Lead/Presenter: Laura Graham, Resource Center - HERC
All Authors: Graham LA (VA Health Economics Resource Center, VA Palo Alto Health Care System, Palo Alto, CA), Illarmo, SS (VA Health Economics Resource Center, VA Palo Alto Health Care System, Palo Alto, CA) Sox-Harris, AH (Center for Innovation to Implementation, VA Palo Alto Health Care System, Palo Alto, CA) Wagner, TH (VA Health Economics Resource Center, VA Palo Alto Health Care System, Palo Alto, CA)
The VA Corporate Data Warehouse (CDW) Vital Signs domain only contains physiologic data elements that were read and recorded by nursing staff while patients are in inpatient wards or clinic settings. Excluded from CDW are data collected directly from patient monitoring devices, such as those in intensive care units (ICUs) or operating rooms (ORs). The omission of device-collected data could bias VA research that depends on vital signs for patients with a high likelihood of visiting these locations, such as surgery or trauma patients. Our objective was to examine the differences between device-collected physiologic data and CDW Vital Signs domain data to understand the potential impact of omitting these data.
We identified 2,765 patients undergoing inpatient surgery at one of six VA hospitals between January 1, 2020, and December 31, 2020 (3,092 procedures). Patients were identified using Picis Clinical Information Suite, a platform used by at least 70 VA hospitals to store device-collected data from ORs, ICUs, or post-anesthesia care units (PACUs). Patients with at least one OR encounter and a length of stay greater than 24 hours were included. We examined five commonly reported postoperative vital signs - systolic blood pressure (SBP), heart rate (HR), respiratory rate (RR), oxygen saturation (O2), and temperature (Fahrenheit). Bivariate statistics were used to compare the frequency, volume, values, and information content (Shannon entropy) of Picis device-collected physiologic data to the more traditionally used CDW Vital Signs data.
96.1% of patients had at least one vital sign recorded in Picis that was not available in the CDW data. Including Picis data in our analyses added an average of 843 SBP (median = 144), 2,026 HR (median = 498), 1,843 O2 (median = 506), 125 RR (median = 38), and 58 (median = 3) temperature observations per patient. Picis vital signs were more likely to be recorded earlier in the postoperative stay â€“ 55.6% of Picis vital signs were recorded on postoperative day 1 or 2 as compared to 32.2% of CDW vital signs. And, the distribution of vital sign values differed by the data source. Average RR, SBP, and temperature were lower in the Picis data (14.8 vs. 17.8 breaths per minute, 123.5 vs. 130.3 mmHg, 97.8 vs. 98.1 Fahrenheit, respectively). Average HR was higher in the Picis data (81.9 vs. 79.1 beats per minute). Lastly, average entropy per patient, a measure of the information content of a data series, was greater for Picis data as compared to CDW data for all vital signs except for temperature.
Surgical inpatients frequently have more device-collected physiologic data that is not included in the CDW Vital Signs domain. These device-collected vital signs from ORs, ICUs, and PACUs occur earlier in the inpatient stay, are significantly different from those found in the CDW Vital Signs domain and provide more information about longitudinal trends.
These differences may lead to information bias in VA studies that use CDW Vital Sign domain data. Studies incorporating only VA CDW vital signs should consider this limitation when designing studies or interpreting study findings.