Rationale for testing recognition of TIA

5.9.1 Bias in clinic-case derived tools – ‘missed’ high risk TIA

The classical presentations of TIA are well known to clinicians and the clinic referral cohort of TIA and mimics of TIA have similar characteristics in regard to these features, such as weakness and speech disturbance. However, there may be cases of TIA sharing common features that are less well known to GPs and therefore may be missed after initial

presentation in primary care. If these particular TIAs are high risk, then they will present a short time afterwards with stroke.

A major bias of the Dawson diagnostic tool for TIA is that it is derived using patients who have been referred to clinics, and as such GPs or ED clinicians have been concerned enough about the underlying diagnosis in these patients to refer for a specialist opinion. Patients who may present to both primary care and ED with high risk TIA but are not considered to have TIA by the treating clinician will not be included in the reference pool of clinic referrals. Such cases may share atypical features (as clinicians may not have

considered the diagnosis) resulting in a biased tool where not all TIAs have been included at derivation. This is particularly problematic if the missed cases are associated with early recurrent stroke.

As OXVASC recruits patients after all vascular events, all patients that present with stroke rather than TIA are also ascertained. All ascertained patients have primary care records examined for relevant risk factors and their control e.g. office blood pressure level, as well as

68 previous presentations to healthcare (either records from primary care or letters from ED) with events that could be due to cerebrovascular disease.

This represents an opportunity to examine in more detail patients who were recruited after a major stroke but may have had a TIA beforehand and presented to primary care after transient phenomena but were not diagnosed as having TIA. By virtue of case definition these transient events are very high risk for recurrent stroke.

The patients that are included are those who had a recurrent stroke within 30 days of a TIA but presented to medical attention in the time window after TIA and before the recurrent stroke. A time to recurrent event of ≤ 30 days after TIA was chosen because it could be argued that some presentations, if atypical, are less likely to represent a high risk TIA if the recurrent stroke occurs beyond this time window.

Although this yet again introduces a bias as patients with high risk events that do not present after TIA are not included, there is no reliable record taken prospectively after the initial TIA in such cases. Furthermore if one makes the assumption that such presentations are associated with a persistent lack of presentation to healthcare then there will not be an opportunity for a clinician to ‘miss’ the diagnosis of TIA prior to presentation with stroke. The care-seeking behaviour of patients in OXVASC who had a recurrent ischaemic stroke within 30 days of a TIA was used to identify those patients who presented to primary care but the diagnosis of TIA was not considered. Other referral routes from primary care were assessed to define ‘missed events’ in primary care.

Data were available for all patients with ischaemic stroke after TIA, ascertained for the first eight years (2002 to 2010) of OXVASC.

5.9.2 What is a ‘missed’ cerebrovascular event?

A liberal definition of a missed event would include all patients who have had a delay in receiving optimal therapy because the diagnosis was not considered at the initial consultation. This could be for a number of reasons after presenting to primary care -

1. The diagnosis was not considered and the patient had a recurrent stroke which resulted in further presentation to healthcare for investigation and treatment.

2. The diagnosis was not considered until a later date before any recurrent events and then a referral was made for investigation and treatment.

69 3. The diagnosis was not considered but a referral was made to other specialist teams

e.g. ophthalmology who then made the diagnosis and referred onwards for investigation and treatment.

Out of the above patient groups, point 1 defines those patients who are likely to be at highest risk of recurrent stroke and also where primary care may have the greatest difficulty in making a diagnosis. Point 2 defines patients at lower risk as there are no recurrent events after initial presentation during any period of delay. Point 3 defines a group where GPs suspect that there is an underlying condition to diagnose but refer to an intermediary

specialist who then either asks the GP to refer on for a TIA assessment or refers the patient directly without further involvement of primary care. This latter group requires an analysis of referral routes to TIA clinic and identification of those patients who arrived via an

intermediary specialist.

5.9.3 ‘Hidden TIA’ and prediction rule development

Patients with TIA can be correctly recognised by a referring clinician yet have a recurrent stroke requiring hospitalisation before being assessed in a TIA clinic. In OXVASC, these patients are ascertained as a stroke, rather than a TIA. Therefore by restricting the derivation of decision rules to patients seen in clinic (not only in OXVASC but also in other centres such as Glasgow in the Dawson tool), this will reduce the assessment of true TIA cases referred by GPs as the total set of GP suspected TIA will be missing the patients with highest early risk.

This creates an interesting bias as ideally the highest risk patients should be identified by a diagnostic or referral support tool but they did not appear in the Dawson derivation set by definition (as only clinic-seen patients were included) and have not been used in the

derivation of decision rules above, partly to allow for a fair comparison with the Dawson tool. The incidence of these high risk cases was assessed to determine impact on clinic- based derivation datasets.

5.9.4 Other missed opportunities – intermediary specialist referral

Given that the effects of transient cerebral ischaemia will result in dysfunction of a regional part of the body associated with a particular area of brain or retina, it is perhaps unsurprising that GPs may refer to intermediary specialists (i.e. not TIA specialists). The referral will be based on the presumption that the affected body part is dysfunctional, rather than the

70 visual symptoms which can be interpreted at first consultation as ocular in origin (due to lenticular, humoural or retinal disease).

Bias could be introduced if such patients had recurrent strokes before being seen in a clinic, or if persisting non-disabling symptoms developed and in OXVASC (and in other clinic- defined cohorts used for TIA diagnosis) were ascertained as a stroke rather than TIA. The extent to which intermediary specialist referral in OXVASC introduces a bias in using clinic- defined TIA populations for diagnostic decision rules will be assessed by estimating the incidence of affected patients.

5.10 Using missed TIA to assess diagnostic decision making - case vignette