Training health professionals in smoking cessation

Provision of advice and support to smokers by health care professionals, in primary care settings, improves cessation rates, although the effects are modest (Silagy 2000a). Even though the rates appear low from the perspective of many clinicians, they could translate into a substantial public health benefit if consistently provided, as approximately 80% of adults have contact with a health care practitioner, usually in primary care, at least once each year (Silagy 1992). It is disappointing, therefore, that the number of patients who report receiving advice on smoking cessation from health professionals is low (Silagy 1992; Wallace 1987). Increasing the amount and quality of interventions from primary care health professionals is frequently cited as a way of realising this potential health gain (Sanders 1992). Providing training in smoking cessation is one possible method for doing this, and a variety of courses and methods are available. However, while individual studies have shown an effect of training on physician's activities, there has been doubt about the extent to which this translates into changes in patient behaviour (Cummings (Priv) 1989; Cummings 1989; Kottke 1989).

We sought to address this issue by systematically identifying and reviewing the evidence from randomised controlled trials that have studied the effects of training and supporting health care professionals to provide smoking cessation advice.

Our a priori hypotheses were

i) that training health care professionals is more effective than no training in increasing the number of smokers who are offered advice about quitting, and who subsequently achieve abstinence.

(ii) the effect of training can be enhanced by either providing prompts and reminders to health care professionals to offer smoking cessation advice to their patients, or encouraging them to offer nicotine replacement therapy as an adjunct to their advice to smokers.

We identified randomised trials in smoking cessation using the Cochrane Tobacco Group search strategy, and scrutinized these trials to identify trials where the principal intervention involved education of health professionals in methods of smoking cessation.

Two reviewers independently extracted data from published reports. Disagreements were resolved by referral to another person. No attempt was made to blind any of these people either to the results of the primary studies or to the intervention the subjects received.

The trials included in the review used cluster randomisation. Outcomes relate to individual patients whilst allocation to the intervention is by provider or practice, and ignoring this may introduce unit of analysis errors. Using statistical methods which assume for example that all patients' chances of quitting are independent ignores the possible similarity between outcomes for patients seen by the same provider. This may underestimate standard errors and give misleadingly narrow confidence intervals, leading to the possibility of a type 1 error (Altman & Bland 1997). Trials may use a variety of statistical methods to investigate or compensate for clustering; we have recorded whether studies used these and whether the significance of any effect was altered. In the absence of consensus on how to pool cluster level data, we decided on a narrative approach to synthesising the data, rather than the formal meta-analysis adopted in an earlier version of this review.

Methodological quality assessment:

We assessed the methodological quality of the studies included in the review according to the extent to which the method of allocation to intervention or control was concealed. Because of the potential importance of cluster effects, we also rated trials according to whether they checked for or adjusted for cluster effects

Ten studies met our criteria for inclusion in the review, randomising practitioners or practice teams. No restriction was placed on the age or sex of smokers who subsequently consulted these practitioners.

Eight of the trials (Cummings 1989; Cummings (Priv) 1989; Kottke 1989; Lennox 1998; Sinclair 1998; Strecher 1991; Wilson 1988; Wang 1994) examined the effect of training in comparison to untrained control groups. In two trials (Cohen (Doc) 1989; Cohen (Dent) 1989), all health care professionals received training, and the randomisation examined the effect of using prompts and nicotine gum, separately and combined, as enhancements to training.

Eight of the ten trials trained medical practitioners, of which one was confined to junior medical staff (Strecher 1991). One also included key non-physician office staff in the practice training (Lennox 1998). One trial trained dental practitioners (Cohen (Dent) 1989), and one community pharmacists (Sinclair 1998). All of the trials were conducted in primary care settings, and directed towards opportunistic intervention. However, because the trials were conducted in more than one country, primary care encompassed a diversity of settings including community practice and pharmacies, hospital-based clinics and health maintenance organisations. The participation rates among those invited ranged from less than 10% in private practice settings to 90% when the participants were members of a residency training programme.

Almost all the training was provided on a group basis in a tutorial or workshop setting. In two studies (Cohen (Dent) 1989; Cohen (Doc) 1989) some of the doctors or dentists had a personal tutorial. The duration of training ranged from a single one hour session to a whole day. A number of different training methods were incorporated in these sessions, including lectures, videos, role plays and discussion. They emphasised minimal contact strategies. The importance of setting quit dates, and offering follow-up, were emphasised in most trials. In only one, however, was a specific follow-up schedule (up to six appointments) recommended (Wilson 1988). Instruction in use of nicotine gum was part of the training in four trials, and two trials examined its effect in addition to training (Cohen (Dent) 1989; Cohen (Doc) 1989). Three trials (Lennox 1998, Sinclair 1998, Wang 1994) based their training on the stages of change model (Prochaska 1983).

All trials reported abstinence as the principal patient outcome. In four trials this was after six months follow-up; in the remainder it was at twelve months or longer.

Nine of the trials reported process variables, one reported only smoking cessation as an outcome (Wang 1994).

Reporting of the method of randomisation was variable. All the trials used cluster randomisation, and all but one of the trials either used the correct unit of analysis, or explored the possibility that clustering of patient outcomes could affect the results. None reported that differences in the method of analysis affected their results.

Effect of training on abstinence rates:
In six of the eight studies with an untrained control group, no effect of training on quit rates was detected. The largest single study (Wilson 1988) found a significant effect of training on sustained abstinence at one year (mean rate by practice in a cluster level analysis adjusted for differences at baseline: 8.8% for the trained group vs 6.1% and 4.4% in the two comparison arms, P <.001). One small study (Wang 1994) found that physicians trained in the stages of change model achieved higher rates than a usual care group, but were not significantly different from a group who received a simple prompt to discuss smoking.

Effect of adjuncts to training on abstinence:
Prompts and reminders to practitioners to deploy smoking cessation techniques were used in addition to a training programme in three trials (Cohen (Dent) 1989; Cohen (Doc) 1989; Strecher 1991). In each, the prompts increased the frequency of health professional intervention, but in only one of the three (Cohen (Doc) 1989), was there a significant improvement in abstinence rates. In one trial, providing nicotine gum in addition to training increased quit rates compared to training alone(Cohen (Dent) 1989).

Effects of training on process variables:
Training increased the smoking cessation activities of health care professionals in almost all the studies. Trained professionals were about 1.5 to 2.5 times more likely to counsel patients about smoking, and to initiate other interventions such as setting a quit date, suggesting a follow-up appointment , and offering self-help materials or nicotine gum.

Programmes designed to train health professionals to provide smoking cessation intervention are effective in increasing process of care outcomes including the number of patients who receive counselling, set a quit date, and are given follow-up appointments, self-help materials and nicotine gum.

We lack strong evidence that training leads to higher quit rates. The encouraging results reported in the largest study (Wilson 1988) were not replicated in the other studies. This may reflect a chance finding, or a difference in the actual intervention. The training in the Wilson study emphasised the importance of follow-up to a far greater degree than any of the other programmes studied. Physicians were trained to challenge the patient at a first appointment, to schedule a separate appointment to set a quit date, and to offer up to four supportive follow-up visits. Moreover, the physicians were paid on a per-item basis for each of these follow-up visits, according to the standard reimbursement scheme in the Canadian health care system. Although in most of the trials some form of follow-up was recommended as part of the training, in none of the others was a specific schedule arranged, and few such visits were scheduled. This may be related to the fact that reimbursement for such visits is generally not forthcoming from third party payers in the United States, where all the remaining studies took place. A possible implication of these findings is that training alone is unlikely to represent a useful investment of resources, unless it is linked to organisational changes which facilitate the intervention. This is consistent with a previous review of fifty randomised controlled trials, covering a wide range of subject areas and types of intervention, which found that practice-enabling educational strategies reinforced by follow-up and reminders were major determinants of successful continuing medical education (Davis 1992).

Our results suggest that adjuncts such as manual or computerised reminders to provide advice on smoking can help, and should form part of such strategies. It is clear from other work that both simple advice and nicotine replacement therapy can improve the chances of quitting in some patients (Silagy 2000a; Silagy 2000b), and should form of part of a primary care strategy for smoking cessation, irrespective of whether more formal training to health care providers is offered.

This approach to data analysis in this review changed substantially in response to comments received from Jeremy Grimshaw, and discussion at a workshop on analysis of clustered data organised by the Cochrane Statistical Methods Working Group.

None

It is not clear from the analysis whether there were unit of analysis errors in the original trials and whether cluster level or patient level outcomes have been used for the analysis. If the pooled analysis is based upon studies with a unit of analysis error, it may overestimate the effectiveness of the intervention. As a minimum, the authors should discuss this issue and its implications for interpreting the results of the review in the discussion section.

Several of the included trials do not appear to meet the stated inclusion criteria.
Eg 1 Types of participants specifies studies 'in which the unit of randomisation was a healthcare practitioner or practice', but one study uses patient randomisation.
Eg 2 Types of outcome measures states that 'studies had to report the outcomes of the intervention both on the performance of the health professionals and on smoking behaviour of patients', but not all trials appeared to report performance measures.

We agree that using patient level data in a meta-analysis when providers were randomised could lead to an overestimation of the results of treatment. We have included a discussion of cluster randomisation, and have reported the way in which studies have analysed their data in relation to clustering. In the absence of a consensus on the best way to pool cluster randomised trials we have removed the meta-analysis, and provided a narrative synthesis of the results instead.

We have excluded one study which randomised patients to trained physicians who provided different types of counselling (Ockene 1991). We have clarified that only smoking cessation is a required outcome for study inclusion.

Tim Lancaster