Article Text

Download PDFPDF

457 Algorithms to identify systemic lupus erythematosus (sle) from electronic health record (ehr) data
  1. R Ramsey-Goldman1,
  2. T Walanus2,
  3. K Jackson3,
  4. A Chung1,
  5. D Erickson4,
  6. K Mancera-Cuevas1 and
  7. A Kho5
  1. 1Division of Rheumatology, Medicine, Chicago, USA
  2. 2Division of General Medicine and Geriatrics, Medicine, Chicago, USA
  3. 3Center for Health Information, Medicine, Chicago, USA
  4. 4Preventive Medicine, Preventive Medicine, Chicago, USA
  5. 5General Medicine, Medicine, Chicago, USA


Background and aims Background: SLE is difficult to diagnose because of the diverse manifestations occurring over time and across care sites. Electronic health records (EHR) present a rich source of patient information which can be mined for diagnosis and identification to improve quality of care or to enrol patients in studies.

Aim Build a rule-based algorithm for each revised 1982/1997 ACR classification criteria for SLE using EHR data.

Methods We included patients from the Chicago Lupus Database (CLD) fulfilling 4 or more of the ACR classification criteria for SLE who also had records in the Northwestern Medicine Electronic Data Warehouse (NMEDW) EHR. ICD-9 codes and lab test results for each ACR SLE criterion were ascertained. We queried patient diagnoses, lab results and used a simple chart string for lab test results from physician notes.

Results Data from 515/783 patients in CLD and the NMEDW EHR were included. When using ICD 9 codes only 8.8% of patients from CLD/NMEDW were identified. With the addition of lab results to the query concordance increased to 54.7%, and a simple text string query to search physician notes for additional lab results increased identification to 57.5%.

Abstract 457 Table 1

Comparing the frequency of Revised ACR Classification Criteria for SLE Identified Two Databases, CLD (disease specific) and NMEDW (EHR).

Conclusion Using ICD codes plus laboratory data from NMEDW increased fulfilment of classification criteria but is still not optimal for patient identification. Additional strategies such as using natural language processing (NLP) or examining fulfilment of SLICC classification criteria for SLE which includes more lab results than ACR may yield an improved rule-based algorithm for the identification of SLE patients in EHR data.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.