(289) Validity of An Electronic Health Record (EHR) and Claims-Derived Proxy Measure of Multiple Sclerosis (MS) Disability

Publication Number: 963

Background: The Expanded Disability Severity Scale (EDSS) quantifies disability in MS and is a key outcome in clinical research. However the EDSS score is not available in claims databases or rarely recorded in the unstructured notes of the EHR, impairing the potential of real-world evidence (RWE) studies in MS. Proxy measures have been proposed based on ambulatory status derived from structured coding of diagnoses and medical aids.

Objectives: To evaluate the validity of ambulatory status derived from US EHR-claims linked records as a proxy measure of MS disability, the correlation and test characteristics were determined in comparison to EDSS scores extracted from the unstructured notes of the EHR.

Methods: A retrospective cross-sectional analysis of an MS patient cohort within the US Optum Market Clarity de-identified EHR and claims linkage database. Eligible patients from Feb 2019-Feb 2020 were required to receive care within an integrated delivery network; to have linked claims data; EHR activity during the observation period; and a combination of ≥3 inpatient or outpatient encounters for MS during the observation period. EDSS scores recorded by clinicians in unstructured notes during the observation period were extracted using natural language processing (NLP). Ambulatory status was derived as a three level category from ICD diagnoses and HCPCS medical aids; 1) fully ambulatory, 2) walks with assistance, and 3) non-ambulatory status. In sensitivity analyses, information from unstructured notes on patients’ ambulatory status extracted by NLP was included, and impact to algorithm performance was evaluated.

Results: Of eligible patients (n=47051), mean age was 53 yrs and 76% were female. MS subtype was extracted using NLP for 18039 patients (38%) with 79% having relapse-remitting MS. EDSS scores were only available for 0.9% (n=444). Median EDSS was 2.5 (IQR 1.0, 6.0). Based on the derived ambulatory status, 68% were fully ambulatory, 21% walked with assistance and 11% were non-ambulatory. Ambulatory status showed moderate correlation with EDSS (0.64, p< 0.001, Spearmans’ rank). When categorising EDSS < 4 as no walking impairment and EDSS≥4 as walking impairment, fully ambulatory status showed 84% sensitivity and 72% specificity in categorising walking status per EDSS. Including ambulatory status recorded in unstructured notes led to modest improvements in correlation with EDSS (0.68, p< 0.001), and higher specificity (85%) in classifying walking status per EDSS, but with lower sensitivity (74%).

Conclusions: Considering the absence of EDSS scores in claims and rarity in EHR databases, ambulatory status may be a relevant and useful proxy indicator of MS disability for the purposes of confounding adjustment for comparative analyses in EHR or claims based studies.