SHM Converge Scientific Abstract ePoster Gallery-Hospitalist Cohort Identification and Validation Within a Large Multi-Hospital Organization

Pdf Summary

This work describes a validated method for identifying hospitalist patients and assigning daily provider-level responsibility within a large multi-hospital health system. The goal is to create a reliable patient cohort and patient-provider attribution system for both retrospective analysis and real-time feedback in a Learning Healthcare System.<br /><br />The authors note that earlier approaches—such as using provider name lists, EMR specialty fields, extracted notes, or billing codes—were often inaccurate and lacked routine validation. To improve accuracy, they built their method around EMR team lists and cross-referenced the provider listed in the EMR with the author of the daily note to confirm the patient-provider pair.<br /><br />Validation was performed against a gold standard of manual clinician chart review for 1,782 patient-days. The resulting cohort identification achieved 99.73% sensitivity and 99.97% specificity for hospitalist patient identification. Daily provider/patient attribution was also highly accurate, with 98.47% sensitivity and 99.80% specificity. Compared with prior methods, the new approach identified 10% more patient-days and removed 2% false positives.<br /><br />The system accurately categorized patient-days, team lists, and patient-provider interactions with 99.5% specificity, and it correctly classified encounter types such as admissions, ICU transfers, discharges, and daily encounters. Accuracy was confirmed not only through manual review but also through a two-year longitudinal check of census counts against previously recorded manual team counts.<br /><br />The method was first validated at a quaternary academic medical center and is being expanded to community and rural hospitals across the health system. Prospective validation is also being implemented to maintain ongoing data quality. Overall, this validated infrastructure supports dashboards, morning census tracking, and measurement of variation in length of stay and daily discharge performance by provider.

Asset Subtitle

Harris L. Carmichael

Meta Tag

Author List Harris L. Carmichael, Linda M. Venner, Marion Gorder, Michael Pirozzi, Monique Mahlum, Rajendu Srivastava

Category Research

Concept Cohort Identification

Concept Hospitalist Quality Metric Attribution

Concept Hospital Medicine Patients

Concept Sensitivity

Concept Specificity

Distinguished Non-Finalist

Presenter Organization Intermountain Healthcare

Presenting Author Harris L. Carmichael

Track Technology in Hospital Medicine

Keywords

hospitalist patients

provider attribution

learning healthcare system

electronic medical record

patient cohort identification

manual chart review

sensitivity and specificity

daily note authorship

team lists

length of stay

Cohort Identification

Hospitalist Quality Metric Attribution

Hospital Medicine Patients

Sensitivity

Specificity