This study used a retrospective, non-experimental, descriptive design and obtained approval from the hospital Institutional Review Board (IRB no.: RA-2019-039). Data for the study were retrieved from electronic medical records of adult patients admitted to the surgical-trauma ICU of a large, urban, private, Magnet hospital and a Level 1 trauma hospital in Hawaii from January 1, 2017 to June 30, 2019.

Patients included in the study sample were 18 years or older, had an ICU length of stay of 48 hours or greater, did not have a PI (any stage) present at hospital admission or within 24 hours of transfer into the ICU where tissue perfusion and Braden scale assessments were conducted at admission and daily during the ICU stay. Pressure injuries include unstageable, Level 1–4, and deep tissue injuries [11]. Measuring these distinctions was not a part of this study. Patients were excluded if there was a PI related to a medical device, moisture associated skin damage, burns, and if death was imminent or the patient was deemed to require comfort measures only. A sample size of 176 patients in each group was needed to power the study to detect a significant difference in the transcutaneous tissue oxygen level between those who developed PI and those who did not.

Data were extracted from the electronic medical records for all patients that met inclusion criteria. Demographic and clinical characteristics included such variables as age, gender, length of stay, Acute Physiology and Chronic Health Evaluation (APACHE) II score, and vasopressor use. Deidentified data about the study patients were kept secure and confidential, and managed according to the requirements of the Health Insurance Portability and Accountability Act.

There were 2 tools used to address the study aim: (1) the noninvasive transcutaneous oxygen monitoring system (Figure 1); and (2) the Braden scale for predicting PI risk (Figure 2) [12]. The noninvasive transcutaneous oxygen monitoring system was used to measure the patient’s perfusion level daily. The system was calibrated prior to every patient’s perfusion assessment by the ICU resident or medical student. After the skin was properly prepared and calibration achieved, a baseline numerical perfusion value was obtained that represented the transcutaneous tissue oxygen level. The patient was then given 100% fraction inspired oxygen for 5 minutes via the mechanical ventilator or through a non-rebreather mask. At that point, the system provided another perfusion value. The numerical difference between the baseline value and the value after 5 minutes was called the delta. A low delta indicated inadequate tissue perfusion and a high delta indicated adequate tissue perfusion. The previously published baseline and delta values had been established for septic patients and were not relevant to this study [13,14]. Therefore, new values were established. The researchers created 441 combinations of tissue perfusion baseline ranges from 20 to 40 and delta ranges from 15 to 35. These combinations were used to define cut-off values. Based on these combinations, a low perfusion baseline and/or low delta was defined as < 30 and considered as tissue perfusion failure. High perfusion values were not pertinent to this study.

The 2^nd tool, the Braden scale, developed in 1985 by Barbara Braden and Nancy Bergstrom, is typically characterized as the gold standard for PI risk assessment as it had a 83% sensitivity and 64% specificity [3,15]. The scale evaluates a patient’s risk for developing PIs based on six sub-scales or risk factors: sensory perception, moisture, activity, mobility, nutrition, and friction and shear. The total Braden score ranges from 6 to 23, with scores less than 18 indicating at-risk patients [6,7].

Patients’ characteristics including demographics and clinical information were summarized using descriptive statistics. For the categorical variables, frequencies and percentages were used, and for the continuous variables, mean and standard deviation or median and interquartile range (IQR) were used. The bivariate association between the variables and presence of PI was tested using Fisher’s exact test for the categorical variables, and using Wilcoxon rank-sum test for the continuous variables. The categorical variables with multiple records per patient were converted into a continuous single observation per patient per category by calculating percentages for each of the categories. Due to the small percentage of patients in the extra-high-risk group, the total Braden scores were combined into 2 categories: low/medium risk and high/extra-high risk. A multivariable logistic model was fitted for the development of PI with adjustments for potential confounders identified from the bivariate association and clinical experiences. All analyses were performed using R Version 3.6.3 (R Core Team) and a p < 0.05 was considered significant.