You are here

Article Text

Article menu

Download PDFPDF

XML
Original research
Modelling care quality for patients after a transient ischaemic attack within the US Veterans Health Administration
  1. Greg Arling1,2,
  2. Jason J Sico3,4,
  3. Mathew J Reeves2,5,
  4. Laura Myers2,6,
  5. Fitsum Baye2,7,
  6. Dawn M Bravata2,8
  1. 1School of Nursing, Purdue University, West Lafayette, Indiana, USA
  2. 2Department of Veterans Affairs (VA) Health Services Research and Development (HSR&D) Precision Monitoring to Transform Care (PRIS-M) Quality Enhancement Research Initiative (QUERI), Richard L Roudebush VA Medical Center, Indianapolis, Indiana, USA
  3. 3Department of Internal Medicine and Neurology, Yale School of Medicine, New Haven, Connecticut, USA
  4. 4Clinical Epidemiology Research Center, VA Connecticut Health System West Haven Campus, West Haven, Connecticut, USA
  5. 5Department of Epidemiology, Michigan State University, East Lansing, Michigan, USA
  6. 6Center for Health Information and Communication (CHIC), Department of Veterans Affairs, Veterans Health Administration, Health Services Research and Development Service CIN 13-416, Indianapolis, Indiana, USA
  7. 7Department of Biostatistics, Indiana University School of Medicine, Indianapolis, Indiana, USA
  8. 8Department of Internal Medicine, Indiana University School of Medicine, Indianapolis, Indiana, USA
  1. Correspondence to Dr Greg Arling; garling{at}purdue.edu

Abstract

Objective Timely preventive care can substantially reduce risk of recurrent vascular events or death after a transient ischaemic attack (TIA). Our objective was to understand patient and facility factors influencing preventive care quality for patients with TIA in the US Veterans Health Administration (VHA).

Methods We analysed administrative data from a retrospective cohort of 3052 patients with TIA cared for in the emergency department (ED) or inpatient setting in 110 VHA facilities from October 2010 to September 2011. A composite quality indicator (QI score) pass rate was constructed from four process-related quality measures—carotid imaging, brain imaging, high or moderate potency statin and antithrombotic medication, associated with the ED visit or inpatient admission after the TIA. We tested a multilevel structural equation model where facility and patient characteristics, inpatient admission, and neurological consultation were predictors of the resident’s composite QI score.

Results Presenting with a speech deficit and higher Charlson Comorbidity Index (CCI) were positively related to inpatient admission. Being admitted increased the likelihood of neurology consultation, whereas history of dementia, weekend arrival and a higher CCI score made neurological consultation less likely. Speech deficit, higher CCI, inpatient admission and neurological consultation had direct positive effects on the composite quality score. Patients in facilities with fewer full-time equivalent neurology staff were less likely to be admitted or to have a neurology consultation. Facilities having greater organisational complexity and with a VHA stroke centre designation were more likely to provide a neurology consultation.

Conclusions Better TIA preventive care could be achieved through increased inpatient admissions, or through enhanced neurology and other care resources in the ED and during follow-up care.

  • quality measurement
  • health services research
  • chronic disease management
  • performance measures
  • statistics

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

https://doi.org/10.1136/bmjoq-2019-000641

Statistics from Altmetric.com

    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.

    INTRODUCTION

    Patients with transient ischaemic attack (TIA) are at high risk of recurrent vascular events, including repeat TIA, stroke and death.1–6 Timely delivery of guideline-concordant cerebrovascular preventive care can reduce risk of these events. Therefore, patients with TIA are ideal candidates for risk reduction interventions.7–10 Timeliness of interventions is an important consideration because preventive actions will achieve maximum benefit if received soon after an index cerebrovascular event.11 Guideline-consistent care for TIA has improved substantially over the last few years; nonetheless, considerable variation remains in care quality between patients and among facilities.12 13 Paradoxically, patients with TIA at highest risk for a recurrent cardiovascular event may be least likely to receive guideline-consistent care.14

    Our study objective was to understand patient and facility factors influencing quality of care provided to patients with TIA treated within the Veterans Health Administration (VHA) System. Quality indicators (QIs), selected from prior research and clinical guidelines, covered diagnostic evaluations including receipt of carotid imaging and brain imaging, as well as treatment interventions including prescription of high or moderate potency statins and antithrombotics.7 15 16 These four QIs were incorporated into a composite measure of care quality. To meet these objectives, we develop and test a multilevel, structural equation model (SEM) for the relationships between patient and facility characteristics and the composite QI measure. Our sample includes patients with TIA either discharged from the emergency department (ED) or admitted to the hospital.

    Methods

    We used VHA electronic health record data to select a sample of 3052 patients with a TIA. The sample represented all admissions to the ED or hospital meeting study criteria who were cared for in the ED or inpatient setting in 110 VHA facilities for the 12-month period from October 2010 to September 2011. Patients with an International Classification of Diseases, Ninth Revision code of 435.x primary discharge code were classified as having a TIA. The diagnosis for this study cohort was found to be accurate when validated with chart reviews.17 For patients who were cared for in the ED only, we selected the primary discharge diagnosis for their ED visit; for patients who were admitted, we selected the primary hospital discharge diagnosis. Patients admitted for observation were coded as an inpatient admission. A total of 3697 patients from 132 VHA facilities were identified as TIA admissions to the ED or hospital during the year. We excluded 22 facilities with fewer than 10 admissions (n=88 excluded patients) of patients with TIA because small numbers could result in unreliable facility estimates. Excluded facilities had higher rates of TIA inpatient admission, and were less likely to be located in an urban area or to have stroke centre designation. Another 536 patients were excluded because they were ineligible for one or more QIs and, as a result, would not have complete composite scores. Thus, all patients included in these analyses were eligible to receive all four quality measures as described in table 1. Institutional review board and VHA Research and Development approvals were received to support this research.

    View this table:
    Table 1

    Operational definitions of quality measures

    The VHA inpatient and outpatient data were the primary data sources. They were supplemented by linked Medicare claim files for the 65% of Veterans in the sample who were Medicare eligible. The VHA files alone or in combination with Medicare files were used to identify past medical history, presenting symptoms and healthcare utilisation (eg, inpatient admission or ED only). Pharmacy Benefits Management data were used to identify medications during the ED visit or inpatient stay; Corporate Data Warehouse data were used to identify vital signs, laboratory data, medication allergies (eg, statin allergies), orders, and consultations.

    The outcome variable was a patient-level composite score based on four inpatient QIs: carotid imaging within 2 days after index event or within 6 months before the index event, brain imaging within 2 days of index event, high or moderate potency statin prescribed within 7 days after discharge and antithrombotic (includes both antiplatelet or anticoagulation medications) prescribed within 7 days after discharge. The two medication-based processes were initially defined as receipt of the medication at the time of hospital discharge, but the definition was changed to within 7 days after discharge based on prior work validating the electronic quality measures versus chart review (eg, if a clinician ordered a medication on the day of discharge, it could take some time for that order to be filled by the pharmacy). Operational definitions of these measures are shown in table 1. Eligible patients were classified on each measure as pass (1) or fail (0). Although guidelines recommend a variety of processes of acute TIA care, these four QIs were chosen because of their clinical relevance (they each have a strong evidence base to support their use as performance measures)7 15; they can be provided across a full spectrum of medical facilities (they do not require specialised services); they include both diagnostic (eg, brain and carotid artery imaging) and therapeutic (eg, antithrombotics, statins) elements of care; the majority of patients with TIA are eligible to receive these processes of care (unlike anticoagulation for patients with atrial fibrillation which is only relevant to a minority of patients or cardiac monitoring which is similarly relevant in only a subgroup of patients with TIA); and they have been demonstrated to be valid for use in electronic quality measurement (compared against the gold standard of chart review).16 17 The composite measure of TIA care quality was defined as the proportion of the four QIs that the patient passed (ie, 0%, 25%, 50%, 75%, 100%), which was treated as a continuous variable for purposes of estimating the SEM.

    We examined the relationship between the composite care quality measure and selected patient and facility characteristics. Variables were chosen because they have been associated in prior studies with inpatient admission or care quality of stroke or patients with TIA. Patient characteristics included presenting symptoms of speech deficit, the Charlson Comorbidity Index (CCI),18 history of dementia and arrival on a weekend (ie, any time on Saturday or Sunday). A timely neurologist consultation was defined as occurring within 1 day of the index event; it could occur in the ED or after the decision to admit. Symptoms of hemiplegia were also significantly related to admission at the bivariate level; however, we excluded them from the model because of high collinearity with speech deficit (results not shown). All patient-level variables were binary with the exception of the CCI.

    At the Veterans Affairs Medical Centres (VAMCs) facility level, we included binary variables (scored 0/1) for falling in the bottom quartile for equivalent (full-time equivalent (FTE)) neurologists per 100 000 patients; stroke centre self-designation of comprehensive stroke centre or limited hours stroke centre versus other facilities that cannot offer consistent stroke care19; high VHA facility complexity level, defined by size, teaching service, and availability of intensive care services20; and high TIA patient volume defined as 25 or more patients with TIA per year.

    We specified and tested a causal model of the effects of patient and facility characteristics, inpatient admission, and neurological consultation on a patient’s QI score. Variable selection and causal structure of the model were determined a priori based on prior research (see discussion section) and available data. We hypothesised that patient characteristics (eg, having dementia, multiple other comorbidities and presenting with a speech deficit) and weekend arrival would influence whether the patient was admitted to the hospital and whether they would receive a neurological consultation, either in the ED or during the inpatient stay. We expected VAMCs with greater availability of neurologists and other stroke care resources (eg, stroke centre designation and greater organisational complexity) would have higher rates of admission and neurological consultation. Correspondingly, we expected that patients admitted to the hospital and receiving neurological consultations would also receive higher quality care.

    We tested these hypotheses with a multilevel path model with a combination of categorical and continuous observed variables using MPlus software.21Figure 1 shows the structure of the model with arrows indicating significant causal paths from the final model. History of dementia, CCI, symptoms of a speech deficit and arrival on the weekend were treated as exogenous variables, which means that they were assumed to have no causal influence on each other, although they could be correlated. We specified causal paths from patient characteristics to inpatient admission and neurological consultation, and a causal path from inpatient admission to neurological consultation. The facility variables, FTE neurologists, stroke centre designation, organisational complexity and TIA volume were also assumed to be exogenous. We specified causal paths from facility variables to inpatient admission and neurological consultation. After testing for the effects of different FTE neurologist quartiles, we settled on a dummy variable of the bottom quartile (compared with quartiles 2–4) as most predictive of admission and neurological consultation.

    Figure 1
    Figure 1

    Path model with patient and facility characteristics, inpatient admission, neurological consultation and composite care quality. Regression (path) coefficients for the structural equation model. Arrows indicate statistically significant (p<0.05) paths with regression coefficients. Standard errors and p values are reported in the table. Exogenous patient variables are dementia, Charlson Comorbidity Index, speech deficit and weekend arrival; exogenous facility-level variables are VMAC facility complexity measure, full-time equivalent neurology staff (bottom two quartiles),>25 TIA patients per year, and designation as a comprehensive stroke centre or limited hours stroke facility; and endogenous patient variables are inpatient admission and neurology consultation. TIA, transient ischaemic attack; VMCA, Veterans Affairs Medical Centre.

    For the statistical modelling, we chose the MPlus Two-Level option with the robust MLR estimator for non-normal data, clustering on VAMC and the EMA optimisation algorithm.21 Fit statistics for the model were based on the difference in log likelihood between the null and full nested models with a scaling correction factor.21

    ​Patient and public involvement

    Patients, clinicians and policymakers are deeply interested in quality of care and the benefits to be derived from this study. Because this was a retrospective study of data from an administrative record, it was not feasible to involve patients in study design or dissemination of findings to individual patients.

    Results

    Among the 3052 patients in the cohort, 95% were men and 45% were younger than age ≤65, 44% age 66–85 and 11% >age 85 (table 2). Relatively, small percentages had a history of dementia (7%) or presented with speech deficits (13%). The mean CCI was 1.77 (SD: 2.10). Eighteen percent of patients arrived on the weekend. Two-thirds of patients were admitted and the remainder discharged directly from the ED. Sixty-three percent of patients had a neurology consultation. Most of the VAMCs were located in urban areas; 69% were classified as complex; 62% had a designation as a comprehensive or limited hours stroke centre19; and 35% had 25 or more patients with TIA per year. The average FTE neurologists per 100 000 patients ranged from 3.06 for the bottom quartile (0.00–5.62) to 17.44 (11.86–26.85) for the top quartile. Only four facilities had no FTE neurologists.

    View this table:
    Table 2

    Patient and facility characteristics

    Pass rates on the individual QIs were as follows: 36% for carotid imaging, 86% for brain imaging, 57% for high or moderate potency statin at discharge and 90% for antithrombotic at discharge. The composite QI pass rates ranged from 0.00 to 1.00. Only 1% of patients had a score of 0.00 (no QIs passed); 11% a score of 0.25; 26% a score of 0.50; 42% a score of 0.75 and 21% a score of 1.00. The mean patient-level composite score was 0.67 with a SD of 0.236.

    The 110 VAMCs had a mean composite facility-level QI score of 0.67. There was considerable variation in their scores with the bottom quintile of facilities ranging from 0.45 to 0.61 and top quintile ranging from 0.74 to 0.85. The interclass correlation was 0.09, indicating that 9% of the variance in scores was between facilities and 91% was among patients within facilities. The facility mean neurological consultation rate was 0.61 (SD=0.27; IQR=0.43–0.84) and the mean admission rate was 0.67 (SD=15; IQR=0.60–0.77)

    Figure 1 shows the path model graphically with statistically significant (p<0.05) regression coefficients. Table 3 presents the regression coefficients, standard errors and level of statistical significance for each variable. A coefficient above 0.00 indicates a positive effect and below 0.00 a negative effect. The ratio of the regression coefficient to the SE in table 3 gives an indication of the strength of relationships. Two patient-level variables had significant positive effects on being admitted to the hospital: speech deficit (1.652) and CCI score (0.115). Weekend arrival (0.247) nearly achieved significance at p<0.059.

    View this table:
    Table 3

    Significant patient and facility variables from the structural equation model

    At the facility level, being in the bottom quartile for FTE neurologists had a negative effect on inpatient admission (−0.411), while having a higher volume of patients with TIA had a positive effect (0.547). Four patient-level variables had significant effects on neurological consultation: admission to the hospital had a strong positive effect (1.522), whereas weekend arrival (−0.327), history of dementia (−0.687) and CCI (−0.056) had negative effects. Three facility-level variables had significant effects on receiving a neurological consultation: bottom quartile FTE neurologists had a negative effect (−0.674); stroke centre designation (0.880) and facility complexity (1.078) had positive effects.

    Two patient characteristics had significant positive effects on QI score: speech deficit (0.033) and CCI (0.009). Inpatient admission had a strong positive effect (0.231) while neurological consultation had a modest effect (0.059) on QI score. In the initial models, we tested for significant effects between facility variables and care quality but found none; hence, facility-level variables did not directly influence the QI score after controlling for patient characteristics, neurological consultation and decision to admit. The log likelihood difference test (−4490, p <.001) indicated a good model fit.

    To summarise, symptoms of a speech deficit and a higher CCI contributed to being admitted to the inpatient setting. Being admitted had a strong positive effect on receiving a neurology consultation. In contrast, weekend arrival, history of dementia and higher CCI score made neurological consultation less likely. Facilities with a greater volume of patients with TIA had higher inpatient admission rates, whereas facilities with greater complexity and facilities with a stroke centre designation had higher rates of neurological consultation. Inpatient admission had the strongest positive effect on care quality. A speech deficit and higher CCI score, and receiving a neurological consultation also had significant positive effects on care quality.

    Discussion

    Prior studies of TIA care quality have dealt primarily with patients with TIA who were admitted to the hospital13 14; although one study by Kapral and colleagues22 examined care quality for patients with TIA who were either admitted to the hospital or discharged directly from the ED. In their sample of patients from the Ontario Stroke registry, admitted patients were significantly more likely to receive recommended evaluations and treatments, including carotid and brain imaging, lipid-lowering medications and antithrombotic therapy.22 They also found wide variation across hospitals in the proportion of patients with TIA admitted. Other studies have found that history of dementia,23 24 multiple comorbidities,25 speech deficit26 and weekend arrival27 have been associated with admissions or improved care quality for stroke or related chronic conditions.

    We expand on prior research by including both patients with TIA who were discharged from the ED as well as those who were admitted to the hospital. We found that diagnosis and treatment of patients with TIA in the ED without admission to the hospital was a common occurrence (33% of patients), and one that had negative implications for care quality. We also expand on prior research by specifying and testing a causal model (SEM) that includes facility as well as patient characteristics. Prior studies have focused on patient-level characteristics, looking separately at patient characteristics as predictors of inpatient admission, or inpatient admission as a predictor of care quality. Our multilevel, SEM approach allowed us to specify and test specific causal relationships between both patient and facility-level variables simultaneously in a single set of equations.28 29 The SEM approach provides more information than a conventional regression equation because it tests a pre-specified causal model. Moreover, we expanded on the typical single-level SEM model by examining effects at both facility and patient levels.

    Admitting patients for evaluation and management of a suspected TIA is a complex decision, involving patient-level, provider-level and systems-level factors.30 Hospitalisation of patients with TIA can lead to more rapid diagnostic evaluation, prompt initiation of appropriate cerebrovascular preventive therapies, and treatment of intracranial and carotid artery atherosclerosis.31 On the other hand, outpatient evaluation and management of TIA can lower healthcare costs and potentially enhance patient convenience. Studies of post-TIA outpatient clinics—primarily from Europe—have shown that requisite evaluation can occur in less than a day for most patients, and only 25% of patients require being hospitalised for more than a day for appropriate therapies (eg, management of atrial fibrillation).32 However, these specialised clinics are not routinely available in the VHA or other US health systems.

    There are several possible explanations for the positive relationship between presenting with a speech impairment and both inpatient admission and care quality. An overt symptom, that is, speech impairment can reduce diagnostic uncertainty about the TIA, contributing to the decision to admit or initiate preventive care.30 Among patients presenting to healthcare providers with TIA within another large, integrated healthcare system (Kaiser Health System), persistent neurological impairment in speech, strength and gait were positively associated with the decision to admit.26 Speech is also a component of the ABCD2 score, which in combination with other factors has been recommend for use in deciding on inpatient admission.33

    The positive relationship between CCI and hospital admission was expected. Providers are more likely to admit ‘medically complex’ patients with TIA to the hospital, with a higher burden of medical comorbidities, and potentially downward healthcare trajectories. On the other hand, CCI had a negative effect on neurological consultation. For patients with comorbid conditions, management of concomitant active medical problems (eg, hypoxia secondary to aspiration pneumonia/pneumonitis; rapid atrial fibrillation) may have taken precedence over neurological consultation, and/or confirming the diagnosis of TIA was less important in this type of patient.

    The positive effect of CCI on QI score is a welcomed finding. Among stroke patients, higher CCI scores have been associated higher rates of disability and mortality outcomes at hospital discharge and at 1-year follow-up.25 Outside of the cerebrovascular population, patients with higher Charlson scores have been found to have higher rates of 30-day hospital readmission.34

    It was interesting to discover that patients with a history of dementia were less likely to have a neurologist consultation, especially given that vascular risk factors are also well-established risk factors for the development and worsening of dementia.35 36 Deficiencies in the quality of care for patients with dementia have been well documented.23 We should point out that dementia had no significant relationship to the decision to admit or the QI score.

    Although several facility-level characteristics have been associated with improved care for patients with stroke,37–39 very little is known about facility factors that may affect care quality for patients with TIA. Our finding that weekend arrival lowered the likelihood of neurological consultation has been described in prior studies. On weekends, hospitals tend to have decreased staffing, and hence, decreased access to personnel to conduct testing (eg, carotid imaging) and evaluation (eg, neurology consultation).40 A systematic review of studies examining off-hour (nights and weekends) admission of patients with acute ischaemic stroke found increased short-term mortality and disability for patients admitted during off-hours (ie, nights and weekends). The influence of off-hours admission among patients with TIA and stroke has been mitigated when facilities have a comprehensive stroke centre40 where there is greater availability of personnel with stroke expertise.41

    Our finding that composite care quality was most strongly affected by hospital admission is consistent with prior research.22 Similarly, we expected a positive relationship between neurologist consultation and care quality. Neurologists often play a pivotal role in the inpatient management of patients with cerebrovascular disease within stroke centres and elsewhere.30 The association between neurology consultation during the acute event, or afterwards in follow-up, and higher quality of care delivery has been demonstrated in the Early Use of Existing Preventive Strategies Stroke study.42 Although this relationship between neurology consultation and both quality of stroke and TIA care16 43 44 and improved patients outcomes43 45–47 has been demonstrated in numerous studies, it remains unclear exactly what aspect of neurology consultation provides these benefits. In a note of caution about access to neurologists, Horner et al48 found a selection effect; stroke patients admitted to a neurology service in their study had better prognostic profiles than a comparison group not admitted to the service. This finding is in line with our study where a higher Charlson morbidity had a negative effect on receiving a neurologist consultation. In addition, our facilities in the bottom quartile of FTE neurologists had low rates of impatient admissions. On the other hand, after taking into account morbidity and other patient characteristics in our SEM model, we found a direct positive effect of neurology consultation on care quality. Based on existing evidence, we would encourage systems caring for patients with TIA to ensure access to neurology consultation whether via traditional in-person or newer telehealth approaches.

    These findings highlight the importance of the organisational context in influencing hospital admission and neurological consultation. As expected, patients in facilities with low FTE neurologists were less likely to receive a neurological consultation. Among other facility-level variables, greater TIA patient volume was positively associated with both hospital admission and receipt of neurology consultation, whereas greater organisational complexity and stroke centre designation were associated with patients with TIA seeing a neurologist. These findings suggest that more experience in caring for patients with TIA combined with greater expertise and care resources influence admission decisions and access to specialised care once admitted. Interestingly, none of the facility-level variables had significant direct effects on QI score, so their influence is likely mediated by their effect on admission and neurologist consultation.

    We acknowledge several study limitations. The generalisability of findings is limited because the VHA is an integrated delivery healthcare system serving a Veteran population that is primarily male. Second, although the process measures were all validated against chart review, these data are drawn from administrative records.17 Despite the extensive use of electronic records in the VHA, recording errors or omissions could affect the results. Third, we focused our analysis on care received either in the ED or inpatient setting after the TIA event. Future work should address longer-term outpatient TIA QIs such as those related to management of hypertension or post-discharge issues such as medication non-adherence.28 Fourth, given that time of symptom onset is not available in the electronic data used in this study, we could not evaluate how differences in the time from symptom onset may influence the observed relationships. A patient who is distant from symptom onset and is no longer symptomatic is more likely to be discharged home from the ED in contrast to a patient who has more recent symptom onset who is still symptomatic. Finally, we acknowledge that the decision to admit a patient is complex and it can be influenced by many factors (eg, patient preference, bed availability) that were unmeasured in our study.30

    Conclusions

    Our findings raise a number of issues about delivery of high-quality care to patients after a TIA. First, there is the decision about inpatient admission. The Veterans Affairs (VA) and other healthcare organisations could strive to increase inpatient admissions for patients presenting with a TIA. Healthcare systems also might use observation units to expedite care for patients with TIA.49 50 But, is the incremental increase in care quality associated with an inpatient admission worth the extra costs, both to the VA system and patients? Alternatively, resources might be directed to better coordinate follow-up care, particularly enhanced coordination between the ED, neurologists and primary care.9 42 Rapid access outpatient clinics, for example, show promise as new models of TIA care, although they are yet to have a presence in the VA system. The VA has recently instituted a teleneurology programme to expand access in rural areas. It is too soon to tell how it will influence admitting practices or care quality. Presenting on the weekend diminished the chance of a neurologist consultation. The VA and other healthcare organisations should consider expanding on-call neurology staff and availability of other resources on the weekend and during off hours. Implications for our research findings on TIA care quality and possibilities for future quality improvement work are shown in table 4. Finally, providers should carefully consider how presenting symptoms and comorbid conditions may influence decisions about inpatient admissions, involvement of neurologists and adherence to guideline concordant preventive care.

    View this table:
    Table 4

    Implications of research findings for care quality and quality improvement initiatives

    References

    1. ↵
      1. Johnston SC,
      2. Gress DR,
      3. Browner WS, et al
      . Short-Term prognosis after emergency department diagnosis of TIA. JAMA2000;284:2901–6.doi:10.1001/jama.284.22.2901
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      1. Wu CM,
      2. McLaughlin K,
      3. Lorenzetti DL,
      4. Hill MD, et al
      . Early risk of stroke after transient ischemic attack: a systematic review and meta-analysis. Arch Intern Med2007;167:2417–22.doi:10.1001/archinte.167.22.2417
      OpenUrlCrossRefPubMedWeb of Science
    3. ↵
      1. Giles MF,
      2. Albers GW,
      3. Amarenco P, et al
      . Early stroke risk and ABCD2 score performance in tissue- vs time-defined TIA: a multicenter study. Neurology2011;77:1222–8.doi:10.1212/WNL.0b013e3182309f91
      OpenUrlCrossRefPubMed
    4. ↵
      1. Allen NB,
      2. Kaltenbach L,
      3. Goldstein LB, et al
      . Regional variation in recommended treatments for ischemic stroke and TIA: get with the guidelines-stroke 2003-2010. Stroke2012;43:1858–64.doi:10.1161/STROKEAHA.112.652305
      OpenUrlAbstract/FREE Full Text
    5. ↵
      1. Fisher M
      . Stroke and TIA: epidemiology, risk factors, and the need for early intervention. Am J Manag Care2008;14:S204–11.
      OpenUrlPubMed
    6. ↵
      1. Amarenco P,
      2. Lavallée PC,
      3. Labreuche J, et al
      . One-Year risk of stroke after transient ischemic attack or minor stroke. N Engl J Med2016;374:1533–42.doi:10.1056/NEJMoa1412981
      OpenUrlCrossRefPubMed
    7. ↵
      1. Kernan WN,
      2. Ovbiagele B,
      3. Black HR, et al
      . Guidelines for the prevention of stroke in patients with stroke and transient ischemic attack: a guideline for healthcare professionals from the American heart Association/American stroke association. Stroke2014;45:2160–236.doi:10.1161/STR.0000000000000024
      OpenUrlAbstract/FREE Full Text
    8. ↵
      1. Cadilhac DA,
      2. Kim J,
      3. Lannin NA, et al
      . Better outcomes for hospitalized patients with TIA when in stroke units: an observational study. Neurology2016;86:2042–8.doi:10.1212/WNL.0000000000002715
      OpenUrl
    9. ↵
      1. Sacco RL,
      2. Rundek T
      . The value of urgent specialized care for TIA and minor stroke. N Engl J Med2016;374:1577–9.doi:10.1056/NEJMe1515730
      OpenUrlPubMed
    10. ↵
      1. Rothwell PM,
      2. Giles MF,
      3. Chandratheva A, et al
      . Effect of urgent treatment of transient ischaemic attack and minor stroke on early recurrent stroke (express study): a prospective population-based sequential comparison. Lancet2007;370:1432–42.doi:10.1016/S0140-6736(07)61448-2
      OpenUrlCrossRefPubMedWeb of Science
    11. ↵
      1. Rothwell PM,
      2. Johnston SC
      . Transient ischemic attacks: stratifying risk. Stroke2006;37:320–2.doi:10.1161/01.STR.0000200555.89117.d2
      OpenUrlFREE Full Text
    12. ↵
      1. Bangalore S,
      2. Schwamm L,
      3. Smith EE, et al
      . Secondary prevention after ischemic stroke or transient ischemic attack. Am J Med2014;127:728–38.doi:10.1016/j.amjmed.2014.03.011
      OpenUrlCrossRefPubMed
    13. ↵
      1. Fonarow GC,
      2. Reeves MJ,
      3. Smith EE, et al
      . Characteristics, performance measures, and in-hospital outcomes of the first one million stroke and transient ischemic attack admissions in get with the Guidelines-Stroke. Circ Cardiovasc Qual Outcomes2010;3:291–302.doi:10.1161/CIRCOUTCOMES.109.921858
      OpenUrlAbstract/FREE Full Text
    14. ↵
      1. O'Brien EC,
      2. Zhao X,
      3. Fonarow GC, et al
      . Quality of care and ischemic stroke risk after hospitalization for transient ischemic attack: findings from get with the Guidelines-Stroke. Circ Cardiovasc Qual Outcomes2015;8:S117–24.doi:10.1161/CIRCOUTCOMES.115.002048
      OpenUrl
    15. ↵
      1. Smith EE,
      2. Saver JL,
      3. Alexander DN, et al
      . Clinical performance measures for adults hospitalized with acute ischemic stroke: performance measures for healthcare professionals from the American heart Association/American stroke association. Stroke2014;45:3472–98.doi:10.1161/STR.0000000000000045
      OpenUrlFREE Full Text
    16. ↵
      1. Bravata DM,
      2. Myers LJ,
      3. Arling G, et al
      . Quality of care for veterans with transient ischemic attack and minor stroke. JAMA Neurol2018;75:419–27.doi:10.1001/jamaneurol.2017.4648
      OpenUrl
    17. ↵
      1. Bravata DM,
      2. Myers LJ,
      3. Cheng E, et al
      . Development and validation of electronic quality measures to assess care for patients with transient ischemic attack and minor ischemic stroke. Circ Cardiovasc Qual Outcomes2017;10.doi:10.1161/CIRCOUTCOMES.116.003157
    18. ↵
      1. Charlson ME,
      2. Pompei P,
      3. Ales KL, et al
      . A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis1987;40:373–83.doi:10.1016/0021-9681(87)90171-8
      OpenUrlCrossRefPubMedWeb of Science
    19. ↵
      1. Damush TM,
      2. Miller KK,
      3. Plue L, et al
      . National implementation of acute stroke care centers in the Veterans health administration (vha): formative evaluation of the field response. J Gen Intern Med2014;29(Suppl 4):845–52.doi:10.1007/s11606-014-3036-1
      OpenUrl
    20. ↵
      1. Szabo C
      . Veterans health administration NLB human resources Committee 2005 facility complexity model. Washington, DC: Veterans Health Administration, 2005.
    21. ↵
      1. Muthen LK,
      2. Muthen BO
      . Mplus user’s guide version 7.4. Los Angeles, CA: Muthen&Muthen, 2016.
    22. ↵
      1. Kapral MK,
      2. Hall R,
      3. Fang J, et al
      . Association between hospitalization and care after transient ischemic attack or minor stroke. Neurology2016;86:1582–9.doi:10.1212/WNL.0000000000002614
      OpenUrlPubMed
    23. ↵
      1. Chodosh J,
      2. Mittman BS,
      3. Connor KI, et al
      . Caring for patients with dementia: how good is the quality of care? results from three health systems. J Am Geriatr Soc2007;55:1260–8.doi:10.1111/j.1532-5415.2007.01249.x
      OpenUrlCrossRefPubMedWeb of Science
    24. ↵
      1. Rothman AA,
      2. Wagner EH
      . Chronic illness management: what is the role of primary care?Ann Intern Med2003;138:256–61.doi:10.7326/0003-4819-138-3-200302040-00034
      OpenUrlCrossRefPubMedWeb of Science
    25. ↵
      1. Goldstein LB,
      2. Samsa GP,
      3. Matchar DB, et al
      . Charlson index comorbidity adjustment for ischemic stroke outcome studies. Stroke2004;35:1941–5.doi:10.1161/01.STR.0000135225.80898.1c
      OpenUrlAbstract/FREE Full Text
    26. ↵
      1. Josephson SA,
      2. Sidney S,
      3. Pham TN, et al
      . Factors associated with the decision to hospitalize patients after transient ischemic attack before publication of prediction rules. Stroke2008;39:411–3.doi:10.1161/STROKEAHA.107.491316
      OpenUrlAbstract/FREE Full Text
    27. ↵
      1. Hsieh C-Y,
      2. Lin H-J,
      3. Chen C-H, et al
      . "Weekend effect" on stroke mortality revisited: Application of a claims-based stroke severity index in a population-based cohort study. Medicine2016;95:e4046.doi:10.1097/MD.0000000000004046
    28. ↵
      1. Muthén B,
      2. Asparouhov T
      . Causal effects in mediation modeling: an introduction with applications to latent variables. Struct Equ Modeling2015;22:12–23.doi:10.1080/10705511.2014.935843
      OpenUrl
    29. ↵
      1. Muthén BO
      . Beyond SEM: general latent variable modeling. Behaviormetrika2002;29:81–117.doi:10.2333/bhmk.29.81
      OpenUrlCrossRef
    30. ↵
      1. Homoya BJ,
      2. Damush TM,
      3. Sico JJ, et al
      . Uncertainty as a key influence in the decision to admit patients with transient ischemic attack. J Gen Intern Med2019;34:1715–23. E-Publication.doi:10.1007/s11606-018-4735-9
      OpenUrl
    31. ↵
      1. Cucchiara BL,
      2. Kasner SE
      . All patients should be admitted to the hospital after a transient ischemic attack. Stroke2012;43:1446–7.doi:10.1161/STROKEAHA.111.636746
      OpenUrlFREE Full Text
    32. ↵
      1. Amarenco P
      . Not all patients should be admitted to the hospital for observation after a transient ischemic attack. Stroke2012;43:1448–9.doi:10.1161/STROKEAHA.111.636753
      OpenUrlFREE Full Text
    33. ↵
      1. Easton JD,
      2. Saver JL,
      3. Albers GW, et al
      . Definition and evaluation of transient ischemic attack. Stroke2009;40:2276–93.doi:10.1161/STROKEAHA.108.192218
      OpenUrlAbstract/FREE Full Text
    34. ↵
      1. Logue E,
      2. Smucker W,
      3. Regan C
      . Admission data predict high Hospital readmission risk. J Am Board Fam Med2016;29:50–9.doi:10.3122/jabfm.2016.01.150127
      OpenUrlAbstract/FREE Full Text
    35. ↵
      1. O’Brien JT,
      2. Markus HS
      . Vascular risk factors and Alzheimer’s disease. BMC Med2014;12:218.doi:10.1186/s12916-014-0218-y
    36. ↵
      1. Barnes DE,
      2. Yaffe K
      . The projected effect of risk factor reduction on Alzheimer's disease prevalence. Lancet Neurol2011;10:819–28.doi:10.1016/S1474-4422(11)70072-2
      OpenUrlCrossRefPubMedWeb of Science
    37. ↵
      1. Alberts MJ,
      2. Wechsler LR,
      3. Jensen MEL, et al
      . Formation and function of acute stroke-ready hospitals within a stroke system of care recommendations from the brain attack coalition. Stroke2013;44:3382–93.doi:10.1161/STROKEAHA.113.002285
      OpenUrlAbstract/FREE Full Text
    38. ↵
      1. Kada A,
      2. Nishimura K,
      3. Nakagawara J, et al
      . Development and validation of a score for evaluating comprehensive stroke care capabilities: J-ASPECT study. BMC Neurol2017;17:46.doi:10.1186/s12883-017-0815-4
    39. ↵
      1. Douglas VC,
      2. Tong DC,
      3. Gillum LA, et al
      . Do the brain attack Coalition's criteria for stroke centers improve care for ischemic stroke?Neurology2005;64:422–7.doi:10.1212/01.WNL.0000150903.38639.E1
      OpenUrlCrossRefPubMed
    40. ↵
      1. Albright KC,
      2. Raman R,
      3. Ernstrom K, et al
      . Can comprehensive stroke centers erase the 'weekend effect'?Cerebrovasc Dis2009;27:107–13.doi:10.1159/000177916
      OpenUrlCrossRefPubMedWeb of Science
    41. ↵
      1. Sorita A,
      2. Ahmed A,
      3. Starr SR, et al
      . Off-hour presentation and outcomes in patients with acute ischemic stroke: a systematic review and meta-analysis. Eur J Intern Med2014;25:394–400.doi:10.1016/j.ejim.2014.03.012
      OpenUrlCrossRefPubMed
    42. ↵
      1. Luengo-Fernandez R,
      2. Gray AM,
      3. Rothwell PM
      . Effect of urgent treatment for transient ischaemic attack and minor stroke on disability and hospital costs (express study): a prospective population-based sequential comparison. Lancet Neurol2009;8:235–43.doi:10.1016/S1474-4422(09)70019-5
      OpenUrlCrossRefPubMedWeb of Science
    43. ↵
      1. Goldstein LB,
      2. Matchar DB,
      3. Hoff-Lindquist J, et al
      . Va stroke study: neurologist care is associated with increased testing but improved outcomes. Neurology2003;61:792–6.doi:10.1212/01.WNL.0000082724.77447.3A
      OpenUrlCrossRefPubMed
    44. ↵
      1. Reeves MJ,
      2. Gargano J,
      3. Maier KS, et al
      . Patient-Level and hospital-level determinants of the quality of acute stroke care: a multilevel modeling approach. Stroke2010;41:2924–31.doi:10.1161/STROKEAHA.110.598664
      OpenUrlAbstract/FREE Full Text
    45. ↵
      1. Bray BD,
      2. Ayis S,
      3. Campbell J, et al
      . Associations between the organisation of stroke services, process of care, and mortality in England: prospective cohort study. BMJ2013;346:f2827.doi:10.1136/bmj.f2827
    46. ↵
      1. Gillum LA,
      2. Johnston SC
      . Influence of physician specialty on outcomes after acute ischemic stroke. J Hosp Med2008;3:184–92.doi:10.1002/jhm.313
      OpenUrlCrossRefPubMed
    47. ↵
      1. Smith MA,
      2. Liou J-I,
      3. Frytak JR, et al
      . 30-Day survival and rehospitalization for stroke patients according to physician specialty. Cerebrovasc Dis2006;22:21–6.doi:10.1159/000092333
      OpenUrlCrossRefPubMedWeb of Science
    48. ↵
      1. Horner RD,
      2. Matchar DB,
      3. Divine GW, et al
      . Relationship between physician specialty and the selection and outcome of ischemic stroke patients. Health Serv Res1995;30:275–87.
      OpenUrlPubMedWeb of Science
    49. ↵
      1. Nahab F,
      2. Leach G,
      3. Kingston C, et al
      . Impact of an emergency department observation unit transient ischemic attack protocol on length of stay and cost. J Stroke Cerebrovasc Dis2012;21:673–8.doi:10.1016/j.jstrokecerebrovasdis.2011.02.017
      OpenUrlCrossRefPubMed
    50. ↵
      1. Wheatley MA,
      2. Ross MA
      . Care of neurologic conditions in an observation unit. Emerg Med Clin North Am2017;35:603–23.doi:10.1016/j.emc.2017.03.007
      OpenUrl

    Footnotes

    • Contributors GA: planning, analysis, manuscript preparation and revision (First Author). FB: planning and analysis. LM: data acquisition and manuscript revision and preparation. JJS: funding, manuscript preparation and revision. MJR: planning, manuscript preparation and revision. DMB: funding, planning, data acquisition, manuscript preparation and revision (Senior Author).

    • Funding Supported by Department of Veterans Affairs (VA), Health Services Research & Development Service (HSRD), Stroke Quality Enhancement Research Initiative (QUERI) Service Directed Project (SDP 12-178; Bravata, Principal Investigator). Dr. Sico is supported by a VA HSRD Career Development Award (HX001388-01A1). Support for VA/Centers for Medicare and Medicaid Service (CMS) data is provided by the Department of Veterans Affairs, VA Health Services Research and Development Service, VA Information Resource Center (SDR 02-237 and 98-004).

    • Competing interests None declared.

    • Patient consent for publication Not required.

    • Ethics approval The manuscript contains a statement: ‘Indiana University Institutional review board and VHA Research and Development approvals were received to support this research’.

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

    • Data availability statement No data are available.

    • Author note All authors made a substantial contribution to the (1) conception or design of the work; (2) acquisition, analysis or interpretation of data and (3) or drafting the work or revising it critically for important intellectual content. All authors have given final approval of the draft. All authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.