|Home | About | Journals | Submit | Contact Us | Français|
No psychometrically established measure of acute hangover symptoms is published and available to use in experimental investigations. The present investigation combined data across three studies of residual alcohol effects to establish the properties of a new Acute Hangover Scale (AHS) based on symptoms supported in previous lab studies.
Professional mariners from a Swedish maritime academy (n = 54) and young adult students/recent graduates of urban U.S. universities (n = 135) participated in one of three within-subjects' studies of residual effects of heavy drinking (M = .114 g% breath alcohol concentration [BrAC]). All drank placebo one evening and alcoholic drinks another evening followed by an 8-hour sleep period before completing the AHS 10-20 min after awakening.
The AHS showed excellent internal consistency reliability the morning after alcohol. The AHS mean score and each item were significantly affected by beverage but not demographics or typical drinking, supporting validity.
The AHS is a reliable and valid instrument for assessing acute hangover symptoms in experimental investigations of residual alcohol effects.
Residual alcohol effects are all effects that occur when BrAC is falling after an episode of heavy drinking while hangover refers to the set of subjective symptoms that peaks about when BrAC reaches 0 g%. Residual alcohol effects on behavior are of particular interest to the extent that they affect safety-sensitive occupational performance (Howland et al., in press) but the extent to which this impairment correlates with or results from hangover symptoms is not known, partly due to the lack of any established measure of acute hangover. In order to conduct experimental investigations of determinants and behavioral correlates of hangover, a reliable and valid measure of hangover symptoms is needed for use in acute situations. Early experimental investigations of hangover using individual signs and symptoms without scale development validated certain subjective symptoms whereas objective signs of hangover were not significantly increased (Chapman, 1970; Seppala et al., 1976; Ylikahri et al., 1974).
Survey instruments (e.g., Slutske et al., 2003) were not developed for use in acute situations and the items were not necessarily derived from experimental investigations. Some experiments have used items assessing withdrawal rather than hangover (e.g., McCaul et al, 1991; Span & Earlywine, 1999) or hangover symptoms plus cognitive reactions such as guilt (e.g., Span & Earlywine, 1999) or hangover symptoms plus a withdrawal sign invalid for hangover (Kruisselbrink et al., 2006) or gave no information about item selection or validity (e.g., Finnigan et al., 2005). The present study developed an Acute Hangover Scale (AHS) based on symptoms supported in experimental investigations of hangover, with psychometric determination of items to retain, reliability and validity. While preliminary results were reported on a smaller separate set of maritime academy cadets (Rohsenow et al., 2006), analyses in this larger sample with a broad range of age and experience will provide more stable estimates of psychometric properties. A related aim was to explore which specific items increased significantly as a function of residual effects of alcohol versus placebo since no other study published statistical analyses at the specific symptom level except a study using only five people and a much lower BrAC (.06 g%; Roehrs et al, 1991).
Data from three studies at two sites were combined. Site 1 included 54 professional mariner deck officers fluent in English at the Kalmar Maritime Academy in Sweden; Site 2 included 135 young adults in or recently graduated from universities around Boston, MA, (two studies). Volunteers reported no history of alcohol or drug problems, no exclusionary medical conditions or medications, at least occasionally drinking ≥ five alcoholic drinks in one day if male or ≥ four if female, and were not pregnant or nursing.
All three studies used a double-blind placebo-controlled two-group (alcohol vs. placebo) within-subjects design in which participants drank alcohol on one night and placebo on the other night (counterbalanced order) with order randomization stratified by gender. The beverages were given in a quantity estimated to yield 0.10 g% BrAC (1.2 g/kg body weight for men and 1.0 g/kg for women) or an equivalent amount of matched placebo. Alcohol beverages for Study 1 and Study 3 (n = 49) were high alcohol (7.2%) beer and for Study 2 were 100 proof vodka (n = 38) or 101 proof bourbon (n = 48, by random assignment) mixed with chilled caffeine-free cola. Matched placebos were nonalcoholic beer or caffeine-free cola plus de-carbonated tonic water in amount equivalent to the alcohol with a few drops of vodka or bourbon floated. If participants randomized to alcohol did not reach the target BrAC 15 min after the last beverage, an additional amount of alcoholic beverage was given based on the ratio of obtained/target BrAC, with extra placebo provided to the same number of placebo participants.
Groups of four to five participated 1-7 days apart. They were told not to have alcohol within the prior 24 hours or food or beverage within the prior 3 (Site 1) or 4 hours (Site 2) and were rescheduled if noncompliant. Beverage administration started at 7:30 p.m. (Site 1) or 8:45 p.m. (Site 2) and ended 60-90 min later. Participants were allowed 8 hours for sleeping and an hour for washing and breakfast the next morning before performance assessments. They completed the hangover measure 10-20 min after being awakened because hangover is most detectable before 10 a.m. (Ylikahri et al., 1974).
Recent drinking was estimated using two items for the past 30 days: about how often they had any beer, wine or liquor (rated from 1 “once a day” to 7 “did not drink” with each point anchored) and about how much they to drink on a typical day that they drank (one drink defined as 12 oz of beer or wine cooler, 4 oz of wine or 1 oz of liquor), with eight choices: 1 to 7 drinks and “8 or more drinks”.
The nine AHS items included all the validated items from Chapman (1970; also supported by Ylikahri et al., 1974) except trouble sleeping (not an experience “right now”) and general malaise (not a term used in lay language but replaced by “hangover”). The answer format used the 0-7 scale of Chapman (1970) with Roehrs et al.'s (1991) four anchors: None (0), Mild (1), Moderate (4) and Incapacitating (7). The general instruction was “Please rate how you feel right now on the following rating scales”.
First, Cronbach's alpha was used for reliability and to determine whether removal of any item would improve the reliability. Second, validity was investigated by comparing the AHS total score and items between the two beverage conditions using paired t-tests. Effect sizes are reported in terms of d (a medium effect is d = .50 to .79; a large effect is ≥ .80). Third, discriminant validity tested predicted non-relationship of AHS score to age, race, gender, professional status, and quantity and frequency of typical drinking (since only acute drinking should determine hangover). (Correlations with obtained BrAC were not considered a valid test due to highly restricted range.)
Participant characteristics are displayed in Table 1 by population type. The professional mariners were significantly more likely to be male (χ2 = 48.46, df = 1, p < .0001), were older (t (57.3) = 9.12, p < .0001, df adjusted for unequal variances), more were White (χ2 = 12.60, df = 1, p < .001), and they drank more frequently (χ2 = 11.65, df = 5, p < .04).
The AHS had a standardized Cronbach's α = .84 after alcohol administration. The best item-total correlations were for “hangover”, “dizziness or faintness” and “nausea” (rs = .66-.67) and the lowest was for “thirsty” (r = .39). No item's removal increased reliability. Excluding the item “hangover” (per Slutske et al., 2003) lowered reliability to α = .81 so it was retained. Therefore, a mean score of all AHS items was computed.
The paired t-tests for AHS mean score and each item were significant (Table 2). Beverage condition had a very large effect size for AHS mean score. Among the individual items, the “hangover” item had the largest statistical effect size, two other items (“thirsty” and “headache”) had large statistical effect sizes, two (“tired” and “dizziness or faintness”) had medium effect sizes, and four symptoms had effect sizes just below the medium range. Only “thirsty” and “tired” approached moderate severity with the rest in the mild range.
Discriminant validity was supported by nonsignificant relationships of the AHS with age (r = -.18), race (white vs. non-white; t(186) = 1.85), gender (t(186) = 0.64), professional mariners vs. university participants, covarying age since groups differed in age (F (2,185) = 1.80), mean drinking frequency (r = .30), or drinking quantity on a typical day (r = -.07), all ps > .20.
The AHS, developed for use in assessing hangover symptoms acutely in laboratory investigations of hangover, was found to be highly reliable and valid, consistent with the smaller analysis previously conducted on a separate limited set of participants (Rohsenow et al., 2006). Only symptoms previously empirically supported in laboratory work on hangover were included. The AHS total (mean) score is valid as it was significantly affected by previous evening's consumption of alcohol versus placebo yet was not affected by demographic variables or recent drinking history. Even though the amount of alcohol administered was probably much lower than many people drink when they report hangovers, even this relatively mild hangover was detectable, consistent with BrAC levels producing increased hangover symptoms in studies by others (e.g., Chapman, 1970; Kruisselbrink et al., 2006).
The contribution of specific items was also of interest for understanding the nature of symptoms most relevant to hangover. Probably because AHS items were selected based on previous laboratory work, all items were ones that were significantly affected by alcohol. The items “Thirsty” and “Tired” had the highest values, the only items scored in the “Moderate” range the morning after drinking alcohol. However, these items also were highest the morning after placebo administration so inspecting values only after alcohol may be misleading; people may be tired and thirsty at that time in the morning anyway. The largest effect sizes as a function of alcohol ingestion were for “hangover”, “thirsty”, and “headache”, in that order. The other items all had medium or just below medium statistical effect sizes. Controversy exists over whether the item “hangover” should be included (Slutske et al., 2003), because it includes a direct attribution to residual alcohol effects. It was selected in part to best reflect the item “malaise” (from Chapman, 1970) in plainer language and partly to investigate the degree to which people perceived themselves as hung over after this amount of alcohol. This item had the largest effect size as a function of alcohol administration and the highest item-total correlation with all other items, so it may be worth retaining. While removing this item would lower the reliability of the scale somewhat, the reliability is excellent without it, so investigators have the option of scoring this instrument either way.
Limitations of this study included alcohol administration only to about .12 g%, and limited populations. Future studies can increase the diversity of the participants and investigate the effects of higher BrACs. The primary strength of this study was that it studied hangover prospectively in a double-blind controlled design and therefore avoided biases that occur in retrospective studies (e.g., recall problems, self-serving attributions) or cross-sectional designs (e.g., self-selection bias), and that it started with symptoms that had already shown some empirical foundation. Future work can use this instrument in laboratory studies to investigate a number of theoretical questions about variables affecting intensity of hangover.
This study was supported by the Harvard Injury Control Research Center with funding from the Centers for Disease Control and Prevention (R49 CCR115279) and from the National Institute on Alcohol Abuse and Alcoholism (RO1 AA 12087) to the Boston University School of Public Health, and by the General Clinical Research Center with funding from the National Center for Research Resources (M01 RR00533) to Boston University Medical Center. Grateful appreciation is expressed to the staff and cadets of Kalmar Maritime Academy, Kalmar, Sweden, and Boston University for their assistance, to the staff of the General Clinical Research Center for their assistance, and to Donald Davies for his data analyses.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Damaris J. Rohsenow, Brown University Center for Alcohol and Addiction Studies.
Jonathan Howland, Youth Alcohol Prevention Center, Boston University School of Public Health.
Sara J. Minsky, Youth Alcohol Prevention Center, Boston University School of Public Health.
Jacey Greece, Youth Alcohol Prevention Center, Boston University School of Public Health.
Alissa Almeida, Youth Alcohol Prevention Center, Boston University School of Public Health.
Timothy A. Roehrs, Department of Psychiatry and Behavioral Neuroscience, Wayne State University.