INTRODUCTION

Patients with a transient ischemic attack (TIA) have a 3–10% risk of stroke within 1 week of the event1,2 and an 8–17% 90-day stroke risk.1,3 In addition to stroke risk following TIA, patients have a 2.6% risk of cardiovascular events and 2.6% risk of death within 90 days.4 Based on a 25.1% overall short-term recurrent vascular event rate, TIA patients should be considered as high risk for adverse events and their care should be delivered urgently and directed toward identification of index event etiology as well as risk factor management. The American Heart Association/American Stroke Association (AHA/ASA)5 and National Stroke Association (NSA) guidelines recommend urgent evaluation of TIA patients including admitting TIA patients if their workup cannot be completed in a timely fashion in the outpatient setting.6,7

Although hospitalization of TIA patients has been shown to improve the quality of TIA care,8,9,10,11 admission rates have been declining. A study from the US Nationwide Inpatient Sample (NIS) database demonstrated decreased hospitalization rates for TIA patients between 2000 and 2010 from 118 to 83 per 100,00 population.12 This reduction in hospital admissions for TIA may be due to: improved risk factor management resulting in a true reduced incidence; the revised definition of TIA in 2009, changing from a time-based definition to a tissue-based definition, thereby patients who would have been classified as TIAs under a time-based definition are now classified as strokes7; or new TIA management alternatives (e.g., TIA clinics, observation units) that obviate the need for hospital admission.12

Unlike acute stroke management, controversy exists regarding whether all TIA patients require care as inpatients because neurological symptoms have resolved.13 Many clinical and societal variables may contribute to the decision-making process regarding hospital admission14 and some clinicians advocate admission of all TIA patients.15 Rationale in support of universal hospital admission for TIA includes hospitalization allowing for timely evaluation, opportunity for intervention if a subsequent vascular event occurs, and rapid initiation of secondary prevention strategies.15,16 Opponents to universal TIA admission argue that alternatives such as urgent TIA clinics or observation units can provide care which is equivalent to inpatient services in terms of timeliness and quality but at a markedly reduced cost and potentially with enhanced patient convenience.17

Given these differences in opinion regarding whether or not to admit patients with TIA, it is not surprising that a high degree of variation exists in admission rates for patients with TIA.18 We sought to identify factors that were part of providers’ decision-making process regarding admitting patients with TIA.

METHODS

We conducted a secondary analysis of a FY 2011 study from the Veterans Health Administration (VHA) which evaluated quality of care for Veterans with TIA and minor stroke.8,19,20 We included three sources of data: qualitative data from semi-structured interviews with facility staff members, administrative data, and medical chart review. Indiana University Institutional Review Board and Roudebush VAMC Research and Development Committee reviewed and approved this research.

Qualitative Interview Data

Interviews with stakeholders to evaluate TIA healthcare practices in the VHA were conducted in 2013–2014.20

Sites

Fourteen interview sites from diverse geographic regions (4 East, 4 Midwest, 2 Southwest, and 4 West) were chosen based on annual volume of ≥ 25 patients with a TIA or minor stroke and willingness to participate in site visits. An adequate volume was needed to ensure sites’ interest in TIA care processes, identify providers to interview who cared for patients with TIA on a regular basis, and plan for a future implementation project which would likely target sites which routinely cared for TIA patients. All 14 study sites had at least one neurologist on staff.

Participants

We interviewed clinical staff involved with TIA care from a broad range of disciplines including emergency medicine, neurology, hospitalist and medicine services, nursing, radiology, ophthalmology, primary care, quality management, cardiology, and pharmacy. The number of interviews per site ranged from 2 to 9 with an average of 5.3 interviews per site.

Data Collection

Our methods have been described elsewhere.20 Briefly, we interviewed participants about TIA processes of care—acute and follow-up—as well as facility characteristics, using a semi-structured interview guide. We obtained written permission to audiotape interviews. Audiotapes were professionally transcribed verbatim, de-identified, and imported into a single NVivo10 project file for data coding and analysis.

Data Analysis: Coding

For this secondary analysis, we focused on the interview question: “Tell me what influences you in the decision to or not to admit TIA patients.” We compared information about influences on decision to admit across sites and among providers. From coded interviews, we identified emergent themes and patterns of factors cited by providers in their decision whether to admit TIA patients into inpatient care. We further categorized these factors as TIA clinical characteristics, patient perspectives, provider preferences, use of ABCD2 score, and uncertainty in three areas (diagnosis of TIA and risk of recurrence, resources, and practice). We illustrated these themes with quotations from respondents.

Administrative Data

Administrative data’s purpose for the main study8 was to identify Veterans with TIA or minor stroke, develop electronic quality measure algorithms, and assess 90-day and 365-day outcomes, including death and recurrent events. Veterans with TIA or ischemic stroke who received care in FY 2011 and 2014 in any VHA Emergency Department (ED) or inpatient setting were identified (FY 2011, n = 3676 with TIA; FY 2014, n = 3264 with TIA). Final sample size for the administrative data cohort was n = 697 (FY 2011) and n = 617 (FY 2014) patients from the 14 interview sites. These data were used to identify the proportion of patients with TIA at each of the 14 sites who were directly admitted or admitted from the ED.

Chart Review Data

Chart review’s primary purpose for the main study8 was to validate electronic process measure algorithms. We identified a random sample of patients from participating facilities by selecting a 35% random sample with a minimum of 25 patients per site. A total of N = 812 chart reviews for patients with TIA or minor stroke was conducted at 39 different VAMCs for the main study. Chart reviews were performed at 12 of the 14 interview sites, but one site had no patients with TIA. Therefore, ABCD2 scores were from a random subset of patients with TIA from 11 sites, as the scores are only available from chart review.

Chart review at these 11 sites was used to obtain the ABCD2 score, a tool to identify risk of an ischemic stroke after TIA, based upon five criteria: age, blood pressure, clinical features, duration, and history of diabetes.4,21 Scoring ranges from 0 to 7; higher scores indicate higher risk of recurrent vascular events. Guidelines recommend admission based upon ABCD2 score (e.g., ≥ 3) and other factors (inability to complete workup within 48 h, focal findings).7

Analysis

Mean ABCD2 score was calculated at each facility for patients who were admitted and compared to those who were discharged from the ED. For each site, we compared providers’ interview statements about use of the ABCD2 score in decision-making with the mean facility ABCD2 scores obtained from chart review data. We used t tests to compare ABCD2 scores for patients who were admitted versus patients who were discharged from the ED.

RESULTS

We completed 70 interviews with clinical staff involved with TIA care at 14 facilities.

Factors Influencing the Decision Whether To Admit

TIA Characteristics

Characteristics of TIA events such as recent onset, presence of TIA symptoms while in the ED, or crescendo TIA were commonly cited by providers as reasons for admission. Providers also cited utilizing the ABCD2 score, which includes TIA clinical features (e.g., muscle weakness), as well as duration of symptoms. A delay in presentation (longer period since symptom onset), resolution of symptoms before or during ED stay, or a single TIA event (versus crescendo TIA) were reported as factors related to the decision not to admit:

If a patient’s symptoms are already resolved, they are lower on the priority list. (Site A-Emergency Medicine)

Patient Perspectives About Admission

Clinician-interviewees identified patient preference as a factor that influenced whether the patient was admitted; that is, providers reported that some patients did not want to be admitted because, for example, they “felt fine.” Lack of availability of services sometimes resulted in Veterans refusing admission because of not wanting to wait over a weekend or holiday for testing to begin.

Some patients leave after being admitted, or instead of being admitted because they are asymptomatic, feel fine and have other things to do. This is especially true over the weekend, when some of the tests like echo and ultrasound are not available until Monday. (Site B-Neurology)

Provider Preferences

Respondents cited their gestalt impression as influencing the decision whether to admit.

Admitting or not admitting a patient is more of a gut decision even after taking into consideration many of the health factors evaluated by the ABCD2 scale (age, diabetes, etc.). (Site A-Neurology)

Services such as availability of magnetic resonance imaging (MRI) and neurology consultation contributed to the decision not to admit both if they were lacking (as noted above) or if they were ample. Interviewees stated that when services were readily available, a workup could be completed from the ED and the patient did not need to be admitted Figure 1, Table 1

Figure 1
figure 1

Acute TIA Quality Processes of Care.

Full size image

.

Availability of observation beds provided clinicians opportunity to monitor patients with TIA and conduct an expedited evaluation. Providers at facilities that had observation units cited this resource option in their decision not to admit patients.

Due to our new 10-bed clinical decision unit, which is located right next to the ED, we are able to admit TIA patients who were previously sent home after testing. (Site C-Emergency Medicine)

Another provider factor was a concern that patients who live a long distance from a VAMC would not be able to return for outpatient testing due to lack of transportation, financial means, or social support.

Our patient population tends to…have fewer resources in terms of family support, transportation, etc. (Site D-Neurology)

However, proximity of the patient’s residence to a VA facility and strong patient support at home contributed to providers’ comfort with discharging a patient from the ED to home.

If a patient lives relatively close by and they had their TIA a week ago we probably wouldn’t admit them. (Site D-Neurology)

Interviewees stated that they were more likely to not admit a TIA patient if the patient’s medical management was already maximized or if the patient recently had an evaluation for a prior TIA or stroke event.

If…event was not very recent and they are on an excellent secondary prevention regimen…they will not be admitted. (Site E-Neurology)

ABCD2

Respondents at five facilities reported that a high ABCD2 score was a driver of the decision to admit; a low ABCD2 score was reported by participants at six facilities as a factor in decisions not to admit.

Admittance is probably based upon ABCD2 score—very low percentile and a quick workup means that the patient may be sent home instead of admitted. (Site F-Neurology)

Diagnostic Uncertainty

In 2009, TIA’s clinical definition was revised which may have contributed to increased diagnostic uncertainty.7,22 To be definitive, the diagnosis of TIA requires use of MRI, which is not always readily available. In addition, stroke mimics (e.g., migraine, nerve compression) add to uncertainty of diagnosis.

The ER doctors wouldn’t want to send a TIA patient home and risk getting burned if it turns into a full-blown stroke a day later. (Site F-Hospitalist)

Resource Uncertainty

Staff reported uncertainty in their facility’s ability to complete workups within 24 to 48 h in the outpatient setting, especially during weekends and holidays.

Discharging someone out of the ED for TIA would require figuring out how to get tests done within a very short period of time, which suggests to me that it’s probably a lot easier to just admit them. (Site A-Hospitalist)

Practice Uncertainty

Uncertainty about local practice was cited as a factor by providers. Although VAMCs were required to develop a stroke care policy after the 2011 Treatment of Acute Ischemic Stroke VHA National Directive, TIA policy was not included. None of the interviewed providers reported that their facility had a policy guiding TIA care.

There is no standard policy for our facility and all clinicians have their own style. (Site F-Neurology)

Belief That All TIA Patients Are Admitted

Staff at 12 of the 14 facilities reported that they admitted all or most TIA patients, yet FY 2011 administrative data showed that 9 of 14 sites (64%) admitted fewer than two thirds of TIA patients (Table 2). Facility TIA admission rates for FY 2011 ranged from 45.5 to 84.1% with a mean admission rate of 66.9% (SD = 47.1%). For FY 2014, the range was slightly higher, 51.6 to 89.5%. The facility with the lowest FY 2011 admission rate (45.5%) was second highest (87.5%) in FY 2014; in this case, the provider’s perception would be in alignment with the more recent period. However, the highest facility admission rate in FY 2011 (84.1%) was seventh highest (75.5%) in FY 2014, in which case the provider’s perception that all were admitted was less in alignment with the observed admission rate for the more recent year. Overall, the national rates were similar between the 2 years (67.7% vs 72.2%); nonetheless, there were differences at the individual facility level. Among facilities in this study, admission rates for TIA patients remained the same at five (36%) facilities, were lower in FY 2014 at one (7%) facility, and were higher at eight (57%) facilities in FY 2014 compared to FY 2011. In both FY 2011 and FY 2014, providers at facilities with the lowest TIA admission rates reported that they admitted “all” or “most.” Although the TIA admissions’ rate increased for approximately two thirds of the facilities in FY 2014, none of the facilities admitted all of their TIA patients and nearly three quarters admitted less than 80%.

Table 1 TIA Risk Factors23,24,25
Full size table

In both years, perception by respondents that most TIA patients who presented to their facility’s ED were admitted was inconsistent with the observed facility TIA admission rate.

TIAs all get admitted. (Site G-Emergency Medicine)

Use of ABCD2 Score

Table 3 illustrates providers’ use of the ABCD2 score from FY 2011 to guide admission decisions. Sites with higher admission rates reported greater reliance on use of the ABCD2 score. One facility at which a staff member stated he used the ABCD2 score had a TIA admission rate of 78.9%. Although staff at nine (64%) facilities stated that they utilized the ABCD2 in the decision to admit, many of the statements included qualifying conjunctions (such as “but” or “although”).

We use the ABCD2 score here sometimes, but not always. (Site H-Neurology)

Table 2 Disconnect Between the Provider’s Belief That All TIA Patients Are Admitted and the Observed Facility TIA Admission Rate
Full size table
Table 3 Use of ABCD2 Score in Decision Whether To Admit Patient with TIA
Full size table

As noted previously, the ABCD2 score was cited as supporting evidence in the decision not to admit. Specifically, clinicians appeared to use the low ABCD2 score to validate their decision to discharge ED patients when other factors were present.26

Other uses of the ABCD2 score have been reported, including for patient education about stroke risk.27

Some patients don’t want to stay in the hospital despite neurologists telling them about their ABCD2 score. (Site A-Neurology)

Eighty-two percent of facilities had a mean ABCD2 score > 4; only one facility had a mean ABCD2 score less than 3. Except for one site, there was no statistical difference in mean ABCD2 score among admitted versus not admitted patients, but sample sizes were small (Table 3). Mean ABCD2 scores for patients who were admitted (mean 4.3, standard deviation, 1.4) were not consistently higher than mean ABCD2 scores for patients who were discharged from the ED (mean 4.0, standard deviation 1.5; p = 0.2991; Table 3).

DISCUSSION

Our study includes a multidisciplinary assessment of provider-reported considerations regarding the decision to admit TIA patients. Uncertainty in clinical diagnosis, ability of the facility to conduct timely workups, patient preferences as reported by providers, and facilities’ lack of a TIA policy were all key drivers of the decision to either admit patients with TIA to the hospital or discharge them from the ED. These results are similar to a prior study which found that because no gold standard diagnostic test for TIA exists, uncertainty in the TIA diagnosis is present in all TIA studies.28 Uncertainty in the TIA diagnosis may lead providers to rely more heavily on their gestalt clinical impression,29 a factor cited by providers in our study as an element of the admission decision-making process.

Although we identified several areas of uncertainties factoring into a provider’s decision to admit a patient, somewhat paradoxically, we regularly heard that all patients with TIA get admitted. However, despite this repeated contention, we found a disconnect between what was reported by providers and actual admission patterns based on administrative data. Like findings from another study,30 providers in our study stated they admitted most or all TIA patients. Yet, when we examined the actual facility TIA admission rates, some facilities admitted only about half of their TIA patients in both FY 2011 and FY 2014. It is unclear whether participants were stating what they believed to be their facility’s admission practice or if responses were influenced by a social-response bias.31 Potential reasons for a disconnect between providers’ perspectives (to admit all TIA patients) and the observed facility TIA admission rate may include neurologists were only aware of patients for whom they cared and were not aware of patients who were discharged from the ED without neurology care and providers may have thought about their most recent TIA patient with a definitive diagnosis, rather than considering the full spectrum of TIA patients, including patients with “uncertain” or equivocal symptoms.

Overestimation of admission rates and descriptions of other details of TIA care could be influenced by recall bias because individual staff members care for a relatively small number of TIA patients per year. Low TIA patient volume may also contribute to uncertainty with new or unfamiliar situations.

Availability of alternative management options may influence provider decision-making regarding inpatient admission. Although urgent TIA clinics and ED observation units may be lower cost alternatives to hospital admission, these are not presently available in many areas of the USA.29,32 Providers may be more likely to admit patients if they have the option of placing a patient in observation status.30 In the absence of observation units or TIA clinics, providers may seek to obtain diagnostic tests in the ED; however, constraints on ED resources which may either be fixed (e.g., lack of MRI) or dynamic (e.g., during ED crowding) may increase the likelihood that a provider will admit a TIA patient.33 Although many TIA patients can be appropriately managed in alternative settings, multidisciplinary care and certain testing services (e.g., telemetry) that are available routinely in inpatient settings (and less so in outpatient settings) may be needed for patients with certain event etiologies (e.g., cardioembolic events) or those with a higher degree of comorbidity.14 Until that level of care is available routinely in outpatient settings, hospital admission will be the “gold standard” in the USA for providing timely TIA care.5

We found that the ABCD2 score was not a major factor in most providers’ decisions to admit TIA patients. Opinions varied about this tool—some participants stated they did not find it useful and some cited concerns about the reliability or accuracy. These impressions are supported by existing literature suggesting that ABCD2 scores may not be accurately calculated by non-stroke specialists34 and that they may be less predictive of stroke risk than other patient characteristics.35 McKesson’s 2014 InterQual36 TIA admission criteria were based on the ABCD2 score, outpatient workup availability within 2 days (computed tomography [CT] or MRI, carotid artery imaging, and cardiac echocardiogram), resolution of neurologic symptoms, and appropriate interventions.7 Application of these criteria for TIA admission did not appear to be consistent across sites, although, the VHA utilizes InterQual criteria for admission decision-making.36 Lack of a facility TIA-specific policy might have contributed to the inconsistent decision-making regarding hospital admission for TIA patients.

Strengths of this study included diversity of facilities and participants who were interviewed and availability of administrative and chart review data to complement qualitative data. The primary limitation of the study is that it was conducted in VAMCs; generalizability of findings to non-VA facilities may be limited. Although administrative data may contain errors in coding, we previously validated the administrative data against chart review for the purpose of identifying patients with TIA; among patients identified as having a TIA by VHA administrative data, 95.0% had a TIA by chart review, 4.1% had a minor stroke by chart review, and 0.9% had neither TIA nor minor stroke by chart review.19 The chart review data used in the current study were obtained from a random sample of patients per facility; not all patients at the study sites received a chart review. Interviews were conducted at one point in time (FY 2013–2014), so we are not able to identify trends over time or to draw causal relationships. Another limitation was that transcripts were not read by interviewees to confirm the accuracy of their statements.

Our results indicate that many factors contribute to the decision whether to admit a patient with TIA; however, a key factor is uncertainty due to lack of guidance at sites without a facility TIA-specific policy, as well as inconsistent use of the ABCD2 score, patient’s relative lack of symptoms, limited resources, and concerns about the ability of the system to complete a timely workup. Further research and quality improvement interventions for TIA care should focus on facility adoption of a TIA protocol to address key elements of the admission decision-making process and to standardize both evaluation of TIA and delivery of secondary prevention strategies.