This retrospective study utilized a de-identified dataset provided by the hospital information technology department. The University of California Irvine Institutional Review Board deemed the protocol IRB exempt.
Data Collection
We extracted data for this study from our perioperative database, SIS (Surgical Information Systems, Alpharetta, GA). The initial data pull included all adult procedures in 20 main operating rooms at UC Irvine Medical Center (UCIMC) from September 1, 2013 through September 1, 2015 that were marked as receiving GDFT in the medical record (our charts include a mandatory GDFT field that must be selected yes/no before the chart can be closed).
Inclusion criteria for the study were adult patients age 18 or over having open or laparoscopic abdominal procedures (colectomy, adrenalectomy, gastrectomy, hepatic resection, Whipple or pancreatic procedures, nephrectomy, cystectomy, abdominoperineal resection, or gynecologic oncology procedures). For each identified case, we pulled: case, date and time; procedure; patient demographics (including gender, height, weight, age, American Society of Anesthesiologists Patient Score); patient comorbidities (including hypertension, congestive heart failure, renal failure, and dialysis); NPO time; whether or not the patient received an epidural or arterial line; intraoperative data (including urine output, estimated blood loss, total crystalloid and colloid, blood administration, median and minimum heart rate, median and minimum mean arterial pressure).
We sorted the data and created two groups distinguished as either AM (0715 surgical start time) or PM (after 1200 surgical start time). Cases starting between 0730 and 1200 were excluded to ensure distinct separation between the two groups. Following this, we applied further exclusions to standardize the patient cohorts and reduce variability due to surgical factors: patients less than 18 or older than 100 years; patients with estimated blood loss (EBL) greater than 500 mL or who received blood products intraoperatively; and those with congestive heart failure (CHF), end-stage renal disease (ESRD), or who were receiving dialysis. We also excluded emergency cases, patients admitted for greater than 24 hours prior to surgery (since intravenous fluid could have been administered during the NPO period), and those who received hypertonic bowel preparations.
Institutional NPO Protocols
We instructed all patients to fast at midnight before surgery. If patients needed any PO medications prior to surgery, we instructed them to take these with a sip of water. Patients arrived in the preoperative holding unit roughly 2 hours prior to the start of surgery. Preoperative nurses obtained intravenous access and started crystalloids (normal saline or Ringer’s lactate) at keep vein open (KVO) flows.
Goal-Directed Fluid Therapy Protocol
The primary GDFT protocol in use at UCIMC during this time period was an adaptation of the stroke volume variation (SVV) protocol and is outlined in Figure 1. Additional protocol information is available in the appended GDFT Protocol and GDFT Overview. We monitored patients who did not have arterial lines either by transesophageal Doppler, non-invasive continuous blood pressure monitoring, or by plethysmograph variability index (Masimo corp, Irvine, CA). Our database unfortunately did not record specific monitoring device.
Statistical Analysis & Outcomes
Primarily we sought to determine whether PM patients required different fluid delivery when compared to AM patients when a GDFT protocol recommended care. Our statistical approach therefore assessed this question from many possible perspectives to determine whether any evidence exists to support different fluid delivery requirements in AM compared to during PM surgery. We calculated and analyzed fluid administration totals with both a 1:1 crystalloid to colloid ratio as well as with a 3:1 crystalloid to colloid ratio.
First, direct group-wise testing was performed on total administered volume (crystalloid + colloid) calculated as total ml, total ml/hr, and total ml/kg/hr between the AM and PM groups. These three different calculations allowed us to check raw total, raw total corrected for duration of case, and raw total corrected for duration of case and size of patient. Second, a linear regression model was run that included all of the baseline covariates that differed between groups as well as the plausible confounding factors for differing fluid needs (ASA class, use of epidural anesthesia, laparoscopic vs. open case). Our model used a 1:1 crystalloid to colloid ratio or a 3:1 crystalloid to colloid ratio, which allowed us to evaluate the marginal influence of group (AM or PM) in light of all of the other covariates. Finally, we combined all patients from both groups and scatterplots of NPO time to total fluid (as ml, ml/hr, and ml/kg/hr) assessed the effect of patient-reported NPO time on fluid administration.
We assumed group size imbalance may exist between AM and PM case starts due to the consistent morning starts in all ORs. We knew if we could pull at least 300 patients into the former and 100 patients into the latter, assuming the typical patient received 2100 ± 450 ml of fluid based on previous work 12, with a power of 0.8 and alpha of 0.05, we would sufficiently power our analysis to detect a difference of approximately 300ml of fluid required between the groups. With a minimum reported NPO time difference of 2.5 hours between groups, this represented the ability to detect a need of as little as 120 ml of ‘deficit’ fluids between groups per hour of NPO time.
We performed statistical analyses using SPSS (IBM, Armonk NY) or R (http://www.R-project.org). We report data as mean ± standard deviation, or as count (percentile) for categorical variables. Because we assessed any possible increase in need for fluid, we considered a p-value of <0.05 significant with no corrections for multiple comparisons made.