Study Temporal Impact of Renal Transplantation on End-Stage Renal Disease Using Clinical Notes and Data-Driven Approach

Ying-Chih Lo, MD, PhD
Department of Medicine
Division of General Internal Medicine and Primary Care
Poster Overview

Rationale: Renal transplantation (RTx) is the optimal treatment option for patients with advanced renal disease. Current knowledge of transplant care derived mostly from structured data. Information within free-text clinical notes were less well studied.

Methods: Our retrospective study included patients received RTx within Mass General Brigham (MGB) between 2007 and 2019. Their clinical data and notes documented between 2000 and 2019 were collected. We analyzed longitudinal clinical documented notes before and after renal transplant operation using topic modeling, a nature language processing technique to automatically generate topics/themes embedded in the notes. Temporal trends of these topics were reviewed by clinicians to identify some interesting patterns related to the impact of RTx on patient conditions and care.

Results: There were totally 2,070 RTx patients included in the final analysis and the median follow-up duration was 125 months. After reviewing, we conceptualized several clinical topics and identified some interesting topic trends. Some topics (e.g. diet control) were more frequently mentioned in pre-RTx period, while others (e.g. rejection, skin cancer) more often occurred in post-RTx period.

Conclusions: This analytic method for unstructured text data is helpful for the healthcare providers to understand some less well-known issues and needs in clinical care.

Scientific Abstract

Rationale: Renal transplantation (RTx) is the optimal treatment for end-stage renal disease. Most of RTx care knowledge comes from structured data analyses with hypothesis-driven approaches. No previous study has used clinical notes to characterize temporal impact of RTx on patient care.

Methods: In a retrospective cohort study, we included patients who received RTx within the Mass General Brigham (MGB) health system between September 2007 and December 2019. We obtained clinical notes from MGB data repository, from which, we discovered stable topics using latent Dirichlet allocation-based topic modeling. Topics were manually reviewed by clinicians and temporal patterns of these topics were examined before and after the RTx procedure.

Results: This study included 2,070 patients with mean age of 52 years old. Median follow-up duration was 48 months before and 77 months after the transplantation. We generated 85 stable topics from 805,268 notes. Topics, such as kidney rejection, cytomegalovirus infection, cardiovascular disease, skin cancer, mental disorders, diabetes care, and diet control, presented unique temporal patterns.

Conclusions: We identified several meaningful topics with unique temporal patterns from clinal notes revealing the impact of RTx. Our approach is feasible to elucidate topics related to clinical care which might be only available in clinical notes.

Clinical Implications
Current knowledge of renal transplant care derived primarily from hypothesis-driven approaches using structured data. We demonstrate the feasibility of using nature language processing to discover topics and temporal patterns from clinical notes which may promote post-transplant patient care and quality.
Research Areas
Ying-Chih Lo, MD, PhD, Liqin Wang, PhD, Joshua R. Lakin, MD
Principal Investigator
Li Zhou, MD, PhD

Explore Other Posters

Comments are closed.