Abstract: FR-PO046
Symptom Trajectories in the Electronic Health Record During the Transition to Dialysis
Session Information
- AI, Digital Health, Data Science - II
November 03, 2023 | Location: Exhibit Hall, Pennsylvania Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Augmented Intelligence, Digital Health, and Data Science
- 300 Augmented Intelligence, Digital Health, and Data Science
Authors
- Wetmore, James B., Hennepin Healthcare System Inc, Minneapolis, Minnesota, United States
- Johansen, Kirsten L., Hennepin Healthcare System Inc, Minneapolis, Minnesota, United States
- Gilbertson, David T., Hennepin Healthcare Research Institute Chronic Disease Research Group, Minneapolis, Minnesota, United States
- Roetker, Nicholas S., Hennepin Healthcare Research Institute Chronic Disease Research Group, Minneapolis, Minnesota, United States
Background
How symptoms recorded in the electronic health record (EHR) change during the transition to dialysis has not been fully explored.
Methods
Using Optum's de-identified Integrated Claims-Clinical Dataset, we identified individuals with CKD stages 4-5 who transitioned to dialysis. We searched clinical notes for symptoms, identified by natural language processing, recorded across weekly intervals in the 6 months before and after dialysis initiation. We estimated changes in the odds of a symptom being recorded with an interrupted time series analysis using segmented logistic regression.
Results
The cohort comprised 728 individuals (mean age 67.7 ± 13.1 years, 44.0% women, 55.9% White and 29.9% Black). During the 6 months prior to dialysis initiation, 83.3% were recorded as having pain, 68.4% had fatigue/weakness, 66.3% had shortness of breath, 60.6% had nausea/vomiting, and 37.1% had difficulty concentrating. Before initiation, odds of an individual being recorded as having pain increased (slope: OR 1.02 per week, 95% CI 1.01 – 1.03); dialysis initiation was associated with a decrease (intercept change: OR 0.70, 95% CI 0.60 – 0.83). After initiation, odds of pain were unchanged (post-dialysis slope: OR 1.00 per week, 95% CI 0.99 – 1.01), although this represented a trajectory change relative to the pre-dialysis period (change in slope: OR 0.98 per week, 95% CI 0.96 – 0.99). For fatigue/weakness, odds increased before initiation (OR 1.03 per week, 95% CI 1.02, 1.04), but decreased upon initiation (OR 0.62, 95% CI 0.52 – 0.76) and thereafter over time (OR 0.98 per week, 95% 0.97 – 0.99), representing a reduction in slope relative to the pre-dialysis period (OR 0.95 per week, 95% CI 0.94 – 0.97). Patterns for shortness of breath, nausea/vomiting, and difficulty concentrating were similar to those of fatigue/weakness.
Conclusion
Natural language processing can be used to study symptom changes recorded in individuals with advanced CKD.
Funding
- NIDDK Support