The present study was based on two cross-sectional self-administered postal smoking surveys among Estonian physicians in 2002 and 2014. Initially, the surveys involved all practising physicians in Estonia. In 2002, physicians were drawn from the database of Estonian Health Insurance Fund. In 2014, sample was based on the data from the Estonian Health Care Professionals Registry. In 2002, the questionnaires were mailed to the physicians’ workplace. Non-respondents received the questionnaire twice. In 2014, the survey materials were mailed to the physicians’ home address. To receive home addresses, data from the Estonian Health Care Professionals Registry were linked with the Population Registry in Estonia. Non-respondents received a reminder letter in a month and another envelope containing survey materials in 2 months.

The questionnaire used in these surveys was originally developed by the WHO and modified according to the Estonian healthcare system.17 In 2014, the questions regarding nicotine dependence were added to the questionnaire. Questionnaires concerned individual characteristics, smoking behaviour, attitudes towards and knowledge about tobacco use but also attitudes towards patients’ smoking.

The initial survey sample size of all practising physicians in Estonia was 4140 in 2002 and 5666 in 2014 (table 1). Number of respondents was 2747 and 2903, respectively. The crude response rate was 66.3% in 2002 and 52.0% in 2014. Corrected response rates (excluding the persons who were unavailable, retired, had wrong address, left Estonia or were dead) were 67.8% and 53.1%, respectively.

The sample for the present study was restricted to physicians who were younger than 65 years (n=2549 in 2002, n=2339 in 2014).

The main outcomes were smoking status and physicians’ attitudes towards smoking.

Smoking status was determined by combining answers to several questions concerning smoking and classified as following:

daily smokers (those who currently smoke every day)

For secondary data analysis, smoking status was dichotomised to current smokers (daily and occasional smokers) and non-smokers (past and never smokers). Smoking, in the present study, was defined as smoking cigarettes.

Attitudes towards smoking were determined with following seven statements:

Smoking is very harmful to health.

In the first statement physicians assessed the harmfulness of smoking using 10-point scale (10=very harmful). Those who had chosen 8–10 were considered as having agreed that smoking is very harmful to the health. Those who had chosen 1–7 were considered as disagreed. In all other statements, possible answers were completely agree/somewhat agree/rather disagree/completely disagree/cannot say. For secondary analysis, answers were dichotomised as agree (completely agree and somewhat agree) and disagree (rather disagree and completely disagree).

Background variables age, ethnicity, place of residence and medical specialty were considered as confounding factors.

Age was measured in full years.

Ethnicity referred to self-determined national identity and was classified as Estonian/non-Estonian (mainly Russian).

Place of residence was determined as Tallinn (capital of Estonia), other city and other (not urban) settlement.

Medical specialty was determined based on self-reported specialty and was analysed in three groups: family physician, specialist doctor and dentist.

The study methodology follows The Strengthening the Reporting of Observational Studies in Epidemiology Statement guidelines for reporting observational studies.

Data were analysed separately for men and women. Mean age of respondents with SD was calculated. Distribution (%) of respondents by background variables and by agreement with seven statements concerning attitudes towards smoking was calculated. χ² test was used to find differences in background variables and in attitudes towards smoking between 2002 and 2014. Significance level was set at 0.05. The age-standardised smoking prevalence with corresponding 95% CIs was calculated using European standard population.18

Multiple binary logistic regression models were used to analyse association of physicians’ attitudes towards smoking with study year and smoking status. The models used dichotomised approvement (agreed vs disagreed) as a dependent variable and study year, smoking status, age, ethnicity, place of residence and medical specialty as explanatory variables. Fully adjusted ORs with corresponding 95% CIs were calculated.

Questionnaires with missing smoking status values (n=11) were excluded from the analysis. In total, 4877 questionnaires were included to the descriptive analysis (n=2539 in 2002, n=2338 in 2014). Questionnaires that lacked information concerning background variables and attitudes towards smoking or wherein the statements ‘cannot say’ was answered were excluded from the χ² tests and logistic regression models.

Data were analysed using statistical package Stata V.11.

An overview of physicians’ main background variables is provided in table 2. The majority of participants were females (83.6% in 2002 and 82.5% in 2014). Among men 50.8% in 2002 and 38.1% in 2014 were younger than 45 years (P<0.001) and among women 45.4% in 2002 and 38.8% in 2014 were younger than 45 years old (P<0.001). Mean age of male physicians was 45.2±9.8 in 2002 and 46.9±10.9 in 2014 and of female physicians 46.4±10.2 in 2002 and 46.7±11.2 in 2014, respectively.

The age-standardised prevalence of daily and occasional smoking was lower, but age-standardised prevalence of never smoking was higher in 2014 than in 2002 (table 3). Age-standardised prevalence of past smoking was similar in 2002 and 2014.

Table 4 summarises physicians’ responses regarding attitudes towards smoking in 2002 and 2014. Compared with 2002, in 2014, the attitudes towards smoking were less favourable among both male and female physicians.

Multiple binary logistic regression demonstrated less approving attitudes towards smoking in 2014 than in 2002 (table 5). Compared with 2002, in 2014, male and female physicians agreed significantly more with the statements:

Smoking is very harmful.

Compared to 2002, in 2014, male and female physicians agreed significantly less with the statements:

To stop smoking is very hard for many people, so it is better for their health to simply continue smoking.

Compared to 2002, in 2014, only female physicians agreed significantly less with the statements:

As many people have smoked for their whole lives until old age and not become ill, smoking is not as dangerous as experts declare.

Agreement with all seven statements was associated with smoking status of male and female physicians.

The age-standardised prevalence of daily smoking decreased 1.6 times among male and 1.4 times among female physicians from 2002 to 2014 (the age-standardised prevalence of occasional smoking 2.3 and 3.0 times, respectively). This result was expected as smoking among physicians (aged between 23 and 86 years) in Estonia decreased between 1982 and 2014.19 20 Among these physicians, age-standardised daily smoking decreased 3.3 times among men and 2.6 times among women19 and current smoking 2.8 and 2.4 times,20 respectively, over three decades. Smoking rates in Estonia have come down among general population as well. In 2002, daily smoking prevalence was 49.6% among men and 20.3% among women.21 In 2014, 31.4% of men and 15.8% of women were daily smokers. Although Estonia is considered to have reached the mature state in terms of smoking epidemic, the prevalence of daily smoking among physicians in 2014 was still comparable with the rates of daily smoking among Finnish doctors in 2002.14

The results of this study showed that attitudes towards smoking were less approving in 2014 compared with 2002.

Agreement with the four statements of seven was associated with study year and smoking among male and female physicians. Agreement with the statements that smoking is very harmful and that it is important to reduce smoking among the population was more prevalent in 2014 and less prevalent among smoking physicians. This finding is in accordance with previous international studies showing that compared with non-smokers, smoking physicians agree less that smoking is harmful.4

Agreement with the statements that to stop smoking is very hard for many people, so it is better for their health to simply continue smoking and that smoking does not damage my health as long as I follow a healthy life style in other fields was less prevalent in 2014.

Agreement with the three statements of seven was associated with study year among female physicians only. Association was found between study year and agreement with the statements that as many people have smoked for their whole lives until old age and not become ill, smoking is not as dangerous as experts declare, that to smoke or not to smoke, that is my personal choice and that smoking is only dangerous to my health if I smoke more than 10 cigarettes a day among male physicians. This confirms results from previous studies, according to which smoking behaviour has different patterns among men and women.22 23 The difference in opinions between genders might be related to the fact that behaviour of men in general is considered to be more risk prone.24 The findings of present study might support the notion that, in Estonia, social acceptability of smoking has decreased, and attention has turned towards prevention and health promotion.

Overall, Estonian physicians’ attitudes towards smoking have improved from 2002 to 2014, and the developments concerning decline of smoking in Estonia have created a supportive environment for that. Estonian Tobacco Act, enforced in 2001, renewed in 2005 and amended since with several legal instruments, sets the requirements for creating a smoke-free environment, availability and pricing of tobacco products. Many of Estonian healthcare institutions have joined the network for tobacco-free health services,25 and despite the fact that in healthcare institutions in Estonia, smoking is still allowed in designated smoking areas, many hospitals promote reducing smoking among staff, have prohibited smoking everywhere in the hospital area and have declared the hospital smoke-free.

Agreement with all seven statements described above was associated with smoking status of male and female physicians. Most drastically, compared with non-smokers, men who smoked, had six times higher odds to agree that smoking is not as dangerous as experts declare and had seven times higher odds to agree that smoking does not damage health if the person leads an otherwise healthy lifestyle. For smokers, the beliefs that smoking is not dangerous to health might be based on the fact that not all persons who smoke will develop adverse effects of smoking like lung cancer or other smoking-related diseases.

The fact that compared with non-smokers, smoking physicians agreed less that it is important to reduce smoking and believe more that to smoke is person’s own choice demonstrates that smoking physicians might feel the need for justification for their own behaviour, especially if they have not developed any smoking-related health problems. Opinions on that matter might also indicate that physicians lessen their role in reducing smoking in the population. According to the worldwide literature, attitudes towards smoking differ by region. In former studies among Italian physicians specialising in public health, 79.6% considered health professionals as behavioural models for patients, and 96.6% affirmed that health professionals have a role in giving advice or information about smoking cessation.26 In Serbia, 60.7% of physicians agreed that healthcare professionals serve as role models for their patients and public.27 However, data from a focus group interview carried out among Armenian doctors revealed that the majority of doctors believed they have no role in patients’ quitting.28Compared with Finnish physicians, Estonian physicians were less conscious of their role as healthy lifestyle exemplars in 2002.14 Authors then argued that there might be a fear to influence other people’s behaviour in Estonia. Results of the present study indicate that the fear still exists.

Although it has been shown before that physicians’ beliefs about smoking-related diseases were consistent with medical evidence,15 the results of present study showed that physicians who smoke may see smoking more favourably. These opinions can affect smoking cessation activities of the physicians as it has been shown that smokers might not anticipate health problems related to smoking.29

Possible limitations of this study should be addressed. First, as the study relied on self-reported data, the bias of self-representation should be considered. Second, the corrected response rates of 67.8% in 2002 and 53.1% in 2014 should be taken into account. The possibility that smokers prevail among persistent non-respondents may have led to the underestimation of smoking prevalence rates. Third, limitations could be related to the cross-sectional nature of the study thus not providing the opportunity to investigate causal relationships. Also, as only two time points were used, merely a general tendency of smoking prevalence and attitudes towards smoking can be observed. Despite these shortcomings, the survey data provides an excellent opportunity to analyse changes in smoking status and attitudes towards smoking as both surveys were nationwide, the survey methods and questionnaires were similar and physicians are considered a very homogenous cohort in terms of their education.