Introduction

Stroke is the third leading cause of death worldwide and a major cause of severe adult disability in developed countries [1], [2]. For acute ischaemic stroke, rapid thrombolysis within 4.5 hours can improve the prognosis, with greater benefit from earlier treatment [3], [4]. An expeditious response to stroke symptoms is critical [5]. Many patients present too late to hospital due to a failure, for a variety of reasons, or inability to call emergency services rapidly [5]–[7].

Mass media interventions are promoted as an effective method of improving awareness of health issues and changing behaviour, including encouraging appropriate use of services [8]. A systematic review of 20 evaluations of mass media interventions for a range of conditions reported a positive effect on health services utilisation [8]. Such interventions targeted at stroke have a positive impact on knowledge, including awareness of the need for an emergency response, but with little impact on public behaviour [9].

In February 2009, the Department of Health in England launched the Face, Arm, Speech, and Time (FAST) mass media campaign, to raise public awareness of stroke, specifically its symptoms and the need for an emergency response [10], [11]. An adapted version of the FAST mnemonic (used to screen for the presence of stroke in clinical practice) was used to convey information on three typical stroke symptoms (F = facial weakness, A – arm weakness and S = speech disturbance) and the desired behavioural response (T = time to call emergency services if you recognise any of one of the stroke symptoms). The damage to the brain from stroke was illustrated with an evocative image of a fire rapidly spreading in the head of an older adult depicted. The need for an expeditious response was further emphasised with the aphorism ‘ACT FAST’ along with the key message “The faster you act the more of the person you save”.

Phase one was launched on 9^th February 2009 and ran until end of March 2009 with television, press, and radio advertisements. A second phase ran from November 2009 to December 2009; a third phase during February 2010 to March 2010; with a fourth phase during March 2011. Fifth and sixth phases ran from February 2012 to March 2012, and during March 2013. The most recent phase took place during March 2014. Whilst the overall cost of the campaign to date is unclear, and unlikely to be the figure of £105 million in funding over three years quoted in a Wellcome Trust article [12], the costs are not insignificant; the Department of Health stated that the more restricted three month advertising campaign in phase 4 cost £740,000 [13].

Few studies have evaluated the impact of the campaign on behaviour, although face-to-face structured interviews with members of the public in England conducted immediately before and after the first phase identified increases in awareness of stroke signs, the FAST mnemonic, and behavioural intentions to seek emergency medical care for stroke symptoms [7], [14]. A questionnaire survey administered in one mixed urban/rural population in England also reported high-levels of awareness of the campaign (70%) and FAST symptoms (other symptoms, leg weakness and visual loss that were not highlighted in the FAST campaign, were poorly recognised as signs of stroke) [15]. An audit of ambulance trusts in England reportedly attributed an average increase of 50% in absolute numbers of emergency calls categorised as stroke according to paramedics between April to June 2008 (N = 2020) and April to June 2009 (N = 3040) to the campaign [16]. No effect of the campaign on speed of presentation or numbers of patients thrombolysed were reported by single site studies in England [17], [18].

An evaluation (using observational data and modelling) commissioned by the Department of Health attributed a range of impacts on public and professional behaviour to the FAST campaign within the first year: (i) increased numbers of stroke-related emergency calls (55% over the first four months); (ii) increased numbers of patients presenting earlier to hospital (9900) and receiving specialist treatment (2500); (iii) reduced stroke-related morbidity and mortality (640 patients); (iv) gains in quality adjusted life years (2200); and (v) a return on marketing investment of £3.20 for every £1 spent on the campaign [19]. However the time frame and assumptions of the modelling process have not been clearly described.

Attributing these findings to the campaign is difficult since they are cross-sectional surveys [15], clinical outcomes are based on modelling as opposed to objective data [19], or only report changes in summary statistics between pre- and post-intervention periods [14], [16]–[18]. Consequently, they fail to provide an estimate of effect that takes into account the influence of time trend [8]; for example thrombolytic treatment rates in England were already increasing rapidly before the FAST campaign due to service developments [20].

In order to account for the influence of the campaign over and above background time trend, we aimed to evaluate the impact of three consecutive phases of the FAST campaign using objective measures of behaviour at the population level in England with a time series evaluation.

Methods

Results

There was no evidence of autocorrelation in the regression models for the main analyses. The results are presented graphically in figures 1 and 2 and summarised in tables 1 and 2. Regression models adjusted for seasonality are shown in Tables S1 and S2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Time series graphs of monthly resource utilisation by the Stroke Association over the study period (May 2007 to February 2011).

A: Absolute numbers of website page views, website hits and information materials dispatched by the Stroke Association. B: Absolute numbers of telephone calls received by the Stroke Association helpline. S1: period before the campaign (May 2007 to Feb 2009). S2: period of no campaign activity after phase one of the campaign (Mar 2009 to Oct 2009). S3: period during phases two and three of the campaign (Nov 2009 to Feb 2010). S4: subsequent period with no campaign activity after phase three (Mar 2010 to Feb 2011). Yellow vertical bars represent the time periods for the different phases of campaign activity.

https://doi.org/10.1371/journal.pone.0104289.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Time series graphs of monthly emergency admissions for stroke and thrombolysis activity over the study period (May 2007 to February 2011).

A: Absolute numbers of emergency admissions with a primary diagnosis of stroke in England. B: Absolute numbers of patients receiving thrombolytic treatment in England (centres that submitted data to the SITS register throughout the study period). S1: period before the campaign (May 2007 to Feb 2009). S2: period of no campaign activity after phase one of the campaign (Mar 2009 to Oct 2009). S3: period during phases two and three of the campaign (Nov 2009 to Feb 2010). S4: subsequent period with no campaign activity after phase three (Mar 2010 to Feb 2011). Yellow vertical bars represent the time periods for the different phases of campaign activity.

https://doi.org/10.1371/journal.pone.0104289.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Interval estimates of segmented regression coefficients for changes in data for different time periods.

https://doi.org/10.1371/journal.pone.0104289.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Summary of statistically significant (p<0.5) changes in data for different time periods.

https://doi.org/10.1371/journal.pone.0104289.t002

Discussion

Statistically significant time trends (monthly increases) before the campaign were reported for emergency hospital admissions for stroke (overall and A&E) and thrombolysis activity from hospitals in England that submitted data to the SITS-UK database, but not for access to information about stroke from the SA. There was also a significant monthly decline before the campaign in numbers of emergency hospital admissions for stroke via general practitioners (GPs) - where the public contacted a primary care practitioner as the first response to stroke symptoms, rather than initiated an immediate 999 call. These underlying trends may in part be explained by (i) the publication of the National Stroke Strategy in December 2007 [10], with subsequent national media attention; (ii) discussions and preparation pertaining to implementation of the National Stroke Strategy within stroke services at the regional level across England; (iii) additional factors influencing the screening of patients for stroke by paramedics (ambulance service national Care Quality Commission audit, and the proposal for emergency response times for suspected stroke occurring less than three hours prior to a 999 call to be changed from category B [serious, but not immediately life-threatening – response time 20 minutes] to a category A [life threatening – response time 8 minutes]) connected with the National Stroke Strategy [26], [27]); and (iv) service development with regards to thrombolytic treatment [20].

We found evidence of increased public access to stroke-related information (website hits, webpage views, information materials dispatched by the SA, and calls to their helpline), increased overall/A&E emergency admissions, decreased emergency admissions via GPs, and increased thrombolysis activity that were attributable to the FAST campaign over and above underlying trends. Although adjusted analyses indicated that the SA datasets may have been influenced by seasonality, emergency hospital admissions for stroke (overall and via A&E) increased significantly immediately after the initial phase, whereas emergency hospital admissions via GPs and thrombolysis activity showed statistically significant decreases and increases respectively after the third phase of the campaign. In contrast to a retrospective audit of stroke patients presenting to a single hospital in England [17], we found a statistically significant increase in thrombolysis activity.

Our findings provide evidence of an enduring impact of the campaign in England on public behaviour (i.e. increased awareness of stroke and more people arriving at secondary care, due to fewer people contacting GPs as the first response to stroke symptoms) and greater numbers of appropriate patients with stroke symptoms arriving at specialist centres within the stroke onset time to treatment window for evidence-based treatments such as thrombolysis.

In contrast to our findings, a time series evaluation of a similar FAST campaign rolled out in Ireland during 2010 was not shown to have a sustained impact on emergency department attendance [28]. Mass media campaigns targeting stroke in the Czech Republic [29] and the USA [30] have also met with limited success.

The reasons for the differences between our findings, and evaluations of other campaigns are complex and multi-factorial, and as previously noted their dual nature (public and professional) makes it difficult to elucidate active components that might explain any reported impact [9]. It has been posited that effectiveness of mass media campaigns targeting stroke may be enhanced by spend on media, media mix, and key messages [14]. Message content (e.g. possibility of acute treatments such as thrombolysis [31]) and the type of language (used by patients/bystanders versus professional) used to describe common stroke symptoms [31] may also impact on effectiveness. Furthermore, failure to present details of theoretically grounded development and piloting of the interventions [9] further compounds identification of the optimal active ingredients of effective campaigns.

The message content of FAST in England did not focus on overcoming barriers to performing the desired response (call an ambulance) or adequately address response efficacy (i.e., beliefs associated with executing the desired response by explicitly referring to the availability of an effective emergency treatment such as thrombolysis). Additional impact may be observed with the addition of these components to the message content.

The measures of emergency admissions and thrombolysis activity directly relate to impact on patients with incident stroke and subsequent outcomes, whereas measures of information seeking (access to information resources from the SA) is more indicative of behaviour of people who wish to know if they (or their services) should have done something different when they (or others) experienced a stroke.

These findings are important, but should also be interpreted with some caution. Time series designs increase the confidence that estimates of effect can be attributed to the intervention [32]. Nevertheless, there is always a chance that patterns in these observational data were due to unobserved confounding variables; not all sites in England enter data onto the SITS database and it is possible that thrombolysis activity at these sites had different trends. The number of hospital sites in England that contributed data to the SITS-UK database during 2007, 2008, 2009, 2010 and 2011 were 50, 84, 112, 124 and 89 respectively. However, analyses based on 27 centres contributing thrombolysis activity data throughout the observation period enable a more accurate estimate of an effect of the campaign on public behaviour (changes in thrombolysis activity in other centres might have been due to internal hospital factors). Furthermore, in line with national guidelines [27] the proposed change in emergency 999 response times for suspected stroke was implemented in March/April 2009 (stroke occurring less than three hours prior to a 999 call was changed to a category A (life threatening – response time 8 minutes), which may have led to increased screening of patients for stroke by 999 dispatchers and paramedics, which resulted in more patients reaching hospital within the timeframe for thrombolysis.

Overall, our analyses show a statistically significant impact of phase one of the FAST campaign over and above underlying time trends on information seeking behaviour on stroke (via the SA data) and emergency admissions for stroke (overall and cases admitted via A&E). Our analyses also provide evidence that subsequent phases supported the maintenance and augmented the impact of the initial phase on information seeking behaviour, emergency hospital admissions via GPs and thrombolysis activity. The statistically significant impact on the latter two outcomes after the third phase is likely due to the close proximity of the second and third phases, in effect a ‘double-dose’ of the intervention.

Nevertheless, our findings for magnitudes of the direct effects for phase one and subsequent phases in terms of their clinical significance are at best modest, indicating that the clinical significance and cost-effectiveness of the campaign was substantially less than that modelled by a previous evaluation [19].

Analyses exploring the impact of time from onset of stroke symptoms to time to arrival at hospital would be important (as earlier treatment with thrombolysis increases likelihood of a good outcome), but data on onset time of the stroke are not available in any routine system. Similarly, data reporting on changes in time from arrival to hospital to time of receipt of thrombolysis (door to needle times) are also not routinely available; however door to needle time is not relevant to the aim of the campaign, which targeted emergency response to symptoms of stroke by members of the public.

The underlying assumption is that if more people with stroke symptoms arrive rapidly at secondary care that outcomes will improve through the increased use of evidence-based treatment such as thrombolysis (as identified by our analysis of emergency admissions and thrombolysis activity, although as stated above not all sites in England enter data onto the SITS database). Using the figures for the statistically significant step change increase (FAST Phase one) between February 2009 and March 2009 for overall emergency stroke admissions in England (505 patients, 95% CI = 75 to 935) as an exemplar allows a very conservative estimate (that takes no account of patients arriving at hospital more rapidly and being treated earlier within the 4.5 hour time window for thrombolysis and the resultant increased likelihood of benefit from treatment) of clinical impact on the campaign. Assuming 85% of these cases will be acute ischaemic stroke (505×0.85 = 429 patients), and further assuming a thrombolysis rate of 12% [33] (429×0.12) then 51 additional patients may have received thrombolysis. The mean number needed to treat for a good outcome (independence) is 8 [3], which equates to an additional (51/8) 6 patients in England who may have had good outcomes from treatment. Applying the confidence interval ranges for the step change between February 2009 and March 2009 results in a lower and upper conservative estimate of 1 and 12 additional patients who may have received a good outcome from treatment with thrombolysis.

The conclusions are strengthened by the inclusion of three independent data sources. Segmented regression analyses enabled us to estimate the effect of FAST on multiple objective measures of behaviour, which further highlights the importance of robust evaluation of mass media interventions [8], [9]. Reliance on findings from evaluative studies reporting on awareness and behavioural intentions, small single site studies, and modelling techniques can lead to false confidence in the effectiveness of this type of complex intervention on health outcome benefits and cost-effectiveness. The findings from our study emphasise the importance of robust evaluation and this should extend to all mass media campaigns that represent significant investment of resources.

The need for continuous advertising to sustain public stroke awareness has been highlighted previously [34], [35]. Additional evaluation studies of subsequent phases of the FAST campaign (phase four during March 2011, with fifth, sixth and seventh phases rolled out during February and March 2012, March 2013 and March 2014 respectively) would be helpful to establish if subsequent phases yielded any additional statistically significant impact or sustain the impact of previous phases on public behaviour.

Theory-based interventions targeting the entire range of stroke symptoms, barriers to calling an ambulance, and specific information on the availability of effective secondary care treatments such as thrombolysis, may be warranted to further augment the impact of the campaign message on clinical outcomes [36].

In order to elucidate causal mechanisms that sustain the impact of mass media interventions on public behaviour, designers, policy makers, clinicians and researchers in the field of stroke (and other disciplines where the design of mass media interventions are planned) should utilise appropriate theory and adhere to a structured development process, and undertake robust evaluation before, during and after the roll-out phases using objective measures of behaviour at the population level.

Conclusions

We used the optimal research design to retrospectively evaluate the impact of the FAST campaign, which enables analysis of impact over and above any underlying time trend (interrupted time series evaluation of appropriately available behavioural data at the population level in England). There was an initial statistically significant impact of phase one on information seeking behaviour of the public about stroke and emergency admissions for incident stroke, with additional statistically significant impact on information seeking behaviour, emergency hospital admissions (reduction in inappropriate care seeking behaviour of the public, i.e. contacting a primary care practitioner as the first response to stroke symptoms), and thrombolysis activity (arguably the most important measure of impact directly related to clinical outcomes). The FAST campaign has been promoted internationally as a great success [19], [37]; although our analysis suggests more modest effects in terms of its clinical impact. Consequently, the clinical impact and cost-effectiveness of the FAST campaign in England remains unclear and future campaigns should be accompanied by formal evaluation of impact on clinical outcomes.

Supporting Information

Acknowledgments

We would like to express our thanks to: Nikki Hill (Deputy Director Communications, The Stroke Association); Kennedy R Lees (Professor of Cerebrovascular Medicine, Acute Stroke Unit, University Department of Medicine and Therapeutics, Gardiner Institute, Western Infirmary and Faculty of Medicine, University of Glasgow); Barbara Coyle (Senior Information and Research Manager, The North East Public Health Observatory); Leanne Woolmer (Administrator, Institute for Ageing and Health, Newcastle University) and Justin Presseau (Lecturer in Health Psychology, Institute of Health and Society, Newcastle University). GAF is supported by an NIHR Senior Investigator award.