵ ; the average ABV last reported in 2016 for wine was 12.27%. We decided to use this estimate for wine and 10.22% for sparkling wine and prosecco. ⁵ If participants self-reported liquor use without naming the brand, we assumed 40% ABV.

Volume

The volumes illustrated in the SOP Appendix were used unless the participant specifically reported a specific drink size or number of ounces in the drink. For example, “1” beer was assumed to be 12 oz if no other information was available. Similarly, a standard glass of wine is defined as 5 oz and a standard single drink of hard liquor is defined as 1.5 oz per standard timeline followback convention. ¹ However, those collecting TLFB data should be instructed to inquire about the volume of drinks rather than assuming the standard volumes. If a different volume was noted the calculation was modified accordingly. It is particularly important if participants are pouring their own drinks at home.

Cocktails

To determine the volume and ABV of each cocktail, we used standard drink recipes reported by the International Bartender’s Association (IBA) ⁶ . We then calculated the total alcohol content by multiplying ABV × volume for each ingredient containing alcohol, and summing these totals for each ingredient to yield the total alcohol content. If drink recipes were not found on the IBA site, other sources were used to identify the recipes needed in order to calculate the total alcohol content (see Supplemental Information for details of additional sources and procedures used). Next, the alcohol content of all the drinks consumed on a given day were added together to yield the total standard drinks for that day.

Results

The creation and adoption of the SOP resulted in meaningful differences in standard drinks compared to those calculated using the standard approach. See Table 1 in the appendices for a demonstration of the differences in the 3 different assessments in estimated alcohol consumption between the existing standard TLFB approach, and that of the detailed SOP (second approach). For example, for one participant who self-reported drinking 3 beers, without using the SOP this would be classified as 3 drinks. Once the ABV and volume were specified, this changed to three 16-oz double IPAs at 7.9% ABV. Using the SOP, the calculation went from 3.0 standard drinks to 6.32 standard drinks. For a participant who self-reported 4 glasses of wine, 4 standard drinks would be calculated using the standard procedure. Using the type of wine specified by the participant (4 glasses of Bogle Cabernet Sauvignon at 14.5% ABV), the value increased to 4.83 standard drinks. Lastly, for self-reported cocktail consumption, 2 Hemingway Specials would have been classified as 2 standard drinks using a standard procedure. In the SOP approach, we modified this quantity to 4.8 standard drinks (based on the IBV recipe).

Table 1.

Comparison of pre-/post-SOP mean drinks per day using simulated TLFB data.

Participant	“What they report”	Approach	Additional interviewer questions	What is written on the TLFB calendar	“Standard Drinks” calculation
A	3 beers	Standard approach	N/A (variable)	Three beers	3.00 drinks
		SOP approach	Brand ABV volume	Three 16 oz., double IPA @ 7.9%	6.32 drinks
B	4 glasses of wine	Standard approach	N/A	Four glasses of wine	4.00 drinks
		SOP approach	Brand ABV volume	Four 12 oz. glasses of Bogle Cabernet Sauvignon @ 14.5%	4.83 drinks
C	2 hemingway specials	Standard approach	Brand ABV volume	Two drinks	2.00 drinks
		SOP approach	Brand ABV volume	Two 3.75 oz. @ 38.4%	4.80 drinks

These are examples of the difference in calculating the number of standard drinks between the standard procedure and the SOP procedure.

Discussion

In this brief report we presented an SOP that was developed to ensure reliable and valid conversion of participants’ self-reported alcohol intake to standard drinks. What sets this SOP apart from previous standardized approaches (eg, Fiellin et al ⁷ ) is its comprehensiveness, both in terms of the variety of alcoholic beverages included and in how the accuracy of alcohol content was determined. In several examples displayed in Table 1, new SOP resulted in a higher number of standard drinks than the prior TLFB method of data collection and conversion. This could have a significant impact on study outcomes, both in terms of the primary outcome (percent heavy drinking days) and secondary measures of alcohol use (eg, the number of drinks per day). Increasing the reliability and validity of measurement improves precision and may impact statistical power.

In addition to psychological outcomes, the SOP may also be useful in clinical trials that measure various physiological outcomes such as hormonal changes related to drinking or liver function. Therefore, the importance of this SOP is emphasized by its wide range of application in alcohol clinical studies that assess variable impacts of alcohol on both the body and mind. Table 2 displays the standard drinks in common cocktails. Using the SOP guidelines, the list can continue to expand with other cocktails.

Table 2

Standard drinks in common cocktails.

Name	Standard drinks
Single drink (if no other information provided)	1.00
Double drink (if no other information provided)	2.00
Aperol Spritz	0.98
Bloody Mary	1.00
Daiquiri	1.33
Hemingway Special	2.40
Dark and Stormy	1.33
Long Island Iced Tea	1.69
Mai Tai	1.70
Manhattan	1.40
Margarita	1.40
Martini (Gin or Vodka)—Single	1.40
Martini (Gin or Vodka)—Double	2.80
Mimosa	0.51
Mojito	1.00
Espresso Martini	1.50
Piña Colada	1.10

One significant limitation to implementing the SOP is the increased amount of time required for data acquisition and conversion, and the importance of adequate and consistent training of staff administering the TLFB. Refer to Supplemental Table 2(a) and (b) for an illustration of uniform data collection. Sufficient time must be allowed to administer the TLFB to gather as much detail as possible regarding the specific types and amount of alcohol consumed. One study found significant variability (between 15 and 135 minutes) in the amount of time taken by different staff members when administering the TLFB. ⁸ Consistent training and monitoring are important to reduce such variability. The conversion from individual beverages to the number of standard drinks also is more time consuming when using the newly developed SOP. Due to these intricacies of its administration, the feasibility of implementing the TLFB in clinical settings may be challenging as clinicians would need to have both training to learn the TLFB and adequate time to administer it during appointments. Implementation of the TLFB may be easier in research settings when there is more time to sufficiently train staff and spend more time with each individual patient. However, it is also possible that a detailed and accurate TLFB interview could have clinical value, influencing an individual’s drinking by increasing awareness of alcohol consumption. The adaptability of our updated TLFB SOP into both research and clinical settings may be the subject of future research.

Conclusions

We developed a SOP to increase the accuracy of data acquisition using the TLFB and facilitate systematic conversion of patient-reported “drinks” to the units of alcohol represented by standard drinks. Future alcohol use disorder studies can utilize a detailed SOP, such as the one presented in this paper, to improve the validity of and interrater reliability for TLFB-based quantification of participants’ drinking behaviors. Beyond clinical research, knowledge of the methods involved might also allow clinicians to measure a patient’s current alcohol use more accurately, improving identification of patients engaged in, or at risk of, drinking behaviors with negative health consequences.

Supplemental Material

sj-docx-1-sat-10.1177_11782218231157558 – Supplemental material for A Systematic Approach to Standardizing Drinking Outcomes From Timeline Followback Data

Supplemental material, sj-docx-1-sat-10.1177_11782218231157558 for A Systematic Approach to Standardizing Drinking Outcomes From Timeline Followback Data by Christina Marini, Nicole S Northover, Noah D Gold, Ursula K Rogers, Kelley C O’Donnell, Babak Tofighi, Stephen Ross and Michael P Bogenschutz in Substance Abuse: Research and Treatment