Franklin Dexter1, Richard H Epstein2, Johannes Ledolter3, Jonathan P Wanderer4. 1. Division of Management Consulting, Department of Anesthesia, University of Iowa, 200 Hawkins Drive, 6-JCP, Iowa City, IA 52242, United States. Electronic address: Franklin-Dexter@UIowa.edu. 2. Department of Anesthesiology, Pain Management and Perioperative Medicine, University of Miami, United States. Electronic address: REpstein@med.miami.edu. 3. Department of Management Sciences, University of Iowa, United States. Electronic address: Johannes-Ledolter@UIowa.edu. 4. Department of Anesthesiology, Vanderbilt University Medical Center, United States; Department of Anesthesiology Biomedical Informatics, Vanderbilt University Medical Center, United States. Electronic address: Jonathan.P.Wanderer@Vanderbilt.edu.
Abstract
STUDY OBJECTIVE: Recent studies have made longitudinal assessments of case counts using State (e.g., United States) and Provincial (e.g., Canada) databases. Such databases rarely include either operating room (OR) or anesthesia times and, even when duration data are available, there are major statistical limitations to their use. We evaluated how to forecast short-term changes in OR caseload and workload (hours) and how to decide whether changes are outliers (e.g., significant, abrupt decline in anesthetics). DESIGN: Observational cohort study. SETTING: Large teaching hospital. MEASUREMENTS: 35 years of annual anesthesia caseload data. Annual data were used without regard to where or when in the year each case was performed, thereby matching public use files. Changes in caseload or hours among four-week periods were examined within individual year-long periods using 159 consecutive four-week periods from the same hospital. MAIN RESULTS: Series of 12 four-week periods of the hours of cases performed on workdays lacked trend or correlation among periods for 49 of 50 series and followed normal distributions for 50 of 50 series. These criteria also were satisfied for 50 of 50 series based on counts of cases. The Pearson r = 0.999 between hours of anesthetics and cases. CONCLUSIONS: For purposes of time series analysis of total workload at a hospital within 1-year, hours of cases and counts of cases are interchangeable. Simple control chart methods of detecting sudden changes in workload or caseload, based simply on the sample mean and standard deviation from the preceding year, are appropriate.
STUDY OBJECTIVE: Recent studies have made longitudinal assessments of case counts using State (e.g., United States) and Provincial (e.g., Canada) databases. Such databases rarely include either operating room (OR) or anesthesia times and, even when duration data are available, there are major statistical limitations to their use. We evaluated how to forecast short-term changes in OR caseload and workload (hours) and how to decide whether changes are outliers (e.g., significant, abrupt decline in anesthetics). DESIGN: Observational cohort study. SETTING: Large teaching hospital. MEASUREMENTS: 35 years of annual anesthesia caseload data. Annual data were used without regard to where or when in the year each case was performed, thereby matching public use files. Changes in caseload or hours among four-week periods were examined within individual year-long periods using 159 consecutive four-week periods from the same hospital. MAIN RESULTS: Series of 12 four-week periods of the hours of cases performed on workdays lacked trend or correlation among periods for 49 of 50 series and followed normal distributions for 50 of 50 series. These criteria also were satisfied for 50 of 50 series based on counts of cases. The Pearson r = 0.999 between hours of anesthetics and cases. CONCLUSIONS: For purposes of time series analysis of total workload at a hospital within 1-year, hours of cases and counts of cases are interchangeable. Simple control chart methods of detecting sudden changes in workload or caseload, based simply on the sample mean and standard deviation from the preceding year, are appropriate.