Identification of smokers, drinkers and risky drinkers by general practitioners

Identification of smokers, drinkers and risky drinkers by general practitioners

Drug and Alcohol Dependence 154 (2015) 93–99 Contents lists available at ScienceDirect Drug and Alcohol Dependence journal homepage: www.elsevier.co...

488KB Sizes 1 Downloads 10 Views

Drug and Alcohol Dependence 154 (2015) 93–99

Contents lists available at ScienceDirect

Drug and Alcohol Dependence journal homepage: www.elsevier.com/locate/drugalcdep

Identification of smokers, drinkers and risky drinkers by general practitioners夽 Jakob Manthey a,∗ , Charlotte Probst a,b , Franz Hanschmidt c , Jürgen Rehm a,b,d,e,f a Institute of Clinical Psychology and Psychotherapy & Center of Clinical Epidemiology and Longitudinal Studies (CELOS), Technische Universität Dresden, Chemnitzer Str. 46, 01187 Dresden, Germany b Center for Addiction and Mental Health, 33 Russell Street, Toronto, ON M5S 2S1, Canada c University of Leipzig, Department of Psychosomatic Medicine and Psychotherapy, Semmelweisstr. 10, 04103 Leipzig, Germany d Addiction Policy, Dalla Lana School of Public Health, University of Toronto, 155 College Street, 6th floor, Toronto, ON M5T 3M7, Canada e Institute of Medical Science, University of Toronto, Faculty of Medicine, Medical Sciences Building, 1 King’s College Circle, Room 2374, Toronto, ON M5S 1A8, Canada f Department of Psychiatry, University of Toronto, 250 College Street, 8th floor, Toronto, ON M5T 1R8, Canada

a r t i c l e

i n f o

Article history: Received 30 March 2015 Received in revised form 1 June 2015 Accepted 8 June 2015 Available online 22 June 2015 Keywords: Tobacco Smoking Alcohol Binge drinking General practitioner Screening

a b s t r a c t Background: Identification of risky substance users by general practitioners (GPs) is important for providing brief interventions or to refer cases to specialized care, but detection rates of risky users are low, with alcohol users being identified less frequently than smokers. Methods: We compared GPs’ assessment and patient self-report concerning tobacco use, number of cigarettes smoked daily, alcohol use, alcohol use disorder, and different risky use definitions of 8476 primary care patients from six European countries. Further, we carried out a logistic regression predicting the GPs perception of the patients’ alcohol problems. Results: GPs identified 88.4% (95% confidence interval (CI): 87.1–89.6%;  = 0.84, 95% CI: 0.83–0.86) of all self-reported smokers but only 64.6% (95% CI: 63.2–65.9%;  = 0.35, 95% CI: 0.33–0.37) of all current drinkers, while they were unable to judge the drinking status of every ninth patient. The GPs’ estimation of number of cigarettes smoked daily was slightly lower than the self-report ( = 0.23 cigarettes/day, p < .001) but both measures were correlated with each other. Of all risky drinkers, defined as having alcohol-related problems or showing risky drinking patterns, 28.7% (95% CI: 25.9–31.4%;  = 0.34, 95% CI: 0.31–0.37) were perceived as having problems with alcohol by the GPs. Patients’ self-reported health and social consequences, as well as drinking patterns predicted the GPs’ perception of alcohol problems. Conclusions: GPs were more accurate in identifying smokers than drinkers. Concerning risky drinkers, GPs failed to diagnose a sizeable proportion but were able to detect other drinkers whom common recognition approaches had not recognized. © 2015 Elsevier Ireland Ltd. All rights reserved.

1. Introduction Smoking and risky alcohol use contribute considerably to burden of disease (Lim et al., 2012). As they are in theory completely preventable, they are important public health issues. In this context, general practitioners (GPs) were recommended to screen for

夽 Supplementary material can be found by accessing the online version of this paper at http://dx.doi.org/10.1016/j.drugalcdep.2015.06.017. ∗ Corresponding author. Tel.: +49 17699076559. E-mail addresses: [email protected] (J. Manthey), [email protected] (C. Probst), [email protected] (F. Hanschmidt), [email protected] (J. Rehm). http://dx.doi.org/10.1016/j.drugalcdep.2015.06.017 0376-8716/© 2015 Elsevier Ireland Ltd. All rights reserved.

patients at risk and to offer help (Brotons et al., 2012), e.g., by giving advice to stop smoking (Stead et al., 2008), by delivering brief interventions for risky drinkers (O’Donnell et al., 2014) or by referring patients to specialists. Although there are effective strategies for treating risky substance use in primary care settings (Ashenden et al., 1997; O’Donnell et al., 2014; Papadakis et al., 2010), the number of risky substance users identified by GPs was reported to be low (Brotons et al., 2012; Drummond et al., 2013). This is of special concern, since identification of risky users is the precursor to secondary and tertiary prevention strategies. 1.1. Definition of risky substance use With regard to smoking, dose–response relationships were found between number of cigarettes smoked daily and incidence

94

J. Manthey et al. / Drug and Alcohol Dependence 154 (2015) 93–99

of several diseases (most importantly lung cancer (Law et al., 1997) and ischemic stroke (Bhat et al., 2008)); thus the number of cigarettes smoked daily can be used as an indicator for risky tobacco use. Because smoking 1–4 cigarettes daily increases the risk of allcause mortality compared to non-smokers (Bjartveit and Tverdal, 2005), smoking any amount of cigarettes is considered risky. With regard to alcohol, two major approaches exist to determine users at risk. Within the first approach, alcohol-related consequences are considered symptoms for alcohol use disorders (AUDs; American Psychiatric Association, 2013; World Health Organization, 1993; for a critique: see Martin et al., 2014). The second approach focuses on patterns and amount of alcohol use, using mean daily intake of alcohol (chronic heavy use for longterm adverse effects; Rehm et al., 2013) and irregular heavy drinking occasions (acute heavy use for short-term adverse consequences/injuries; Rossow et al., 2013) as key measures to identify risky users, referred to below as chronic and acute heavy drinking, respectively. 1.2. Identification of risky substance users by GPs GPs were repeatedly shown to identify smokers accurately (Szatkowski et al., 2012; Wilson et al., 2000) but missed more than half of all patients with DSM-IV (American Psychiatric Association, 1994) AUD patterns as identified by structured interview techniques (Mitchell et al., 2012). Detection rate of heavy drinkers was similar in another study where 55% of all heavy drinkers (at least 21 units/week) were recognized by the GPs (Mant et al., 2000). The relatively low rates of GPs in detecting risky drinkers may be due to a general tendency to focus on tobacco rather than on alcohol when detecting users, obtaining health-related information and promoting health (Aira et al., 2004; Geirsson et al., 2005; McAvoy et al., 1999). Further, GPs face several barriers (e.g., lack of time, fear of straining the relationship) in consulting patients about their alcohol use (Drummond et al., 2013), and their conception of risky drinking was reported to be incoherent and potentially impacted by their own drinking habits (Arborelius and Thakker, 1995). In another study (Anderson et al., 2014), GPs were found to manage risky drinkers securely, yet were not committed to doing so, which was reflected in low numbers of actual patients managed for heavy drinking. 1.3. Measuring GPs ability to identify risky users Research has predominantly regarded patient self-report as the benchmark for the calculation of GP detection rates. However, concerns about the eligibility of patient self-report as a benchmark have been raised by survey studies: Compared to bio-medical markers, prevalence rates were underestimated, probably due to social desirability and diverging concepts of ‘drinker’ and ‘smoker’ (Del Boca and Darkes, 2003; Gorber et al., 2009). These factors may play a less important role in GPs’ assessment of the patients’ risky substance use if based on a trusting relationship and professional training. However, GPs cannot be regarded as benchmarks either because their judgments are limited by degree of training, interest in the patients’ substance use and restricted availability of data sources (relying on the patients’ disclosure). Consequently, it may be wise to just look at the association of different data sources without considering one or the other as a benchmark. A systematic analysis of patient and GP agreement on licit substance use – especially concerning risky use – is lacking. This study aims to present new figures of GP detection of tobacco and risky alcohol users, and to elaborate on potential reasons for lower detection rates of risky drinkers, taking into account different definitions of risky use.

2. Material and methods 2.1. Design and data collection We obtained data from the ‘Alcohol Dependence in Primary Care’ study (Manthey et al., in press; Rehm et al., 2015). This study collected data from 358 GPs and their patients across six European countries (Germany, Hungary, Italy, Latvia, Poland, and Spain) between January, 2013 and January, 2014. While regionally representative GP samples were drawn in Germany, Italy, Poland and Spain, GP samples for Hungary and Latvia were nationally representative (GP refusal rate = 56.4%). GPs were sampled from complete registers in all countries except Italy where GPs were recruited from a local GP association and from a sample of GPs who participated in previous studies. GPs were first asked to fill in brief questionnaires, with a focus on alcohol use and related problems, for all patients aged 18–64 within a given time period. A subsample of all patients assessed by the GP was then drawn to be interviewed: In Hungary and Spain, all GP assessed patients were interviewed. In all other countries, subsamples were determined based on GP assessment, aiming at sampling all patients with alcohol problems, in addition to stratified (Italy) or random subsamples (Germany, Latvia, and Poland) of the remaining patients (overall patient refusal rate = 17.8%; N = 8476). The subsample was interviewed using the fully standardized Composite International Diagnostic Interview (CIDI; Wittchen, 1994) in order to assess alcohol use, alcohol problems and AUD diagnoses, in addition to other questions, including two questions on current smoking. Details on other instruments used and sampling strategies can be found elsewhere (Manthey et al., in press; Rehm et al., 2015). Because the sampling scheme involved oversampling patients with alcohol problems for interviews, all analyses were adjusted for by using probability weights using Stata 12.0 (StataCorp, 2011). The GP questionnaire assessed the patients’ smoking (“Smoking: Yes (number of cigarettes/day)/Not in years/Never”), alcohol use (“Does the patient drink alcohol? Yes/No/Can’t judge”), alcohol problems (“Based on your judgment, does the patient have a problem with alcohol? Yes/No”) and current AUD diagnoses (“Do you believe your patient would be diagnosed with alcohol dependence/abuse? Yes/No”). Uncertain answers (“Can’t judge”) and missing values were analyzed separately. The patient interview assessed the patients’ current smoking behavior (“Do you currently smoke regularly? Yes/No”), number of cigarettes smoked (“How many cigarettes a day?”), alcohol use (CIDI: “In the past 12 months did you have at least 12 drinks of any kind of alcoholic beverage? Yes/No”), DSM-5 (American Psychiatric Association, 2013) AUD diagnosis, alcohol problems (meeting at least one AUD symptom presented during CIDI, see also Supplementary Table1 for full list of symptoms), chronic heavy drinking (mean daily alcohol intake in grams/day), acute heavy drinking (number of irregular heavy drinking occasions in the past 30 days, defined as use of 60 g of pure alcohol on one occasion). The last measure was not assessed in Germany, thus excluding 1356 patients from all analyses involving this measure. We were able to make the following comparisons of GP assessment vs. patient self-report based on the variables assessed: current smoking, number of cigarettes currently smoked daily, current/past year drinking, current/past year AUD, and current/past year alcohol problems. For the last comparison, we contrasted the GPs’ perception of alcohol problems with four different definitions of risky use,

1 Supplementary material can be found by accessing the online version of this paper at http://dx.doi.org/10.1016/j.drugalcdep.2015.06.017.

J. Manthey et al. / Drug and Alcohol Dependence 154 (2015) 93–99

namely (1) alcohol-related problems (see Web Table for definition, 0–20 possible problems), (2) chronic heavy drinking (mean alcohol intake in g/day) (3) acute heavy drinking (number of irregular heavy drinking occasions in past month) and (4) alcohol-related problems or chronic heavy drinking or acute heavy drinking.

Table 1 Contingency table of GP judgment and patient-derived information on current smoking and number of cigarettes per day. Patient self-report Current smoking

2.2. Statistical analysis The agreement between dichotomous measures of two different sources can be determined asymmetrically and symmetrically (Agbaje et al., 2012), where asymmetrical measures (sensitivity, specificity, positive and negative predictive value) compare one judgment against a benchmark judgment while symmetrical measures (kappa statistics (␬)) treat both data sources equally. We calculated sensitivity and specificity using the patient self-report as benchmark and positive and negative predictive value using the GP judgment as benchmark (Florkowski, 2008). Both data sources were treated as two unique raters (patient self-report vs. GP judgment) in order to use  statistics, which compares the observed agreement against the expected agreement by chance (Landis and Koch, 1977).  values range between 0 (not different from pure chance) and 1 (perfect agreement). For number of cigarettes smoked per day, we looked at mean differences between GP assessment and patient self-report. In addition, a negative binomial regression between both variables served as proxy for degree of agreement (Sribney, 2005). Continuously scaled measures of risky alcohol use (number of problems, chronic and acute heavy drinking) were recoded into binary variables. We considered at least one patient self-reported alcohol problem, mean daily alcohol intake above 40 g/day for female and above 60 g/day for male (see World Health Organization, 2000) and at least five occasions with at least 60 g alcohol intake in the past 30 days (i.e., at least weekly, similar definition as in Rossow et al., 2013) as best cutoffs, in line with other studies. One combined variable, comprising all cases meeting one or more of the cutoffs, was compared with the GP perception as well. Sensitivity analyses on missing values were carried out. First, to find out whether missing values in one variable were at random, we looked at the distribution of the same aspect derived from the counterpart data set, i.e., from GP assessment for missing patient self-reported data and vice versa. For example, if GPs did not provide information on the patients’ smoking status, we compared the self-reported smoking status of patients with provided GP assessment with the self-reported smoking status of patients with missing GP assessment. In order to test the difference between both groups, we ran logistic regressions for dichotomous measures and linear regressions for continuously scaled variables, with binary variables as predictor contrasting presence and absence of missing values. Second, the GPs were allowed to explicitly state their uncertainty of the patients’ drinking status in the questionnaire (“I can’t judge”). This option was not given in any other item and was analyzed separately, in addition and analogous to other missing values. Lastly, two logistic regression models – including 20 selfreported alcohol problems, mean daily alcohol intake (chronic heavy drinking), and number of irregular heavy drinking occasions (acute heavy drinking) as independent variables – predicted the GPs’ perception of their patients as having alcohol problems (dependent variable), while controlling for age and sex (see Supplementary Table2 for a complete list of variables). The first model included all measures and served as reference for the second model that employed a backwards selection process (Bursac et al., 2008).

2 Supplementary material can be found by accessing the online version of this paper at http://dx.doi.org/10.1016/j.drugalcdep.2015.06.017.

95

GP judgment No (weighted %) Yes (weighted %) Total Sensitivity (95% CI) Specificity (95% CI) Positive predictive value (95% CI) Negative predictive value (95% CI) Kappa value (95% CI)

No

Yes

Total

5233 (64.9) 259 (3.1) 5492 88.4% (87.1–89.6%) 95.4% (94.9–96.0%) 90.1% (89.0–91.3%)

300 (3.7) 2352 (28.3) 2652

5533 2611 8144

94.6% (94.0–95.2%) 0.84 (0.83–0.86) Number of cigarettes smoked per day (N = 7897)

Patient self-report: mean GP judgment: mean Correlation coefficientb

4.27 4.04a .18 (p < .0001)

Note. GP, general practitioner. 95% CI, 95% confidence interval, based on standard error. Sensitivity, specificity, likelihood ratios and percentage agreement are based on weighted percentages. a GP judgment mean statistically different from patient self-reported mean. b Based on negative binomial regression.

The full model was screened for factors with p > .025 (smaller significance level employed to adjust for multiple testing). Starting with variables with greatest p-values, variables were excluded stepwise if their exclusion did not result in changes in coefficients greater than 20% of retained variables, which indicates a confounding factor. 3. Results Patient characteristics are described elsewhere (Manthey et al., in press; Rehm et al., 2015). The sample of 8476 patients was predominantly female (60.0%, 95% confidence interval (CI): 59.0–61.1%), employed (62.2%, 95% CI: 61.1–63.2%) and on average 44.4 years (95% CI: 44.1–44.7 years) old. 3.1. Smoking measures Table 1 presents contingency tables and agreement measures of GP judgment and patient self-report on current smoking. Current smoking prevalence was similar if determined by self-report (32.0%) or by GP assessment (31.4%). GPs correctly identified nine out of ten patients reporting current smoking. The other way round, nine out of ten patients agreed with the GPs’ observation of current smoking. According to  statistics, high overall agreement of both data sources was found on current smoking ( = 0.84). The number of cigarettes smoked per day was estimated at similar but statistically different levels by both data sources. Both measures were also significantly correlated with each other, according to negative binomial regression. For current smoking, GPs failed to provide information on 328 cases (4%) while only 4 patients did not provide information on their current smoking behavior. For the number of cigarettes smoked, we encountered 567 missing values (6.8%) from the GP and 24 missing values (0.3%) from the patient interview. According to sensitivity analyses, cases without valid GP assessment reported a significantly higher mean of cigarettes smoked daily (6.5) than cases with valid GP assessment (4.3, p < .001), meaning that GPs have failed to assess

96

J. Manthey et al. / Drug and Alcohol Dependence 154 (2015) 93–99

Table 2 Contingency tables of GP judgment and patient-derived information on past year alcohol use, DSM-5 AUD and alcohol problems. Patient self-report Past year alcohol use

GP judgment No (weighted %) Yes (weighted %) Total Sensitivity (95% CI) Specificity (95% CI) Positive predictive value (95% CI) Negative predictive value (95% CI) Kappa value (95% CI)

Past year DSM-5 AUD

No

Yes

Total

No

Yes

Total

1801 (25.8) 715 (9.0) 2516 64.6% (63.2–65.9%) 74.1% (72.4–75.8%) 82.4% (81.2–83.5%) 52.8% (51.1–54.4%) 0.35 (0.33–0.37)

1659 (23.1) 3343 (42.1) 5002

3460 4058 7518

7507 (89.9) 435 (4.3) 7942 32.4% (28.5–36.4%) 95.5% (95.0–95.9%) 30.9% (27.2–34.6%) 95.8% (95.3–96.2%) 0.29 (0.25–0.32)

337 (4.0) 194 (1.9) 531

7844 629 8473

Patient self-report Chronic heavy alcohol use a

At least one alcohol problem

GP judgment No (weighted %) Yes (weighted %) Total Sensitivity (95% CI) Specificity (95% CI) Positive predictive value (95% CI) Negative predictive value (95% CI) Kappa value (95% CI)

No

Yes

Total

No

Yes

Total

7034 (84.3) 425 (4.3) 7459 28.9% (26.2–31.7%) 95.1% (94.7–95.6%) 43.4% (39.7–47.0%) 91.3% (90.6–91.9%) 0.30 (0.27–0.33)

683 (8.1) 331 (3.3) 1014

7717 756 8473

7517 (90.2) 603 (6.0) 8120 42.1% (36.9–47.4%) 93.8% (93.3–94.3%) 21.1% (18.1–24.1%) 97.6% (97.3–98.0%) 0.24 (0.20–0.27)

191 (2.2) 153 (1.6) 344

7708 756 8464

Patient self-report Acute heavy alcohol use b

GP judgment No (weighted %) Yes (weighted %) Total Sensitivity (95% CI) Specificity (95% CI) Positive predictive value (95% CI) Negative predictive value (95% CI) Kappa value (95% CI)

Any risky drinker c

No

Yes

Total

No

Yes

Total

6322 (91.7) 523 (6.0) 6845 43.7% (36.3–51.1%) 93.8% (93.3–94.4%) 14.1% (11.3–16.9%) 98.6% (98.3–98.9%) 0.20 (0.16–0.24)

86 (1.3) 93 (1.0) 179

6408 616 7024

5844 (83.7) 298 (3.4) 6142 28.7% (25.9–31.4%) 96.1% (95.7–96.6%) 52.4% (48.4–56.5%) 90.1% (89.3–90.8%) 0.34 (0.31–0.37)

649 (9.2) 329 (3.7) 978

6493 627 7120

Note. GP, general practitioner. DSM-5, Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition. AUD, alcohol use disorder. 95% CI, 95% confidence interval, based on standard error. Sensitivity and specificity are based on weighted percentages. ‘GP judgment’ refers to current drinking and alcohol dependence/abuse in the first two contingency tables (first row). For the remaining contingency tables (second and third row), GPs’ perception of the patients’ alcohol problems was contrasted with four different operationalizations of risky alcohol use. a At least 40 g and 60 g daily intake of pure alcohol for female and male drinkers, respectively. b At least five occasions of at least 60 g pure alcohol intake per occasion; was not assessed in Germany. c Collapsed category: at least one self-reported alcohol problem, chronic or acute heavy alcohol use.

more frequent smokers rather than non-smokers. No other differences were observed between valid and invalid responses on smoking measures. 3.2. Alcohol measures Contingency tables and agreement measures of different alcohol assessments can be found in Table 2. Overall, GPs identified six out of ten self-classified drinkers, every third AUD case and about three to four out of ten risky drinkers. In terms of recognition of risky drinkers, GPs’ perception of problematic alcohol use corresponded slightly more to chronic and acute heavy drinking than to patients with at least one alcohol related problem. Conversely, of those patients being perceived to have alcohol problems by the GP, less than half reported at least one problem themselves and even fewer reached the threshold of chronic and acute heavy drinking. When collapsing all patients meeting at least one risky use threshold into a single category, GPs recognized about three out of ten cases as having alcohol problems. In total, GPs perceived

12.9% of all patients to have alcohol problems while only 7.1% of all patients met at least one respective threshold. Slight to fair overall agreement was indicated by  statistics for all six comparisons. GPs expressed uncertainty about the patients’ current drinking status (“I can’t judge”) in 926 cases (11.3%), in addition to 22 (0.3%) missing GP assessments on this aspect. According to sensitivity analyses, patients with uncertain GP judgment were less likely to classify themselves as drinkers than those with certain GP judgment (59.9% vs. 65.2%, p = .002). For the remaining alcohol measures, no missing values were encountered in GP assessed variables, and less than 0.1% missing values were found in patient self-reported variables, except for number of irregular heavy drinking occasions (96 missing values, 1.4%). No significant differences were observed between non-missing and missing responses. 3.3. Factors predicting GPs’ perception of alcohol problems In order to determine factors predicting the GPs perception of patients with alcohol problems, two logistic regressions were

J. Manthey et al. / Drug and Alcohol Dependence 154 (2015) 93–99

carried out (see supplementary Table3 ). The full model (Model 1) contained all variables (20 alcohol problems, acute and chronic heavy drinking) and identified three DSM-IV alcohol dependence symptoms (health problems when drank less, social objections, occupational problems), the ICD-10 harmful use criterion, chronic and acute heavy drinking measures as well as age and sex as significant predictors for the GP judgment. The reduced Model 2 contained the same significant predictors from Model 1, with ‘health problems due to alcohol’ found to be the strongest predictor (OR = 4.73), followed by ‘social objections’ (OR = 3.22) and mean daily alcohol intake (OR = 1.01). Both models predicted the GPs judgment precisely according to Goodness-of-fit tests, and explain about one fourth of the observed variance. 4. Discussion This paper deals with primary care patients’ licit substance use behaviors comparing two data sets derived by GP questionnaire and patient self-report. We identified 8476 cases with valid GP questionnaire and patient interview from six European countries and contrasted both assessments of different smoking and alcohol measures, focusing on risky alcohol use definitions. 4.1. Main findings Our findings suggest that GPs’ knowledge of their patients’ tobacco use is overall very accurate, including the number of cigarettes smoked per day. However, GPs have more difficulties in identifying self-classified drinkers, with 35% of current drinkers being left out. Additionally, the drinking status of about every ninth patient could not be judged by the GPs. For AUDs and risky drinking, lower agreements were found. GPs identified one out of three AUD cases as determined by patient selfreport and recognized three out of ten patients with self-reported alcohol problems and four out of ten patients with self-reported chronic or heavy drinking patterns. The GPs’ perception of the patient as having alcohol problems was predicted by somatic, social and occupational issues as well as patterns of heavy drinking but not by psychological, legal and financial problems. 4.2. Integration of main findings First, precise GP assessments of patients’ smoking behavior corroborate other studies (e.g., Szatkowski et al., 2012; Wilson et al., 2000) that found high agreement for smoking but contrast the study by Mant et al. (2000) who reported similar agreement measures for both smoking and drinking. Our figures indicated that GPs are less precise in ascertaining the patients’ use of alcohol than of tobacco, which is highly consistent with other studies reporting that GPs prefer assessing smoking rather than drinking (Aira et al., 2004; Geirsson et al., 2005; McAvoy et al., 1999) due to barriers they face (Arborelius and Thakker, 1995; Geirsson et al., 2005) or their lack of commitment (Anderson et al., 2014). Second, the observed gap in recognizing users also applies to different risky use patterns: While the GPs’ estimate and patient self-reported number of cigarettes smoked daily was very similar, much lower degrees of agreement could be observed on all risky drinking measures. It must be noted that the risk of smoking is much easier to assess than risky drinking. For tobacco, the number of cigarettes smoked daily is a reliable single measure to assess the smoker’s risk while for risky drinking it involves the number of

3 Supplementary material can be found by accessing the online version of this paper at http://dx.doi.org/10.1016/j.drugalcdep.2015.06.017.

97

drinks, drinking patterns and context as well as problems related to alcohol. Additionally, the GPs’ preference for assessing smoking rather than alcohol use (Aira et al., 2004; Geirsson et al., 2005), including barriers (Arborelius and Thakker, 1995; Geirsson et al., 2005) and stigmatization (Schomerus et al., 2011), may also explain lower degrees of agreement compared to smoking. 4.3. Explaining GPs’ perception of alcohol problems If GPs’ assessments were regarded as diagnostic tests, they would be discarded in the light of our findings because of very little agreement with the benchmark, thus missing out on many risky drinkers (false negatives) and incorrectly labeling many patients as problematic drinkers (false positives). However, a more symmetrical view of both data sources may help to understand the high rates of false positives and negatives. While we determined risky drinkers via alcohol-related problems (derived from DSM-IV/5), drinking levels and patterns, GPs usually refer to physical examination and ICD-10 criteria, which include a diagnosis related only to health damaged by psychoactive substance use (harmful use). Thus, they may miss (false negative) risky drinkers who have not yet developed physical problems. However, GPs’ approach may also enable them to identify risky drinkers that would otherwise remain undetected by clinical interviews (false positives). Common problems caused or exacerbated by alcohol use (such as certain types of cancer, hypertension, and ischemic heart disease) are not explicitly listed in the CIDI and may only be recognized as alcohol problems by the GP. Results from the logistic regression support this assumption: The strongest predictor was related to health problems due to alcohol, but only a relatively low proportion of variance was explained by both models. Although GPs also seem to be referring to criteria not part of the ICD-10 (e.g., drinking patterns), they supposedly use other criteria to determine alcohol problems, relying on somatic indicators. The heavy stigma attached to AUDs (Schomerus et al., 2011) could be another reason for such a high rate of false positives. Patients may find it difficult to admit their alcohol problems in a standardized interview setting rather than in a private conversation with their GP. A trustful, long-lasting relationship with the GP might even facilitate the disclosure of these problems. Consequently, GPs seem to be able to detect somatic alcoholrelated problems but have more problems with detecting other consequences. Following our considerations, risky drinking levels and patterns may remain undetected by GPs if not accompanied by physical complications. Further, we regard both data sources as equally dependable, and we indicated reasons why so-called false positives can be considered problematic users as well (see also Rehm et al., 2015). 4.4. Conclusions Our results suggest that GPs’ knowledge of the patients’ legal substance use is much more precise for smoking than for drinking. We showed that certain risky drinkers were identified by GPs, mostly via health indicators. However, other groups of risky drinkers, such as patients with predominant psychological problems, may be overlooked by the current diagnostic routine of the GPs. 4.5. Implications Based on the findings of this study, GPs should be supported in their efforts to screen for risky drinkers. However, a recent study found postgraduate training in managing alcohol problems in primary care to be low but positively related to the number of heavy drinkers that GPs managed (Anderson et al., 2014). Expansion of

98

J. Manthey et al. / Drug and Alcohol Dependence 154 (2015) 93–99

this training should not only include increasing perceived effectiveness (Geirsson et al., 2005) and reducing fear (Arborelius and Thakker, 1995), but – considering our findings – it should also elaborate on different dimensions of risky alcohol use, including the role of mental problems and risky drinking patterns. An instrument to briefly but effectively screen for risky drinkers is the AUDIT-C. Making the assessment of drinking levels and patterns part of the daily routine (e.g., with AUDIT-C (Bush et al., 1998)) could contribute to better screening and brief interventions for risky drinkers, after all. Yet our findings also suggest that common approaches of identifying risky drinkers might fail to detect certain subgroups. More studies should compare patient self-report and GP assessment of licit substance use to bio-medical markers as gold standard. They should also examine the GPs’ approach of assessing alcohol problems and the extent to which they consider drinking levels and patterns. Finally, it should be investigated whether training can bring about higher detection rates of risky drinkers in primary care settings. 4.6. Study limitations Our study cannot claim representativeness for all of Europe but only for selected regions. Further, we encountered relatively high refusal rates among GPs, which should be considered when generalizing our results. We want to acknowledge that our findings are limited by means of patient self-report and GP assessment. It should also be noted that our questionnaire allowed GPs to express their uncertainty concerning the patients’ alcohol use but not concerning other aspects. Substantially lower missing rates in all other items were observed and may be attributable to the lack of such a category. Role of funding source The study was financially supported by an investigator initiated grant to the last author and the GWT-TUD (Gesellschaft für Wissens- und Technologietransfer der TU Dresden mbH – company with limited liabilities for transferring knowledge and technology of the Dresden University of Technology) by Lundbeck (grant number 414209). The study sponsor has no role in study design; collection, analysis, and interpretation of data. The study sponsor also had no role in writing of the report; and the decision to submit the paper for publication. The corresponding author confirms that the authors had full access to the data in the study at all times, and had final responsibility for the decision to submit for publication. The corresponding author hereby states that no author has been reimbursed for writing this manuscript. Contributors JR conceptualized the APC study and served as PI. JM conceptualized the data analyses and conducted the quantitative and qualitative analyses. JM wrote a first draft of the paper, and all authors contributed to it. We confirm that the manuscript has been read and approved by all named authors and that there are no other persons who satisfied the criteria for authorship but are not listed. We further confirm that the order of authors listed in the manuscript has been approved by all of us. Conflict of interest JM, CP, and FH have no potential conflict of interest stated. JR reports grants from GWT-TUD during the conduct of the study and grants, personal fees and being board member (Nalmefene) for Lundbeck outside the submitted work.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.drugalcdep.2015. 06.017 References Agbaje, J., Mutsvari, T., Lesaffre, E., Declerck, D., 2012. Measurement, analysis and interpretation of examiner reliability in caries experience surveys: some methodological thoughts. Clin. Oral Invest. 16, 117–127. Aira, M., Kauhanen, J., Larivaara, P., Rautio, P., 2004. Differences in brief interventions on excessive drinking and smoking by primary care physicians: qualitative study. Prev. Med. 38, 473–478. American Psychiatric Association, 1994. Diagnostic and Statistical Manual of Mental Disorders, fourth ed. American Psychiatric Association, Washington, DC. American Psychiatric Association, 2013. Diagnostic and Statistical Manual of Mental Disorders, fifth ed. American Psychiatric Association, Washington, DC. Anderson, P., Wojnar, M., Jakubczyk, A., Gual, A., Reynolds, J., Segura, L., Sovinova, H., Csemy, L., Kaner, E., Newbury-Birch, D., Fornasin, A., Struzzo, P., Ronda, G., van Steenkiste, B., Keurhorst, M., Laurant, M., Ribeiro, C., do Rosário, F., Alves, I., Scafato, E., Gandin, C., Kolsek, M., 2014. Managing alcohol problems in general practice in Europe: results from the European ODHIN Survey of General Practitioners. Alcohol Alcohol 49, 531–539. Arborelius, E., Thakker, K.D., 1995. Why is it so difficult for general practitioners to discuss alcohol with patients? Fam. Pract. 12, 419–422. Ashenden, R., Silagy, C., Weller, D., 1997. A systematic review of the effectiveness of promoting lifestyle change in general practice. Fam. Pract. 14, 160–176. Bhat, V.M., Cole, J.W., Sorkin, J.D., Wozniak, M.A., Malarcher, A.M., Giles, W.H., Stern, B.J., Kittner, S.J., 2008. Dose–response relationship between cigarette smoking and risk of ischemic stroke in young women. Stroke 39, 2439–2443. Bjartveit, K., Tverdal, A., 2005. Health consequences of smoking 1–4 cigarettes per day. Tob. Control 14, 315–320. Brotons, C., Bulc, M., Sammut, M.R., Sheehan, M., Manuel da Silva Martins, C., Bjorkelund, C., Drenthen, A.J., Duhot, D., Gorpelioglui, S., Jurgova, E., Keinanen-Kiukkanniemi, S., Kotanyi, P., Markou, V., Moral, I., Mortsiefer, A., Pas, L., Pichler, I., Sghedoni, D., Tataradze, R., Thireos, E., Valius, L., Vuchak, J., Collins, C., Cornelis, E., Ciurana, R., Kloppe, P., Mierzecki, A., Nadaraia, K., Godycki-Cwirko, M., 2012. Attitudes toward preventive services and lifestyle: the views of primary care patients in Europe. The EUROPREVIEW patient study. Fam. Pract. 29 (Suppl. 1), i168–i176. Bursac, Z., Gauss, C.H., Williams, D.K., Hosmer, D.W., 2008. Purposeful selection of variables in logistic regression. Source Code Biol. Med. 3, 17. Bush, K., Kivlahan, D.R., McDonell, M.B., Fihn, S.D., Bradley, K.A., Ambulatory Care Quality Improvement, P., 1998. The AUDIT alcohol consumption questions (AUDIT-C) – an effective brief screening test for problem drinking. Arch. Intern. Med. 158, 1789–1795. Del Boca, F.K., Darkes, J., 2003. The validity of self-reports of alcohol consumption: state of the science and challenges for research. Addiction 98, 1–12. Drummond, C., Wolstenholme, A., Deluca, P., Davey, Z., Donoghue, K., Elzerbi, C., Gual, A., Robles, N., Goos, C., Strizek, J., Godfrey, C., Mann, K., Zois, E., Hoffman, S., Gmel, G., Kuendig, H., Scafato, E., Gandin, C., Reynolds, J., Segura, L., Colom, J., Baena, B., Coulton, S., Kaner, E., 2013. Alcohol interventions and treatments in Europe. In: Anderson, P., Braddick, F., Reynolds, J., Gual, A. (Eds.), Alcohol Policy in Europe: Evidence from AMPHORA. AMPHORA, Barcelona, pp. 65–80. Florkowski, C.M., 2008. Sensitivity, specificity, receiver-operating characteristic (roc) curves and likelihood ratios: communicating the performance of diagnostic tests. Clin. Biochem. Rev. 29, S83–S87. Geirsson, M., Bendtsen, P., Spak, F., 2005. Attitudes of Swedish general practitioners and nurses to working with lifestyle change, with special reference to alcohol consumption. Alcohol Alcohol 40, 388–393. Gorber, S.C., Schofield-Hurwitz, S., Hardt, J., Levasseur, G., Tremblay, M., 2009. The accuracy of self-reported smoking: a systematic review of the relationship between self-reported and cotinine-assessed smoking status. Nicotine Tob. Res. 11, 12–24. Landis, J.R., Koch, G.G., 1977. The measurement of observer agreement for categorical data. Biometrics 33, 159–174. Law, M.R., Morris, J.K., Watt, H.C., Wald, N.J., 1997. The dose–response relationship between cigarette consumption, biochemical markers and risk of lung cancer. Br. J. Cancer 75, 1690–1693. Lim, S.S., Vos, T., Flaxman, A.D., Danaei, G., Shibuya, K., Adair-Rohani, H., Amann, M., Anderson, H.R., Andrews, K.G., Aryee, M., Atkinson, C., Bacchus, L.J., Bahalim, A.N., Balakrishnan, K., Balmes, J., Barker-Collo, S., Baxter, A., Bell, M.L., Blore, J.D., Blyth, F., Bonner, C., Borges, G., Bourne, R., Boussinesq, M., Brauer, M., Brooks, P., Bruce, N.G., Brunekreef, B., Bryan-Hancock, C., Bucello, C., Buchbinder, R., Bull, F., Burnett, R.T., Byers, T.E., Calabria, B., Carapetis, J., Carnahan, E., Chafe, Z., Charlson, F., Chen, H., Chen, J.S., Cheng, A.T., Child, J.C., Cohen, A., Colson, K.E., Cowie, B.C., Darby, S., Darling, S., Davis, A., Degenhardt, L., Dentener, F., Des Jarlais, D.C., Devries, K., Dherani, M., Ding, E.L., Dorsey, E.R., Driscoll, T., Edmond, K., Ali, S.E., Engell, R.E., Erwin, P.J., Fahimi, S., Falder, G., Farzadfar, F., Ferrari, A., Finucane, M.M., Flaxman, S., Fowkes, F.G., Freedman, G., Freeman, M.K., Gakidou, E., Ghosh, S., Giovannucci, E., Gmel, G., Graham, K., Grainger, R., Grant, B., Gunnell, D., Gutierrez, H.R., Hall, W., Hoek, H.W.,

J. Manthey et al. / Drug and Alcohol Dependence 154 (2015) 93–99 Hogan, A., Hosgood 3rd, H.D., Hoy, D., Hu, H., Hubbell, B.J., Hutchings, S.J., Ibeanusi, S.E., Jacklyn, G.L., Jasrasaria, R., Jonas, J.B., Kan, H., Kanis, J.A., Kassebaum, N., Kawakami, N., Khang, Y.H., Khatibzadeh, S., Khoo, J.P., Kok, C., Laden, F., Lalloo, R., Lan, Q., Lathlean, T., Leasher, J.L., Leigh, J., Li, Y., Lin, J.K., Lipshultz, S.E., London, S., Lozano, R., Lu, Y., Mak, J., Malekzadeh, R., Mallinger, L., Marcenes, W., March, L., Marks, R., Martin, R., McGale, P., McGrath, J., Mehta, S., Mensah, G.A., Merriman, T.R., Micha, R., Michaud, C., Mishra, V., Mohd Hanafiah, K., Mokdad, A.A., Morawska, L., Mozaffarian, D., Murphy, T., Naghavi, M., Neal, B., Nelson, P.K., Nolla, J.M., Norman, R., Olives, C., Omer, S.B., Orchard, J., Osborne, R., Ostro, B., Page, A., Pandey, K.D., Parry, C.D., Passmore, E., Patra, J., Pearce, N., Pelizzari, P.M., Petzold, M., Phillips, M.R., Pope, D., Pope 3rd, C.A., Powles, J., Rao, M., Razavi, H., Rehfuess, E.A., Rehm, J.T., Ritz, B., Rivara, F.P., Roberts, T., Robinson, C., Rodriguez-Portales, J.A., Romieu, I., Room, R., Rosenfeld, L.C., Roy, A., Rushton, L., Salomon, J.A., Sampson, U., Sanchez-Riera, L., Sanman, E., Sapkota, A., Seedat, S., Shi, P., Shield, K., Shivakoti, R., Singh, G.M., Sleet, D.A., Smith, E., Smith, K.R., Stapelberg, N.J., Steenland, K., Stockl, H., Stovner, L.J., Straif, K., Straney, L., Thurston, G.D., Tran, J.H., Van Dingenen, R., van Donkelaar, A., Veerman, J.L., Vijayakumar, L., Weintraub, R., Weissman, M.M., White, R.A., Whiteford, H., Wiersma, S.T., Wilkinson, J.D., Williams, H.C., Williams, W., Wilson, N., Woolf, A.D., Yip, P., Zielinski, J.M., Lopez, A.D., Murray, C.J., Ezzati, M., AlMazroa, M.A., Memish, Z.A., 2012. A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet 380, 2224–2260. Mant, J., Murphy, M., Rose, P., Vessey, M., 2000. The accuracy of general practitioner records of smoking and alcohol use: comparison with patient questionnaires. J. Public Health Med. 22, 198–201. Manthey, J., Gual, A., Jakubczyk, A., Pieper, L., Probst, C., Struzzo, P., Trapencieris, M., Wojnar, M., Rehm, J., in press. Alcohol use disorders in Europe: a comparison of general population and primary health care prevalence rates. J. Subst. Use. Martin, C.S., Langenbucher, J.W., Chung, T., Sher, K.J., 2014. Truth or consequences in the diagnosis of substance use disorders. Addiction 109, 1773–1778. McAvoy, B.R., Kaner, E.F., Lock, C.A., Heather, N., Gilvarry, E., 1999. Our Healthier Nation: are general practitioners willing and able to deliver? A survey of attitudes to and involvement in health promotion and lifestyle counselling. Br. J. Gen. Pract. 49, 187–190. Mitchell, A.J., Meader, N., Bird, V., Rizzo, M., 2012. Clinical recognition and recording of alcohol disorders by clinicians in primary and secondary care: meta-analysis. Br. J. Psychiatry 201, 93–100. O’Donnell, A., Anderson, P., Newbury-Birch, D., Schulte, B., Schmidt, C., Reimer, J., Kaner, E., 2014. The impact of brief alcohol interventions in primary healthcare: a systematic review of reviews. Alcohol Alcohol. 49, 66–78.

99

Papadakis, S., McDonald, P., Mullen, K.A., Reid, R., Skulsky, K., Pipe, A., 2010. Strategies to increase the delivery of smoking cessation treatments in primary care settings: a systematic review and meta-analysis. Prev. Med. 51, 199–213. Rehm, J., Allamani, A., Elekes, Z., Jakubczyk, A., Landsmane, I., Manthey, J., ˜ J., Pieper, L., Probst, C., Snikere, S., Struzzo, P., Vedova, R.D., Moreno-Espana, Voller, F., Wittchen, H.-U., Gual, A., Wojnar, M., 2015. General practitioners recognizing alcohol dependence: a large cross-sectional study in 6 European countries. Ann. Fam. Med. 13, 28–32. Rehm, J., Marmet, S., Anderson, P., Gual, A., Kraus, L., Nutt, D.J., Room, R., Samokhvalov, A.V., Scafato, E., Trapencieris, M., Wiers, R.W., Gmel, G., 2013. Defining substance use disorders: do we really need more than heavy use? Alcohol Alcohol 48, 633–640. Rossow, I., Bogstrand, S., Ekeberg, O., Normann, P., 2013. Associations between heavy episodic drinking and alcohol related injuries: a case control study. BMC Public Health 13, 1076. Schomerus, G., Lucht, M., Holzinger, A., Matschinger, H., Carta, M.G., Angermeyer, M.C., 2011. The stigma of alcohol dependence compared with other mental disorders: a review of population studies. Alcohol Alcohol 46, 105–112. Sribney, B., 2005. How can I estimate Correlations and their Level of Significance with Survey Data?, http://www.stata.com/support/faqs/statistics/estimatecorrelations-with-survey-data/ (accessed on 07.01.15). StataCorp, 2011. Stata Statistical Software: Release 12. StataCorp LP, College Station, TX. Stead, L.F., Bergson, G., Lancester, T., 2008. Physician advice for smoking cessation. Cochrane Database Syst. Rev., CD000165. Szatkowski, L., Lewis, S., McNeill, A., Huang, Y., Coleman, T., 2012. Can data from primary care medical records be used to monitor national smoking prevalence? J. Epidemiol. Commun. Health 66, 791–795. Wilson, A., Manku-Scott, T., Shepherd, D., Jones, B., 2000. A comparison of individual and population smoking data from a postal survey and general practice records. Br. J. Gen. Pract. 50, 465–468. Wittchen, H.U., 1994. Reliability and validity studies of the WHO – Composite International Diagnostic Interview (CIDI): a critical review. J. Psychiatric Res. 28, 57–84. World Health Organization, 1993. The ICD-10 Classification of Mental and Behavioural Disorders: Diagnostic Criteria for Research. World Health Organization, Geneva. World Health Organization, 2000. International Guide For Monitoring Alcohol Consumption And Related Harm. World Health Organization, Geneva, Switzerland.