A machine learning algorithm can accurately detect traumatic intracranial hemorrhage using data collected before patients arrive at the hospital. Published in the study Clothing network open.
The researchers created a pre-hospitalized trial using data paramedics, including patient age, gender, systolic blood pressure, heart rate, body temperature, respiratory rate, consciousness, student abnormalities, post-traumatic stress disorder, vomiting, hemiplegia, and clinical deterioration. Whether the head trauma is caused by strong force or pressure and the patient has multiple injuries.
The study analyzed the electronic health records of 2,123 patients with head injuries who were taken to Tokyo Medical and Dental University Hospital from April 1, 2018 to March 31, 2021. The machine learning model detected traumatic intracranial hemorrhage with 74% sensitivity and 75% specificity using a pre-hospital data.
By comparison, a prediction model using the National Institute for Health and Care Excellence (NICE) guidelines, calculated after consulting physicians, had a sensitivity of 72% and a specificity of 73%, which was not statistically different from the pre-hospital model. .
“Although conventional screening equipment requires testing by a physician, our proposed models require only pre-transport patient information, which is readily available,” the study authors wrote.
“The results suggest that our proposed prediction models may be useful in creating a triage system that can be used to evaluate the best organization where a head injury patient should be transported. More validity is needed with potential and multicenter data sets.”
Why it matters
Researchers say that assessing head injuries in cases can improve outcomes for patients. The current system for head injuries requires paramedics to bring patients to the hospital if they decide it is necessary, where a doctor will evaluate whether the patient needs a CT scan. After the scan, the patient may need to be taken to another hospital.
Adding field trials allows ambulances to take patients to the best site for first aid care, reducing treatment time.
“Effective outcomes for head injury patients are worse when their transport is delayed, so third-time transportation time should be reduced by creating a reliable field trim tool,” the researchers wrote.
As the use of artificial intelligence expands in healthcare, Experts and studies point to the importance of observation for bias, which can exacerbate existing health inequalities.
AI developers are also needed Conduct thorough testing to make sure the model works in all environments. Researchers in this head trauma study noted that this is a limitation of their research, as it focuses on a single site in Japan.
“Since this was a single-center study and only included patients who were hospitalized and underwent head CT, our data sets could not represent the general population of patients with head injuries,” they wrote.
“Furthermore, we suggest that our model may underestimate high-risk patients based on calibration plots. To apply our model to clinical practice, we should verify predictive accuracy using a potential external validation set and investigate the optimal cutoff value.” “