Advantages
- The prediction system without specialists of hyponatremia
- Reducing the frequency of monitoring s-Na levels
Background and Technology
Hyponatremia is a relative excess of total body water to sodium and is seen in a variety of medical conditions including congestive heart failure, liver disease, the syndrome of inappropriate antidiuretic hormone, and as a result of medications (eg, thiazide diuretics, psychotropic agents, and chemotherapeutic agents). It is reported that the risk of death during hospitalization is increased by more than 50% in patients admitted with hyponatremia compared with normonatremia. In particular, severe hyponatremia causes severe CNS symptoms such as impaired consciousness and convulsions. Neurologic symptoms secondary to cerebral oedema require urgent therapeutic care.
Severe hyponatremia is treated by administrating bolus of saline for correction of Na. However, rapid correction of s-Na is known to cause osmotic demyelination syndrome (ODS), which is a severe, irreversible neurologic disorder that is difficult to recover from. To minimize the risk of ODS, it is recommended to careful monitoring of the s-Na levels and optimal s-Na correction for each patient. On the other hand, frequent monitoring per 3h and night time blood test is required, These are burden for patients and medical workers.
Researchers start to develop machine learning algorithms for predicting s-Na level. At first, they chose seven predictors from the longitudinal data of hyponatremia patients undergoing intensive care in a tertiary care hospital. The selected seven parameters (total intravenous fluid infusion and oral intake, Na and K content of infusions, urine volume, and serum Na, K, and Cl levels) are easy to obtain clinically over time.
Hyponatremia is also considered a risk in sports such as swimming, marathon running, and triathlon, and cases of death have been reported. Optimal hydration is also considered an effective strategy for maintaining performance, and this technology is expected to have potential for use in the sports field.
Reference and Patent
- AMIA2022 Clinical Informatics Conference
- Patent pending (unpublished)
Principal Investigator
Dr. Shintaro Oyama（Center for Healthcare Information Technology (C-HiT), Nagoya University, Associate Professor）
Current stage and Partnering
- We performed ten-fold cross validation with shallow machine learning algorithms, and each model showed good validation accuracy. The SVR model showed good performance with RMSE of 0.0048081 and R2 of 0.94. They plan to validation using more large number of patient data from other hospitals.
- S-Na levels and infusion contents are managed by an electronic health record system. We think this model combines with electronic health record system and enable to alert the doctors in the event of any abnormalities. We hope to see further development in collaboration with companies such as electronic health record system and medical information system manufacturers.
- We think this model could be linked to an infusion control system, which is expected to improve safety and reduce the burden on medical workers through automatic control of dosage and other means. We hope for further development in collaboration with infusion pump manufacturers and others.
- We expect this model to be used in the field of sports, but they don’t obtain such data at this stage. If you are interested in joint development from the initial stage, we would like to talk with companies interested in sports science.
Project No: TT-04272