Implementation strategies to increase smoking cessation treatment provision in primary care: a systematic review of observational studies

Abstract

Background

Internationally, there is an ‘evidence-practice gap’ in the rate healthcare professionals assess tobacco use and offer cessation support in clinical practice, including primary care. Evidence is needed for implementation strategies enacted in the ‘real-world’. Aim: To identify implementation strategies aiming to increase smoking cessation treatment provision in primary care, their effectiveness, cost-effectiveness and any perceived facilitators and barriers for effectiveness.

Methods

‘Embase’, ‘Medline’, ‘PsycINFO’, ‘CINAHL’, ‘Global Health’, ‘Social Policy & Practice’, ‘ASSIA Applied Social Sciences Index and Abstracts’ databases, and grey literature sources were searched from inception to April 2021. Studies were included if they evaluated an implementation strategy implemented on a nation-/state-wide scale, targeting any type of healthcare professional within the primary care setting, aiming to increase smoking cessation treatment provision. Primary outcome measures: implementation strategy identification, and effectiveness (practitioner-/patient-level). Secondary outcome measures: perceived facilitators and barriers to effectiveness, and cost-effectiveness. Studies were assessed using the Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I) tool. A narrative synthesis was conducted using the Expert Recommendations for Implementing Change (ERIC) compilation and the Consolidated Framework for Implementation Research (CFIR).

Results

Of 49 included papers, half were of moderate/low risk of bias. The implementation strategy domains identified involved utilizing financial strategies, changing infrastructure, training and educating stakeholders, and engaging consumers. The first three increased practitioner-level smoking status recording and cessation advice provision. Interventions in the utilizing financial strategies domain also appeared to increase smoking cessation (patient-level). Key facilitator: external policies/incentives (tobacco control measures and funding for public health and cessation clinics). Key barriers: time and financial constraints, lack of free cessation medications and follow-up, deprioritisation and unclear targets in primary care, lack of knowledge of healthcare professionals, and unclear messaging to patients about available cessation support options. No studies assessed cost-effectiveness.

Conclusions

Some implementation strategy categories increased the rate of smoking status recording and cessation advice provision in primary care. We found some evidence for interventions utilizing financial strategies having a beneficial impact on cessation. Identified barriers to effectiveness should be reduced. More pragmatic approaches are recommended, such as hybrid effectiveness-implementation designs and utilising Multiphase Optimization Strategy methodology.

Protocol registration

PROSPERO:CRD42021246683

Background

Smoking remains one of the leading preventable causes of illness and premature death in the UK [1] and worldwide [2]. Most adult smokers want to quit smoking [3,4,5] but quit attempts have a low success rate because fewer than a third use evidence-based treatment [3, 5]. For example, the current stop-smoking interventions recommended in the UK are: behavioural support, nicotine replacement therapy (NRT), bupropion, varenicline, and nicotine-containing electronic cigarettes [6]. Healthcare practitioners can trigger and aid quit attempts, increasing cessation likelihood by up to three times [7]. Primary care is suitable for addressing cessation because smokers frequently attend, and it is an opportunistic and trustworthy setting [8]. The World Health Organization (WHO) recommends that “cessation support and treatment is provided in all health care settings and by all health care providers” [9], especially in primary health care systems as this infrastructure already exists in most countries and has a high population coverage. Clinical guidelines recommend addressing patients’ tobacco use by giving “brief advice” to all patients [10]. The first model for this was the ‘5As’ [11, 12] and in some countries now, the ‘3As’ or ‘Very Brief Advice’ (VBA) is recommended [13] (Appendix 1).

Cancer Research UK recently modelled that if GPs intervened at least once a year with all smokers who attended an appointment (referring smokers to a stop smoking service (SSS) and prescribing a cessation medication), national smoking prevalence in 2030 in England would be 2.5% lower than if current rates of brief advice were continued [11]. Despite evidence-based recommendations and guidelines, internationally there is an ‘evidence-practice gap’ in the rates at which healthcare professionals assess tobacco use and offer support in clinical practice in the real-world [14]. A systematic review of primary care physicians in 17 countries found the following average rates: “65% for ‘Ask’, 63% for ‘Advise’, 36% for ‘Assess’, 44% for ‘Assist’, and 22% for ‘Arrange’ [14].

Integrating smoking cessation treatment into routine clinical care and infrastructures is difficult [5]. Implementation science argues that focused efforts are required to facilitate the movement of evidence-based practices (EBP) (e.g., 5As/VBA) into clinical practice because the contexts that the EBP aims to enter are complex and variable [15]. ‘Implementation strategies’ are “methods or techniques used to enhance the adoption, implementation, and sustainability of a clinical program or practice” [16, 17], e.g., “remind clinicians”, “fund and contract for the clinical innovation” [18]. The Expert Recommendations for Implementing Change (ERIC) programme defined 73 distinct ‘implementation strategy’ categories organised into nine implementation strategy domains [18, 19] (Appendix 2).

A recent Cochrane review [20] evaluated randomised and cluster-randomised controlled trials of implementation strategies designed to increase the rate and quality of delivery of the 5As/VBA to adult primary healthcare patients, when delivered in addition to ‘standard’ cessation support or ‘usual care’. Their primary outcome measure was smoking abstinence at long‐term follow‐up (at least 6 months) and their secondary outcomes were practitioner performance in the 5As and quit attempts. They found moderate-certainty evidence for adjunctive counselling (counselling delivered by a health professional other than the primary care physician), free stop-smoking medications, and tailored print materials increasing quit rates. They found no clear evidence for biomedical feedback, provider training, or provider incentives increasing quit rates. For secondary outcomes, they found some evidence that adjunctive counselling increased cessation medication provision, quit attempts, and arranging patient follow‐up by physicians; free stop-smoking medications increased quit attempts; and mixed results for tailored print-materials regarding quit attempts. They found evidence that provider training increased smoking status recording, cessation advice provision, cessation counselling, and providing self-help materials, but mixed results for participants setting a quit date, cessation medication provision, quit attempts, and arranging patient follow‐up. For multi-component interventions, adjunctive counselling combined with free stop-smoking medications, and adjunctive counselling combined with provider smoking cessation training increased quit rates. Combining provider training with flow sheets to aid physician decision-making also increased quit rates; but the results for secondary outcomes for smoking status recording, cessation medication provision, and physicians arranging patient follow‐up were mixed. Lastly, combining provider training with outreach facilitation had no effect on quit rate, recording smoking status, providing cessation medication, or quit attempts; but had some beneficial effect on participants setting a quit date, providing self-help materials, and arranging patient follow-up.

The Cochrane review did not include observational studies; hence the current review focuses on studies which evaluated the impact of implementation strategies enacted without researcher input, in the ‘real-world’, as a national or state-wide policy or change to clinical guidelines. This review complements the Cochrane review and sought to identify differences in the findings which may be due to barriers to implementation in the real-world, to help explain the evidence-practice gap.

Review questions

The aim of this systematic review was to identify implementation, effectiveness and cost-effectiveness of implementation strategies aiming to increase smoking cessation treatment provision in the primary care setting. Given the evidence-practice-gap, it is important to identify potential facilitators and barriers to implementation. As a secondary outcome, we therefore extracted the proposed facilitators and barriers to effectiveness (qualitative outcomes) from the studies which assessed the effectiveness of implementation strategies on quantitative outcomes.

• RQ1 (Primary): What implementation strategies aiming to increase smoking cessation treatment provision in the primary care setting have been implemented on a national or state-wide scale?

• RQ2 (Primary): Which implementation strategies were effective (practitioner-level and patient-level outcomes) in increasing smoking cessation treatment provision in the primary care setting?

• RQ3 (Secondary): What explanations (perceived facilitators and barriers) have been proposed to explain why certain implementation strategies to increase the provision of smoking cessation treatment in primary care settings were/were not effective?

• RQ4 (Secondary): What is the cost-effectiveness of effective implementation strategies to increase the provision of smoking cessation treatment in primary care settings?

Methods

Protocol and registration

The systematic review protocol was registered on PROSPERO on 1 April 2021  (ID: CRD42021246683), it follows the PRISMA statement [21] (Appendix 3).

Amendments made after protocol registration were: interventions were only included if they involved an implementation strategy enacted on a national/state-wide scale; PhD theses were excluded; key contacts and organisations were not contacted to identify publications not retrieved by the search strategy.

Search strategy

The searches were carried out on 7 April 2021. ‘Smoking’, ‘smoking cessation’ and ‘primary care’ subject headings and key words were used. MEDLINE, EMBASE, CINAHL, PsycINFO, ASSIA Applied Social Sciences Index and Abstracts, Global Health, and Social Policy & Practice were searched for published journal articles. OpenGrey, Social Care Online, and Healthcare Management Information Consortium (HMIC) Database were searched for grey literature. The King’s College London library service and the authors of the related Cochrane review [20] were consulted in developing the search strategy (Full search strategy: Appendix 4).

Forward and backward direct citation tracking was conducted using Web of Science [22]: studies published before 7 April 2021 which cited the included studies, and studies referenced by the included studies, were screened against the inclusion/exclusion criteria.

Article selection

Inclusion/exclusion criteria

Participants/population

The target of the intervention(s) was any type of healthcare professional within the primary care setting. ‘Primary care setting’ was defined as family medicine or general medical practice [20]. Excluded are public health interventions delivered outside primary care practices and interventions delivered in dental settings or pharmacies. Studies including the whole practice patient population were included, as well as those which included specific sub-populations in primary care settings (e.g., people with chronic obstructive pulmonary disease (COPD), diabetes, adolescents, pregnant women). Studies were excluded if outcome data could not be extracted exclusively for the primary care setting.

Intervention/exposure

Articles were included if they evaluated an ‘implementation strategy’ [16,17,18] aiming to increase smoking cessation treatment provision in the primary care setting which was implemented on a national or state-wide scale. The focus of this review was specifically on implementation strategies which were implemented nation-wide or state-wide because we were interested in the scalability of implementation strategies. Articles which assessed local-scale (i.e.: ‘non-national’ or ‘non-state-wide’) implementation strategies were excluded.

Control

Control could be usual care, any other intervention, or before and after designs. Cross-sectional studies without a comparison/control were excluded.

Outcome measures

Articles that assessed any of the primary outcome measures were included in the review.

Primary outcome measures:

• 1. Implementation strategy identification: Description of the implementation strategy was extracted from the article.

• 2. Implementation strategy effectiveness:

• 2.a. Practitioner-level:

• Practitioner performance in 5As/VBA – definitions used by the original studies were accepted:

• • Ask (ask patients about smoking at every visit)

• • Advise (advise all tobacco users to quit)

• • Assess (assess smokers’ willingness to try to quit)

• • Assist (assist smokers’ efforts with treatment and referrals, e.g.: ‘discuss medications’, ‘prescribe medications’, ‘set a quit date’, ‘provide counselling’)

• • Arrange (arrange follow-up contacts to support cessation efforts)

• 2.b. Patient-level:

• • Smokers entering into cessation programmes, facilitated by healthcare professionals in primary care (e.g.: attending smoking cessation clinic or behavioural support appointments; filling prescriptions; calling quit telephone helpline)

• • Smokers setting a quit date or quit attempts

• • Smoking cessation

Secondary outcome measures:

• 1. Facilitators and barriers to effectiveness: Explanations (perceived facilitators and barriers) offered by the original study authors to explain why certain implementation strategies aiming to increase the provision of smoking cessation treatment in primary care settings were/were not effective.

• 2. Implementation strategy cost-effectiveness: Any measures relating to cost-effectiveness or economic indicators of the intervention in the study (which may have involved one or multiple implementation strategy categories).

Date

Date restrictions regarding publication were not applied.

Study design

Non-randomised designs, including comparative observational study designs such as cohort (prospective and retrospective) studies, case–control studies, interrupted time series studies.

Language

English.

Publication type

Published studies and reports were included. PhD theses, conference abstracts, protocols, reviews, systematic reviews, letters, editorials, commentaries, and studies with only qualitative outcomes were excluded.

Screening process

The search results were imported into www.cadima.info and duplicates were removed. Articles were screened by BT at two stages (title/abstract, and full text). Reasons for exclusion were documented at the full text level. The PICO checklist used during screening was piloted between BT and a second screener (PP-H). The second screener (PP-H) screened 200 records at the title/abstract screening stage, and 30 records at the full text screening stage. Inconsistencies were discussed and the inclusion and exclusion criteria were clarified by the two screeners and a third reviewer (LB).

Data extraction

Data from included studies were extracted into a pre-piloted form (Appendix 5). BT performed the data extraction for the 49 included studies, where there were uncertainties about the outcomes, BT consulted LB. Data extraction was performed by BT with oversight from LB. Authors were contacted to provide missing data. Where these data were not provided, they are reported as “missing”.

Risk of bias assessment

The ROBINS-I (Risk Of Bias In Non-randomized Studies of Interventions) tool was used to evaluate the risk of bias in non-randomised observational studies [23,24,25]. BT performed the risk of bias assessments. After the first five studies were assessed, LB also assessed these, and BT and LB compared ratings. The risk of bias assessment ratings and justifications are included in Appendix 6.

The tool assesses risk of bias in seven domains [23]:

• Pre-intervention: (1) confounding, (2) selection of participants into the study.

• At intervention: (3) classification of interventions.

• Post-intervention: (4) deviations from intended interventions, (5) missing data, (6) measurement of outcomes, (7) selection of the reported result.

Then an overall risk of bias rating is decided for each study: low, moderate, serious, or critical risk of bias, or no information available [23].

Synthesis methods

Due to heterogeneity in study populations and outcome measures, a narrative synthesis was used.

Based on the descriptions provided in the included studies, the key aspects of the interventions under investigation were coded to the nine implementation strategy domains and 73 categories developed by the ERIC program [18, 19] (Appendix 2).

Perceived facilitators and barriers, extracted from the studies, were mapped to the determinants in the Consolidated Framework for Implementation Research (CFIR) [26] (Appendix 7).

Results

Study selection

The database search strategy yielded 12,527 records. After de-duplication and screening, 42 studies met the inclusion criteria. Forward and backward direct citation tracking identified an additional seven papers, resulting in 49 papers being included in this review (Fig. 1).

Fig. 1

PRISMA flow diagram. PRISMA flow chart showing the number of papers identified through the search strategy and the study selection process

Study characteristics

Table 1 shows the characteristics of the 49 included studies. Studies were set in the UK (n = 23) [27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49], USA (n = 13) [50,51,52,53,54,55,56,57,58,59,60,61,62], Ireland (n = 4) [63,64,65,66], the Netherlands (n = 3) [67,68,69], Australia (n = 2) [70, 71], Turkey (n = 1) [72], Poland (n = 1) [73], Finland (n = 1) [74], and one [75] compared different policies in the Germany and the UK.

Table 1 Study characteristics

Thirteen were cohort studies [31, 40, 44, 48, 54, 56, 60, 62, 65, 66, 68, 71, 73]. One was a controlled before-and-after study [69]. Three were cross-sectional with a comparator [58, 59, 75]. The other 32 studies were repeated cross-sectional studies. Ten of these used advanced analytical techniques: interrupted time series design [27, 33, 37, 46, 67], segmented regression design [34,35,36], regression discontinuity design [45], difference-in-differences and triple differences design [57]. Thirteen of the repeated cross-sectional studies tested for statistical significance between pre- and post-intervention measurements [32, 42, 43, 47, 49,50,51, 53, 55, 70, 72, 74], and nine only described the pre- and post-intervention measures [28,29,30, 38, 39, 41, 61, 63, 64].

Risk of bias

One study [71] had ‘critical’ risk of bias and was not included in the narrative synthesis as per Cochrane guidance [25]. Twenty-four studies [28,29,30, 32, 36, 38, 39, 41,42,43,44, 49, 51, 52, 55, 61, 63, 64, 66, 69, 70, 72, 74, 75] had ‘serious’ risk of bias, predominantly due to receiving a poor rating for the ‘bias due to confounding’ domain. Twenty studies [27, 33,34,35, 37, 45,46,47, 50, 53, 54, 56,57,58,59,60, 65, 67, 68, 73] had ‘moderate’ risk of bias; these had more sophisticated study designs which controlled for time-varying confounders but did not have a pre-specified/pre-registered analysis plan, scoring a poor rating for the ‘bias in selection of reported result’ domain. Four studies [31, 40, 48, 62] were at ‘low’ risk of bias – these were cohort studies which controlled for various confounders and had pre-specified analysis plans (Appendix 6).

RQ1: Implementation strategies that were implemented

Interventions in six studies [37, 50, 51, 56, 67, 72] used multiple implementation strategies; interventions in the other 42 studies used one key implementation strategy category only (Fig. 2). We did not identify studies for all possible implementation strategy domains and categories which are outlined in the list developed by the ERIC program [18, 19] (Appendix 2). The domains in which implementation strategies were identified were ‘Utilize financial strategies’ (Domain 8., 34 studies), ‘Change infrastructure’ (Domain 9., 14 studies), ‘Train and educate stakeholders’ (Domain 5., three studies), and ‘Engage consumers’ (Domain 7., three studies). More details of the implementation strategy domains and categories are given in Table 2 and summarised below when discussing outcomes for RQ2 and RQ3.

Fig. 2

Implementation strategy categories identified in the included studies. The interventions in the 49 included studies were coded to the implementation strategy domains (1 to 9) and categories (1 to 73) developed by the Expert Recommendations for Implementing Change (ERIC) program [18, 19]. Each column represents one of the 73 implementation strategy categories. A shaded cell indicates the specific strategy that the intervention under investigation in the study involved. Only the four domains that were identified are displayed in this figure. The other domains were not included in any of the studies: Use of evaluative and iterative strategies (Domain 1), Provide interactive assistance (Domain 2), Adapt and tailor to context (Domain 3), Develop stakeholder inter-relationships (Domain 4), Support clinicians (Domain 6). (Wright, 2018) [71] was excluded from narrative synthesis as it was at critical risk of bias

Table 2 Results

RQ2: Effectiveness & RQ3: Perceived Facilitators and Barriers

For conciseness and clarity, we present the effectiveness findings and the key facilitators and barriers proposed by the included studies’ authors together in this section, organised by implementation strategy domain and category. Details can be found in Table 2 and a summary of the facilitators and barriers in Table 3. The extracted quantitative outcomes are included in Appendix 8.

Table 3 Summary of perceived facilitators and barriers

Utilize financial strategies (Domain 8)

Thirty-four studies [27,28,29,30,31,32, 37,38,39,40, 42,43,44,45, 47,48,49,50, 54, 56,57,58, 60,61,62,63,64,65,66,67, 69, 72, 73, 75] evaluated interventions using an implementation strategy that increased funding towards the provision of smoking cessation treatment in primary care.

Fund and contract for the clinical innovation (Category 57)

The six studies in this category investigated policies where primary care practices received funding to deliver national cardiovascular disease prevention programs (including health checks). Two studies were at low risk of bias [31, 48], two moderate [65, 73], and two serious [66, 69].

Effectiveness. For practitioner-level outcomes: two studies showed an increase in smoking status recording [48, 73]; two indicated an increase [31, 69] and one no effect [73] on the provision of cessation advice; and one increased [31] and one had no effect [73] on cessation medication prescribing. For patient-level outcomes, three studies indicated an increase [31, 65, 66] in cessation while one showed no effect [73].

Facilitators/barriers. A perceived barrier was that health check programs focused on the ‘risk factor identification’ and not the ‘intervention’ aspects of cessation treatment [31, 73]. Another barrier was time constraints and insufficient financial recompense for physicians to deliver cessation treatment [31, 66, 69, 73]. Authors noted that there was selection bias in the type of patients who respond to an invitation for a health check [31, 48], but that the value of opportunistic health checks should not be underplayed [31]. A proposed facilitator to increase effectiveness was improved linkages to community-based programmes and support [65, 69] or improved mechanisms for follow-up/monitoring of cardiovascular risk factors in primary care [31, 66].

Place innovation on fee for service lists/formularies (Category 59)

The 10 studies in this category examined changes in insurance schemes which included aspects of smoking cessation treatment [54, 57, 58, 60,61,62, 67] or the introduction of a new smoking cessation medication [37, 63, 64]. One study was at low risk of bias [62], six moderate [37, 54, 57, 58, 60, 67], and three serious [61, 63, 64].

Effectiveness. For practitioner-level outcomes, the introduction of a new cessation medication onto a country’s prescription scheme – NRT in Ireland in 2001 [63, 64], and varenicline in England in 2006 [37] – increased the prescription of the new medication, but did not change overall prescribing of smoking cessation medications. For practitioner-level outcomes, in the USA, increasing access to health insurance coverage which included smoking cessation treatment, increased smoking status recording (multi-state, Oregon) [61, 62], cessation advice provision (Colorado) [58] and cessation medication prescribing (Oregon, multi-state) [54, 60]. In the Netherlands [67], increasing health insurance coverage for smoking cessation also increased cessation medication prescribing. For patient-level outcomes, in the USA, one study (Massachusetts) found no difference in quit attempts [57] but two studies (Oregon, and multi-state) found a positive effect on smoking cessation following the increases in medication prescribing [54, 60]. The Dutch study [67] indicated increased cessation, but evidence for this was less robust. Patient-level outcomes were not measured in the studies assessing the introduction of new medications.

Facilitators/barriers. Perceived barriers were that physician confidence in, and patients’ awareness of, cessation medications was too low [37]. A proposed facilitator of increasing access to health insurance coverage was that this increases access to medications and primary care services [54, 60], which in turn increase the odds that services like smoking status assessment would be performed [62]. Other proposed facilitators included structural characteristics, such as providing sufficient education/training about the 5As/VBA [63], delivering 5As/VBA as an organisational priority and allocating sufficient physician time for it [57, 58, 63].

Alter incentive/allowance structures (Category 60)

Of the 16 studies in this category, two studies in the USA (Oregon, and multi-state) [50, 56] investigated the ‘Meaningful use’ (MU) scheme, which included the introduction of incentive payment for physicians to record their patients’ smoking status and offer cessation assistance alongside other measures (such as changing recording systems) from 2011. The other 14 studies examined various amendments of the Quality and Outcomes Framework (QOF), a pay-for-performance scheme in the UK which financially incentivised GPs to perform certain interventions.

• 13 studies [28,29,30, 32, 38,39,40, 42,43,44,45, 47, 49] investigated the 2004 QOF, which set the following targets: every 15 months record smoking status for patients who have coronary heart disease, diabetes mellitus, COPD, transient ischaemic attack or stroke, asthma, or hypertension; and every 15 months offer cessation advice or referral to a cessation service for these co-morbid patients who smoke.

• One study [30] investigated the 2004, 2006 and 2008 QOF. 2006 amendment: record smoking status in patients without smoking-related morbidity every 27 months rather than ‘ever’. 2008 amendment: chronic kidney disease, schizophrenia, bipolar disorder, and other psychoses were added to the list of smoking-related conditions which required recording of smoking status and cessation advice every 15 months.

• One study [27] investigated the 2012 QOF amendment: offer referral to the National Health Service Stop Smoking Services (NHS SSS) and prescribe pharmacotherapy to all people who smoke, regardless of their smoking-related medical history.

In this category, one study [40] was at low risk of bias, five moderate [27, 45, 47, 50, 56], and ten serious [28,29,30, 32, 38, 39, 42,43,44, 49]; most of the latter did not account for underlying secular trends.

Effectiveness. For practitioner-level outcomes, several studies in the UK for the 2004 QOF found increased smoking status [29, 30, 32, 38, 40, 42,43,44, 47, 49] and cessation advice recording [30, 32, 40, 42,43,44, 49] in primary care for all patients and those who had a QOF-targeted-morbidity. However, one study of survey participants with a QOF-targeted-morbidity found that the recall of receiving cessation advice by patients did not increase significantly [45]; one study found that there was an increase in cessation advice provision to pregnant women who smoked [not direct targets of this policy] but this increase was not sustained long term [39]; and one study which compared the rate of cessation advice recording in primary care electronic health records with the rate of patient recall of receiving cessation advice found mixed results (increase in the former, no effect on the latter) [28]. Two studies found that there was no effect of the 2004 QOF on cessation medication prescribing [an indirect target of the policy] [32, 45], while one found an increase in cessation medication prescribing [40]. The study which assessed the 2006 and 2008 revisions of the QOF only examined practitioner-level outcomes and found no significant effect on smoking status recording or cessation advice provision (but the outcome for the 2008 QOF is less robust) [30]. The one study investigating the 2012 QOF amendment also only examined practitioner-level outcomes and found an increase in the provision of cessation advice and referrals to NHS SSS, but no increase in cessation medication prescribing [27]. In the USA, the two studies (Oregon, and multi-state) [50, 56] examined the introduction of incentive payments via the ‘Meaningful Use’ scheme, however as this intervention included several measures from multiple implementation strategy domains, it is not possible to disentangle individual effects. These studies found an increase for practitioner-level outcomes: an increase in smoking status recording [50, 56], cessation counselling [50] and cessation medication prescribing [50]. For patient-level outcomes, one study indicated an increase in cessation too [50]. In contrast, the only study assessing a patient-level outcome of the QOF 2004 found no effect on cessation [40].

Facilitators/barriers. A suggested barrier to effectiveness on the cessation medication prescribing outcome was incorrect wording/electronic coding of clinical targets [27, 30, 32] – authors recommended that the clinical behaviours and outcome measures targeted are made clearer [27]. Another proposed barrier to effectiveness was the way the implementation outcomes are measured: some authors suggested that any observed increase in cessation advice-giving may not reflect an increase in ‘real life’, but rather more complete recording of advice GPs were already giving [28, 32]. Alternatively, GPs may have increased their provision of cessation advice, but patients were not recognising it as ‘advice’ [28, 39], perhaps due to improper practitioner training on smoking cessation and the delivery of the 5As/VBA [39]. Hence, uncertainty regarding real-world cessation advice provision may be the reason for the mixed effect observed for the cessation medication prescribing outcome. A proposed facilitator was to combine financial incentives with other quality improvement initiatives, such as active dissemination of cessation guidelines and ongoing training [39] and support for front-line staff, within a comprehensive tobacco control strategy [44].

Use capitated payments (Category 65)

The two studies [72, 75] in this category assessed capitated payments (where providers of care are given a set amount of money per patient for delivering clinical care). One [72] (described in more detail in Category 66 below) assessed other measures in addition to capitated payments so it is not possible to disentangle individual effects there. Both studies in this category were at serious risk of bias.

Effectiveness. The two studies [72, 75] found no effect on cessation advice provision, and one [75] found no effect on cessation medication prescribing. No patient-level outcomes were measured.

Facilitators/barriers. A proposed barrier was regarding cultural factors within health infrastructures, one study suggested that physicians did not consider cessation treatment to fit with the “traditional curative model of medicine” and that physicians assume they know the barriers which prevent their patients from quitting [75].

Change infrastructure (Domain 9)

Fourteen studies [34,35,36,37, 41, 50, 52, 53, 55, 56, 59, 67, 68, 72] evaluated an intervention using an implementation strategy that involved infrastructure change aiming to increase the provision of smoking cessation treatment in primary care.

Mandate change (Category 66)

The two studies [36, 72] assessed aspects of wider infrastructure change which aimed to increase smoking cessation treatment provision in primary care. One study [72] investigated a broad national health infrastructure change occurring between 2003–2010 in Turkey (‘Health Transformation Program’) alongside other measures, the other study [36] investigated a change in 2013 to the public health commissioning infrastructure in England (where responsibility for commissioning cessation services was transferred to regional budgets). Both studies were at serious risk of bias.

Effectiveness. Only practitioner-level outcomes were measured. The Turkish study [72] found no effect on the provision of cessation counselling. The English study [36] found a negative effect on the prescribing of any and dual NRT to pregnant women who smoke.

Facilitators/barriers. A proposed barrier for effectiveness was that external policies which indirectly result in the decommissioning of cessation services decrease the stimulus for GPs to discuss smoking cessation and directly prescribe NRT in primary care [36].

Change record systems (Category 67)

The two studies in this category both assessed multiple implementation strategy domains, so it is not possible to disentangle individual effects for this category. Both studies have been described earlier: one examined the ‘Meaningful use’ (MU) scheme in Oregon, USA which included changing recording systems alongside other measures [50], and the other was the broad health infrastructure change in Turkey which included changing recording systems as one of its measures [72]. One study was at serious risk of bias [72] and one moderate [50].

Effectiveness. For practitioner-level outcomes, the Turkish study found no effect on the provision of cessation counselling [72]. The Oregon study [50] found increased smoking status recording, cessation counselling and prescribing of cessation medications. Patient-level outcomes were only assessed in the Oregon study which also indicated an increase in cessation [50].

Facilitators/barriers. A proposed facilitator from the Oregon study which indicated effectiveness was that the change to the recording system aligned well with an existing practice in the clinic (having smoking status as a ‘vital sign’) [50].

Create or change credentialing and/or licensure standards (Category 69)

The seven studies in this category investigated either an expansion to the indications for NRT to new patient populations in primary care [34, 35, 41] or the publication of new/updated national guidelines regarding smoking cessation treatment [37, 53, 55, 67]. Five studies were at moderate risk of bias [34, 35, 37, 53, 67], two serious [41, 55].

Effectiveness. Most of the effects were measured for practitioner-level outcomes. In the UK, the expansion of indications for NRT in 2005 did not increase prescribing of NRT to pregnant women who smoke [41], to adolescents who smoke [34], or to patients who have cardiovascular disease who smoke [35]. Publication of the national guideline related to varenicline in 2007 in the UK increased prescribing of varenicline but had no effect on the overall prescribing rate for cessation medications [37]. In the USA (multi-state), the release (1996) and update (2000) of the national guidelines for the treatment of tobacco use had no impact on the recording of smoking status or cessation advice [53]. In the USA (multi-state), the 2013 national guideline recommendation to provide low-dose computed tomography for lung cancer screening for certain patients who smoke led to an increase in cessation counselling recording and referral to smoking cessation programs, and increased smoking cessation medication prescribing [55] – however the outcome measure for this study may have been confounded by the re-released 2015 national guideline recommendation for clinicians to offer cessation support to smokers. In the Netherlands, the introduction of the first national tobacco treatment guideline in 2007 did not have any significant immediate or long-term trend impact on primary care prescriptions of smoking cessation medications or dispensed prescriptions [67]. The only study to assess a patient-level outcome was the Dutch study, which found no significant effect of this intervention on cessation [67].

Facilitators/barriers. A proposed barrier for the lack of increase on NRT prescribing to patients who have cardiovascular disease who smoke was that external factors – perhaps even the increase in the prescription of varenicline – led to a widespread decrease in prescribing for NRT [35]. Regarding guideline publications, a proposed facilitator to achieve effectiveness was that future guideline changes should be accompanied by other measures which target the time barriers that clinicians face, such as systems-level interventions that can identify patients’ smoking status and support clinicians’ efforts by facilitating referral to resources outside the physician’s office [53].

Change accreditation or membership requirements (Category 71)

Of the five studies in this category, two studies [52, 72] investigated accreditation programs for primary care physicians and were at serious risk of bias. The other three [56, 59, 68] investigated changing accreditation standards for primary care practices and were at moderate risk of bias.

Effectiveness. For the accreditation programs for physicians [52, 72], only practitioner-level outcomes were measured. The program in Turkey (already described) found no effect on the provision of smoking cessation counselling [72]. The other study in the USA (multi-state) [52] found that patient-recalled cessation advice increased significantly post-intervention. For the accreditation standards for practices [56, 59, 68], most of the effects were measured for practitioner-level outcomes. In the two multi-state studies in the USA, there was an increase in the recording of smoking status [56, 59] and provision of cessation interventions [59] following the change of standards which applied to community health centres. However, in one of the studies [56] (described above), incentive payments were also introduced as the change to standards occurred so the effects of individual implementation strategies cannot be disentangled. In the Netherlands [68], the accreditation program for primary care in 2005 had a mixed effect on smoking status recording (no effect for COPD patients, but positive effect for cardiovascular patients) and an uncertain effect on cessation advice provision. This study had no effect on the patient-level outcome: cessation [68].

Facilitators/barriers. A proposed facilitator was that quality improvement interventions may be effective if they are compatible with and integrated into the clinics’ usual culture and systems of care [52]. These may be attractive to physicians because they can take ownership of the tailored improvement plans, but the intervention should be simple [68].

Engage consumers (Domain 7)

Three studies [33, 46, 51] evaluated an intervention which used an implementation strategy that involved engaging people who smoke to raise awareness about the availability of cessation treatment in primary care. The facilitators and barriers are discussed for the domain.

Prepare patients/consumers to be active participants (Category 54)

Of the two studies in this category, one examined the introduction of smoke-free legislation in England [33] because an indirect target was to increase smoking cessation treatment in primary care. The other study in Delaware, USA was the ‘Ask and Act’ program which displayed patient materials in primary care clinics designed to engage patients who smoked, alongside other measures [51]. One study was at moderate risk of bias [33] and one serious [51].

Effectiveness. One study found that prescribing of all smoking cessation medications increased in the months leading up to the introduction of smoke-free legislation, but this increase was not sustained [33]. The other study found an increase in cessation advice recording (practitioner-level outcome) and an increase in cessation (patient-level outcome) following the program which engaged patients who smoked [51], but the authors implied the effect was more likely due to the other ‘educating healthcare professionals’ components of the intervention (Domain 5, below).

Use mass media (Category 56)

The only study [46] in this category evaluated the impact of anti-tobacco mass media advertising and pharmaceutical company-funded smoking cessation medication advertising in England, and was at moderate risk of bias.

Effectiveness. The only relevant finding was that neither intervention had a significant effect on NRT prescribing in primary care (practitioner-level outcome) [46].

Facilitators/barriers: A proposed facilitator was that when engaging consumers, the intervention needs to be sustained for longer durations [33, 46] or that consumers need to be engaged in multiple ways, considering other contextual factors and social norms around tobacco use [33].

Train and educate stakeholders (Domain 5)

Three studies in this domain [51, 70, 74] evaluated an intervention which used an implementation strategy that involved training and educating healthcare professionals in primary care who deliver cessation treatment. The facilitators and barriers are discussed for the domain.

Distribute educational materials (Category 40)

Of the two studies in this category, one examined the ‘Ask and Act’ program in Delaware, USA (mentioned above) which included a measure where cessation materials were distributed to physicians [51]. The other study evaluated an intervention where an educational pack designed to prompt the delivery of smoking status assessment and cessation advice was distributed to GPs in Victoria, Australia [70]. Both studies [51, 70] were at serious risk of bias.

Effectiveness. For practitioner-level outcomes, one study [70] found no effect on smoking status recording but both studies found an increase in cessation advice provision [51, 70]. For patient-level outcomes, one study [51] indicated increased cessation – however, the intervention also involved ‘Conduct educational meetings’ (Category 42) and ‘Prepare patients/consumers to be active participants’ (Domain 7, Category 54), but the authors implied that the effect can likely be attributed to Domain 5.

Conduct educational meetings (Category 42)

Of the two studies in this category, one examined the ‘Ask and Act’ program in Delaware, USA (mentioned above) which included a measure which delivered continuing medical education programs for physicians [51]. The other study evaluated the Finnish ‘National Programme for Chronic Bronchitis and COPD 1998–2007’ where training events were organised for primary healthcare personnel [74]. Both studies [51, 74] were at serious risk of bias.

Effectiveness. For practitioner-level outcomes, one study [74] found a positive effect on smoking status recording and the other study [51] found increased cessation advice recording. The Delaware study [51] also found an increase in cessation (patient-level outcome), but, as aforementioned, this intervention covered multiple implementation strategy categories.

Facilitators/barriers: For this domain, proposed facilitators were the simplicity of the educational material the physicians received [70] and it was suggested that educating physicians in smoking cessation treatment can lead to physicians feeling more comfortable with delivering and billing for cessation counselling [51].

RQ4: Cost-effectiveness of implementation strategies

Some studies [28, 30, 32, 38, 43, 45, 47, 66, 69] included the cost of the interventions but none investigated cost-effectiveness.

Discussion

Summary of evidence

This systematic review aimed to find evidence for the adoption of implementation strategies on a national/state-wide scale, and effectiveness and cost-effectiveness regarding smoking cessation treatment provision and patient smoking outcomes in real-world primary care settings. The 49 included studies assessed only four out of nine implementation strategy domains. The majority of studies identified in this review did not measure patient-level outcomes. We found some evidence for interventions which utilized financial strategies having a beneficial impact on cessation. There were 34 studies which investigated interventions utilizing financial strategies, with only four being at low risk of bias. These appeared to increase the recording of smoking status and cessation advice, but the effect on cessation medication prescribing was mixed. Only one study assessed quit attempts and it found no effect, but seven out of nine studies which assessed smoking cessation found an increase. There were 14 studies which investigated interventions changing infrastructure, none at low risk of bias. These had mixed results for smoking status recording, cessation advice provision and cessation medication prescribing. No studies measured quit attempts, and one out of three studies which assessed smoking cessation found an increase. Only three studies, all at serious risk of bias, investigated interventions which trained and educated stakeholders. These indicated a beneficial impact on smoking status and cessation advice recording, and smoking cessation, but should be interpreted with caution because the evidence was low-quality. There were three studies which investigated interventions engaging consumers, none at low risk of bias. Two studies showed no effect on cessation medication prescribing in primary care. One study assessed cessation advice provision and cessation (both increased), but the intervention also involved implementation strategy categories which involved training and educating stakeholders and the effectiveness was attributed to this latter domain by the study authors. No studies assessed cost-effectiveness.

Authors of the included studies suggested a range of barriers and facilitators. Some key facilitators were the simplicity of the intervention and external policies/incentives which were complementary to the smoking cessation aims of the intervention (such as, wider tobacco control measures and funding for public health and cessation clinics) and having the ability for physicians to refer smokers to cessation programs or community-based support. Some of the key barriers included time and financial constraints, lack of free cessation medications and follow-up, deprioritisation and unclear targets in primary care, lack of knowledge of healthcare professionals, and insufficient messaging to patients about available cessation support options. Some of the key barriers identified were similar to those identified recently by the UK Royal College of Physicians [76].

This review complements the findings of a recent Cochrane review [20] which evaluated randomised and cluster-randomised trials of similar interventions but in controlled environments. There appears to be a ‘gap’ between the implementation strategies that have been enacted on a national/state-wide scale (identified by this review) and those demonstrating efficacy in trials [20]. While trials indicated efficacy of adjunctive counselling and tailored print materials on quit rate [20], no studies have assessed these interventions in national implementation.

Trials found a beneficial impact of adding cost-free medications to standard cessation support on smoking quit rates and quit attempts [20]. Regarding real-world implementation, this review found some evidence that increasing access to health insurance which included coverage for smoking cessation treatment had a beneficial impact on the recording of smoking status, the provision of cessation advice and cessation medications, and cessation. The only study which assessed quit attempts found no effect [57]. Where new free cessation medications were introduced, prescribing of the new medication increased but there was no change in overall prescribing for cessation medications (other outcomes were not assessed).

Trials found no clear evidence that provider incentives could increase smoking cessation [20]. In real-world implementation of financial incentives studied in this review, cessation outcomes were only assessed in two out of 16 studies (one showed an increase [50], one no effect [40]). We found evidence that a nationally implemented financial incentive for GPs was effective in increasing the recording of smoking status and cessation advice, and (in one study [27]) referral to cessation services; however, there was a mixed effect on cessation medication prescribing and smoking cessation. We also identified studies where primary care practices received funding to deliver national cardiovascular disease prevention programs (including health checks); these overall indicated increased smoking status recording, cessation advice and cessation medication provision, and cessation. There was no robust evidence regarding capitated payments.

Trials found some evidence for provider training, either individually or in combination with other interventions: the former having some beneficial impact on smoking status recording, cessation advice provision, cessation counselling, and providing self-help materials; the latter, a beneficial impact on quit rates and some outcomes of cessation assistance (setting a quit date, providing self-help materials, and arranging patient follow-up) [20]. We identified some low-quality evidence of provider training as a ‘real-world’ intervention (three studies, all at serious risk of bias), having a beneficial impact on smoking status recording, cessation advice recording, and cessation.

Strengths and limitations

A robust approach was used to identify and synthesise relevant literature using a pre-registered protocol and a comprehensive search strategy. However, the search strategy may not have identified all relevant papers, because different terminology exists internationally for ‘primary care setting’ and no effective observational study filter exists [77]. To mitigate against this, the search terms from a recent Cochrane review [20] were used, the search strategy was piloted, and backward and forward citation tracking of included studies was conducted. A limitation is that only articles in English were included.

This systematic review investigated the scalability of national and state-wide policies, where policies were implemented without researcher input over large geographical areas, potentially diverse in patient and provider characteristics. This review evaluated observational studies which, whilst at risk of bias and unable to demonstrate causality, can provide evidence of real-world implementation. A large number of studies were included in the evidence synthesis, however, only half were at moderate or low risk of bias. Despite an international scope, most studies were set in the UK and the USA. In six studies, the intervention involved multiple implementation strategy categories and it was challenging to disentangle their individual effects.

Implications and recommendations

Our findings indicate that during the development of future implementation strategies, a significant consideration should be given to the current demands of the primary care setting, such as existing time constraints and clinical priorities; future implementation strategies should better align with existing technologies and the routine systems in place; and the clinical outcomes which are targeted should be clearly communicated. We recommend profiling, both in the clinic and in government papers, that smoking cessation is a key priority and that various cessation support is available.

Future research could investigate the five implementation strategy domains not identified by this review (‘Use of evaluative and iterative strategies’, ‘Provide interactive assistance’, ‘Adapt and tailor to context’, ‘Develop stakeholder inter-relationships’, ‘Support clinicians’) and the strategies that were efficacious in the controlled-trial setting [20]: adjunctive counselling and tailored print materials. However, we recommend that the perceived facilitators and barriers identified by this review are considered when designing interventions.

We advise that hybrid effectiveness-implementation designs [15] are used, where studies robustly assess both the effectiveness of implementation strategies on (practitioner-level) provider performance as well as (patient-level) smoking outcomes. Additionally, we recommend measuring ‘advice provision about e-cigarettes’ as an additional outcome – due to the relative novelty of e-cigarettes being recommended as harm reduction tools in clinical guidelines (in 2021 in the UK [6] and Australia [78]), none of the studies in this review investigated this. Lastly, we recommend using methods such as Multiphase Optimization Strategy (MOST) [79], which consider the time and resource constraints of clinical settings, and verify that all the components of the 5As/VBA or the proposed implementation strategy interventions are optimised and cost-effective.

Conclusions

This systematic review aimed to find evidence for the adoption, on a national or state-wide scale, of implementation strategies aiming to increase smoking cessation treatment provision in real-world primary care settings. The implementation strategies identified involved utilizing financial strategies, changing infrastructure, training and educating stakeholders, and engaging consumers. The first three strategies appeared to increase the rate of smoking status recording and cessation advice provision in primary care. The most amount of evidence was identified for the utilizing financial strategies domain, which also appeared to increase smoking cessation.