Abstract
The aim of this study was to bring to light the meanings attributed to four concepts denoting problematic drinking—“alcoholism,” “alcohol dependence,” “alcohol misuse,” and “risky drinking”—and to investigate whether individuals’ preferences for using these concepts vary depending on their demographic characteristics, their stakeholder group membership, and their own drinking habits. These questions were examined by means of a vignette experiment, in which Swedish social workers, health care staff, journalists, researchers, and the general public (n = 863) read a series of 10 randomly constructed fictitious cases (vignettes; n = 8,630) and made judgments about which of the four concepts best portrayed the case descriptions. Results based on multilevel multinomial logistic regression showed that alcoholism was the concept chosen for the “heaviest” and most problematic drinking, whereas risky drinking was most often employed to describe moderate drinking habits that resulted in few negative consequences. The concepts alcohol misuse and alcohol dependence were positioned between alcoholism and risky drinking, and while misuse was linked to drinking large quantities and to more negative consequences, dependence was more frequently used to describe vignettes that involved depictions of craving. Moreover, the respondents’ conceptual choices were affected by their own gender, age, drinking habits, and stakeholder group membership. Taken together, the study’s findings reveal both consensus and variation in the respondents’ choice of concepts, indicating that while these concepts may to some extent be employed for making meaningful distinctions between different types of problematic drinking, there is substantial variation in the way the concepts are employed.
Introduction
Since the turn of the millennium, the care and treatment of people diagnosed with alcohol problems has come to attract renewed interest among Swedish politicians and central decision makers. Thus, as an effect of the successive abandonment of the traditional Swedish alcohol policy, focused on limiting availability, which followed in the wake of the country’s entry into the EU, a number of new initiatives have emerged, primarily aimed at rendering individual treatment and secondary prevention more effective. This process has involved the publication of a number of inquiries, directives, and guidelines. These include national guidelines for the care and treatment of misuse and dependence (National Board of Health and Welfare [NBHW], 2007, 2017), the implementation of these guidelines (https://skl.se; Swedish Association of Local Authorities and Regions, 2018), the Substance Misuse Inquiry (Government Official Report [GOR], 2011, No. 35), and a national strategy for policies and provisions in the areas of alcohol, narcotics, doping, and tobacco (ANDT; Skr. 2015/16:86, https://data.riksdagen.se/). By and large, these documents focus on desirable goals and on the alleged or proven effectiveness (“evidence”) associated with various interventions (e.g., NBHW, 2007, 2017). Less is said about the character and definitions of the problems that the proposed measures are intended to resolve. The fact that there are currently major disagreements among researchers when it comes to defining, understanding, and explaining problematic or deviant drinking (cf. Fraser, Moore, & Keane, 2014; Heim, 2014) is barely mentioned. 1 The absence (in the above-cited reports and guidelines) of a comprehensive research-based discussion about the nature of problematic drinking is quite remarkable given that the ways of managing and solving a specific problem tend to be contingent upon the definition and explanation of the problem in question (e.g., Gusfield, 1981; Rochefort & Cobb, 1994).
In point of fact, the lack of unanimity in the field of alcohol research is neither new nor unknown but has a long history (cf. Blomqvist & Wallander, 2017). In 1969, the alleged “chaos” in researchers’ endeavors to define and explain problematic drinking was concisely summarized by the Norwegian criminologist Nils Christie and the Finnish social researcher Kettil Bruun in their contribution to the 28th International Congress on Alcohol and Alcoholism in the United States: From the outset, the conceptual framework seems to be all chaos. The words employed are as many as pebbles on the beach: alcohol, alcoholism, drug, drug dependence, excessive drinkers, symptomatic drinkers, addiction, habituation, narcotics, use, abuse, chemicals, stuff, mind-altering substances, sickness, sin, crime, treatment, punishment, help, and so on. When one moves from one article or book to the next and puts the key concepts together, the end result bears a considerable resemblance to a psychedelic picture. (Christie & Bruun, 1969, p. 65)
Needless to say, the lack of a science-based consensus about the definition or nature of the phenomena that we frequently refer to as “alcohol addiction,” “substance use,” “alcoholism,” “alcohol dependence,” “alcohol misuse,” or “risky drinking” produces real-world challenges at a number of levels. For example, policy makers working with the central planning of alcohol policy measures need clear definitions of problematic drinking in order to accurately estimate the number of alcohol consumers potentially in need of society’s help (Amundsen, 2010). Furthermore, policy makers’ choices of definitions and explanations (from among the many available) are crucial, since these choices will determine who is given responsibility for the care and treatment of alcohol problems, and will indirectly also affect the actual content of treatment provision (cf. GOR, 2011, No. 35). In the same vein, at the clinical level, whether a caregiver perceives “alcohol dependence” as a “chronic brain disease” or as a largely transient “way of life leading to predicaments” (Drew, 1986; Fingarette, 1988), will result in very different help recommendations for a given problem drinker (cf. Wallander & Blomqvist, 2005, 2008). Finally, since the dominant ways of representing problematic drinking in research, official reports and—in particular—the media are likely to inform popular perceptions of alcohol problems and the ways in which problematic users define themselves (e.g., Hacking, 1991), fuzzy or one-sided conceptions may restrict individuals in their efforts to find a suitable way out of problematic drinking habits.
Against this background, the aim of this study is to bring to light the meanings (cf. the connotations, the definitions) attributed to a number of the most frequently employed concepts denoting problematic drinking (alcoholism, alcohol dependence, alcohol misuse, and risky drinking; in Swedish: alkoholism, alkoholberoende, alkoholmissbruk och riskdrickande). 2 More precisely, our intention is to investigate (a) which meanings are attributed to the four concepts, (b) whether the concepts are used interchangeably (as “fat words”) or whether they are attributed distinct meanings (being “useful concepts”), and (c) whether the respondents’ preferences for using these concepts vary depending on who they are (demographic characteristics and stakeholder group membership) and on their own drinking habits. These questions are examined by means of a vignette experiment, a factorial survey (Rossi & Nock, 1982), in which each respondent reads a series of fictitious cases (vignettes) and then makes judgments about whether the case description would be best portrayed as alcoholism, alcohol dependence, alcohol misuse, or risky drinking (or none of these). By randomly varying the contents of the vignettes, we are able to investigate whether (and if so how) the four concepts distinguish between cases involving different people (in terms of demographic and socioeconomic factors), variations in drinking patterns, and a number of negative consequences of the drinking.
The study’s respondents are drawn from a number of groups that may be regarded as stakeholders in the field and are therefore of special interest in trying to establish the “Swedish answer” to the question of how problematic drinking is actually defined. These include professionals working with problematic substance use (1) in the municipal social services (social workers) and (2) in local primary care (health care staff, i.e., medical doctors and nurses); (3) journalists specializing in lifestyle, health, social problems, and similar subjects; (4) researchers in the field of substance use; and (5) the general public. The reasons for choosing these respondent groups are that they either generate or explore perceptions, explanations, and/or “evidence” about problematic drinking (the researchers), apply such understandings in their professional work, and represent the “professional stance” in this field (the social workers and the health-care staff), 3 disseminate information about alcohol via the media (the journalists), or generally employ perceptions and explanations of problematic drinking to communicate with others and to understand themselves (the general public). By bringing to light the meanings ascribed by these stakeholders to central concepts linked to problematic drinking, this study aims to pave the way for intellectual clarity in the field and to improve the quality of communication between actors.
The article is structured as follows: In the next section (Background), we briefly describe the four concepts that are investigated in the study and summarize some of the conclusions from previous studies about conceptions of problematic drinking. The subsequent section (Methods) describes the research design, sampling of respondents, and data collection. In the Results section, we describe our analytic strategy and present results from a number of multilevel multinomial logistic regression analyses based on 8,410 vignette judgments made by 856 respondents. We close with a discussion of the main conclusions of the study.
Background
This study aims to explore the meanings attributed by stakeholders to the four concepts most frequently employed in Sweden to define problematic drinking: alcoholism, alcohol dependence, alcohol misuse, and risky drinking (in Swedish: alkoholism, alkoholberoende, alkoholmissbruk, riskdrickande). Since these concepts were selected due to their widespread use as descriptors of problematic drinking in the public debate and in professional domains, they do not include “addiction,” for which there is no linguistic equivalent in Swedish, nor the recently introduced concept “substance use disorder” (Diagnostic and Statistical Manual of Mental Disorders ([DSM]; https://www.psychiatry.org, American Psychiatric Association, 2018), which has as yet not become established in the professional discourses nor obtained public recognition. 4 In the section below, we briefly outline the historical roots and contemporary usage of each of the four concepts (primarily focusing on the Swedish context), as well as critically discussing their most important formal definitions (i.e., academic/professional/official definitions). A couple of these definitions involve the diagnoses outlined in the DSM (https://www.psychiatry.org, American Psychiatric Association, 2018) and the Classification of Mental and Behavioural Disorders (International Classification of Diseases; ICD; http://www.who.int, World Health Organization [WHO], 2018a). It is worth noting that in Sweden, these diagnostic manuals do not constitute part of the knowledge of the general public but instead function as tools for certain professions, including primarily health-care staff (ICD) and researchers and psychiatrists (DSM). However, as the definitions outlined in these manuals have been singled out by the Swedish NBHW, and incorporated into the national guidelines for the care and treatment of misuse and dependence (NBHW, 2007, 2017), they cannot be excluded from the discussion. Finally, we will present a brief overview of the results from some previous studies on public and professional notions, images, and understandings of problematic drinking.
Alcoholism
The oldest and most long-standing concept associated with problematic drinking is indisputably “alcoholism.” Originally, the term alcoholism was coined in the influential treatise Alcoholismus Chronicus (1849–1851) by the Swedish doctor Magnus Huss, who described the detrimental physiological, psychological, and social consequences of excessive and prolonged alcohol consumption. Together with partly similar, earlier formulations by Rush (1785/1943) in the USA and Trotter (1804/1988) in England, Huss’s work served to constitute the first “disease theory of alcoholism.” The common denominator among these early formulations of the disease model was that excessive alcohol consumption ought not to be seen as a moral deficiency or as a sign of a wicked society, but as a malfunctioning of individuals. The subsequent version of the disease theory, which received popular acceptance in the USA in the 1930s (e.g., Beauchamp, 1980) and was effectively disseminated by Marty Mann and the so-called Alcoholism Movement (see Roizen, 2004) and later by the AA network, viewed chronic drunkenness not as an inescapable consequence of ingesting alcohol but rather as a characteristic of certain people who were disposed to being unable to handle alcohol in moderation.
In 1952, the WHO issued an official definition of alcoholism that rapidly became the international standard. According to this definition, “Alcoholics are those excessive consumers whose dependence on alcohol has attained such a degree that it shows a noticeable mental disturbance or an interference with their bodily and mental health, their interpersonal relations and their smooth economic functioning, or who show the prodromal signs of such a development. They therefore need treatment” (WHO, 1952, p. 16). As has been noted by Cameron (1995), this definition leaves most of its key terms unexplained (e.g., “dependence,” “mental disturbance,” “smooth economic functioning”) and therefore places the definitional power in the hands of whoever uses it.
Despite the fact that disease theories associated with the concept of alcoholism have lived on until fairly recently as by far the most common way of defining and explaining problematic drinking in the Western world, they have never achieved the same levels of acceptance in Sweden. For example, in the early decades of the 20th century, the liberal physician and social commentator Ivan Bratt (1909) was influential in his attempts to describe excessive drinking in terms of socially and culturally determined habits, 5 and in the 1940s, the Care of Alcoholics Committee stated that problematic drinking was the result of “a complex interaction between congenital and acquired personal characteristics, possible states of disease, and environmental conditions” (GOR, 1948, No. 27, p. 40). Even when taking into account a number of waves of medicalization during the 20th century, Sweden has—at least until fairly recently—been characterized by a nonmedical approach to problematic drinking. Notwithstanding this, and despite the fact that a definition of alcoholism is not included in the currently dominant diagnostic systems (DSM-V; ICD-10), the concept of alcoholism is still in use, primarily in the rhetoric of the AA (Alcoholics Anonymous World Services, Inc, 2001; available in Swedish), but occasionally also by researchers and practitioners in the medical sphere (e.g., Högberg, Pålsson, Wells, Larsson, & Spak, 2017; Lundin, Mortensen, Halldin, & Theobald, 2015).
Alcohol Dependence
The term “dependence” was introduced around 40 years ago, when Edwards (Edwards, 1977; Edwards & Gross, 1976) presented the concept “alcohol dependence syndrome.” This concept was created in an attempt to reconcile the languishing alcoholism concept with new empirical findings about the heterogeneity and changeability of alcohol problems in the population (Bergmark & Oscarsson, 1987). It is true that the concept of dependence, which rather quickly gained wide acceptance in the field, increased precision in that it required the fulfillment of three of seven (DSM) or three of six (ICD) specific criteria during a 12-month period. However, given that these definitions have since been altered recurrently and that the most up-to-date definitions bear little resemblance to the original syndrome (cf. Room, 1998), their alleged exactness appears less impressive. According to the DSM-IV (the most recent DSM manual covering this concept; see Note 4), substance dependence presupposes the simultaneous fulfillment of three of the following criteria (in short): (1) tolerance, (2) withdrawal symptoms, (3) drinking more/longer than intended, (4) having unsuccessfully tried to cut down/stop, (5) having spent a lot of time drinking/being sick/aftereffects, (6) having given up important/interesting/pleasurable activities in order to drink, and (7) having continued drinking despite depression/other health problems. 6 As has been noted by Dawson, Grant, and Hartford (1995), there are 99 different ways of combining three or more of these seven criteria, a fact that might be interpreted to mean that there are 99 “types” of dependences according to DSM-IV.
In Sweden, the dependence concept constitutes an established part of the health professions’ (e.g., medical doctors and nurses) vocabularies and also manifests itself in the term used to designate the part of the health-care system that deals with problematic drinking and drug use (cf. “beroendevården”; in English “dependency care”). A recent complication with the dependence concept, however, is that it has undergone a “conventionalization” in everyday and media parlance (Hellman, 2010), and its use has expanded to denote all and sundry kinds of strong interests and preferences (Blomqvist, 2012a). This may at least in part be linked to the emergence of a new “treatment market,” in which new entrepreneurs promote their cures for “dependencies” on sex, love, shopping, exercise, internet use, and so on (Blomqvist, 2012a).
Alcohol Misuse
In the contemporary Swedish legislation, in the traditions of the Swedish social work profession, which bears the primary responsibility for dealing with problematic drinking at the individual level, and in the public discourse, “alcohol misuse” has long been the most well-established concept denoting problematic drinking. Here, we may note that while we translate the Swedish term “missbruk” as “misuse,” it is also very common (not least in official contexts) to use the English term “abuse.” While we have decided to primarily use the English term that is considered the least judgmental, and which is increasingly employed by scholars and practitioners, both misuse and abuse constitute correct translations of the Swedish term “missbruk.” Accordingly, we will refer to “abuse” whenever the term is explicitly used in the context of legislation, diagnostic systems, and so on (see below).
The Swedish term “missbruk” (cf. misuse/abuse) has a central position in the Social Services Act (2001/453http://www.riksdagen.se) and is one of the legal criteria for subjecting individuals to coercive care in accordance with the Care of Abusers (Special Provisions) Act (1988/870http://www.riksdagen.se). Interestingly, neither of these two pieces of legislation is founded on a clear definition of the concept. At the basic level, because misuse (of something) only signifies a deviation from an allegedly normal and accepted use (of something), the definitional power ultimately lies with the person who uses the concept to make a diagnosis. In the dominant diagnostic manuals, abuse has until recently been defined as a “milder” form of dependence. Thus, in DSM-IV (see Note 4), abuse is defined as fulfilling at least one of the following four criteria in the course of a period of 12 months (in short): (1) having problems with work/family/school due to drinking, (2) getting into potentially harmful situations due to drinking, (3) having legal problems due to drinking, and (4) continuing to drink despite it causing trouble with family and friends. A similar if not identical definition is employed by the WHO (although it is not included as a diagnostic term in the ICD-10) under the heading “hazardous use” (WHO, 2018b).
In Sweden, the concept and understanding of alcohol misuse has fairly recently been debated at the policy level. In the first official guidelines for care and treatment in this area, which were introduced in 2007, the NBHW advised against using the concept “missbrukare,” not as a result of its imprecision but because it might potentially be perceived as pejorative and stigmatizing. Moreover, the same document (NBHW, 2007) stated that it could not be determined—on the basis of the available knowledge—whether or not misuse should be regarded as a disease. Only 4 years later, however, in a report produced by the Substance Misuse Inquiry (GOR, 2011, No. 35), it was concluded that both alcohol misuse and alcohol dependence are to be considered as “states of disease” (GOR, 2011, No. 35, p. 24). This latter conclusion is very interesting, particularly when viewed in light of the fact that the dimensions typically associated with alcohol misuse primarily include various negative social consequences of drinking (see, e.g., the DSM criteria above).
Risky Drinking
Another concept that has become increasingly influential, both in official documents and in the media, is “risky drinking” (see Note 2). In Sweden, this concept was first launched by the Family Medical Institute in 2004 in connection with a new project aimed at developing competence among primary care staff in dealing with patients with presumed drinking problems. In 2006, the project was transferred to the Swedish National Institute of Public Health, and subsequently it became part of the national ANDT strategy for preventing substance use and similar problems (Prop. 2010/11:47, Skr. 2015/16:86, https://data.riksdagen.se/). As of today, risky drinking has become institutionalized as a category in its own right, as indicated by the identification of “risk drinkers” as a main target group for interventions by the Substance Misuse Inquiry (GOR, 2011, No. 35), and the fact that officials and other actors have started providing guidelines for allegedly “safe” drinking habits (GOR, 2011, No. 35).
As has often been pointed out however, there is no consensus, either nationally or internationally, on the definition of “safe” or “hazardous” drinking habits (e.g., Espman & Allebeck, 2011; Montonen et al., 2016). In addition, it is far from clear which types of risks suggested limits to drinking are intended to prevent (e.g., Herring, Berridge, & Thom, 2008). Rather, it remains fundamentally unclear whether the defined “risky drinking” patterns are linked to detrimental physiological harm, psychological distress, social marginalization, or increased problem drinking (Herring, Berridge, & Thom, 2008). For this reason, most current guidelines have chosen to define “risky drinking” in purely quantitative terms, either in terms of the number of standard drinks per week or in terms of the number of standard drinks consumed on one occasion—two measures that do not necessarily correspond with one another. In Sweden, for example, although there is no official definition, “risky drinking” among men has as a rule been measured as drinking 14 standard drinks or more per week or consuming more than 4 standard drinks per drinking occasion (Andréasson & Allebeck, 2005). The corresponding amounts for women are nine standard drinks or more per week and/or more than three standard drinks on one occasion (Andréasson & Allebeck, 2005). These quantitative definitions are associated with several limitations. First, the legitimacy of the chosen “threshold” separating safe from hazardous drinking is questionable since both the content (measured in pure alcohol) of a “standard drink” and the number of standard drinks chosen to indicate risky drinking vary across different countries (e.g., Espman & Allebeck, 2011; Montonen et al., 2016). Second, there is substantial individual and subgroup variation in the experienced frequency of negative consequences following the consumption of a certain amount of alcohol (e.g., Selin, 2004).
To summarize, our brief synopsis shows that the formal definitions and understandings of each of the four most frequently used concepts denoting problematic drinking have varied across time and/or space and that most of the definitions leave substantial leeway for subjective interpretation.
A Short Overview of Some Previous Research
While this is the first study to investigate and compare the meanings attributed to the above described concepts of problematic drinking, some previous research has devoted itself to the conceptual, historical, and empirical analysis of terms generally associated with drinking. A number of these studies provide support for the conclusion that while the terminology used to describe or just talk about drinking is very extensive (e.g., Cameron et al., 2000; Levine, 1981), the meaning of many of the terms or concepts employed tends to be “fuzzy” or “slippery” (e.g., Herring et al., 2008). Further, previous research in the field suggests that we might find both consensus and variation between different groups in their perceptions, beliefs, and attitudes about problematic drinking. For example, while several studies have shown that the dominant perceptions of the severity of addiction-related problems (for the individual and for society), the assumed nature of addiction, and the rated options for recovery from addiction vary considerably, both across substances of addiction, populations, and countries (e.g., Blomqvist, 2009, 2012b; Blomqvist, Koski-Jännes, & Cunningham, 2014; Hirschovits-Gerz et al., 2011; Holma et al., 2011), few significant differences have been found between treatment professionals’ and lay people’s perceptions (Koski-Jännes, Hirschovits-Gerz, & Pennonen, 2012; Koski-Jännes, Pennonen, & Nyyssönen, 2012; Samuelsson, Blomqvist, & Christophs, 2013). In addition, several studies have aimed to trace the levels of relative support in the general population for various notions of problematic drinking including the views that these problems constitute a disease, a bad habit, a sin, a form of wrongdoing, an expression of weakness of will, or an escape from personal, psychological, and/or social hardships (e.g., Blomqvist & Christophs, 2005; Blomqvist, Cunningham, Wallander, & Collin, 2007; Cunningham, Blomqvist, & Cordingley, 2007; Cunningham, Sobell, & Sobell, 1996). Some conclusions from these studies are that the disease notion of problematic drinking occupies a stronger position in North American culture than in Scandinavia (Blomqvist, 1998; Blomqvist et al., 2014) and that people with no personal experience of problematic drinking may endorse a more “moralistic” view than people who have such experiences (cf. Blomqvist, 2012b). In one cross-cultural study exploring the meanings and interpretations of the alcohol dependence criteria included in ICD-10, the authors (Schmidt & Room, 1999) found that descriptions of dependence symptoms tend to vary across sites characterized by different drinking cultures.
Taken together, the results from these studies suggest that the terms, notions, and understandings associated with problematic drinking are not only numerous but are also contingent on both culture and context, and that they may also be influenced by the individual’s own experiences of problematic drinking.
Method
With the aim of exploring the meanings that social workers, health-care staff, journalists, researchers, and the general public ascribe to four frequently used concepts of problematic drinking, this study employs a factorial survey, which is an experimental method for collecting data about individuals’ beliefs and judgments (e.g., Auspurg & Hinz, 2015; Rossi & Nock, 1982; Wallander, 2009). In the Methods section, we will (1) describe our factorial survey design and (2) provide details of our approach to sampling respondents and collecting data.
The Factorial Survey Design
The factorial survey is a hybrid method that combines the strengths of the survey (external validity) and of the experiment (internal validity) in the study of individuals’ beliefs and judgments (Auspurg & Hinz, 2015; Rossi & Nock, 1982; Wallander, 2009). The core of the method is the vignette (a fictive description of a person or a situation), which is made up of one level (value) from a number of dimensions (variables), and to which the study participants are asked to respond. By using a large sample of vignettes (see below), and by letting the included dimensions vary across the vignettes, it is possible to analyze the independent effects of each dimension on the respondents’ judgments. Accordingly, it is possible to draw conclusions about consensus and variation in the respondents’ judgment principles, that is, about how they draw on the information provided in the vignettes in making judgments about the vignettes (Wallander, 2009).
In a typical factorial survey, the respondents are requested to judge multiple vignettes, all of which have been randomly sampled from a large population of vignettes (Wallander, 2009). In the present study, each respondent was asked to read 10 unique vignettes describing problematic drinking and to make judgments about whether the case descriptions were indicative of risky drinking, alcohol dependence, alcohol misuse, alcoholism, or none of these (altogether, 8,630 unique vignettes were assessed by the respondents). Figure 1 provides two examples of a vignette and the rating scale.

Examples of a vignette and the Rating Scale.
These vignettes include one level from each of 12 dimensions, which together describe a person in terms of demography (gender and age), social factors (social status and social network), drinking (duration, type of drink, quantity), and negative medical and/or social consequences of the drinking (craving, abstinence, medical problems, mismanaged work/studies/important commitments, lost social contacts). The selection of the drinking-related dimensions and of the dimensions portraying the consequences of drinking was based on a brief review of earlier research, including (but not restricted to) that focused on the above-discussed formal definitions of the concepts. The demographic and social dimensions were included with the aim of investigating whether any or several of the concepts of problematic drinking—as defined and used by our stakeholders—might be associated with specific categories of drinkers (in terms of gender, age, social status, and social network; cf. Samuelsson & Wallander, 2014). 7 Table 1 summarizes our factorial survey design in terms of dimensions and levels, the wordings of the levels, and the percentages of levels included in the total sample of vignettes.
The Vignette Dimensions, Levels, Wordings, and Inclusion Percentage in the Vignettes.
Note. N = 8,630. The nonvaried vignette text is displayed in italics.
aIf the vignette person is younger, she or he is studying to become a lawyer/medical doctor/business executive. bPercentages for the four quantity levels, aggregated for all types of drinks.
The vignette universe of our factorial survey design consists of 262,144 unique vignettes, a figure which is derived by multiplying the number of levels for all the 12 dimensions. By drawing a random sample of 10 vignettes per respondent (8,630 vignettes in total) from the full population of vignettes, we arrived at a vignette sample which is representative of the vignette universe (for more details about sampling in factorial surveys, see Auspurg & Hinz, 2015). Our sampling strategy generated a vignette sample with approximate balance across the dimension levels (which means that the dimension levels appear with equal frequency in the vignettes; see Table 1), and with approximate orthogonality between pairs of vignette dimensions (the strongest correlation between any two dimensions is r = .030). These characteristics of our vignette sample provide our analyses with strong statistical power and allow for conclusions about causal relationships between each vignette dimension and the choices of concepts made by the respondents to describe the vignettes.
Respondents and Data Collection
The stakeholders who comprise the respondents in the study consist of the two main professional groups working with problematic drinking, that is, (1) social workers and (2) primary health-care staff (medical doctors and nurses), and also of (3) journalists, (4) researchers in the field of substance use, and (5) the general public.
Social workers were recruited via The Union for Professionals (Akademikerförbundet SSR; www.akademssr.se), the predominant trade union for social workers in Sweden. The union provided us with e-mail addresses for a random sample of 1,000 members employed by the municipal social services and simultaneously promoted the study on their web page. This yielded 185 valid responses.
Medical doctors and nurses employed in primary health-care centers were recruited by the Norwegian section of IMS Health (the name of the company has since been changed to IQVIA, www.iqvia.com), which keeps a record of medical staff in the Swedish health-care system. IMS Health distributed our questionnaire to a random sample of 250 medical doctors and 250 nurses, a procedure that—after repeated reminders—provided us with 49 valid responses from 28 medical doctors and 19 nurses (other function = 2).
In order to recruit journalists who write in the daily press or in periodicals about issues pertaining to lifestyle, health, social problems, and similar subjects (including substance use), we made use of Pressregistret (www.pressregistret.se)—a company specializing in media communication (i.e., in providing various customers with the necessary contacts to distribute information via the media). Pressregistret provided us with a list of e-mail addresses, and after scrutinizing the list for journalists with the relevant orientation, the questionnaire was distributed to 864 journalists, which resulted in 65 valid responses.
With the purpose of recruiting researchers focused on substance use, we approached networks of researchers associated with the Centre for Social Research on Alcohol and Drugs (www.sorad.su.se), the Swedish Society for Research on Alcohol and Drugs (www.sadforskning.se), the Swedish Council for Information on Alcohol and Drugs (www.can.se), and Karolinska Institutet www.ki.se). By informing about the study using e-mail lists, and by making contact with researchers directly via their e-mail addresses (as available on the Internet), we asked 104 researchers in the area of substance use about their willingness to participate in the study. Of these, we were able to recruit 55 researchers from within the social sciences (n = 30), the medical sciences (including epidemiology; n = 16), and the behavioral sciences (n = 8; item nonresponse = 1).
Finally, a stratified (according to gender and age) random sample of 765 individuals from the general population was selected from a number of web panels, managed by the web survey company Alstra (www.alstra.se), which also administered all of the data collection. This procedure resulted in 509 respondents, aged 17–79 years. Altogether, our sample consists of 863 respondents aged 17–82 years (mean = 44, SD = 13.7), of whom 61% are women and 39% men. Although the majority of our subsamples were randomly selected, the response rates are on average very low, which means that we must be careful when generalizing the results from the study to larger populations.
The data collection, which was administered by Alstra (see above), was carried out in the form of a web questionnaire distributed via the respondents’ e-mail addresses. The survey was open between April and August 2016 and involved between one and two reminders (the number of reminders depended on the response rates in the different subsamples). The web questionnaire consisted of 10 randomly selected vignettes, after which followed questions about the respondent’s gender, age, and drinking habits (measured using the Alcohol Use Disorders Identification Test (AUDIT; Babor, de la Fuente, Sanders, & Grant, 2011). An additive index was constructed on the basis of the individual’s responses to the 10 AUDIT-questions, resulting in a variable ranging between 0 and 31 (mean = 4.3, SD = 4.3), with higher values indicating more extensive alcohol consumption and/or more negative consequences of this consumption.
Results
In this section, we begin by describing the distribution of the vignette judgments across the five options on the vignette rating scale. Thereafter, we introduce multilevel multinomial logistic regression analysis and give examples of how to interpret results from such analyses. Subsequently, we present the results from six regression models, starting with the vignette variables and concluding with the respondent variables.
Introduction to the Statistical Analyses
As described above, our respondents were each asked to evaluate 10 vignettes and to make judgments about which of the four concepts of problematic drinking best described the vignette (the response categories also included the option “none of these”). A simple frequency table shows that the respondents chose to describe 888 of the 8,630 vignettes (10%) as risky drinking, 2,083 vignettes (24%) as alcohol dependence, 2,503 vignettes (29%) as alcohol misuse, 2,973 vignettes (34%) as alcoholism, and 183 vignettes (2%) as none of these. Because the last option (“none of these”) was infrequently employed by the respondents, these vignettes were excluded from the multivariate analyses.
In order to analyze the dimensionality of the four concepts—that is, how they are used as descriptors of various instances of problematic drinking—a number of multilevel multinomial logistic regression analyses were conducted, in which the choice of problematic drinking concept constituted the dependent variable, while the vignette dimensions and a number of respondent variables were included as independent variables. Multinomial logistic regression constitutes an extension of the more well-known binomial logistic regression analysis, and it is used in a similar way with the intention of predicting the odds for a certain outcome (1) in comparison with another outcome (0 = reference category) on the basis of a number of independent variables. Because the dependent variable in our analyses includes four qualitatively different outcomes, each analysis includes three models, in which each of three outcomes is compared with a fourth (=reference category). To enable pairwise comparisons of all outcomes, three multilevel multinomial logistic regression analyses were conducted, in which alcoholism (Models 1–3), alcohol misuse (Models 4–5), and alcohol dependence (Model 6) served as the reference categories.
Table 2 presents odds ratios and p values for all six regression models. The odds ratio is the odds for a certain outcome divided by the odds for a different outcome. Using Model 1 as an example, the odds ratios express the odds of a certain vignette characteristic being classified as risky drinking, divided by the odds of this characteristic being classified as alcoholism. Identical odds for the two outcomes result in an odds ratio of one (1), and higher and lower values can be interpreted as increases or decreases (in percent) in the odds that a vignette characteristic will be classified as risky drinking rather than alcoholism. As an illustration, the vignette variable age (in Model 1) is represented by three dummy variables (younger, lower middle age, higher middle age) and a reference category (elderly). The odds ratio of 1.71 means that if the vignette person is younger rather than elderly, the odds that the vignette will be classified as risky drinking rather than as alcoholism increase by 71%. For the continuous respondent variable AUDIT (see the final part of the model), the odds ratio (1.14) indicates that each additional point on the AUDIT Scale results in an increase of 14% in the odds that the risky drinking concept will be given preference over the alcoholism concept.
Multilevel Multinomial Logistic Regression Models Examining the Effects of Vignette Variables and Respondent Variables on Respondents’ Choices of Concepts (Risky Drinking/Alcohol Dependence/Alcohol Misuse/Alcoholism).
Each of the respondents in the study has assessed 10 vignettes, yielding a hierarchical data set consisting of 8,410 observations at Level 1 (vignettes and judgments) and of 856 observations at Level 2 (respondents). 8 Since it cannot be assumed that the vignette judgments are uncorrelated within respondents, the study has employed so-called multilevel (multinomial logistic regression) analysis (see Hox, Moerbeek, & van de Schoot, 2017), which corrects for autocorrelation and also allows for the exploration of any potential unexplained variance at Level 2 (Hox, Moerbeek, & van de Schoot, 2017). As Table 2 includes as many as six models, all of which include quite a few numbers, we will not provide a detailed interpretation of all the results in the table but will rather (1) summarize the dimensionality of the four concepts at an overarching level and (2) account for the significance of the respondent variables. Further, since the number of observations at Level 1 is so large (>8,000), even small odds ratios yield small p values (cf. Salkind, 2010). In order not to overinterpret the numerical estimates but rather highlight the most prominent results, we will only discuss Level-1 odds ratios accompanied by p values close to <.001. At Level 2, however, where the number of observations is <1,000, we will consider odds ratios associated with p values of <.05 as statistically significant.
Vignette Variables
Starting with the results associated with the demographic vignette dimensions, Table 2 shows that the respondents’ choices of concepts are not guided by the vignette person’s gender. However, the person’s age matters in the sense that alcoholism is more often used to describe elderly people than younger people. With regard to the social factors embedded in the vignettes, the analyses indicate that the choice of concept is not related to the vignette person’s social network but with her or his social status. Thus, the results include a difference between risky drinking and alcohol dependence on the one hand and alcoholism on the other hand, whereby there is a preference for using the first two concepts to describe people with good to normal social status (operationalized by means of various occupations), whereas alcoholism is more often reserved for people with a very low social status (such as those lacking a permanent source of income).
The vignette dimensions that describe drinking habits, that is, duration, type of drink, and quantity, are all significantly related to the respondents’ choices of concepts. When it comes to the duration and the quantity of the drinking, it is noticeable that alcoholism is generally associated with a very high consumption over a considerable period of time, whereas risky drinking is primarily employed to describe people who have been drinking at moderate levels, and only recently. Alcohol dependence and alcohol misuse are utilized for describing drinking patterns that lie between alcoholism and risky drinking, and they differ only with regard to the quantity of drinking, with alcohol misuse more often being associated with larger quantities than dependence. The results illustrating the relationships between various types of drinks and the four concepts of problematic drinking are not clear-cut, and only some of them are displayed in Table 2, in which liquor (cf. strong spirits) is used as the reference category. 9 At the general level, there are patterns indicating that alcoholism is more often associated with liquor and illegal spirits, whereas the concepts of risky drinking and alcohol dependence are more often used to describe people who drink strong beer or wine. The pattern seen for the odds of classifying vignettes involving a certain type of drink as alcohol misuse is bidirectional, with these odds increasing for both wine and strong spirits, depending on whether the comparison is made in relation to the concept of alcoholism or that of risky drinking.
The vignettes include five dimensions that refer to negative consequences of drinking, and these may be further categorized into physical consequences (craving, abstinence, medical problems) and social consequences (mismanaging job/studies/important commitments, having lost important social contacts). Analogously with the drinking habits results, the analyses show that the alcoholism concept is linked to most of these negative consequences, whereas risky drinking is used to describe vignettes characterized by an absence of such physical or social consequences of drinking. And as was also the case in the results focused on drinking habits, alcohol dependence and alcohol misuse can be found halfway between these two “extreme” concepts, although with some important exceptions. Thus, craving (i.e., experiencing a strong urge to drink more) is associated with alcohol dependence as often as it is with alcoholism but is less often described using the alcohol misuse concept. In a similar vein, medical problems in the form of an ulcer are frequently linked with both alcohol misuse and alcoholism but are less often judged as indicating alcohol dependence. Finally, with regard to the social consequences of drinking, the alcohol misuse concept is close to the alcoholism concept in the sense that the relevant vignette dimensions do not significantly differentiate between the two concepts, whereas alcohol dependence is associated with the absence of social consequences to a much greater extent than alcohol misuse.
Respondent Variables
In order to explore differences between respondents with regard to their use of the four concepts of problematic drinking, we included four respondent variables as predictors in our analyses. Starting with the variable indicating respondent group membership, it is notable that there are considerable differences between researchers and the general population with regard to their employment of risky drinking, alcohol dependence, alcohol misuse, and alcoholism as descriptors of the vignettes. As is shown in Table 2 (Models 1 and 2), researchers are more than 70 times as prone as the general population to choose the risky drinking or alcohol dependence concepts over the alcoholism concept and are more than 40 times as likely to prefer alcohol misuse to alcoholism. These large differences are due to the fact that the researchers in the study only classified 4% of the vignettes as indicating alcoholism, whereas the general population chose alcoholism as a descriptor for 38.5% of the vignettes. Journalists, on the other hand, are even more inclined than the general population to classify the vignettes as indicating alcoholism (when compared to risky drinking and alcohol dependence). When compared to the general population, the journalists are also more prone to judge the vignettes as indicating alcohol misuse rather than risky drinking or alcohol dependence. If we look at the two respondent groups that represent the “professional stance” with regard to the issues examined in the study (i.e., social workers and health-care staff), the differences between these groups are in most cases negligible, and the pattern of results (which are somewhat more clear-cut for the social workers) indicates that they tend to favor the concepts alcohol misuse and alcohol dependence.
In two of the models (Models 4 and 5), the respondents’ gender and age are significantly associated with their choices of concept. Thus, for example, men are more inclined than women to classify the vignettes as risky drinking or alcohol dependence rather than as alcohol misuse. In Model 5, there is a very weak but statistically significant effect of age, indicating that the older the respondent becomes, the more inclined she or he is to describe the vignettes as involving alcohol dependence rather than alcohol misuse. Last but not least, in all the models shown in Table 2, the respondents’ personal drinking habits, measured as points on the AUDIT Scale, are significantly associated with their choices of concepts. More specifically, the analyses show that the more an individual drinks, and the more negative consequences she or he is experiencing from her or his own drinking, the more likely she or he is to choose one of the “lighter” concepts of problematic drinking. This means that she or he is less likely overall to use alcoholism as a vignette descriptor, is more likely to choose alcohol dependence over alcohol misuse, and is also more likely to prefer risky drinking to alcohol dependence or alcohol misuse.
Finally, all the variance components presented at the bottom of Table 2 are statistically significant, which means that there remains some unexplained between-respondent variance for all pairwise comparisons of concepts (cf. Hox et al., 2017). This suggests that the respondents’ choices of concepts might be further explained by respondent variables not included in the analyses and/or that there is some random variation in the way individual respondents employ the concepts of risky drinking, alcohol dependence, alcohol misuse, and alcoholism as descriptors of problematic drinking.
Discussion
With the aim of investigating the meanings that social workers, health-care staff, journalists, researchers, and the general public ascribe to the concepts alcoholism, alcohol dependence, alcohol misuse, and risky drinking, this study employed a factorial survey in which 863 respondents classified (i.e., chose a concept/descriptor for) 10 unique vignettes portraying varying cases of problematic drinking (in all, the study includes 8,630 unique vignettes). By means of multilevel multinomial logistic regression analysis, we were able to analyze patterns of agreement among the respondents regarding the dimensional makeup of the concepts as well as variance across groups of respondents in their preferences for using the concepts as descriptors of problematic drinking. In the section below, we will summarize and discuss the most important findings from the study, take some time to contemplate the findings against the background of Christie and Bruun’s (1969) notion of “fat words,” and also describe some limitations associated with the research design.
Starting with the results relating to the ascribed meaning of the four concepts under study, it is worth noting that all but two of the dimensions included in the vignettes had a bearing on our respondents’ choice of concepts. The most influential dimension—according to the magnitude of the odds ratios—was the “quantity of drink,” which was followed by the “duration of drinking.” These two drinking-related dimensions produced an extremely clear distinction between two of the concepts involved—a distinction that was further confirmed by the results relating to the negative consequences of drinking. Thus, the concept denoting the “heaviest” or most problematic drinking was undoubtedly alcoholism, which is most commonly used in descriptions of cases characterized by the largest quantities, the longest durations, and the regular prevalence of all of the negative consequences referred to in the vignettes. At the other end of the spectrum, we find “risky drinking,” a concept typically chosen to describe cases involving moderate quantities, short durations, and in which the drinking produced no (or few) negative consequences. Between these two extremes, we found alcohol misuse and alcohol dependence—two concepts with somewhat dissimilar connotations. Even though there were no significant differences between misuse and dependence with regard to duration, type of drink, abstinence, or any of the vignette individuals’ sociodemographic characteristics, there were several results indicating that the meaning associated with the alcohol misuse concept (when compared to dependence) lies closer to that associated with alcoholism. Thus, alcohol misuse was more often used to describe cases involving large quantities of drinking and negative consequences such as medical problems, mismanagement of work or studies, or lost social contacts. The finding that our respondents on the whole considered the misuse concept to be more descriptive of cases involving a more problematic use than dependence is interesting, particularly in view of the fact that the diagnostic system DSM (IV; https://www.psychiatry.org, American Psychiatric Association, 2018) previously classified abuse (cf. misuse) as a milder form of dependence. One exception to this pattern was that a craving for more alcohol following a drinking day was more frequently classified as dependence than misuse, suggesting that the use of the concept dependence is more strongly related to a biomedical perspective, according to which the central understanding and explanation of the individual’s urge for more alcohol involves the workings of physical mechanisms located in the brain (cf. Leshner, 1997).
Considering the above-described results, and adding the fact that the alcoholism concept was more often employed when the vignette person was elderly, had lower social status, and consumed strong liquor, we may conclude that alcoholism (followed by alcohol misuse) is used to describe both the heaviest and indeed the most deviant (i.e., the least socially accepted) cases of problematic drinking. In contrast, risky drinking (and in some sense also alcohol dependence) was more often used to describe people with good to normal social status, who drank fairly moderate quantities of wine, and who experienced fewer (or no) negative consequences—that is, people who have their heavy drinking “under control” and who can therefore lead “normal” lives. The finding that the alcoholism concept is generally used to describe the heaviest and least socially acceptable cases of problematic drinking might not be very surprising. However, if we take into account the fact that alcoholism is the oldest (and perhaps also the most “outdated”) of our four concepts and that the international communities of both researchers and professionals (cf. ICD and DSM) as well as the Swedish legal system and professional associations have long given preference to other concepts denoting problematic drinking (e.g., dependence and abuse/misuse, see above), it is noteworthy that alcoholism was the most frequently employed concept among the respondents in our study (it was chosen for every third vignette). Even though this latter result might to some extent be explained by the composition of the vignette contents (i.e., by the fact that a large proportion of the vignettes portrayed very heavy and/or deviant cases of problematic drinking), it nonetheless reveals that the currency of the alcoholism concept remains strong in contemporary Swedish thinking about problematic drinking.
Turning from the identified patterns of consensus regarding the meaning of the four concepts under study to the variance identified between different groups of respondents, we may note that all the respondent characteristics included in the analyses as independent variables had an impact on the respondents’ choice of concepts. Most clearly, their choices were linked to which stakeholder group they belonged to. This was seen in that members of the general population, and even more so journalists, were more inclined than the other three groups to classify the vignettes as alcoholism or alcohol misuse; the two professional groups, and particularly the social workers, tended to favor the concepts alcohol misuse and alcohol dependence; and the researchers preferred to use the concepts alcohol dependence and risky drinking to a greater extent. This implies inter alia that the research-based concepts—that is, dependence and risky drinking—have not yet permeated the public discourses on problematic drinking. The relative proximity between the journalists and the general public (when compared to the other groups), combined with the fact that the journalists and the researchers comprise the two stakeholder groups that are furthest apart in their classifications of problematic drinking, could indeed be taken to mean that the media has failed in its task of communicating scientific knowledge to the general public. On the other hand, the journalists’ preferences for employing the two concepts that portray the heaviest (and most deviant) drinking might be viewed as a sign of a strategy based on the well-researched media logic of “sensationalism” (cf. Grabe, Zhou, & Barnett, 2001) and thus of using “lay-” concepts, the meanings of which are well established (or thought to be well established) in the public discourse. When it comes to the two groups that represent the “professional stance” in this study (the social workers and the health-care staff), these appear to prefer the two concepts that have for some time enjoyed the greatest legitimacy in Sweden, either by being included in the Swedish legislation (misuse) or in terms of being the preferred scientific concept (dependence), over a concept which they might regard as being somewhat outdated and pejorative (alcoholism) and a concept which has yet to spread beyond the domain of the researchers (risky drinking). On the whole, the above-described variance across stakeholder groups contrasts with the results of previous studies (e.g., Samuelsson et al., 2013; see above), which have found few significant differences between treatment professionals’ and lay people’s perceptions of addiction. However, while our results support the suggestion that individuals’ preferences for using particular problematic drinking concepts are indeed associated with their stakeholder group membership, they do not reveal whether social workers, health-care staff, journalists, researchers, and members of the general public actually ascribe very different meanings to these concepts. This latter question will be further scrutinized in a future article, in which we will systematically compare the meanings ascribed to the different concepts by the five groups of stakeholders.
Another result of great significance is the stable relationship found between an individual’s own alcohol consumption and his or her conceptual choices. The fact that drinking more (and with more negative consequences, as measured by the AUDIT Scale) implied choosing “lighter” concepts as descriptors for the vignettes indicates that “heavy drinkers” tend to prefer risky drinking to the other three concepts, to favor dependence over misuse and alcoholism, and to choose misuse before alcoholism. Perhaps the finding that an individual’s own experiences of and involvement in a particular phenomenon is associated with his or her evaluation of that phenomenon as being more or less problematic should not surprise us. However, it suggests the importance of being careful when choosing how to portray problematic drinking in the media, in research, and in official reports. If “heavy” and to some extent “stigmatizing” concepts (such as alcoholism, see above) are frequently and mechanically used to describe very varied types of problematic drinking, this might be detrimental to problematic users’ understandings of themselves and might consequently restrict them in their efforts to find alternative ways of life.
Given the results of the study summarized above, is it possible to provide an answer to the question of whether alcoholism, alcohol dependence, alcohol misuse, and risky drinking constitute “fat words” with little content (which can be used interchangeably; cf. Christie & Bruun, 1969) or whether they are indeed useful definitions (which differentiate between various patterns/forms of problematic drinking)? The answer must inevitably be both. The identified patterns of consensus described and discussed above suggest several important and meaningful distinctions between the four different concepts. On the other hand, the findings showing that the choice of concept to some degree depends on who you are, in terms of gender, age, stakeholder group, and alcohol consumption, and that there remains some unexplained between-respondent variance, imply that there is both socially structured disagreement and unexplained (and perhaps inexplicable) disagreement about the actual connotations of these four concepts. As noted in the introduction to this article, such potential disagreement may possibly lead to poor communication between actors and may produce real-world challenges not only at the level of policy-making but also in clinical practice and in individuals’ efforts to find suitable ways out of problematic drinking habits. Not long ago, in an attempt to come to terms with the negative consequences associated with cross-cultural variation and rapid fluctuations in the terminology and understanding of problematic drinking and drug use, a number of leading researchers in the field suggested that the term “heavy use over a period of time” could be employed to define all substance use disorders (Rehm et al., 2013). Their main arguments were that heavy use over time is in itself responsible for the other physiological, psychological, and social characteristics commonly associated with addiction, dependence, or misuse, and that “heavy use” implies a continuum of (more or less problematic) drinking—a continuum which might serve to reduce the levels of stigmatization frequently attached to behaviors categorically labeled as, for example, dependence or misuse. Interestingly, the empirical findings of our study, which suggest that the quantity and duration of drinking constitute the two most important factors when distinguishing between concepts of problematic drinking, provide some support for the view that “heavy use over a period of time” may indeed constitute an appropriate and useful description of problematic drinking.
When drawing conclusions from this study, it is important to consider some limitations associated with our research design. One such limitation is that the inflexibility of the experimental design—which is inevitable due to the necessity of standardization—means that only a restricted number of dimensions and levels can be incorporated into the vignettes, and that the wordings (i.e., operationalizations) of the vignettes must be adapted so as to allow for the combination of all levels from all dimensions. In addition, even though the experiment enjoys high levels of internal validity and allows us to identify consensus and variance in the meanings attributed to the concepts under study, it does not provide us with empirical data that may be employed for understanding and interpreting the identified patterns. By supplementing our factorial survey study with a smaller focus group study, we hope to be able to shed more light on and dig deeper into the question of how members of different stakeholder groups in Sweden conceptualize problematic drinking.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study is part of a larger project (funded by the Swedish Research Council, project number 2014-14224-111923-43) that explores historical and contemporary conceptions and definitions of problematic drinking. The project has been reviewed by the Ethical Review Board in Stockholm, Sweden (number 2016/256-31).
