Readmission for Stroke and Quality of Care among Patients Hospitalized with Transient Ischemic Attack (TIA): Findings from Get With The Guidelines (GWTG)-Stroke Emily C. O’Brien1, Xin Zhao1, Gregg C. Fonarow2, Eric E. Smith3, Lee H. Schwamm4, Deepak L. Bhatt5, Ying Xian1, Jeffrey L. Saver2, Mathew J. Reeves6, Eric D. Peterson1, Adrian F. Hernandez1 1Duke Clinical Research Institute, Durham, NC; 2Ronald-Reagan UCLA Medical Center, Los Angeles, CA; 3Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta, Canada; 4Massachusetts General Hospital, Harvard Medical School, Boston, MA; 5VA Boston Medical Center, Harvard Medical School, Boston, MA; 6Michigan State University, East Lansing, MI Presenter Disclosure Information Name: Emily O’Brien, PhD Title: Readmission for Stroke and Quality of Care among Patients Hospitalized with Transient Ischemic Attack (TIA): Findings from Get With The Guidelines (GWTG)Stroke DISCLOSURE INFORMATION: Disclosures: The American Heart Association and the American Stroke Association fund Get With The Guidelines-Stroke. The program has been supported in part by unrestricted educational grants to the American Heart Association by Pfizer, Inc., New York, NY, and the Merck-Schering Plough Partnership (North Wales, PA). Background • Transient Ischemic Attack (TIA) is associated with a markedly elevated risk for ischemic stroke • Comprehensive risk prediction tools encompassing readily available variables may enhance identification of patients at risk for recurrent stroke • Evidence-based management of TIA may reduce the risk of recurrent stroke • The benefit of optimal clinical management in the context of underlying risk has not been fully explored Objectives 1. Estimate risk of one-year admission for ischemic stroke after TIA hospitalization 2. Assess receipt of quality-of-care metrics by baseline readmission risk 3. Characterize the association between quality-of-care metrics provided during TIA hospitalization and one-year risk of ischemic stroke readmission Hypotheses • Patients with higher baseline ischemic stroke readmission risk are less likely to receive evidence-based care • Receipt of evidence-based care is associated with lower rates of readmission for all baseline risk subgroups • Data Sources Methods – Get With The Guidelines-Stroke Hospital-based quality improvement initiative Trained personnel abstract demographic, clinical, and event information at participating sites – Probabilistic linkage to Medicare inpatient claims using indirect identifiers • Study population – Starting population: N=108,527 TIA patients from 1326 GWTG sites – Exclusions: Not linked to CMS data, or non-index records (N = 9145) Transferred out, hospice, death, or no documented discharge destination (N =3471) CMO (N = 435) Not enrolled in Medicare FFS at hospital discharge (N = 4721) Admitted after 2008 (N = 22,690) CMS discharge date after 2008 (N = 173) – Final study population: N=58,809 GWTG Readmission Risk Score • Predicted probability of readmission based on patient baseline characteristics – Demographics – Comorbidities • Cox proportional hazards modeling with backwards selection (stay criterion of p=0.05) • Discriminative performance of the model examined using c-statistics and ROC curves • Patients categorized into quintiles of predicted readmission risk Evidence-Based Care • Individual – – – – – Antithrombotics by hospital day 2 Anticoagulation for patients with atrial fibrillation Antithrombotics at discharge Lipid-lowering medications at discharge Smoking cessation counseling • Global – TIA defect-free care: receipt of all measures for which the patient was eligible Outcomes – Primary Hospitalization for ischemic stroke – Secondary All-cause mortality Statistical Analysis • Baseline characteristics compared using Pearson Chi-squared and Wilcoxon rank sum tests • Cox proportional hazards model to estimate risk for ischemic stroke readmission over one year in derivation cohort, with performance evaluation in validation cohort • Censoring at death or loss of Medicare eligibility • Cox models to estimate association between DFC and readmission within score quintiles Baseline Characteristics (%) • One-year risk of ischemic stroke hospitalization=5.7% Stroke Readmission within 1 year (N=3,318) No Stroke Readmission within 1 year (N=54,854) P-Value Age, median 80.0 79.0 <.0001 Male gender 39.2 39.0 0.78 Black Race 10.3 7.4 <.0001 Prior stroke 44.6 34.4 <.0001 CAD/Prior MI 37.1 32.8 <.0001 Carotid stenosis 6.0 5.5 0.27 Diabetes 31.7 26.7 <.0001 PVD 6.6 5.2 .0004 Hypertension 81.8 81.2 0.40 Heart Failure 3.4 2.3 <.0001 Smoker 9.5 8.1 0.006 Dyslipidemia 39.3 45.1 <.0001 Variable In-Hospital Quality Measures (%) Stroke Readmission within 1 year (N=3,318) No Stroke Readmission within 1 year (N=54,854) P-Value Early Antithrombotics 96.8 96.3 0.28 Discharge Antithrombotics 95.9 95.7 0.71 Anticoagulants for AF 85.9 89.1 0.02 Statin (LDL>100 or ND) 59.4 61.4 0.04 Smoking Cessation 88.4 85.4 0.16 Defect-Free Care† 62.2 63.9 0.04 Variable † Defect-free care=receipt of all TIA achievement measures for which the patient was eligible GWTG 1-Year IS Readmission Risk Score Model Variable Hazard Ratio 95% CI Age (per 10 year increase) 1.01 1.01, 1.02 White Race 0.79 0.72, 0.86 Atrial Fibrillation/Flutter 1.46 1.34, 1.58 Previous Stroke/TIA 1.45 1.35, 1.55 CAD/Prior MI 1.15 1.07, 1.23 Diabetes Mellitus 1.26 1.16, 1.36 Smoking 1.35 1.19, 1.52 Dyslipidemia 0.87 0.81, 0.93 Arrival Mode (EMS vs. Other) 1.20 1.12, 1.28 Ambulate Independently at Discharge 0.82 0.76, 0.88 C-statistic=0.603 1-Year IS Readmission Overall 1-year Risk • Death: 11.8% • IS Readmission: 5.7% Observed Risk (%) 25 20 15 IS Readmission Death 10 5 0 1st 2nd 3rd 4th 5th GWTG Readmission Risk Score Quintile *‡ GWTG Readmission for Stroke Risk Score estimated from age, gender, race, history of stroke/TIA, prosthetic heart valve, CAD/Prior MI, carotid stenosis, diabetes mellitus, PVD, hypertension, HF, smoker, dyslipidemia, hospital size, hospital type, and geographic region (c-statistic 0.59) Results 100 90 Early antithrombotics Antithrombotics at discharge 80 % Anticoagulation for AF Statins 70 Smoking Cessation 60 DFC 50 1st 2nd 3rd 4th 5th GWTG Readmission Risk Score Quintile *‡ GWTG Readmission for Stroke Risk Score estimated from age, gender, race, history of stroke/TIA, prosthetic heart valve, CAD/Prior MI, carotid stenosis, diabetes mellitus, PVD, hypertension, HF, smoker, dyslipidemia, hospital size, hospital type, and geographic region (c-statistic 0.59) DFC and IS Readmission HR (95% CI) 1.4 1.0 0.7 Unadjusted 1st 2nd 3rd 4th GWTG Readmission Risk Quintile 5th Limitations • Results may be influenced by residual confounding • Information about medication use after discharge not available • Percent of TIA patients admitted at each GWTG hospital not known • DFC is a composite measure and appears to be driven largely by the statin use measure in this population • Results may not be applicable to broader TIA patient population Conclusions • Patients who were readmitted for ischemic stroke within one year of TIA had a greater comorbidity burden than patients who were not readmitted • TIA patients with a high baseline risk of readmission for IS are less likely to receive defect-free care than low-risk patients, largely due to lack of statin treatment • Standardized risk assessment and delivery of optimal inpatient care for TIA may help to reduce this apparent risk-treatment mismatch Acknowledgements • The authors would like to thank the staff and participants of the GWTG-Stroke Registry for their important contributions to this work Thank you Candidate Risk Score Variables Age (per 10 year increase) Female Race (White vs. Other) Medical History of Atrial Fibrillation/Flutter Medical History of Prosthetic Heart Valve Medical History of Previous Stroke/TIA Medical History of CAD/Prior MI Medical History of Carotid Stenosis Medical History of Diabetes Mellitus Medical History of PVD Medical History of Hypertension Medical History of Smoking Medical History of Dyslipidemia Medical History of HF Academic Region – NE vs. W Region – MW vs. W Region – S vs. W Risk Treatment Paradox • Documented for heart failure and acute MI (possibly improving over time) • Uncertainty about the risk: benefit ratio in patients at higher risk who are generally under-represented in randomized trials • Information gaps in administrative datasets (i.e., lack of data on confounding clinical and functional variables that the clinician must weigh in making clinical decisions but which are not captured in administrative databases) • Should also consider overutilization of nonevidence based therapies ABCD2 Score -Validated for 2, 7, and 90 day IS but not long-term IS Readmission Risk Prediction • Meta-analysis of 26 unique risk-prediction models • C-statistics: 0.55-0.65 • Two of five found that addition of functional or social variables improved discrimination • Limitations include lack of information on hospital and systems-level factors, do not assess HRQOL, do not account for preventable readmissions JAMA. 2011;306(15):1688-1698