Using K-Means Clustering to Identify Physician Clusters by Electronic Health Record Burden and Efficiency

Objectives: Electronic health records (EHRs) have transformed the way modern medicine is practiced, but they remain a major source of documentation burden among physicians. This study aims to use data from Signal, a tool provided by the Epic EHR, to analyze physician metadata in the Montefiore Health System via cluster analysis to assess EHR burden and efficiency. Methods: Data were obtained for a one-month period (July 2020) representing a return to normal operation post-telemedicine implementation. Six metrics from Signal were used to phenotype physicians: time on unscheduled days, pajama time, time outside of 7 AM to 7 PM, turnaround time, proficiency score, and visits closed the same day. k-Means clustering was employed to group physicians, and the clusters were assessed overall and by sex and specialty. Results: Our results demonstrate the partitioning of physicians into a higher-efficiency, lower-time outside of scheduled hours (TOSH) cluster and a lower-efficiency, higher-TOSH cluster even when stratified by sex and specialty. Intra-cluster comparisons showed general homogeneity of physician metrics with the exception of the higher-efficiency, lower-TOSH cluster when stratified by sex. Conclusions: Taken together, the clusters uniquely reflect the EHR efficiency-burden of the Montefiore Health System. Applying k-means clustering to readily available EHR data allows for a scalable, efficient, and adaptable approach of assessing physician EHR burden and efficiency, allowing health systems to examine documentation trends and target wellness interventions.

Effect of Telemedicine and the COVID-19 Pandemic on Medical Trainees' Usage of the Electronic Health Record in the Outpatient Setting

OBJECTIVES

This study aimed to (1) determine the impact of COVID-19 (coronavirus disease 2019) and the corresponding increase in use of telemedicine on volume, efficiency, and burden of electronic health record (EHR) usage by residents and fellows; and (2) to compare these metrics with those of attending physicians.

METHODS

We analyzed 11 metrics from Epic's Signal database of outpatient physician user logs for active residents/fellows at our institution across three 1-month time periods: August 2019 (prepandemic/pre-telehealth), May 2020 (mid-pandemic/post-telehealth implementation), and July 2020 (follow-up period) and compared these metrics between trainees and attending physicians. We also assessed how the metrics varied for medical trainees in primary care as compared with subspecialties.

RESULTS

Analysis of 141 residents/fellows and 495 attendings showed that after telehealth implementation, overall patient volume, Time in In Basket per day, Time outside of 7 a.m. to 7 p.m., and Time in notes decreased significantly compared with the pre-telehealth period. Female residents, fellows, and attendings had a lower same day note closure rate before and during the post-telehealth implementation period and spent greater time working outside of 7 a.m. to 7 p.m. compared with male residents, fellows, and attendings (p < 0.01) compared with the pre-telehealth period. Attending physicians had a greater patient volume, spent more time, and were more efficient in the EHR compared with trainees (p < 0.01) in both the post-telehealth and follow-up periods as compared with the pre-telehealth period.

CONCLUSION

The dramatic change in clinical operations during the pandemic serves as an inflection point to study changes in physician practice patterns in the EHR. We observed that (1) female physicians closed fewer notes the same day and spent more time in the EHR outside of normal working hours compared with male physicians, and (2) attending physicians had higher patient volumes and also higher efficiency in the EHR compared with resident physicians.

Physician Electronic Health Record Usage as Affected by the COVID-19 Pandemic

OBJECTIVES

To utilize metrics from physician action logs to analyze volume, physician efficiency and burden as impacted by telemedicine implementation during the COVID-19 (coronavirus disease 2019) pandemic, and physician characteristics such as gender, years since graduation, and specialty category.

METHODS

We selected 11 metrics from Epic Signal, a functionality of the Epic electronic health record (EHR). Metrics measuring time spent in the EHR outside working hours were used as a correlate for burden. We performed an analysis of these metrics among active physicians at our institution across three time periods-prepandemic and telehealth implementation (August 2019), postimplementation of telehealth (May 2020), and follow-up (July 2020)-and correlated them with physician characteristics.

RESULTS

Analysis of 495 physicians showed that after the start of the pandemic, physicians overall had fewer appointments per day, higher same day visit closure rates, and spent less time writing notes in the EHR outside 7 a.m. to 7 p.m. on patient scheduled days. Across all three time periods, male physicians had better EHR-defined "efficiency" measures and spent less time in the EHR outside working hours. Years since graduation only had modest associations with higher same day visit closure rates and appointments per day in May 2020. Specialty category was significantly associated with appointments per day and same day closure visit rates and also was a significant factor in the observed changes seen across the three time periods.

CONCLUSION

Utilizing EHR-generated reports may provide a scalable and nonintrusive way to monitor trends in physician usage and experience to help guide health systems in increasing productivity and reducing burnout.

Rapid Implementation of a Multidisciplinary COVID-19 Cytokine Storm Syndrome Task Force

Objective

Patients with coronavirus disease 2019 (COVID‐19) can progress to a state of unregulated inflammation called cytokine storm syndrome (CSS). We describe formation and operation of a COVID‐19 multidisciplinary consultation service that was allowed to individualize treatment for critically ill patients with COVID‐19 during the pandemic.

Methods

Institutional experts from different subspecialties formed a COVID‐19 CSS task force at Montefiore Medical Center, Bronx, NY. They agreed on a set of four clinical and six laboratory parameters that can help early identify COVID‐19 CSS. We describe the formation and implementation of the COVID‐19 task force. The case series description of the COVID‐19 CSS consultation cohort highlights consultation volume, baseline characteristics, clinical and laboratory parameters, and how biologic treatments were allocated to these patients.

Results

Between April 4,2020, and May 7,2020, the COVID‐19 CSS task force was formed, consisting of adult and pediatric rheumatologists and allergy and immunology physicians. The task force evaluated a total of 288 patients, of whom 197 (68%) were male, the median (interquartile range [IQR]) age was 62 (51‐70) years, 122 (42%) were Hispanic, and 88 (31%) were Black or African American. The common presenting symptoms in all referred patients were dyspnea (85%) and diarrhea (80%). Thirty‐one patients who received biologic therapy were younger, with a median (IQR) age of 53 (32‐63) years, as opposed to 62.5 (52‐70) years in the nonbiologic group (P = 0.008). A higher proportion receiving biologics was in the critical care setting (26 [84%] vs 151 [59%]; P = 0.006).

Conclusion

To the best of our knowledge, this is the first multidisciplinary collaborative effort to provide individualized patient recommendations for evaluation and treatment of patients with COVID‐19 who may have CSS. This working model helped to devise an approach that may have identified patients who were most likely to benefit from biologic therapy in the absence of evidence‐based guidelines.

Misoprostol and Prednisone Treatment of Lupus Nephritis

Fourteen patients were enrolled in a placebo-controlled double-blind randomized trial of 8 weeks of treatment for active lupus nephritis. Seven patients received prednisone at a dose of 1 mg kg(minus sign1) day(minus sign1) plus misoprostol at a dose of 200 &mgr;g P.O. Q.I.D.; 7 patients received prednisone plus placebo. The patients included 12 females, 2 males; 3 African-Americans, 3 Asians, 5 Hispanics, and 3 Caucasians. There were no serious side effects associated with prednisone or misoprostol treatment during the 8-week study period. Laboratory measures obtained at baseline, 4, 8, and 12 weeks included complete blood count (CBC), ESR, C reactive protein (CRP), serum creatinine, creatinine clearance, 24-h urine protein excretion, C3, C4, and anti-double stranded DNA (anti-dsDNA). Statistical analysis was conducted assessing change in measures over time in the entire study group by paired t-tests. The effect of treatment on change over time was examined by analysis of covariance. Log transformation of the variables was performed prior to statistical analysis. For the entire study group, the mean levels of ESR, CRP, 24-h protein excretion, C3, C4, and anti-dsDNA were improved at 4, 8, and 12 weeks. The mean ESR at baseline was 70 plus minus 8 compared to 42 plus minus 8 at 12 weeks (p < 0.01). The mean CRP at baseline was 0.6 plus minus 0.2 compared to 0.2 plus minus 0.1 at 12 weeks (p < 0.01). The 24-h protein excretion was 4367 plus minus 769 mg at baseline compared to 2512 plus minus 709 mg at 12 weeks (p = 0.02). The mean C3 at baseline was 40 plus minus 4 mg dl(minus sign1) compared to 60 plus minus 4 mg dl(minus sign1) at 12 weeks (p < 0.01). The mean C4 at baseline was 14 plus minus 1 mg dl(minus sign1) compared to 23 plus minus 2 mg dl(minus sign1) at 12 weeks (p < 0.01). The mean anti-dsDNA at baseline was 4268 plus minus 1780 compared to 316 plus minus 111 at 12 weeks (p < 0.001). The baseline serum creatinine (1.12 plus minus.15 mg dl(minus sign1)) and creatinine clearance (82 plus minus 15 ml min(minus sign1)) were not significantly changed at 12 weeks (1.10 plus minus 0.2 mg dl(minus sign1) and 80 plus minus 17 ml min(minus sign1), respectively). Comparing the misoprostol treatment group to the placebo group, there were no statistically significant differences for ESR, CRP, creatinine, creatinine clearance, 24-h protein excretion, C3, C4, or anti-dsDNA at any time points. However, comparing the misoprostol treatment group at 4 weeks to the placebo group, a more rapid decrease in anti-dsDNA (reduction of 3297 plus minus 2374) was observed in the misoprostol treatment group versus 304 plus minus 409 in the placebo group), as well as lower mean anti-dsDNA levels (464 plus minus 140) in the misoprostol treatment group versus 4118 plus minus 3834 in the placebo group). Also, the C3 and C4 levels at 12 weeks in the misoprostol treatment group (67 plus minus 5 and 27 plus minus 3 mg dl(minus sign1), respectively) were greater than levels in the placebo group (52 plus minus 4 and 19 plus minus 3 mg dl(minus sign1), respectively). The data demonstrate that oral prednisone is effective in reducing proteinuria and improving the biological markers of disease activity (i.e., ESR, CRP, C3, C4, and anti-dsDNA) following short-term treatment of active lupus nephritis. Additionally, the more rapid change in anti-dsDNA levels and superior normalization of complement levels in the treatment group, although not statistically significant, are consistent with a biologic effect of misoprostol on lymphocyte function and the production of a pathogenic autoantibodies.

The role of the neutrophil in rheumatoid arthritis

Neutrophils acquire lysosomal granules and assembly systems for producing soluble mediators of inflammation during the process of maturation in the bone marrow. Subsequently, they emigrate from the circulation when they become attracted to joint spaces of rheumatoid arthritis patients by chemoattractants such as the complement split product, C5a, and leukotriene B4. Exposure to immune complexes, rheumatoid factor, and cytokines in the synovial fluid results in neutrophil activation with release of granule contents, toxic oxygen metabolites, and proinflammatory products of the arachidonic acid cascade. This process is analogous to a local Arthus reaction in which activation of the complement system is a central event. Some of the inflammatory materials released by this reaction contribute to the cartilage destruction seen in rheumatoid arthritis. Because others are chemoattractants themselves, they perpetuate intraarticular inflammation and permit the predominance of acute inflammatory cells in lesions maintained by chronic inflammation.