Portfolio

Problem Definition

Correlation, measurement and experimental studies. Random assignment of mobile phone RF/Air interface geolocation parameters to subjects across the study. Assessment & Prediction of Clinical risk encountered with social distancing of synthetic ICU patients.

Null Hypothesis: There is zero relationship between risk of contracting Covid 19 and Social distancing. Quantify the relationship.

Inputs

Project Duration: 3 months

Inputs: Mimic iv ICU cross sectional dataset ( https://mimic.mit.edu/iv/ ), deidentified EHR data: 53,423 critical care admissions, 26 tables, 324 variables Charted Events, laboratory measurements (LOINC), over 2million rows of unstructured data (provider notes) coded with SNOMED CT, ICD-9,ICD-10 and LOINC codes. SAPS ii score value set, NEWS Score value set & MS-Project project plan.

Algorithm

Algorithm: R, python, Data Extraction, inclusion/exclusion criteria, NLP, tokenization, Word2Vec, lemmatization, simulation of mobile parameters (pathloss, rssi, geolocation coordinates), project management: Trello

Outputs

Outputs: Aggregate dataset with randomly assigned synthetic user mobile geolocation parameters, distribution of clinical risk, frequency table, relationship between clinical risk and social distancing, statistical significance, geolocation map.

Risk Exposure

Following a normal distribution of the population, the sample size indicated a larger percentage of patients were at low clinical risk based on the NEWS score computed from their vital signs.

Distribution of Clinical Risk associated with social distancing.

Clinical risk for patients situated less than 6 ft away from other users with similar clinical diagnoses and vital signs increased by about 10%.

Unstructured Word cloud

Output from the NLP module creates a distribution of the words used in the provider notes. The size of the words are correlated with the frequency of the words in the corpus. An indication of priority in the interaction between patient and provider, correlation between concepts, terminology, terms with corresponding clinical diagnoses.

Aggregate dataset with distributed cluster of patients with random geolocation coordinates

Project composite dataset contains unstructured variables engineered from unstructured input data, structured data from ICD 9, ICD 10, LOINC,SNOMED, CPT codes merged with assigned SAPS SAPS and NEWS score. User mobile air interface parameters and geolocation coordinates assigned randomly.

User geolocation map and aggregate dataset

‍

GITHUB

Covid -19 Contact Tracing Management System

Input:

Algorithm:

Output: