A Systematic Literature Review of Self-Reported Smoking Cessation Counseling by Primary Care Physicians

Tobacco consumption is a risk factor for chronic diseases and worldwide around six million people die from long-term exposure to first- or second-hand smoke annually. One effective approach to tobacco control is smoking cessation counseling by primary care physicians. However, research suggests that smoking cessation counseling is not sufficiently implemented in primary care. In order to understand and address the discrepancy between evidence and practice, an overview of counseling practices is needed. Therefore, the aim of this systematic literature review is to assess the frequency of smoking cessation counseling in primary care. Self-reported counseling behavior by physicians is categorized according to the 5A’s strategy (ask, advise, assess, assist, arrange). An electronic database search was performed in Embase, Medline, PsycINFO, CINAHL and the Cochrane Library and overall, 3491 records were identified. After duplicates were removed, the title and abstracts of 2468 articles were screened for eligibility according to inclusion/exclusion criteria. The remaining 97 full-text articles reporting smoking cessation counseling by primary care physicians were assessed for eligibility. Eligible studies were those that measured physicians’ self-reported smoking cessation counseling activities via questionnaire. Thirty-five articles were included in the final review (1 intervention and 34 cross-sectional studies). On average, behavior corresponding to the 5A’s was reported by 65% of physicians for “Ask”, 63% for “Advise”, 36% for “Assess”, 44% for “Assist”, and 22% of physicians for “Arrange”, although the measurement and reporting of each of these counseling practices varied across studies. Overall, the results indicate that the first strategies (ask, advise) were more frequently reported than the subsequent strategies (assess, assist, arrange). Moreover, there was considerable variation in the items used to assess counseling behaviour and developing a standardized instrument to assess the counseling strategies implemented in primary care would help to identify and address current gaps in practice.


Introduction
Tobacco consumption is a preventable risk factor for non-communicable diseases such as chronic obstructive pulmonary disease (COPD) and cardiovascular disease.Each year, around six million people die from long-term exposure to first-or second-hand smoke worldwide [1].Globally, one of the guiding instruments for tobacco control is the World Health Organization Framework Convention on Tobacco Control (WHO FCTC) [2].The convention gives specific recommendations for a number of different tobacco control strategies that should be implemented, such as developing comprehensive smoking cessation guidelines and introducing warning labels on cigarette packages [2].One approach to reduce tobacco consumption that is recommended in guidelines for the treatment of tobacco dependence is to offer smoking cessation counseling in the primary care setting [3][4][5].Smoking cessation counseling by general practitioners (GP's) has been shown to increase quit rates [6].The general practice is an appropriate setting for smoking cessation counseling for a number of reasons [7].First, GP's have suitable access to the target group because around 80% of the German population visit their GP at least once per year [8].Second, regular personal contact builds trust between GP's and patients and facilitates the provision of quit advice [9].Third, face-to-face contact allows for the delivery of individual smoking cessation advice [10].
The clinical practice guideline of the US Public Health Service contains a comprehensive approach to smoking cessation counseling in primary care settings, which specifies individual counseling steps such as asking about tobacco consumption and recommending the use of pharmacological aids; the 5A's strategy [5].The 5A's refer to a sequence of 5 different counseling strategies: "Ask" (ask all patients about tobacco use), "Advise" (advise all tobacco users to quit), "Assess" (assess the willingness to quit), "Assist" (assist with quitting) and "Arrange" (arrange follow-up) [5].Examples of other approaches are the ABC model (Ask about and document smoking status, give brief advice and encourage the use of cessation support) and the recommendation of the American Association of Family Physicians (AAFP; Ask about tobacco use, advise to stop using tobacco products and provide behavioral interventions).We focus on the 5 A's strategy because it distinguishes between 5 counseling steps and is therefore more inclusive than the ABC model and AAFP recommendation which describe 3 counseling steps [4,5,11].
Although smoking cessation counseling is effective and recommended in clinical guidelines, it is not fully implemented in primary care [10,12].In order to further understand and address the discrepancy between evidence and practice, an overview of current counseling practices is needed.A systematic literature review has examined the frequency of behavioral counseling by physicians for multiple behaviors (tobacco consumption, physical activity and nutrition) and found that the use of educational materials and referral to smoking cessation courses were frequently reported counseling strategies [13].For example, educational materials were recommended by 58% of Scottish physicians and 61% of Canadian physicians [14,15].However, comparability of quantitative information on smoking cessation counseling practices was limited and it is not clear which counseling steps are implemented in practice.Therefore, the aim of this literature review is to systematically assess physician-reported smoking cessation counseling in primary care by classifying counseling practices according to the 5A's strategy.

Protocol registration and search strategy
The present literature review was reported according to the PRISMA statement [16].A protocol for this literature review was not registered.The database search was limited to studies published between 2000 and June 2015 because counseling behavior was categorized according to the 5A's strategy, which was published by the US Public Health Service in the year 2000 and the literature search was conducted in June 2015.A preliminary search of Medline via OVID served to identify relevant keywords.The core search strategy was developed using the keywords and relevant synonyms to capture smoking cessation counseling by physicians and reviewed by a librarian experienced in database searches.Among the main keywords used in the search were: "smoking cessation", "counseling" and "primary care physician".The search strategy was then adapted in order to perform an extensive literature search in the following databases: Embase, Medline and PsycINFO via OVID as well as CINAHL and the Cochrane library (Please see S1 File for the detailed search strategy per database).Further articles were identified by screening the bibliographies of articles retrieved from the initial search.

Study selection
Two independent reviewers (ALBa and JN) performed the title and abstract screening of the studies identified by the search based on a checklist for the inclusion/exclusion of studies framed in relation to PICOS (Population, intervention, comparator, outcome and study design; see Table 1).The reliability of the checklist was tested on a random selection of 100 articles prior to the title and abstract screening.Studies for which all items on the checklist were answered with "yes" were included in the review.There was substantial agreement (Cohen's Kappa = 0.66) between the two reviewers (JN and ALBa), who then performed the full-text screening independently.Any discrepancies were resolved by discussion until consensus was reached.

Study inclusion/exclusion criteria
Studies were included in the review if they were published in either English or German language and measured physicians' self-reported smoking cessation counseling activities via questionnaire.The main outcome was the proportion of physicians reporting smoking cessation counseling.We excluded studies that reported the proportion of patients receiving smoking cessation counseling because it has been shown that physician and patient-reported data differ [17].Research articles that defined the majority of the study population as either general practitioners, family physicians or internists (referred to as physicians in the following) were included in the review [18].Research has shown that physicians provide smoking cessation counseling to young patients more frequently than to adults.We therefore excluded articles with study populations consisting only of paediatricians or young patients (below age 18) [19].

Data extraction
A data extraction sheet was pilot-tested and adjusted as necessary in order to obtain relevant data from the articles.One reviewer (ALBa) performed the initial data extraction and a second reviewer (BS) checked whether the data were extracted accurately.The following information was extracted: study identifiers (Author, year, and country), the sample (physician characteristics), the methods (data collection method) and results (response rate, sample characteristics and smoking cessation counseling behavior).The proportion of physicians providing smoking cessation counseling to patients was extracted and categorized according the 5A's.[5].

Quality assessment
Quality assessment was conducted using the Mixed Methods Appraisal Tool (MMAT) to assess the quality of various study types (e.g.quantitative studies, mixed methods studies etc.) [20].The tool's validity and reliability have been established [21,22].An overall quality score is determined using criteria that vary by study design.Quality assessment was performed independently by two reviewers (ALBa and BS) and the quality assessment for each study is shown in Table 2.

Data synthesis
The main outcome was the proportion of physicians reporting counseling and in cases where multiple items were used to measure smoking cessation counseling behavior, the item that best represented the respective 5A's strategy was chosen.In the case of rating scales, we reported the responses to the end point of the scale.For example, for the response options "Never", "Occasionally", "Almost always" and "Always", the proportions of physicians that selected "Always" were reported.If multiple forms of assistance were reported (e.g.prescribing NRT, handing out leaflets, referral to expert), we reported only the most frequently offered form of assistance.

Study selection
Overall, 3491 records were identified by the electronic database search.After duplicates were removed, the title and abstracts of 2468 articles were screened according to inclusion/exclusion criteria.The remaining 97 full-text articles were assessed for eligibility.Of these articles, 62 were excluded (see S3 File for a list of the reasons for exclusion).The remaining 35 articles were included in the final review.The process for including/excluding studies is illustrated in the flow diagram in Fig 1 [16].Response options in bold correspond to the percentages reported for the 5 A's.doi:10.1371/journal.pone.0168482.t002

Description of included articles
In total, 35 articles were included in the review and reported studies from 17 countries (see Table 2 for details).Sample sizes from studies included in the review ranged from 37 to 4074 (mean = 809).Thirty-one studies reported response rates, with a mean response rate of 64% (range: 18% to 95%).Thirty-four studies used survey designs and 1 article reported an intervention study (single group, pre-post design) [9] of which baseline data were included.Surveys were mainly distributed via post (16 studies) or in person (7 studies).Descriptive data of all studies included in the review are shown in Table 2.

Smoking cessation counseling
Of the 35 articles, 8 articles reported counseling behavior corresponding to all 5 counseling strategies (ask, advise, assess, assist and arrange) [7,23,26,29,31,32,36,46].The strategies "Ask" and "Advise" were most frequently reported."Ask" was reported in 29 articles and on average 65% (range: 7% to 100%) of physicians asked about their patient's smoking behavior.Twenty-five articles reported "Advise" and on average, 63% of physicians (range: 13% to 99%) advised their patients to stop smoking.Fourteen articles reported behavior corresponding to "Assess" with an average of 36% (range: 11% to 72%) of physicians assessing their patient's smoking status.Twenty-three articles reported "Assist", with an average of 44% (range: 2% to 98%) of physicians providing assistance.Fourteen articles reported "Arrange" and on average 22% (range: 2% to 54%) of physicians arranged follow-up consultations.Table 2 shows the proportions of physicians offering smoking cessation counseling, categorized according to the 5A's.
The items and response options used in each study are shown in Table 2. Whereas some researchers used a single item to measure counseling behavior, others used multiple items.For example, counseling behavior corresponding to the "Ask" strategy was assessed with the item "During a consultation, do you ask patients whether they smoke?" in one study [44], whereas another study [35] used the items "I ask patients at their annual visits if they use tobacco", "When patients come in for unrelated problems, I ask them about tobacco use" to measure behavior.When multiple items were used to measure counseling behavior in a single article, the responses to the item that most accurately represented the strategy of interest was reported.For example, the items ''Do you assess the smoking status when a patient attends the practice for the first time?"and "Do you routinely assess the smoking status when a patient attends the practice?"[9] both concern asking patients about tobacco use and therefore relate to the strategy "Ask" (systematically identify all tobacco users at every visit).However, the second item corresponds to the strategy more closely because it concerns asking all patients about tobacco use (not only upon the first visit to the practice office).Therefore, the responses to the second item are reported in the review.
The questionnaires used also differed between studies and some researchers used adapted versions of existing questionnaires to measure counseling behavior.For example, one study used a 45-item questionnaire that was based on a questionnaire by the World Health Organization [28,56].In contrast, other studies included questionnaires developed by the researchers during project group discussions or expert consultations (e.g.in-depth interviews).Further differences were seen in the response options of questionnaire items.Dichotomous response scales ("Do you ask patients whether he or she smokes?";"yes","no") [37] were used in some studies, whereas others used rating scales ("During a consultation, do you ask patients whether they smoke?";"never", "occasionally", "almost always" and "always") [44].When a rating scale was used to assess counseling behavior, only data from the end point of the scale was reported.For example, in order to extract counseling behavior for the strategy "Ask", we reported the proportion of physicians who responded "always" to the item "During a consultation, do you ask patients whether they smoke?"

Quality of included articles
The quality assessment is included in Table 2. Details of the quality assessment are shown in S4 File.There was moderate agreement between the two raters on the assessment of study quality (Cohen's Kappa = 0.59).Any discrepancies were resolved by discussion.The methodological quality of the included studies varied and the population sizes were between N = 37 to N = 4074.Also, information on the instruments used to measure physicians' smoking cessation counseling was often insufficient or lacking and the information available is displayed in Table 2.However, no study was excluded due to its methodological quality.

Discussion
The aim of this review was to systematically assess the frequency of behavioral counseling for smoking cessation in primary care and the 5A's strategy was used to structure smoking cessation counseling.Main findings are that the first strategies "Ask" and "Advise" were more often reported than the subsequent strategies "Assess", "Assist" and "Arrange".This finding corresponds to research that found the last two strategies (assist, arrange) to be least reported [57].However, it is not possible to say whether individual steps were not performed by physicians or simply not measured or reported by the researchers.Secondary findings are that the proportions of physicians reported differed considerably per 5A strategy.For example, 81% (95% CI = 74%, 88%) of physicians reported to give advice in one study [29], whereas only 13% (10%, 16%) of physicians gave advice according to another study [50].While differences in proportions may result from variations in sample characteristics and settings, they may also be due to the wording of questionnaire items used to assess counseling behavior [13].More physicians may have indicated advising patients in the first study [29] because the item ("Do you advise the patient to quit smoking?") can be understood to refer to only some or all patients.In contrast, physicians in the second study [50] were specifically asked if they offered advice to all of their patients.These differences could also explain why fewer physicians indicated giving advice in the second study.
Also, there are different forms of assistance, such as providing supporting information, referring to quitlines or recommending the use of pharmacotherapy.According to a literature review, assistance was most frequently offered in form of educational materials [13].In contrast, the results of the present review suggest that nicotine replacement therapy (NRT) was the most frequently offered form of assistance.For example, one study reported that 61% (57%, 65%) of physicians recommended NRT, whereas only 25% (22%, 28%) provided selfhelp materials [25].However, it is difficult to clearly determine which form of assistance was offered most frequently because not all articles included in this review distinguished between different assistance types.

Limitations
Risk of bias is seen in the self-report questionnaires used to measure physicians' smoking cessation counseling activities.Self-report measures may introduce social desirability or recall bias [58].In particular, physicians may overestimate their adherence to guidelines and the frequency with which they deliver preventive services [13,17].Therefore, a limitation is that studies presenting physician-reported but not patient-reported data were included in this review [17].Besides, it has been proposed that instruments to assess the delivery of preventive services should take into account the perspective of both, physicians and patients [59,60].Also, physicians who participate in surveys on smoking cessation counseling may be more interested and hence more engaged in counseling activities than the overall population of physicians.Nevertheless, self-report measures make it possible to measure many participants and are therefore an appropriate research tool for this population [61].While other methods such as videobased observation can be used to conduct research in primary care, these methods complicate the research process due to technical requirements and confidentiality and privacy aspects [62].Moreover, differences in the items used to assess counseling behavior limited the comparability of study results.A systematic review on behavioral counseling for cardiovascular disease found that physician-reported smoking cessation counseling differed, depending on the phrasing of survey questions [13].Moreover, the number of items and type of response option (dichotomous or rating scale) may influence reports of counseling behavior [43].Further, it should be noted that the 5A's strategy used to structure smoking cessation counseling was initially presented by the US public health service and may be less familiar outside of the US [5].Although the strategy is mentioned in the guidelines of other countries, such as Canada [63], and Germany [64], specific recommendations may differ between countries.We still chose the 5A's strategy because it distinguishes between 5 counseling steps and is therefore more comprehensive than other frameworks such as the ABC model [4,5,11].Also, only articles in English and German language were included in the review, which means that relevant studies published in other languages may have been missed.

Implications
While this review summarizes physicians' perspective of smoking cessation counseling in practice, it does not provide information on the reasons why physicians do or do not offer counseling.Changes in health care policy and clinical guidelines affect provider behaviour and may influence the frequency and type of smoking cessation counseling.According to previous studies, self-efficacy beliefs, prior training and patient characteristics determine whether physicians offer counseling [13,65].For example, physicians may not approach patients who appear unmotivated or do not intend to quit smoking due to fear of harming the physician-patient relationship [66].A systematic literature on the barriers to discussing smoking cessation with patients exists [66] and future studies could explore a possible association between attitudes towards counseling and actual counseling behavior [13].This would help to further understand and address common barriers.Also, future studies could include patient-reported or more objective data, in order to gain further insights into the implementation of smoking cessation counseling in practice.
In concluding, more research is needed in order to monitor changes and draw firm conclusions about provider behavior.The present results suggest that there is need for a standardized instrument to assess counseling behavior that differentiates between the possible forms of advice and assistance (e.g.NRT, referral to specialist) because this would help to identify current gaps in practice.While instruments to assess smoking cessation counseling have been developed in relation to the 5A's strategy, these instruments require additional evaluation.Most instruments remain unpublished and data on reliability and validity are often missing [57].

**
Met 50% of MMAT criteria; *** Met 75% of MMAT criteria; **** Met 100% of MMAT criteria), see S4 File for further information; GP = General practitioner; FP = Family practitioner; N = Sample size; M = Mean; n.r.= not reported.a Questions in quotation marks are reported in their exact wording; and all other questions are reported semantically.b

Table 2 . Descriptive data of included articles. Reference & Country Sample & Response rate (%) Gender (%) & Age Data collection Overall Physician-reported counseling (% and items a ) Response question a Ask Advise Assess Assist Arrange options b
Ask: Frequently; Sometimes; Don't ask, Advise: More than 30; 11-30; 1-10; no patients (Continued)

Table 2 .
(Continued) never Legend.More asterisks indicate a higher quality grade according to the Mixed Methods Appraisal Tool (MMAT; * Met 25% of MMAT criteria;