Variation and Trends in the Documentation of National Institutes of Health Stroke Scale in GWTG-Stroke Hospitals

Anything that uses the National Institutes of Health Stroke Scale for measurement is almost worthless because of subjectivity.
http://circoutcomes.ahajournals.org/content/8/6_suppl_3/S90.abstract?etoc

1. Mathew J. Reeves, PhD,
2. Eric E. Smith, MD, MPH,
3. Gregg C. Fonarow, MD,
4. Xin Zhao, MS,
5. Michael Thompson, PhD,
6. Eric D. Peterson, MD,
7. Lee H. Schwamm, MD and
8. DaiWai Olson, PhD

+ Author Affiliations

1. From the Department of Epidemiology and Biostatistics, Michigan State University, East Lansing (M.J.R., M.T.); Department of Clinical Neurosciences, Hotchkiss Brain Institute, University of Calgary, Calgary, Canada (E.E.S.); Division of Cardiology, Geffen School of Medicine, UCLA (G.C.F.); Duke Clinical Research Centre, Durham, NC (X.Z., E.D.P.); Department of Neurology, Massachusetts General Hospital, Boston (L.H.S.); and Department of Neurology and Neurotherapeutics, University of Texas Southwestern, Dallas (D.O.).

1. Correspondence to Matthew J. Reeves, PhD, Department of Epidemiology, Michigan State University, 900 W Fee Rd, East Lansing, MI 48824. E-mail reevesm@msu.edu

Abstract

Background—Although National Institutes of Health Stroke Scale (NIHSS) is an important prognostic variable, it is often incompletely documented in clinical registries, such as Get With The Guidelines (GWTG)–Stroke. We describe trends in NIHSS documentation by GWTG-Stroke hospitals, identify patient-level and hospital-level factors associated with documentation, and determine the degree to which the reporting of NIHSS is potentially biased.

Methods and Results—We analyzed NIHSS documentation in 1 184 288 patients with acute ischemic stroke admitted to 1704 GWTG-Stroke hospitals between 2003 and 2012. We used multivariable logistic regression models to identify hospital-level and patient-level predictors of NIHSS documentation. We examined the relationship between hospital-level NIHSS documentation rates and observed NIHSS scores to determine whether the reporting of NIHSS data was subject to selection bias. The overall NIHSS documentation rate was 56.1%; the median NIHSS was 4 (interquartile range, 2–9). Between 2003 and 2012, mean hospital-level NIHSS documentation increased dramatically from 27% to 70% (P<0.0001). Documentation was higher in patients who arrived by ambulance, who arrived soon after onset, and were treated at larger, primary stroke centers. Hospital-level NIHSS documentation rates and NIHSS scores were modestly inversely correlated (r=−0.207; P<0.0001), suggesting that NIHSS data from hospitals with low documentation were shifted toward higher values. In sensitivity analysis, the degree of bias in NIHSS reporting was reduced in more recent years (2011–2012) when NIHSS documentation was noticeably better.

Conclusions—Documentation of NIHSS is higher in patients who are thrombolysis candidates. Evidence of modest bias in NIHSS scores was observed but this has lessened as the documentation of NIHSS has improved in recent years.