Person:
Singh, Karandeep

Loading...
Profile Picture

Email Address

AA Acceptance Date

Birth Date

Research Projects

Organizational Units

Job Title

Last Name

Singh

First Name

Karandeep

Name

Singh, Karandeep

Search Results

Now showing 1 - 1 of 1
  • Publication
    A Concept-Wide Association Study of Clinical Notes in Search of New Predictors of Kidney Failure
    (2015-09-29) Singh, Karandeep; Bates, David W.; Curhan, Gary C.; Waikar, Sushrut S.
    Importance: Identifying modifiable risk factors for kidney disease progression is critical towards the development of strategies to prevent the onset of kidney failure. Existing epidemiologic approaches are unable to translate unstructured clinical information into knowledge about potential risks. Objective: The objective of this study is to use clinical notes to identify novel risk factors for kidney failure. Design: We conducted a retrospective study consisting of evaluating the end stage renal disease (ESRD) using individual concepts extracted from clinical notes in the year prior to the first outpatient general nephrology visit. We treated death as a competing risk in this analysis and used proportional subdistribution hazards regression. We determined statistical significance in univariate and multivariate models using a 5% threshold for false discovery rate (q-value < 0.05). The multivariate models included the Tangri score, an established predictive score for ESRD. Setting: A tertiary care center. Participants: Patients were included if they were seen by an adult outpatient nephrologist at a Brigham and Women’s Hospital-affiliated clinic between January 1, 2004 and June 18, 2014. Patients were excluded if they had visits only with renal transplant (and not general nephrology), known ESRD prior to the first nephrology visit, the absence of any clinical notes prior to the first nephrologist visit, no follow-up at BWH after the nephrology visit, or no creatinine values following the nephrology visit. After considering 9,817 patients seen in nephrology clinic, we identified 4,682 patients who met the inclusion/exclusion criteria. Intervention(s) for clinical trials or Exposure(s) for observational studies: Not applicable. Main Outcome(s) and Measure(s): Time-to-ESRD from the date of first outpatient general nephrology visit, with death treated as a competing risk. Results: We identified 107 concepts contained within clinical notes prior to the first nephrology visit with hazard ratios greater than 1 (q < 0.05). After adjusting for the Tangri linear predictor, 31 concepts were found to have hazard ratios greater than 1 (q < 0.05), 22 of which were deemed plausible. Conclusions and Relevance: This is the first study to demonstrate the feasibility of using concepts derived from clinical notes to identify new risk factors for disease. Given the variability in clinical notes and multiple hypothesis testing, validation at a separate center is needed. Trial Registration: Not applicable.