Article, Neurology

Initiation of the ABCD3-I algorithm for expediated evaluation of transient ischemic attack patients in an emergency department

a b s t r a c t

Background: The use of ABCD3-I score for Transient ischemic attack evaluation has not been widely inves- tigated in the ED. We aim to determine the performance and cost-effectiveness of an ABCD3-I based pathway for expedited evaluation of TIA patients in the ED.

Methods: We conducted a single-center, pre- and post-intervention study among ED patients with possible TIA. Accrual occurred for seven months before (Oct. 2016-April 2017) and after (Oct. 2017-April 2018) implementing the ABCD3-I algorithm with a five-month wash-in period (May-Sept. 2017). Total ED length of stay , admissions to the hospital, healthcare cost, and 90-day ED returns with subsequent stroke were an- alyzed and compared.

Results: Pre-implementation and post-implementation cohorts included 143 and 118 patients respectively. A total of 132 (92%) patients were admitted to the hospital in the pre-implementation cohort in comparison to 28 (24%) patients admitted in the post-implementation cohort (p b 0.001) with similar 90-day post-discharge stroke occurrence (2 in pre-implementation versus 1 in post-implementation groups, p N 0.05). The mean ABCD2 scores were 4.5 (1.4) in pre- and 4.1 (1.3) in post-implementation cohorts (p = 0.01). The mean ABCD3-I scores were 4.5 (1.8) in post-implementation cohorts. Total ED LOS was 310 min (201, 420) in pre- and 275 min (222, 342) in post-implementation cohorts (p N 0.05). Utilization of the ABCD3-I algorithm saved an average of over 40% of total healthcare cost per patient in the post-implementation cohort.

Conclusions: The initiation of an ABCD3-I based pathway for TIA evaluation in the ED significantly decreased hos- pital admissions and cost with similar 90-day neurological outcomes.

Introduction

Transient ischemic attack is a common and high-risk clinical syndrome with 15% of Cerebrovascular accidents (CVA) preceded by TIA, half of them within 48 h [1-3]. The definition of TIA has been cate- gorized as either time or tissue-based [4,5]. Time-based TIA is a clinical diagnosis defined as acute Neurological deficits lasting for b24 h, regard- less of neuroimaging findings. Tissue-based TIA is defined with negative diffusion-weightED magnetic resonance imaging (MRI), and has re- cently been thought to be superior to the time-based definition due to significant increases in the rates of subsequent strokes of tissue- positive relative to tissue-negative TIAs [6].

TIA patients are commonly seen in the Emergency Department (ED). Prior risk stratification tools have been proven inadequate in the ED

* Corresponding author at: Baylor University Medical Center, 3500 Gaston Ave., Dallas, TX 75246, United States of America.

E-mail address: [email protected] (H. Wang).

setting with respect to TIA evaluations. Investigations of ABCD2, one of the most widely studied risk stratification tools for TIA, have been proven inaccurate at any cut point for prediction of short-term stroke risk [7-9]. This lack of effective risk stratification has resulted in high Hospital admission rates and healthcare cost [10,11]. In recent years, given the current healthcare climate in the United States, question has been raised on whether TIA patients can be directly discharged from the ED. Some studies recommended discharging patients from the ED with rapid outpatient follow-up and considered it to be a safe and effec- tive strategy, while others preferred hospital admission for most TIA pa- tients, even those thought to be at low risk for stroke [11-13]. Such significant discrepancy is due to a lack of a reliable tool for risk stratifi- cation among TIA patients in the ED.

With the introduction of the ABCD3-I score in 2010, studies have re- ported higher sensitivity with the ABCD3-I score than with the ABCD2 score for predicting 90-day stroke events [14,15]. Low-risk patients (ABCD3-I, 0-3) or those with negative neuroimaging tended to have very low rates of post-discharge CVA within 90 days [16-18]. Therefore,

https://doi.org/10.1016/j.ajem.2019.06.018

the ABCD3-I approach may create the opportunity for ED physicians to safely discharge tissue-based TIA patients directly from the ED. Though the use of MRI is not popular in EDs across the nation, the trend of using MRI for TIA evaluation has become more common and has the potential to improve efficiency [10,19]. However, current literature is still lacking data to support this hypothesis.

We hypothesized that implementing an ABCD3-I approach to TIA patients could potentially minimize unnecessary hospitalizations, re- duce CT utilization, decrease ED boarding, and save healthcare cost. Therefore, we aim to investigate the benefits of discharging tissue- based TIA patients by applying the ABCD3-I algorithm in our ED.

Methods

Study design and setting

We performed a single-center observational study from Oct. 1, 2016 until April 30, 2018. A pre-intervention phase was conducted from Oct. 1, 2016 until April 30, 2017. A post-intervention phase was conducted from Oct. 1, 2017 until April 30, 2018 with a five-month wash-in period in between. Due to the nature of this study design, patients were not randomized, the study was not blinded, and sample size estimation and sensitivity analysis were not required. The pre-intervention phase was conducted retrospectively, and the post-intervention phase was conducted prospectively. This study took place at a large, urban commu- nity hospital ED with approximately 110,000 annual patient visits. It is deemed to be a Quality improvement project with the waived approval of local Institutional Review Board.

Study participants

Consecutive patients who presented in the study ED and diagnosed with TIA during the study period were eligible and included in this study. We excluded the following patients: 1) patients aged b18 years,

2) prisoners, 3) pregnant patients, and 4) patients who had contraindi- cations to neuroimaging (e.g. MRI).

Study intervention

Pre-intervention

Pre-intervention evaluation of patients who present to the ED with a suspected TIA included a CT of the head without contrast followed by further evaluation in the hospital with neuroimaging (e.g. MRI/MRA/ CTA), at the discretion of the treating physicians. All patients were followed up within 90 days after their index ED visit or hospital discharge.

Intervention

The study ED initiated a standard protocol utilizing the ABCD3-I al- gorithm to evaluate ED TIA patients. Any patient who presented with a clinical TIA was evaluated using this algorithm. The detailed clinical al- gorithm is shown in Fig. 2. Briefly, an emergent CT head was replaced by emergent MRI/MRA head and neck in the ED, with the agreement of the three relevant departments (ED, neurology, and radiology). MRI/MRA findings were emergently reported by radiologists. An ABCD3 score was calculated on each patient by the ED physician in conjunction with the on-call neurologist via phone consultation. Based on the imag- ing findings and ABCD3-I scores, patients were either admitted to the hospital for further evaluation and treatment or directly discharged from the ED with outpatient clinic follow-up.

Post-intervention

All patients enrolled during the study period who followed this new clinical algorithm were followed up at least 90 days after ED or hospital discharge.

Study protocol

This is a pre- and post-intervention quality improvement project. The study included two phases. During the pre-intervention phase, we performed a retrospective chart review on patients who presented to the ED and were diagnosed with TIA from Oct. 1, 2016 to April 30, 2017. From Oct. 1, 2017 to April 30, 2018, patients who followed the new ABCD3-I algorithm were prospectively enrolled in the post- intervention group. Two independent researchers trained on data ex- traction reviewed the charts, abstracted data using a structured chart re- view data collection sheet and collected prospective data. Basic patient characteristics, clinical outcomes, and ED/hospital clinic metrics were compared between these two groups.

Study variables and outcome measurements

Patient characteristics included age, gender, race, Insurance type, and mode of arrival. Clinical variables included Imaging results (CT, CTA, MRI, MRA). ED/hospital operational metrics included ED length of stay , ED disposition, ED returns, and patient healthcare cost. In order to protect patient privacy, patient ED and hospital charges were converted from dollar values into relative value units (RVU). We intended to compare RVU differences and calculate percent savings be- tween pre- and post-intervention cohorts. Our primary outcome was to determine the differences on hospital admissions, ED LOS, and healthcare cost between the two groups. The secondary outcomes were patient ED returns and the occurrence of CVA within 90 days from the index ED/hospital discharge.

Data analysis

Continuous variables were summarized using mean with standard deviation (SD) and median with interquartile ranges (IQR) and com- pared using Student t-test for mean comparisons and Wilcoxon Rank- sum test for median comparisons. Categorical variables were summa- rized using frequency and percentage and compared using Fisher’s exact tests. All descriptive and statistical analyses were performed using Stata 14.2 (College Station, TX). A p value b0.05 was considered statistically significant.

Fig. 1. Study patient flow diagram.

Table 1 Study patient general characteristics.
	Pre-Intervention (N = 143)	Post-Intervention (N = 118)	p
Age — Year, Median (IQR)	68 (57, 76)	64 (53, 76)	0.23
Mean (SD)	67 (14)	64 (16)	0.12
Gender — Male, n (%)	64 (45)	49 (42)	0.60
Female, n (%) Race — n (%) Caucasian	79 (55) 91 (64)	69 (58) 79 (67)	0.55
African American	43 (30)	35 (30)
Other Insurance type — n (%) Commercial	9 (6) 45 (31)	4 (3) 36 (31)	0.14
Medicare	82 (57)	56 (47)
Medicaid	6 (4)	6 (5)
Self-pay	8 (6)	15 (13)
Other ED mode of arrival — n (%) Ambulance	2 (1) 59 (41)	5 (4) 48 (41)	0.99
Non-ambulance	83 (58)	69 (58)
Unknown	1 (1)	1 (1)

Abbreviations: n (number); IQR (Interquartile Range); SD (Standard Deviation).

Results

A total of 261 patients were analyzed in this study with 143 patients in the pre-intervention and 118 patients in the post-intervention group. A very small percentage of patients were either missing important in- formation such as outcome data or did not have 90-day discharge follow-up. Since this makes data imputation less meaningful, we ex- cluded these patients from the final analysis (see Fig. 1). Study patient general characteristics are shown in Table 1 with no significant differ- ences between pre- and post-intervention cohorts (p N 0.05).

After the intervention, there was a significant reduction in hospital admissions, with admission rates decreasing from 92.3% (132/143) to 23.7% (28/118) (p b 0.001, Table 2). During 90-day follow up after the index visit, only one patient had a CVA occurrence in the post- intervention group, compared to two patients having CVAs in the pre- intervention group (p N 0.05, Table 2). This indicates that patients with low-to-moderate risk ABCD3-I scores and negative neuroimaging findings can be safely discharged directly from the ED, with excellent outcomes.

Meanwhile, other outcomes were compared between pre- and post- intervention groups. The use of the ABCD3-I pathway reduced CT utili- zation from 99% in pre-intervention to 37% in post-intervention cohort (p b 0.001, Table 2). The use of the ABCD3-I score saved an average of

60 RVUs per patient (an average savings of over 40%) from the total healthcare cost (e.g. ED and in-hospital). Patient index CVA occurrence, 90-day follow-up CVA occurrence, ED 7-day, 30-day, and 90-day returns, and ED total length of stay (LOS) had no statistically significant differences between these two groups (Table 2).

Discussion

ABCD3-I represents a highly sensitive risk stratification tool for TIA patients, and our study demonstrated strong prospective performance of an ABCD3-I pathway in an emergency department setting. Given the traditionally high utilization of resources by TIA patients and the current healthcare climate, there exists a significant need to care for these patients in a more efficient, cost-effective manner. Our interven- tion significantly reduced admissions, CT utilization, and Total cost, while showing similar clinical outcomes in comparison to patients who were traditionally placed into the hospital for completion of TIA evaluation [8,15].

Similar to other ABCD2 versus ABCD3-I comparison studies reported previously, we yielded similar results in short-term CVA occurrences during the index ED visits and better CVA predictions with the applica- tion of ABCD3-I scores. TIA patients with low risk ABCD3-I scores tended to have very low short-term (7-day) and long-term (90-day) CVA

Fig. 2. ED ABCD3-I pathway flow diagram during post-intervention phase.

Table 2 Outcome comparison between pre- and post-intervention groups.
	Pre-intervention	Post-intervention	p
	(N = 143)	(N = 118)
ABCD2 — Median (IQR)	5 (4, 6)	4 (3, 5)	0.01
Mean (SD)	4.5 (1.4)	4.1 (1.3)	0.01
ABCD3-I — Median (IQR)		4 (3, 6)
Mean (SD)		4.5 (1.8)
Index CVA – n (%)	9 (6.3)	7 (5.9)	0.90
ED Total LOS — Min, Median (IQR)	310 (201, 420)	275 (222, 342)	0.08
Mean (SD)	378 (314)	288 (97)	b0.01
ED Disposition — n (%)
Discharged	11 (7.7)	90 (76.3)	b0.001
Placed into hospital	132 (92.3)	28 (23.7)
ED CT Utilization — Yes, n (%)	141 (98.6)	44 (37.3)	b0.001
ED Returns — Yes, n (%)
7-day	4 (2.8)	3 (2.5)	0.90
30-day	5 (3.5)	2 (1.7)	0.37
90-day	3 (2.1)	4 (3.4)	0.52
90-day CVA occurrence — Yes, n (%)	2 (1.4)	1 (0.9)	0.68
ED charge — RVUs, Median (IQR)	18 (18, 21)	16 (15, 22)	0.74
Mean (SD)	19 (7)	20 (17)	0.47
Total charge — RVUs, Median (IQR)	188 (158, 232)	108 (98, 135)	b0.001
Mean (SD)	210 (105)	150 (146)	b0.001

Abbreviations: IQR (Interquartile Range); SD(Standard Deviation); CVA(Cerebrovascular Accident); CT(Computerized Tomography); RVU(Relative Value Unit).

occurrences [8,15,20]. In addition, our study also questions the value of using the ABCD2 score to determine candidates for ED discharge. Cur- rent literature shows that tissue-positive events can occur with low ABCD2 scores. Similarly, their C-statistics ranged from 0.5 to 0.7 in dif- ferent studies indicating poor-to-weak differentiation and poor predict- ability of subsequent CVA events [6,18,21]. Therefore, we believe the ABCD2 score may be valuable on the appropriate guidance to different Levels of care when hospitalized (e.g. admission to Observation Units for low-to-moderate risk ABCD2 score patients and hospital full admis- sion for high risk ABCD2 score patients) [22-24], whereas, ABCD3-I could be used for the guidance of ED discharge among TIA patients.

To our knowledge, this study represents the first prospective appli- cation of an ABCD3-I based pathway to TIA patients. On literature search, one similar report using ABCD2 and negative diffusion- weighted imaging yielded a 64% discharge rate directly from the ED and only one patient had CVA at one year follow-up [25]. However, this study lacked a control group [25]. Our study used a pre- and post- intervention cohort comparison and was conducted within the same system. Such an approach has ED and healthcare system advantages by reducing hospitalizations and minimizing cost, while resulting in op- timal patient-centered care outcomes (e.g. ED LOS, ED returns, 90-day CVA occurrence). Our future research will be focused on further identi- fying clinical and ED operational outcomes among high risk versus low risk tissue-based TIA in a large-scale patient size.

Although these results are promising, this study has several limita- tions. First, due to the nature of this study design, patient selection bias and missing or Inaccurate data are considerations. We did not per- form longer-term follow-up beyond 90 days, and we are also unable to know patient outcomes if they visited other Hospital systems. Second, our post-intervention group tended to have younger patients, with slightly less severe ABCD2 scores, which may have resulted in a lower index CVA occurrence. Third, the study hospital ED has the capability to perform neuroimaging testing (e.g. MRI/MRA) quickly in the ED, which makes our study feasible. Such approach usually requires hospi- tal administrative support and mutual agreement among different de- partments (e.g. radiology, neurology). In this study, we did not adjust healthcare cost for patient comorbidities, nor other confounders that might increase patient total healthcare cost. Last, this is a pilot quality improvement project with limited sample size. However, given the promising results, in order to reach generalizability, a multi-center large-scaled prospective study is warranted.

Conclusion

The initiation of an ABCD3-I based pathway for suspected TIA pa- tients in the ED significantly reduced hospital admissions and Healthcare costs with similar 90-day neurological outcomes. This path- way provided cost-effective risk stratification and enables safe dis- charge from the Emergency Department for most TIA patients.

Funding sources/disclosures

N/A

Acknowledgments

N/A

References

AJEM