Summary: A new study shows routine hospital blood tests can help predict spinal cord injury severity and survival chances. Researchers applied machine learning to thousands of patient records and found that patterns in common blood markers — including electrolytes, immune cells and other routine measures — forecasted outcomes as early as one to three days after hospital admission.
Unlike neurological exams that require patient cooperation, this blood-based approach provides objective, timely information that could strengthen emergency decision-making, triage and resource allocation for spinal cord injury (SCI) patients across diverse clinical settings.
Key Facts
- Predictive value: Trajectories of routine blood markers forecast injury severity and in-hospital mortality.
- AI insights: Machine learning models identified hidden patterns and improved accuracy as additional test data accumulated over time.
- Practical use: Routine blood tests are widely available, inexpensive and more accessible than MRI or specialized molecular biomarkers.
Source: University of Waterloo
Routine blood samples taken and tracked during early hospitalization can serve as dynamic biomarkers to predict outcomes after spinal cord injury, according to research led by the University of Waterloo. Using large-scale clinical data and advanced analytics, the study evaluated whether serial blood measurements — routinely collected in most hospitals — could reliably signal the presence, severity and survival likelihood following traumatic SCI.
Traumatic spinal cord injury affects millions worldwide and presents with highly variable clinical courses. The World Health Organization estimates more than 20 million people were living with SCI in 2019, with roughly 930,000 new cases annually. Because initial neurological exams can be limited by patient consciousness and other factors, there is a critical need for objective tools that support early prognosis and care planning.
Dr. Abel Torres Espín, a professor in Waterloo’s School of Public Health Sciences, explains: “Routine blood tests could provide clinicians with affordable, objective information to help predict the risk of death, confirm the presence of a spinal cord injury and estimate how severe it might be.”
The research team analyzed hospital records from more than 2,600 patients in the United States. They applied machine learning to millions of individual data points drawn from routine lab measurements collected during the first three weeks after injury. By modeling how multiple biomarkers change over time, the researchers uncovered trajectories that correlate with clinical outcomes.
Key findings include robust early prediction of mortality and injury severity without relying on early neurological assessments. The models were able to distinguish motor complete versus incomplete injuries, and forecast in-hospital mortality, with meaningful accuracy as early as one to three days after admission. Predictive performance improved as more serial blood tests were recorded over subsequent days and weeks.
“A single biomarker at one time point can be informative, but the real predictive power comes from combining multiple markers and tracking their trajectories,” said Dr. Marzieh Mussavi Rizi, a postdoctoral scholar in Torres Espín’s lab. The dynamic, longitudinal approach captures evolving physiological responses — such as shifts in electrolytes, inflammation and immune cell counts — that static measures miss.
While imaging (MRI) and specialized molecular assays can provide objective data, these resources are not equally available in all hospitals. Routine blood testing is universally accessible, cost-effective and easily repeatable, making it a practical solution for emergency departments and intensive care units worldwide.
The study demonstrates that data-driven blood test trajectories can complement clinical assessment, especially in early stages when neurological exams are unreliable. Implementing such predictive tools may help clinicians prioritize interventions, plan monitoring strategies and allocate limited critical care resources more effectively.
“Predicting injury severity within the first days after trauma is clinically valuable but challenging using neurological assessment alone,” Torres Espín said. “Our results suggest routine blood data can offer early, actionable insights and that prediction accuracy improves as time and additional tests accumulate. This foundational work opens possibilities for better-informed treatment decisions and optimized critical-care resource use for people with serious physical injuries.”
About this spinal cord injury and neurology research news
Author: Ryon Jones
Source: University of Waterloo
Contact: Ryon Jones – University of Waterloo
Image: Image credited to Neuroscience News
Original Research: Open access. Title: “Modeling trajectories of routine blood tests as dynamic biomarkers for outcome in spinal cord injury” by Abel Torres Espín et al., published in npj Digital Medicine.
Abstract
Modeling trajectories of routine blood tests as dynamic biomarkers for outcome in spinal cord injury
Routinely collected blood tests can reflect underlying pathophysiological processes. This study demonstrates that the dynamics of common, routinely collected blood measurements hold predictive validity in acute spinal cord injury (SCI).
Using the MIMIC dataset (n = 2,615) for model development and the TRACK-SCI study (n = 137) for validation, researchers identified multiple distinct trajectories among routine blood markers. They developed machine learning models to make dynamic predictions of in-hospital mortality, to detect the presence of SCI among spine trauma patients, and to classify SCI severity (motor complete versus incomplete).
Performance metrics from the study include: an out-of-train ROC-AUC of 0.79 [0.77–0.81] for in-hospital mortality prediction on day one post-injury, improving to 0.89 [0.88–0.89] by day 21. For detecting SCI following spine trauma, the best ROC-AUC reached 0.71 [0.69–0.72] by day 21. By day seven, the ROC-AUC for predicting SCI severity was 0.81 [0.77–0.85]. The full dynamic models outperformed the SAPS II severity score after seven days of hospitalization.
These results suggest that routinely obtained blood tests, when analyzed longitudinally with machine learning, can serve as practical, widely available biomarkers to inform early prognosis and clinical decision-making in spinal cord injury.