Data source

A retrospective study was performed using data, provided by Statistics Netherlands. We selected data about highly educated Dutch citizens, physician registrations, clinical SUD diagnoses, psychiatric and somatic comorbidity, general functioning, alcohol consumption patterns, and sociodemographic characteristics. These data were available from five different registers:

1)

Demographics register containing demographics (gender, year of birth, country of birth, educational level, and educational direction) of all legally residing citizens of The Netherlands from 2011 up to and including 2019 (Statistics Netherlands [30, 32]. Statistics Netherlands derives these data from the municipal population registers, educational level registers, and the Labor Force Survey (a rotating panel that is surveyed every quarter).

2)

Individual Healthcare Professions register containing data from the Central Information Point for Healthcare Professions [31]. This register includes dates of registration and deregistration, medical profession, and medical specialty.

3)

Mental healthcare claims register containing data about diagnoses in Dutch mental healthcare from 1 January 2011 to 31 December 2016 [28, 29]. These diagnoses are based on the Diagnostic and Statistical Manual of Mental Disorders 4th Edition (DSM-IV).

4)

Public Health Monitor register containing data on (determinants of) health, social situation, and lifestyle in a sample of Dutch citizens in 2012 and 2016 [6, 7]. The Public Health Monitor is conducted once every four years by Community Health Services, Statistics Netherlands, and the National Institute for Public Health and the Environment.

5)

Health Survey register containing data on health, medical contacts, lifestyle, and preventive behavior in a sample of Dutch citizens from 2014 up to and including 2019 [33]. The Health Survey is an annual survey conducted by Statistics Netherlands, which is part of the Dutch Lifestyle Monitor data collection [9].

Study population

In the Netherlands, a medical graduate receives a Master’s degree and can either start residency directly, work as a physician-not-in-training (temporary supervised clinical work before residency), or start a PhD trajectory, with only a minority obtaining a PhD [22]. Thus, physicians with a registration in the Individual Healthcare Professions register can either have a Master’s degree or a PhD degree. The demographics register was used to select all Dutch citizens aged between 25 and 65 years with a high educational level (Master or PhD degree) in the period from 2011 up to and including 2016. This population was defined as the “reference population”. Citizens with an active registration as physician between 1 January 2011 and 31 December 2016 were identified as “physicians”, based on the Individual Healthcare Professionals register (Total population). In physicians and the reference population, SUD patients were identified based on DSM-IV coding of the Mental healthcare claims register (SUD patients). The same selections were made for the Public Health Monitor and Health Survey registers in the period from 2012 up to and including 2019 to identify drinkers among the reference population and physicians (Questionnaire respondents).

Sociodemographic characteristics

Available sociodemographic characteristics included gender, age, country of birth, medical specialty, and educational background. The continuous variable age was recoded into a categorical variable (25 to 34 years, 35 to 44 years, 45 to 54 years, and 55 to 65 years) and country of birth was categorized into three categories (The Netherlands, European, and Non-European). For physicians, medical specialties were divided into five categories: (1) general practice; (2) (psycho) social medicine; (3) contemplative somatic medicine; (4) surgical and supportive medicine; and (5) no specialty, see Additional file 1: Table S1 [12]. Educational background was presented in eight categories for the reference population: (1) education; (2) humanities and arts; (3) social sciences, business and law; 4) science, mathematics and computing; (5) engineering, manufacturing and construction; (6) agriculture and veterinary; (7) health and welfare (including medicine); and 8) services.

From the Public Health Monitor and Health Survey registers information was also available on working hours per week (not working or less than 1 h, 1 to 12 h, 12 to 31 h, and 32 or more hours) and household income quintile (1st (lowest income), 2nd, 3rd, 4th, and 5th (highest income) quintile).

Definition of SUD diagnoses and alcohol consumption patterns

SUD patients and accompanying comorbidity and functioning were identified by a clinical diagnosis of SUD in the Mental healthcare claims register. Substances of abuse or dependence and comorbid psychiatric disorders were identified by DSM-IV codes on substance-related disorders (Additional file 1: Table S2). DSM-IV codes of comorbid somatic disorders were recoded into a dichotomous variable (“complex” + “singular” versus “none”). DSM-IV codes of start and end scores on the Global Assessment of Functioning (GAF) were divided into three categories: (1) persistent danger to major impairment (GAF 0-40); (2) serious to moderate symptoms (GAF 41-60); and (3) mild to no symptoms (GAF 61-100).

Drinkers were identified by the Public Health Monitor and Health Survey registers as those who reported having consumed at least one alcohol unit in the past 12 months [6, 7],Statistics Netherlands et al. [33]. Among drinkers, those compliant with the alcohol consumption recommendation, heavy drinkers, and excessive drinkers were identified. Compliance with the alcohol consumption recommendation was defined as drinking up to maximum one unit of alcohol per day, in line with the recommendation of the Health Council of the Netherlands [6, 7], Statistics Netherlands et al. [33]. Heavy drinking was defined as consuming six (males) or four (females) or more units of alcohol per day at least once a week in the last 6 months, in line with the definition of the public Health Monitor and the Health Survey [6, 7], Statistics Netherlands et al. [33]. Consuming more than 21 (males) or 14 (females) units of alcohol per week was defined as excessive drinking [6, 7],Statistics Netherlands et al. [33]. The groups of heavy and excessive drinkers were not mutually exclusive and therefore we did not present a group of moderate drinkers, which results in row percentages that do not add up to 100%.

Data analysis

The registry data allowed us to censor clinical SUD diagnoses and alcohol consumption patterns in the reference population and physicians. First, we used descriptive statistics to perform benchmark analyses between the reference population and the physicians with regard to SUD patients. The prevalence of clinical SUD diagnoses was calculated by dividing the number of (reference or physician) citizens with a clinical SUD diagnosis between 2011 and 2016 by the total number of (reference or physician) citizens between 2011 and 2016. Mean years of clinical SUD diagnosis between 2011 and 2016 were calculated by dividing the total number of clinical SUD diagnoses between 2011 and 2016 by the total number of (reference or physician) SUD patients between 2011 and 2016. Next, clinical SUD diagnoses, psychiatric and somatic comorbidity, and general functioning were compared between reference and physician SUD patients.

Second, respondents of the Public Health Monitor and the Health Survey were taken together and representatives from the reference population and the physicians were identified. First, characteristics of the total sample of questionnaire respondents and drinkers were benchmarked. Next, we performed benchmark analyses for the distribution of alcohol consumption patterns (compliance with the alcohol consumption recommendation, heavy drinkers, and excessive drinkers) within drinkers.

We decided not to test group differences statistically, since p-values are very dependent on sample sizes and may lead to misleading conclusions. Due to large numbers very small differences will become statistically significant (p < 0.001), even if these differences are considered as not relevant. In turn relevant differences in small groups may become not significant, even if differences are considered relevant. We therefore argue and recommend to focus on differences in means and proportions instead of statistical significance. Also in other fields, this is a recommended approach to benchmark and analyze large datasets [2, 25].

Small numbers (< 5) are not reported to prevent disclosure of physicians, in some cases the second smallest cell had to be cleared to avoid retracing. Analyses were performed using the Statistical Package for Social Sciences (SPSS), version 25 for Windows (IBM Corporation, Amonk, NY).