At the wake of the COVID-19 health crisis, we are facing a severe shortage of ICU beds. The conventional logistics of the healthcare industry are not efficient enough for risk analysis and for that reason, efficient allocation of ICU beds is not possible for the high-risk patients. It has been recorded that 34% of all lives lost could have been saved if there were appropriate mechanisms for allocation of ICU beds. We need to know how critically ill a patient is to assess the priority. 86% of mistakes made in the healthcare industry are due to mishandling of cases and inappropriate allocation of beds
is one of the main reasons for that.
Doctor's suggestion during visits, key medical parameters and various clinical notes can be analysed for building effective healthcare system that can prioritize high-risk patients for appropriate allocation of ICU beds. For this, Named-Entity Recognition (NER) technique in NLP can be used to extract data from medical records at the time of admission to predict important keywords for symptom analysis. This instantly classifies keywords in the medical notes and segregate the patients into 4 categories viz. Respiratorial, Close Monitoring, Cardiac, and Infection.
After segregation of patients, the vitals are monitored and a regular analysis on patient mortality at real-time is conducted with accuracy weightage analysis. We want to minimize false positive rate, at every point of prediction so we will show associated accuracy rate calculated over ensembling of different quality scores for n-th day of stay: Tags will be assigned which shall be shown in dashboard where the patients can be monitored. Everything will be in a scalable web interface.
- Using machine learning algorithm, we will train given dataset and predict whether person will in house die or survive based on data provided.
- Using appropriate regression model, we will predict the length of stay in the ICU.
Risk factor will be calculated live over a linear classifier with user-based features over time and acuity scores. This will further help in reducing False-Positive rate.
- Machine Learning
- Basics of NLP
- MERN Stack
1. Nadeau, D. and Sekine, S., 2007. A survey of named entity recognition and classification. Lingvisticae Investigationes, 30(1), pp.3-26.
2. McMillan, S., Chia, C.C., Van Esbroeck, A., Rubinfeld, I. and Syed, Z., 2012, September. ICU mortality prediction using time series motifs. In 2012 Computing in Cardiology (pp. 265-268). IEEE.