Context

Local general practitioners (GPs) in primary-care or accident and emergency departments can refer any patient they suspect had a TIA but whom they did not consider required immediate hospital admission, to the rapid TIA service at Charing Cross Hospital, Imperial College NHS Healthcare Trust, London (UK) (figure 1). These patients or the caregiver at home (usually by telephone) are then contacted by our team to arrange a clinic appointment within 24 hours of referral received. Our TIA clinic is organised to provide a standardised assessment to all our patients. On the same day, blood tests (lipid profile, red blood cell, white blood cell, platelet count, concentration of glucose, haemoglobin A_1c, C reactive protein, creatinine), ECG, brain imaging (usually CT), and carotid ultrasound imaging and a clinical assessment by a stroke physician are obtained. Patients are discharged at home immediately after the assessment, unless the treating physician believes the patient requires urgent admission to our hyper acute stroke unit (HASU). Patients who are discharged with a diagnosis of TIA from our service are then contacted by our Cardiology Department, Imperial College NHS Healthcare Trust, to have a 24h-Holter ECG placed during a scheduled appointment.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Time from symptom onset to telephone call and to examination in our TIA clinic. ACEM, automated continuous ECG monitoring; TIA, transient ischaemic attack.

Local challenges in implementation

The project had several implementation challenges with multiple stakeholder involvement for the large-scale use of ACEM in our TIA service. According to an internal audit conducted on 136 patients with TIA assessed in our TIA clinic between 1 July and 31 December 2018 (phase 1), the median time interval to initiate the 24h-Holter ECG was 62 days (IQR 13–111). The rate of newly detected AF was 2.21% (3/136) and 7.4% of the patients (10/136) had a recurrent stroke/TIA within 6 months. We decided to use ACEM technology to significantly improve the delay to initiate the cardiac monitoring and to enhance the yield of AF detection in these patients and to reduce the 6-month risk of recurrent stroke/TIA.

Brainstorming meetings with all the TIA team members helped to develop a protocol and to find dedicated clinic space. Stroke clinical nurse specialists were trained to initiate the ACEM-Holter and to remove them after the recording. Stroke physicians were trained to use these new automated health tools, to send the traces via the secure internet connection and to read and interpret the results. We involved physicians and all the other health professionals to participate in this novel care delivery strategy process from very early on. This helped to create ideas to avoid the need for additional resources. Information Governance Department and the Information and Communication Technology Department at Imperial College NHS Healthcare Trust, London (UK) made a secure internet connection to send the ACEM-Holter traces to a central server (Apoplex Medical Technologies, Pirmasens, Germany) in order to respect privacy, security and data confidentiality of the patients involved in this project. They also ensured that all our infrastructure was in place to integrate this new technology.

Study intervention

Between 1 January and 30 June 2019 (phase 2) an ACEM-Holter was applied for 24 hours to all the patients with TIA immediately after their assessment in our clinic at the stroke physician’s office without sending the patient to the Cardiology Department. The day after patients returned in our clinic and the stroke physicians (LD or DK) sent the ACEM-Holter traces via a secure internet connection to a central server (Apoplex Medical Technologies, Pirmasens, Germany) where an automated AF episode detection algorithm was applied using the SRAclinic software. This software uses an algorithm that detects QRS complexes of the ECG data and then classifies them as being of atrial or ventricular origin and finally creates a list of R–R-intervals.15 To detect episodes of AF, the software performs a time series analysis of multiple mathematical parameters that are typical for an absolute arrhythmia during AF. The SRA algorithm consists of three main steps in identifying possible AF. In the end, this results in a final classification into one of the following three groups: (1) no evidence of AF; (2) increased risk of AF or (3) presence of manifest AF. SRA clinic detection of manifest AF episodes includes only episodes lasting >30 s. In the case of detected AF, the SRA service provides source of ECG rhythm strips which were reviewed by a trained stroke physician for confirmation of AF. SRA clinic has a 99% sensitivity and specificity compared with Holter ECG in the presence of manifest AF on ECG data. The final outcome of the traces analysis was sent from the central server to a secure email address and available to be read within few minutes. The clinical management as well as diagnostic procedures of all the patients with TIA, with and without AF, then followed routine practice of the treating physicians. Our team evaluates the risk of stroke and TIA at 6 months (in both phases) via the follow-up clinics during face-to-face interviews or through telephone calls. If the patient could not be contacted, a close relative or their family doctor is interviewed.

Patients

The following inclusion criteria had to be fulfilled in both the two phases of the study: (1) age ≥18 years old; (2) TIA diagnosis confirmed by a trained stroke physician; (3) no history of AF; (4) no AF on ECG performed in the TIA clinic; (5) no implanted pacemaker. Exclusion criteria were: (1) intracranial haemorrhage (2) TIA mimics17 (migraine aura, seizures, syncope, peripheral vestibular disturbance, transient global amnesia, functional/anxiety disorder, amyloid spells, subarachnoid haemorrhage, structural brain lesion and paroxysmal symptoms due to demyelination). We extracted event characteristics, patient demographics, risk factors and medical history from medical records. Data of consecutive patients were extracted using a prespecified case report file that encompassed patient characteristics including age, vascular risk factors, laboratory results, relevant medical history and medications. CHA₂DS₂-Vasc score was calculated for each patient.

Definitions

TIA was defined as a brief episode of neurological dysfunction caused by focal brain or retinal ischaemia, with clinical symptoms typically lasting less than 1 hour, and without evidence of acute infarction.18 By accepted convention, an episode lasting at least 30 s documented was considered as diagnostic for AF.19 High blood pressure was defined as systolic pressure ≥140 mm Hg and/or diastolic pressure ≥90 mm Hg, use of antihypertensive medication, and/or previous diagnosis of high blood pressure. Carotid stenosis was defined according to the Trial of ORG 10172 in Acute Stroke Treatment20 criteria as narrowing of the internal carotid artery lumen of ≥50% on carotid duplex ultrasound or angiography. Diabetes mellitus was defined as a history of diabetes that was confirmed in medical records, and/or use of insulin/oral hypoglycaemic agents, and/or random non-fasting blood glucose concentration ≥11.1 mmol/L. Hypercholesterolaemia was defined as fasting total cholesterol serum level ≥5.18 mmol/L (200 mg/dL), and/or fasting low-density lipoprotein cholesterol serum level of ≥4.14 mmol/L (160 mg/dL), and/or use of lipid-lowering medications. Coronary heart disease was defined by a history of acute myocardial infarction, or angina pectoris, or coronary artery bypass graft, or percutaneous coronary intervention. The duration of the cardiac monitoring was automatically provided (hours) with the routine and ACEM-Holter analysis. The time to initiate the cardiac monitoring was calculated as the total time (days) from the point of the assessment in TIA clinic to the point of activation of the cardiac monitoring.

Statistical analysis

Data analysis was conducted using SPSS (V.26.0, IBM). Characteristics of the study population were described using medians and IQR or mean and SD or numbers with percentages. P<0.05 was considered as significant. T-test or Mann-Whitney test were used to compare continuous variables. For categorical variables, a χ² or Fisher exact test was used to compare distributions.