Construction of a Multidimensional Score for the Functional Prognosis of Cerebral Infarctions

Overview

In France, stroke is the leading cause of non-traumatic acquired motor disability in adults, the second leading cause of major cognitive impairment, and the third leading cause of death in men and women. The average age of onset is 73 years (70 years for men, 76 years for women). All types combined, approximately 100,000 patients are hospitalized for stroke each year, approximately 40,000 people die and 30,000 have serious after-effects at one year. The spectrum of functional sequelae ranges from motor and sensory impairment to cognitive impairment; moreover, 30 to 50% of patients have a recurrence within 5 years. Data from the Dijon Stroke Registry showed in 2011 that only 36% of people who had a stroke between 2000 and 2009 were symptom-free 1 month after the event; 22% of patients had mild or moderate disability according to the modified Rankin Scale (mRS); and 42% were unable to walk without assistance or had died. Based on 2009 self-reported data, more than one in two (51%) of those with a history of stroke with sequelae reported significant difficulty or inability to walk 500 meters; 45% had difficulty with at least one activity of daily living. The mortality rate was 44.7 per 100 000 persons in 2013. Cerebral infarctions (CIs) account for the majority of strokes (70-75%). In 2014, despite a 12.5% decrease from 2008, hospital case fatality for CIs remained high at 9%. In 2015, the case fatality rate was 10.7% at 30 days and 11.9% at 1 year. Thirty-day mortality alone concentrated nearly half (47%) of the 1-year mortality.

Data regarding the validity of poor prognostic criteria for CI are still scarce and remain of imperfect quality, particularly because of heterogeneous populations and variable definitions of "poor prognosis." Prognostic assessment must take into account both neurological and non-neurological elements. Biomarkers have been one of the avenues explored for several years. For example, hyperglycemia in the acute phase of CI is associated with a more severe prognosis in terms of survival and functional recovery, and with an increased risk of bleeding after thrombolysis, independently of age and severity of the CI. Other biological markers have been studied, but the methodology of the studies is very variable and their quality is moderate, which does not allow us to conclude that these markers are independent predictive factors for the prognosis of CI. Their predictive "added value" on simple clinical data, such as the National Institutes of Health Stroke Scale (NHISS) severity score and age, has not been established and their clinical significance remains uncertain. The mRS 3 months after the event is the criterion most frequently used to assess the functional prognosis of patients after a heart attack. However, being able to anticipate the mRS at discharge from hospital, as early as admission, appears to be a major challenge in order to adapt therapies and to choose the best care pathway as soon as possible after hospitalization. Structured management of stroke in neurovascular units (UNV), with highly specialized medical and paramedical staff, has a proven benefit in preventing death and dependence. The era of "precision" medicine, based on the identification, evaluation, organization and analysis of a multitude of variables obtained for each individual, now holds great potential for the management of CI. The deployment of a computerized patient record (CPR) in health care institutions is a real opportunity in this respect. The detailed medical data collected from the moment the patient enters the hospital and throughout his or her hospitalization could be used to construct a composite score combining information from the patient's history, clinical data, and the results of additional examinations (especially biological and imaging). Beyond structured data, information expressed in textual form could also be considered using innovative artificial intelligence methodologies. A neurovascular intensive care unit (NICU), to which a neurovascular unit (NVU) is attached, has existed within the Groupe Hospitalier Paris Saint-Joseph (GHPSJ) since 2006. The institution currently counts more than 700 stays per year for IC management, a figure that is constantly increasing (600 in 2016, 727 in 2019); 98% of these patients made a passage in the USINV in 2019. In addition, the CIO was deployed from 2014 in the facility, and during 2015 specifically for the Neurology-Neurovascular service. This gives the facility the opportunity to conduct a single-center pilot study to test the contribution of new data, combined with the simple clinical data usually used, to assess the functional prognosis of patients with CI. The objective of this first retrospective work is to evaluate the interest of the data available in the CIO and their processing via artificial intelligence methods to characterize the functional prognosis of patients with a CI, at admission and then during hospitalization.