Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TELII: Temporal Event Level Inverted Indexing for Cohort Discovery on a Large Covid-19 EHR Dataset

Yan Huang

TL;DR

TELII is engineered to pre-compute and store the relations along with the time difference between events, thereby providing fast and accurate temporal query capabilities, and is implemented for the OPTUM de-identified COVID-19 EHR dataset, which contains data from 8.87 million patients.

Abstract

Cohort discovery is a crucial step in clinical research on Electronic Health Record (EHR) data. Temporal queries, which are common in cohort discovery, can be time-consuming and prone to errors when processed on large EHR datasets. In this work, we introduce TELII, a temporal event level inverted indexing method designed for cohort discovery on large EHR datasets. TELII is engineered to pre-compute and store the relations along with the time difference between events, thereby providing fast and accurate temporal query capabilities. We implemented TELII for the OPTUM de-identified COVID-19 EHR dataset, which contains data from 8.87 million patients. We demonstrate four common temporal query tasks and their implementation using TELII with a MongoDB backend. Our results show that the temporal query speed for TELII is up to 2000 times faster than that of existing non-temporal inverted indexes. TELII achieves millisecond-level response times, enabling users to quickly explore event relations and find preliminary evidence for their research questions. Not only is TELII practical and straightforward to implement, but it also offers easy adaptability to other EHR datasets. These advantages underscore TELII's potential to serve as the query engine for EHR-based applications, ensuring fast, accurate, and user-friendly query responses.

TELII: Temporal Event Level Inverted Indexing for Cohort Discovery on a Large Covid-19 EHR Dataset

TL;DR

TELII is engineered to pre-compute and store the relations along with the time difference between events, thereby providing fast and accurate temporal query capabilities, and is implemented for the OPTUM de-identified COVID-19 EHR dataset, which contains data from 8.87 million patients.

Abstract

Cohort discovery is a crucial step in clinical research on Electronic Health Record (EHR) data. Temporal queries, which are common in cohort discovery, can be time-consuming and prone to errors when processed on large EHR datasets. In this work, we introduce TELII, a temporal event level inverted indexing method designed for cohort discovery on large EHR datasets. TELII is engineered to pre-compute and store the relations along with the time difference between events, thereby providing fast and accurate temporal query capabilities. We implemented TELII for the OPTUM de-identified COVID-19 EHR dataset, which contains data from 8.87 million patients. We demonstrate four common temporal query tasks and their implementation using TELII with a MongoDB backend. Our results show that the temporal query speed for TELII is up to 2000 times faster than that of existing non-temporal inverted indexes. TELII achieves millisecond-level response times, enabling users to quickly explore event relations and find preliminary evidence for their research questions. Not only is TELII practical and straightforward to implement, but it also offers easy adaptability to other EHR datasets. These advantages underscore TELII's potential to serve as the query engine for EHR-based applications, ensuring fast, accurate, and user-friendly query responses.

Paper Structure

This paper contains 8 sections, 1 equation, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. Event Relation Extraction
  4. Inverted Index using Event Relations
  5. TELII Temporal Query with MongoDB
  6. Experimental Results
  7. Discussion
  8. Conclusion

Figures (5)

  • Figure 1: Computational time consumption for query "Find patients who had Diagnosis of U071 COVID-19 before Diagnosis of R52 Pain" by using record-based data schema or ELII.
  • Figure 2: The process of building TELII for an EHR dataset.
  • Figure 3: Result1. The query performance of ELII-DIAG, TELII-DIAG, and TELII on testing task 1 query 1-6. The response time is in seconds.
  • Figure 4: Result2. The query performance of ELII-DIAG, TELII-DIAG, and TELII on testing task 2 query 1-5. The response time is in seconds.
  • Figure 5: Result3. The query performance of ELII-DIAG, TELII-DIAG, and TELII on testing task 3 query 1-6. The response time is in seconds for ELII-DIAG, and in milliseconds for TELII-DIAG and TELII.