Alcohol use disorder severity and reported reasons not to seek treatment: a cross-sectional study in European primary care practices

GPs and their patients were representatively sampled in nine regions from six European countries (Italy, Germany, Hungary, Latvia, Poland and Spain) using complete registers or local GP associations. In some countries, GPs were randomly sampled from registers while in others sampling was stratified according to pre-selected criteria (e.g. urbanity, size of GP practice). 56.4 % of contacted GPs refused study participation. The majority of GPs refused their participation for reasons unrelated to this study’s characteristics, e.g. tight schedules, insufficient space in the PC facility, lack of interest in research activities generally or current participation in other studies. For countries with populations below 40 million inhabitants (Latvia and Hungary) primary care practices were sampled to be nationally representative. In the remaining countries primary care practices were regionally representative (Italy: Friuli-Venezia Giulia and Tuscany region; Germany: Saxony and Berlin state; Poland: Łódź and Podkarpackie province; Spain: Catalonia autonomous community). In Hungary and Spain, all patients with GP assessment were interviewed. In the remaining countries, patients were selected for interview either randomly or based on stratification (oversampling AUDs cases as judged by the GP), with an overall patient refusal rate of 17.8 %. Ethical approval was obtained in each of the participating countries and the respective reference numbers are shown in Additional file 1: Table S1 as Additional file.

GPs were asked to fill in a brief questionnaire for all patients (18–64 years of age) coming to their practice on a given preselected day or several days. Information on the patients’ demographics, health status, reasons for consulting the GP, current alcohol use and AUDs, and information regarding treatment for AUDs were determined by the GP, based on his/her best knowledge of the patients’ problems. GPs were allowed to use the patients’ medical charts. Trained study personnel interviewed the patients after their appointment at the GP’s premises or on a subsequent day at the study center. Informed consent was obtained by all interviewed patients.

The patient interview included the AUDs section of the Composite International Diagnostic Interview (CIDI); [30] assessing information on alcohol use and AUDs criteria according to the fourth as well as the fifth revision of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV; [31], DSM-5); [32]. To assess reasons for not seeking treatment, patients reporting any DSM-5 AUDs symptom, were subsequently asked if they sought help. If they responded negatively they were asked an open ended (‘If you did not ask for help, please describe why’; lifetime and 12 months) as well as a set of closed questions (asking for shame, stigma, dissatisfaction with treatment system and wish not to stop drinking completely; referring to the past 12 months only). Additionally, the patient interview comprised questions on socio-demographics as well as a number of validated questionnaires.

A coding scheme was developed in a directed content analysis [33]: First a literature review identifying key reasons for not seeking treatment was performed to generate the basic coding scheme. Then all answers on the open ended question were read to identify meaningful subcategories and include reasons not yet covered by the basic coding scheme. After coding a sample of 100 answers the authors discussed and adapted the scheme where necessary to capture all reported reasons. The coding scheme (see Additional file 1: Table S2) was then applied independently by two raters. All cases of non-agreement were revisited and discussed between the authors to come to a final decision.

The scheme allowed for multiple coding on the following broader categories (number of subcategories): ‘lack of problem awareness’ (6), ‘cope alone’ (4), ‘stigma or shame’ (3), ‘other prominent problem’ (2), ‘encounter barriers’ (7), and ‘not codable’ (2). Answers were classified as ‘not codable’ when the patient did report seeking professional help or when the answer was incomplete or incomprehensible.

Comprehensive variables were then generated including the information on treatment seeking assessed in closed variables as well as the coded answers from open ended questions (see Additional file 1: Table S2). For the evaluation of Saunders’ theoretical framework [17] we summarized the reported reasons in the following four categories: (1) ‘denial of the problem’; (2) ‘minimization of problem severity’; (3) ‘self-change’; and (4) ‘encounter barriers’. See Table 1 for an overview on broader categories, subcategories and classification according to Saunders’ framework.

Inter-rater agreement was evaluated using kappa statistics. Kappa was calculated for accordance on the broader category (e.g. any kind of lack of problem awareness was coded vs. no lack of problem awareness coded) as well as accordance on subcategories within broader categories [34].

Frequencies and percentages were calculated for all patients who were diagnosed with an AUD by their GP or by CIDI (at least a mild AUD as per DSM-5), had not received treatment, and had a valid answer in the questions concerning treatment seeking. Exploratory logistic regressions were calculated to investigate patient characteristics (age, sex, years of education, employment status, and AUDs severity) predictive of providing a valid answer in the questions concerning treatment seeking. Percentages and frequencies of broader categories by country were calculated for the past 12 months. To investigate country specific outliers, standardized residuals [35] were calculated. A standardized residual > |3| was considered as outlier [35]. The country specific results are reported in the Additional file 1: Table S3 and should be interpreted with caution, because of the small sample size.

To investigate associations with AUDs severity, the reason for not seeking treatment ordered according to Saunders’ framework (see above) served as dependent variable. AUDs severity was operationalized as defined in DSM-5 (sub threshold (1 criterion), mild (2–3 criteria), moderate (4–5 criteria), severe (6 or more criteria)). As the proportional odds assumption did not hold, multinomial logistic regression was performed introducing AUDs severity as a factorial variable and adjusting for age. Based on the resulting model, predicted probabilities were calculated for all levels of AUDs severity. All analyses were conducted in STATA 12, considering 12 months and lifetime (excluding 12 months) variables separately. Sample weights were applied in all analyses to account for different sampling probabilities [27, 28].

The inter-rater accordance was assessed based on all coded answers (N = 552 for past 12 months questions and N = 1,107 for lifetime), irrespective of the respondent’s treatment status and AUDs diagnosis (i.e. sub threshold cases with only one criterion included). Multiple answers per patient were possible. The two raters showed a substantial to perfect accordance as shown in Table 2.

The majority of the 9,098 patients (3,715 males and 5,383 females) interviewed came from Hungary (N = 2,306), followed by Spain (N = 1,994). The remaining patients came from Germany (N = 1,356), Latvia (N = 1,302), Poland (N = 1,197) and Italy (N = 943). The mean age was 44.3 years (standard deviation 13.3 years), ranging between 18 and 64 years. On average the patients received 12.7 years of education (standard deviation 3.6 years) and 13.3 % of the sample was unemployed. Details on patients demographics are reported elsewhere [27].

As shown in Fig. 2, 1,008 patients were diagnosed with an AUD (12 months). 80 % of those patients did not receive any professional treatment [28] and of those 251 reported a reason why they did not receive treatment. Of the 1,774 patients with a lifetime AUD diagnosis (12 months excluded), 77 % did not receive treatment and 664 reported a reason for not seeking treatment. For questions referring to the past 12 months AUDs, unemployed patients were more likely to report a reason for not seeking treatment (OR = 2.24, 95 % confidence interval (CI) 1.36-3.70) and patients with a severe AUD were less likely to report a reason for not seeking treatment (OR = 0.53, 95 % CI 0.31-0.92). For lifetime AUDs no patient characteristic was predictive of reporting a reason for not seeking treatment.

Frequencies (multiple answers per patient possible) and weighted percentages including CIs of categories and subcategories are shown in Table 3. For past 12 months cases, the most frequent reason for not seeking treatment was ‘lack of problem awareness’ (55.3 %). Patients reporting ‘lack of problem awareness’ most frequently, did not consider their drinking as a problem or stated that no help was needed. The second most common reasons was ‘stigma or shame’ (28.6 %), closely followed by ‘encounter barriers’ (22.8 %) and ‘cope alone’ (20.9 %). The specific barrier reported most frequently was the wish not to stop drinking (72.9 %). Among patients fulfilling at least four DSM-5 criteria (moderate to severe AUDs), the share of patients reporting ‘lack of problem awareness’ was 52.1 % (12 months, 95 % CI 39.1-65.1; not shown in Table 3).

Country specific analyses showed a higher than expected prevalence for Germany (‘lack of problem awareness’), Latvia (‘cope alone’), and Spain (‘shame or stigma’ and ‘encounter barriers’) as indicated by standardized residuals above three. Lower than expected prevalence was only observed in Spain (‘lack of problem awareness’). Country specific percentages and standardized residuals can be found in Additional file 1: Table S3.

For lifetime AUDs 78.3 % of the patients named ‘lack of problem awareness’ as a reason for not seeking treatment. The second most common answer was ‘cope alone’ (17.2 %), while the remaining categories were all below 5 %.

Multinomial logistic regression was conducted to test the hypothesis that severity of AUDs (12 months) was associated with the reported reason for not seeking treatment, categorized according to Saunders’ theoretical framework (1= ‘denial of the problem’, 2 = ‘minimization of problem severity’, 3 = ‘self-change’, 4 = ‘encounter barriers’; see Table 1). The analysis included sub threshold cases fulfilling only one DSM-5 criterion and therefore included more patients than the descriptive analyses reported above. Overall 379 patients were included in the analysis of which 192 had a sub threshold, 140 a mild, 40 a moderate and 25 a severe AUD. The results are shown in Table 4. Based on the model probabilities for reporting a certain reason for not seeking treatment were calculated for each level of severity as defined by DSM-5. The results shown in Fig. 3 indicate lower probabilities of reporting ‘denial’ and higher probabilities to report ‘encounter barriers’ as AUDs severity increases. However, both trends were discontinued for patients with severe AUDs.

As also reflected in the odds ratios from the multinomial regression, the predicted probabilities of reporting ‘minimization of problem severity’ or ‘self-change’ were overall relatively low and did not show a clear association with AUDs severity.

The present study investigated reasons for not seeking treatment on a sample of European primary care patients that were diagnosed with an AUD but did not receive any professional AUDs treatment. We were able to show that the reasons for not seeking treatment were related to AUDs severity. However, half of the patients that fulfilled at least four DSM-5 AUDs criteria in the past 12 months still stated that they did not have a problem with alcohol use requiring treatment. The finding that the main reason for not seeking treatment was a lack of acknowledging the problem was in line with studies from the US [26]. As discussed below, the results indicate a huge potential for prevention when the health care system would address those patients with mild AUDs that are already at increased risk but do not (yet) perceive a need to seek treatment [36].

The present study showed once more that patients are unlikely to recognize and acknowledge problems related to their alcohol use themselves. Thereby a continuous monitoring of patients’ alcohol use could be useful not only for prevention of a number of physical health consequences but also for the primary as well as secondary prevention of AUDs. A Swedish study investigated frequencies of life style advice in GP settings in a large sample of primary care patients [37]. Across all age groups patients were least likely (only 4.7 % of the patients) to receive advice about alcohol compared to other life style areas like diet, smoking, or exercise.

The screening and monitoring of patients’ blood pressure could serve as a model for routine alcohol use assessment in primary care [38]. The latter practice has furthermore been argued to decrease stigmatization of AUDs [39]: instead of classifying patients into two distinct categories of ‘healthy and diseased’ each patients’ alcohol use behavior would be allocated on a continuum of related risk.

As reported in more detail in other publications [27, 28] the present study showed that GPs were more likely to detect AUDs when they became evident in decreased liver functioning and other physical indicators. Again, with a routine assessment of patients alcohol use, GPs might be able to offer brief interventions before more severe consequences emerge. Brief interventions (in primary care) have been shown to be highly effective and useful in prevention of progression into more severe stages of AUDs, alcohol-related mortality and societal costs [40‐43].

The present study also showed that for more severe cases of AUDs, treatment barriers were reported by the majority of the patients as reasons not to seek treatment. This indicates that even once patients acknowledged a need for treatment they might face barriers as a lack of information or high costs. Next to drinking cultures and the extent of stigmatization of AUDs, the treatment systems vary across Europe [44, 45, 9]. Thereby treatment barriers should be investigated in the context of the specific health care system.

The wish not to stop drinking completely was particularly frequent among the treatment barriers reported in this study. Treatment approaches that emphasize goals other than abstinence [46, 47] could be reconsidered for those patients. For example, ‘guided self-change’ (GSC) is a brief cognitive-behavioral motivational intervention designed to assist in recognizing and resolving alcohol-related problems [48]. It includes goal setting, self-monitoring of drinking behavior, analysis of drinking situations, and learning alternate coping skills [49].

This study was one of the largest undertaken to investigate reasons for not seeking treatment on representative samples of primary care patients, with a good refusal rate of below 20 % on the individual level. We were able to investigate reasons for not seeking treatment on a sample that had not yet received treatment for AUDs and thereby varied with respect to AUDs severity and other patient characteristics that tend to be more homogenous among patients in specialized care [50]. Furthermore, this study combined information from open ended and closed questions thus including yet unknown or less investigated reasons. When we compared this study with the only study with a comparable sample size [3], the present study was able to also investigate 12 months answers which were found to be somehow different from lifetime answers and less prone to memory biases.

The main limitations of the present study were the cross-sectional design, neither allowing any causal interpretation nor clear assessment of temporal sequences as suggested by Saunders’ model. Although the GP’s refusal rate of 56.4 % was considerable, as other studies with register-based random sampling usually have lower response rates [51, 52] it seems overall acceptable. As reported above, the majority of GPs refused study participation for reasons unrelated to this study. However, some GPs reported doubts about releasing patient data despite approval from research ethics boards and assurance of full anonymity and very few GPs were afraid of losing patients given the nature of the study on patients’ alcohol use which is a stigmatized subject [9]. Therefore it cannot be excluded completely that those GPs who refused study participation have a different patient population than the participants of this study. Furthermore, due to the low response rate in the open ended questions we can only speculate about those who did not give any reasons. While we could not find any patient characteristics predictive of reporting a reason for not seeking treatment for lifetime AUDs, participants with a severe AUDs diagnosis were less likely to report a reason as compared to persons with a light AUDs diagnosis. It is possible that those patients with severe AUDs did not report a reason because they did not acknowledge the problem or perceived a higher stigmatization. Overall, as reflected in the decreased likelihood to report any reason for not seeking treatment as well as in the discontinuation of observed trends in the reported reasons, this group of patients seems to be a special group and the respective results should be interpreted with caution.