METHODS TO DERIVE A CROHN'S DISEASE ACTIVITY INDEX FROM ELECTRONIC HEALTH RECORDS DATA

Abstract

Listen

Translate article

Abstract BACKGROUND The Crohn’s Disease Activity Index (CDAI) is currently the gold-standard for measuring disease activity in Crohn’s disease (CD) patients. It is the most commonly used outcome measure in clinical trials, however, due to the burden of completion, it is infrequently used in routine care. Real world data, including electronic health records (EHR), are useful for studying longitudinal disease and treatment outcomes in the real world, but most EHR databases do not include CDAI. Building off the methods presented by Rudrapatna et al. 2021, we calculated a derived CDAI using data elements available in EHRs. METHODS Data were derived from the OM1 PremiOM UC Dataset (OM1, Boston, MA), a multisource real-world database with linked healthcare claims and EHR data on US patients with CD (2013-present). Weight, body temperature and hematocrit were sourced from structured vital signs and laboratory data. Ideal body weight was calculated from height and gender using the Devine formula. Extraintestinal manifestations/complications were evaluated based on the presence or absence of ICD diagnosis codes for each condition. Abdominal mass was abstracted from clinical notes. Number of stools in the past 7 days, average daily abdominal pain, and general well being were extrapolated from single-day measures extracted from clinical notes rather than based on 7-day diaries. Each component value was multiplied by the weights outlined in Best et al. 1976. Components recorded within 4 months were added to calculate a derived CDAI. Multiple CDAIs could be calculated for each patient over time, but each component was allowed to contribute to only one CDAI score. RESULTS Among 13,159 CD patients, only 12 had a CDAI recorded in their notes. After applying the algorithm, we were able to calculate a derived CDAI for 1,378 patients. 489 patients had > 1 score derived. Figure 1 shows the distribution of derived CDAIs. Patients with a derived CDAI were younger and a higher proportion were treated with biologics and corticosteroids than those who could not have a CDAI calculated (Table 1). CONCLUSIONS A derived CDAI can be calculated from data that may be routinely recorded in EHRs. Efforts to standardize the collection of CDAI components in EHRs would further increase the availability of derived CDAIs. Further work to validate these derived scores against a gold standard and to replicate the findings in other real-world datasets should be considered. Validation was not feasible in this dataset due to the lack of recorded CDAIs. Derived CDAIs could potentially be used to study longitudinal changes in disease status and treatment effectiveness in settings where observed CDAIs are unavailable.