Introduction
Prof. Nils Gunnar Wahlgren (Chair), Department of Clinical Neuroscience Karolinska Institute, Sweden. Dr. Myzoon Ali (VISTA Coordinator) University of Glasgow www.vistacollaboration.org vista.plus@glasgow.ac.uk Development and Validation of an Ordinal Prognostic Model for Stroke Proposed Investigators: Darin B. Zahuranec, MD Jeffrey Wing, PhD MPH
Background
The overall aims of the project will be to develop and pilot test a novel web-based communication tool to support surrogate decision makers for patients with moderate to severe stroke. While countless prognostic models for stroke have been developed, few are in actual use in practice and none have been developed with a focus on the needs of patients and their families. Existing models tend to focus on prediction of a single dichotomous outcome, which may not be flexible enough for patients and families. We propose to use the VISTA dataset to develop and validate a novel prognostic model for stroke that will use data early in the hospitalization to predict the full ordinal range of outcomes of stroke based on the modified Rankin Scale. This prognostic model will be incorporated into web-based tool developed under a separate grant mechanism that will translate the ordinal Rankin scale into a format that is accessible to a diverse range patients and surrogate decision makers.
Hypotheses
It will be feasible to develop a prognostic model for stroke predicting the full range of outcomes with the mRS
Methods
The study population will include ischemic stroke and intracerebral hemorrhage patients. To the extent possible, this should be based on registry data (e.g. VISTA-Plus) rather than clinical trial data to limit the possibility of bias from a healthy trial population. The primary outcome will be the modified Rankin Scale at 90 days. Depending on data availability, we may be interested in investigating 6 month and 1 year Rankin as secondary outcomes, or possibly Barthel index as well. Specific covariates needed will be discussed with the VISTA team and will depend on data availability. The mRS at 90 days will be modeled using cumulative ordered logistic regression via the proportional odds model with baseline age, sex, comorbidities, and initial stroke severity (National Institutes of Health Stroke Scale) as predictors. Separate models will be developed for ischemic stroke and ICH patients, with partial proportional odds models used if the proportionality assumption does not hold. The final model will construct five response functions to calculate predicted probabilities for each mRS category (e.g. this patient has a 5% probability of mRS 1, a 12% probability of mRS 2, etc.), with calibration assessed with the Hosmer and Lemeshow goodness-of-fit test. The model will be internally validated using boostrap methods and externally validated[1] with a separate stroke dataset that Dr. Zahuranec has used extensively in prior publications.[2-6] Model output will be presented following expert recommendations[7, 8] for prognosis by describing the likely best case (e.g. 90th percentile) and worst case scenario (10th percentile) to account for variability in treatment and allow surrogates to “hope for the best,” but “prepare for the worst.”[9] In order to meet our institutional standards, the dataset received will need to meet US government HIPAA requirements for a de-identified dataset (see http://www.hhs.gov/hipaa/for-professionals/privacy/special-topics/de-identification/#standard) for details. This will require that the dataset contain no dates or personally identifiable information. Any age over 89 (e.g. 90 and older) will need to be removed from the dataset and coded as a categorical variable (such as "age 90 or older").
References
1. Altman DG, Vergouwe Y, Royston P, Moons KG. Prognosis and prognostic research: validating a prognostic model. BMJ 2009;338:b605. 2. Zahuranec DB, Sanchez BN, Brown DL, et al. Computed tomography findings for intracerebral hemorrhage have little incremental impact on post-stroke mortality prediction model performance. Cerebrovasc Dis 2012;34:86-92. 3. Zahuranec DB, Morgenstern LB, Sanchez BN, Resnicow K, White DB, Hemphill JC, 3rd. Do-not-resuscitate orders and predictive models after intracerebral hemorrhage. Neurology 2010;75:626-633. 4. Lisabeth LD, Sanchez BN, Baek J, et al. Neurological, functional, and cognitive stroke outcomes in Mexican Americans. Stroke 2014;45:1096-1101. 5. Kerber KA, Brown DL, Skolarus LE, et al. Validation of the 12-item stroke-specific quality of life scale in a biethnic stroke population. J Stroke Cerebrovasc Dis 2013;22:1270-1272. 6. Fletcher JJ, Morgenstern LB, Lisabeth LD, et al. A population-based analysis of ethnic differences in admission to the intensive care unit after stroke. Neurocrit Care 2012;17:348-353. 7. Back A, Arnold RM, Tulsky JA. Mastering communication with seriously ill patients: balancing honesty with empathy and hope. Cambridge [England] : New York: Cambridge University Press, 2009. 8. Discussing Prognosis "ADAPT" [online]. Available at: http://www.vitaltalk.org/sites/default/files/quick-guides/ADAPTforVitaltalkV1.0_0.pdf. Accessed 5/31/2015. 9. Holloway RG, Arnold RM, Creutzfeldt CJ, et al. Palliative and end-of-life care in stroke: a statement for healthcare professionals from the American Heart Association/American Stroke Association. Stroke 2014;45:1887-1916.
Proposed Funding Arrangement Details:
Dr. Zahuranec has funding to support the £2,000 contribution for data analyses. Due to the timing of grant funding, it would be preferable if this contribution could be made prior to January 31, 2017 (rather than at the conclusion of the project as is typical). However, even if the contribution is made after January 31, 2017, Dr. Zahuranec has discretionary research funds that he can use to support the contribution. Analysis will be conducted by Dr. Wing, who has extensive experience in prognostic modeling.