Article Text

This article has a correction. Please see:


Can a media campaign change health service use in a population with stroke symptoms? Examination of the first Irish stroke awareness campaign
  1. Lisa Mellon1,2,
  2. Anne Hickey1,
  3. Frank Doyle1,
  4. Eamon Dolan3,
  5. David Williams2
  1. 1Department of Psychology, Division of Population Health Sciences, Royal College of Surgeons in Ireland, Dublin, Ireland
  2. 2Department of Stroke and Geriatric Medicine, Royal College of Surgeons in Ireland, Dublin, Ireland
  3. 3Department of Medicine for the Elderly, Connolly Hospital, Dublin, Ireland
  1. Correspondence to Lisa Mellon, Department of Psychology, Division of Population Health Sciences, Royal College of Surgeons in Ireland, Dublin 2, Ireland; lisamellon{at}


Background Mass media campaigns for stroke awareness encourage the public to recognise stroke symptoms and respond to stroke in a timely manner. However, there is little evidence to suggest that media messages can influence behaviour after stroke onset. The F.A.S.T. (Face Arm Speech Time) test is a common stroke recognition tool used in public education campaigns.

Objective To assess the impact of the F.A.S.T. campaign on health service use in Ireland, which has had no previous exposure to a F.A.S.T. media campaign.

Methods An interrupted time series design was used to detect behaviour change after the introduction of the first Irish F.A.S.T. campaign in presentations of patients with suspected stroke to two emergency departments (EDs), serving a population of about 580 000.

Results There was a significant change in ED attendance of patients with reported stroke symptoms after the introduction of the F.A.S.T. campaign (β=0.84, 95% CI 0.43 to 1.24; p<0.001), although this was not sustained. ED presentation within 3.5 h was associated with emergency medical services activation (OR=3.1, p<0.001) and self-referral to the ED (OR=2.67, p<0.001).

Conclusions This first Irish F.A.S.T. campaign had an initial impact on ED attendance of patients with stroke symptoms. However, the campaign effects were not sustained in the long term. Results indicate that prehospital delay in accessing acute stroke services is a complex process with involvement of factors other than stroke knowledge and intention to call 911.

  • stroke
  • pre-hospital
  • thrombolysis

Statistics from


Thrombolysis with tissue plasminogen activator within 3–4.5 h after onset of ischaemic stroke is associated with improved patient outcome and reduction in disability.1 The chances of a favourable outcome fall twofold for every 90 min delay in treatment of acute ischaemic stroke.2 However, despite a large body of evidence supporting its efficacy, thrombolysis rates have remained consistently low in Ireland3 and internationally.4 Studies have sought to identify barriers to thrombolysis use, with patient delay in seeking medical attention highlighted as the main prehospital factor accounting for its underuse.5 ,6 In comparison with any other intervention (such as the use of stroke teams or faster emergency medical services (EMS) triage),7 reducing prehospital delay should increase the use of thrombolytic agents and remains a significant challenge in acute stroke care.

Media campaigns are used to increase stroke recognition and change health behaviour because they can reach large target audiences with behaviourally focused messages. Evidence supports the role of mass media in inducing changes in health services use.8 Media-driven stroke campaigns are commonly used to increase stroke knowledge and awareness. The principal outcome of increasing public stroke knowledge is to reduce prehospital delay. Evaluations of previous media-driven stroke awareness campaigns have shown that a campaign has demonstrable effects on knowledge.9 ,10 However, stroke awareness campaigns may only have a limited direct effect on behaviour by reducing prehospital delay and increasing activation of EMS.11 This suggests that improved stroke knowledge is not consistently associated with intention to call 911, therefore a ‘disconnection’ may exist between the translation of symptom recognition into appropriate action during acute stroke onset.12

A deficit in the knowledge of stroke within the Irish population, which has about 10 000 strokes a year (one every 60 min) has been highlighted previously and the need for a stroke awareness campaign has been identified.13 The Irish Heart Foundation in 2010 funded the Act F.A.S.T. (Face, Arm, Speech, Time) stroke awareness campaign, the first of its kind in Ireland.14 The original F.A.S.T. message was created as a short, easy to remember mnemonic for stroke education in 1999, and is based on the Cincinnati Pre-Hospital Stroke Scale, developed for EMS personnel.15 It highlights the common warning signs of stroke and outlines the correct behavioural response after stroke recognition: to call emergency services, and to do so immediately. The F.A.S.T. message is considered to be an appropriate message for stroke education given its brevity.16 However, it has been criticised owing to its inability to identify 100% of patients with a stroke and its poorer performance in cases of haemorrhagic stroke.17

An intervention to change behaviour must have behaviour, or a measurable consequence of behaviour, as its end point.18 We aimed to assess if exposure to this first national stroke awareness campaign using the F.A.S.T. message had an impact on health service use, and if it was associated with target behaviours promoted by the media campaign, particularly EMS transportation and rapid action upon recognition of stroke symptoms.


The F.A.S.T. media message is widely used as a public stroke education message. The first Irish targeted public F.A.S.T. campaign for stroke was broadcast between May 2010 and June 2011, through national television and low-level regional radio advertising. The campaign used the Act F.A.S.T. message, using a voiceover with a local accent. There were three major waves of the media campaign during the study period with continuous television advertising for 3-week periods in May 2010, August 2010 and January 2011. The campaign was high volume with an average 73.4 gross rating points (GRPs) for the study period. The GRP is a measurement of the size of the audience exposed to a particular media message. It is expressed as a percentage and is calculated by multiplying the reach of the campaign (the percentage of homes or people viewing the campaign) by the frequency (the number of times it was aired). A higher GRP indicates greater population exposure to the media campaign.

Ethical approval was granted for a retrospective anonymous study of all emergency department (ED) attendances to two large teaching hospitals in north Dublin city during 1 year. The study hospitals serve a population of about 580 000 and both provide a routine thrombolysis service. All ED attendances were manually screened from ED registers to identify cases where the patient presented with symptoms of a stroke, or where they perceived their symptoms to be a stroke. Interhospital transfers from peripheral regional centres were excluded from the screening process as information on initial symptom onset was not available. The following key words were used: slurred speech, facial droop, query stroke, collapse, numbness/weakness/power loss in arm/leg, dizziness, confusion, visual disturbance and previous stroke. Trained ED clerical personnel recorded the reason for attendance on the ED register, from first-hand patient information, EMS documentation or general practitioner correspondence. Information on each presentation relevant to stroke was extracted from ED case notes, including demographic details, time of symptom onset, mode of transport to hospital, symptoms, neuroimaging, thrombolysis administration, stroke confirmation and destination after discharge from ED. Diagnosis of ischaemic stroke, haemorrhagic stroke and transient ischaemic attack were confirmed using hospital discharge codes. Time of symptom onset was recorded from ED notes. For patients with wake-up stroke, the time of symptom onset was taken as the time the patient was last seen well. Onset-to-door time (OTD) was defined as the time from symptom onset to time of ED presentation. Full data were not available for all ED presentations owing to lack of documentation during ED admission.

Statistical analysis

Descriptive statistics and χ2 test for categorical variables were used to examine the association between patient characteristics and OTD, which was categorised into OTD ≤1 h, OTD >1–3.5 h, OTD >3.5–6 h, OTD >6–24 h, OTD >24 h and unknown time of symptom onset. A predictive model of OTD ≤3.5 h was developed by entering all factors into a univariate and multivariate model of logistic regression analysis. The length of time, 3.5 h, was chosen as the thrombolysis eligibility window based on guidelines which recommend a realistic maximum of 60 min from ED presentation to thrombolysis administration.19 ,20 A p value<0.05 was considered statistically significant. Data for wake-up stroke were not included in the regression analysis owing to unreliability of the OTD time.

An interrupted time series design using a segmented Poisson regression model was used to detect if there was a change in the number of ED attendances of patients with stroke symptoms after the introduction of each wave of the F.A.S.T. campaign. The increase or decrease in slope of the trend after each wave of the campaign was also examined. This approach was also used to assess changes in both the level and trend (slope) of each behavioural indicator of campaign impact: (1) ambulance arrival and (2) presentation within the thrombolysis window (<3.5 h), after introduction of each wave of the F.A.S.T. campaign. Data were tested for autocorrelation, and negative binomial regression was used where likelihood ratio tests indicated that the data were overdispersed. All data analysis was conducted using Stata Release V.11.0.


From 1 March 2010 to 28 February 2011, 870 patients with reported stroke symptoms were admitted to the ED. Of these, 434 (49.9%) were confirmed stroke/ transient ischaemic attack on clinical presentation, neuroimaging and discharge diagnosis. OTD times were available for 803 (92.3%) of the sample. Fourteen patients were documented as awakening with symptoms, of whom 10 had a stroke diagnosis. Figure 1 outlines diagnoses for the sample. The mean age (±SD) for the sample was 65.0 (±17.2).

Figure 1

Classification of emergency department (ED) presentations by diagnosis. TIA, transient ischaemic attack.

Three hundred and sixty-seven (42%) of 870 ED presentations of patients with stroke symptoms were within the thrombolysis eligibility window (≤3.5 h). Of patients with a known OTD time (n=803), 115 (14.3%) arrived within 1 h, 252 (31.4%) presented at >1–3.5 h, 95 (11.8%) presented at >3.5–6 h, 140 (17.4%) at >6–24 h and 201 cases (25.o%) arrived more than 24 h after symptom onset. Of those with a confirmed stroke diagnosis and a documented OTD time, 141 (52%) presented within 3.5 h. The thrombolysis rate for the sample was 8.7%. Arrival by ambulance accounted for 55.1% of all ED presentations. Of all ED presentations, 716 (82.3%) of those reporting stroke symptoms required hospital admission. The most common symptom reported for all ED presentations was speech disturbance, reported in 431 cases (50%). Unilateral weakness was reported in 46% of cases, followed by facial droop (37%). Sample characteristics for patients with a documented OTD time are presented in table 1 by OTD.

Table 1

Patient characteristics according to onset-to-door time

Factors related to OTD ≤3.5 h

Univariate logistic regression analysis is presented in table 2. Age (p<0.001), ambulance arrival (p<0.001), self-referred presentations (p<0.001) and stroke diagnosis (p=0.001) were significant predictors of OTD ≤3.5 h at the univariate level of analysis. There was no significant association with gender (p=0.12), weekend presentation (p=0.06) and exposure to the F.A.S.T. campaign (p=0.21). A multivariate model of OTD ≤3.5 h showed a significant association for ambulance arrival (p<0.001) and self-referred presentations (p<0.001).

Table 2

Factors associated with onset-to-door time (OTD) ≤3.5 h in univariate and multivariate logistic regression (N=803)

Interrupted time series analysis

Before the start of the campaign there was an average weekly admission rate of 11 patients with reported stroke-like symptoms. There was a significant moderate-level trend for ED activity after initiation of the first wave of the F.A.S.T. campaign (β = 0.84, 95% CI 0.43 to 1.24; p<0.001) and a significant change in the slope after wave 1 (β =−0.079, 95% CI −0.14 to −0.01, p<.05), with a drop in ED activity after withdrawal of the advertising (figure 2). During wave 1 of the campaign the average weekly admission rate increased to 31.3 patients a week, with a post-campaign period rate of 19.6 patients a week. There was no evidence of a level trend or slope effect for waves 2 and 3 of the campaign. Two separate Poisson regression models were fitted for campaign effect on the number of ED arrivals by ambulance and presentation within the thrombolysis window. In both behavioural indicants, the model did not detect significant changes in the level and trend of presentation (data not shown).

Figure 2

Emergency department (ED) activity trends over the F.A.S.T. (Face Arm Speech Time) campaign period.


This study illustrated the limited efficacy of media campaigns in translating knowledge and intention into appropriate action. Previous research has similarly reported that media campaigns for stroke awareness have been related to increased understanding of stroke symptoms, but no sustained lowering of response times or mortality rates has been demonstrated.12 The F.A.S.T. message outlines two correct behavioural responses when stroke occurs: to call 911/999 for EMS assistance, and to do so immediately. In our analyses, there was no change in hospital ED activity trends for either behaviour, and campaign exposure was not associated with presentation within the thrombolytic window. It might be that even though individuals did carry out the correct action by telephoning for an ambulance, there was still a time delay and hence the campaign had little impact on overall emergency response.

An intervention to change behaviour, such as a media campaign, must have behaviour, or a measurable consequence of behaviour, as its end point. Few studies such as ours have examined the direct impact of a media campaign on hospital activity, the most clinically relevant indicator of campaign efficacy. It is difficult to identify a direct relationship between a media campaign and individual ED presentations, although a population level change in service use for stroke may indicate appropriate behaviour change. From an acute stroke treatment perspective, both behaviours promoted by the campaign—EMS transportation and rapid action—need to occur in order to increase the speed of ED presentation and maximise the potential for thrombolysis. Perhaps there is a poor public awareness of the time dependency for stroke. Research on stroke awareness in Ireland reported that 90% of patients were aware of drug treatments available for stroke, although only 1% could specifically name thrombolysis or clot-busting drugs, and >75% of the sample were unable to name emergency treatments that might reduce the effect and extent of stroke.21 Possibly, patients recognise that there is a need for fast reaction in the case of stroke onset, but do not understand why speed is so essential. Additional media campaigns should convey that treatments for stroke are time dependent and perhaps emphasise the ‘time is brain’ imperative in order to reinforce the need for rapid action in combination with EMS activation.

Our findings suggest that prehospital delay is a complex problem, with other factors in addition to knowledge about stroke interacting with the decision-making process during stroke onset. Previous studies have highlighted influential factors, such as perception of the seriousness of the situation, embarrassment at calling EMS, bystander intervention and perceived control over the situation.22 Understanding the complex decision-making process during an acute event is crucial for developing effective interventions to reduce time to hospital arrival. Community initiatives for stroke awareness may facilitate sustained behaviour changes by minimising barriers to health services, providing appropriate cues to action and dealing with misconceptions and fear surrounding stroke onset and treatments.23 A recent Irish study demonstrated that a community-based educational intervention improved stroke knowledge, and also provided a forum for discussion of stroke management and treatment with at-risk groups .24 This type of intervention may be more effective than mass media campaigns at promoting appropriate help-seeking behaviour and translation of knowledge to action.

Our study had some limitations. The baseline period with no F.A.S.T. exposure and the time period after the campaign were short, limiting the level of meaningful comparison between campaign waves. Distance from the ED was not noted, which might have affected delay time. Other factors that might have contributed to prehospital delay times were not recorded in this study owing to non-standardised documentation in ED notes. Stroke severity, symptom-related impairment, location at onset of symptoms and bystander intervention have been found to influence delay times in previous studies.25


In this analysis we have shown that the first Irish F.A.S.T. campaign had an initial impact on ED attendance of patients with stroke symptoms, but no sustained effect. Activation of EMS services and self-referral were associated with faster presentation during stroke onset. The evaluation of campaign efficacy, using behavioural end points such as thrombolysis rates, OTD reduction rates or increased EMS calls, results in accurate estimation of the clinical significance of the campaign as measured by its effect on the morbidity and mortality of patients with a stroke. Future public campaigns need to emphasise the availability of thrombolytic treatments, emphasising the ‘time is brain’ response to stroke.


We thank Dr Peadar Gilligan, Dr Aidan Gleeson, Dr Patricia Houlihan, Ms Joan McCormack, Beaumont Hospital and Dr Joe McKeever, Dr Mick Molloy, Dr Emily O'Connor, Connolly Hospital for access to data. We also thank Magdalena Bastiansen, Ciana Maher, Ronan Murphy and Delong Zeng for assistance with data collection and Professor Kathleen Bennett for statistical advice.


View Abstract


  • Contributors All authors: study design, manuscript preparation; LM: data collection, statistical analysis

  • Funding LM is an HRB PhD scholar in health services research funded by the HRB in Ireland under grant no. PhD/2007/16.

  • Competing interests None.

  • Ethics approval Beaumont Hospital research ethics committee and Connolly Hospital research ethics committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • Correction
    BMJ Publishing Group Ltd and the British Association for Accident & Emergency Medicine