Electronic health records for the diagnosis of rare diseases

Kidney Int. 2020 Apr;97(4):676-686. doi: 10.1016/j.kint.2019.11.037. Epub 2020 Jan 14.


With the emergence of electronic health records, the reuse of clinical data offers new perspectives for the diagnosis and management of patients with rare diseases. However, there are many obstacles to the repurposing of clinical data. The development of decision support systems depends on the ability to recruit patients, extract and integrate the patients' data, mine and stratify these data, and integrate the decision support algorithm into patient care. This last step requires an adaptability of the electronic health records to integrate learning health system tools. In this literature review, we examine the research that provides solutions to unlock these barriers and accelerate translational research: structured electronic health records and free-text search engines to find patients, data warehouses and natural language processing to extract phenotypes, machine learning algorithms to classify patients, and similarity metrics to diagnose patients. Medical informatics is experiencing an impellent request to develop decision support systems, and this requires ethical considerations for clinicians and patients to ensure appropriate use of health data.

Keywords: artificial intelligence; education; electronic health record; pediatric nephrology; rare diseases.

Publication types

  • Research Support, Non-U.S. Gov't
  • Review

MeSH terms

  • Algorithms
  • Electronic Health Records*
  • Humans
  • Machine Learning
  • Natural Language Processing
  • Rare Diseases* / diagnosis
  • Rare Diseases* / epidemiology