(P-546) Early Assessment of the Impact of the COVID-19 Era on Charlson Score

Background: COVID-19 research has primarily focused on understanding, treating, and preventing the virus. Medical care utilization has also been affected, potentially impacting research that uses diagnosis claims to define cohorts, assess cohort characteristics, and adjust for comorbidities.

Objectives: To characterize the impact of the COVID-19 era on Charlson score calculations.

Methods: Using a 5% random sample from Optum’s de-identifed Clinformatics® Data Mart Database (2007-2020), we identified 445,319 patients with continuous medical and drug coverage between January 2018 to June 2020, allowing a 45-day gap. Charlson comorbidity scores (Quan 2011) were calculated for 6-month time periods from January to June in each year for each patient. Patients with a score >0 during the 2018 or 2019 periods were used to assess changes in 2020. Among this population, we compared period prevalence of each Charlson condition. Descriptive statistics (mean [SD], N-tiles) were calculated to assess person-level changes in score, change in number of encounters, and number of encounters during each period.

Results: Among 445,319 patients in the cohort, 130,761 (29.4%) had a score >0 in 2018 or 2019. Mean Charlson score in this group increased from 1.58 [1.52] in 2018 to 1.75 [1.60] in 2019 but decreased to 1.50 [1.74] in 2020. On average, patients experienced a decrease in Charlson score from 2019 to 2020 (mean change -0.25 [1.55]) vs an increase from 2018 to 2019 (0.17 [1.66]). Number of medical encounters during the period were more strongly correlated with 2020 Charlson scores (r= 0.400 [0.396, 0.0405]) than in prior years (2019: 0.347 [0.342, 0.352]; 2018: 0.338 [0.333, 0.343]). Though mean number of encounters per period were highly variable, most patients had fewer encounters in 2020 than in 2019 (Median change [IQR]= -1 [-6, 2]). Median change was 0 [-4, 5] in the prior year. While prevalence was relatively constant among severe conditions (metastatic solid tumor, AIDS, severe liver disease, dementia), prevalence declined in 2020 from 2019 among other Charlson conditions, ranging from -6.1% to -35.5%.

Conclusions: Early evidence suggests that patient-level Charlson scores decline during the COVID-19 era and may be attributable to decreases in US health care encounters, as higher scores are correlated with a higher number of encounters. Charlson scores calculated over a longer period (e.g., 1 year) may have a different effect. Future work should examine longer follow-up period as more data are available, and future research should include assessments of COVID-period effects of Charlson scores within disease cohorts of interest.