The Turku University Hospital adult medical-surgical ICU has 24 beds and serves a population of 700 000. All surgical and medical intensive care patients in the region except patients with major burns and patients undergoing solid organ transplantation are treated in this tertiary hospital. Bed occupancy was approximately 79% during the study period.
Patients. On admission, patients are classified based on their treatment needs by the treating physician, who determines the main admission diagnosis and other diagnoses and is responsible for the input of pertinent patient data into the ICU electronic clinical documentation and information system (Clinisoft, GE Healthcare, Buckinghamshire, UK). Nurses with special training in using the modified Jackson/Cubbin (mJ/C) scale for PU/I risk assessment and in wound identification and care enter relevant information into the database, including the type and frequency of skin cleansing, skin integrity, PU/I presence, and National Pressure Ulcer Advisory Panel/European Pressure Ulcer Advisory Panel (NPUAP/EPUAP) PU/I classification.21 All laboratory results and medication data also are entered. The data system calculates both the mJ/C and SOFA scores.
Patient data from all ICU admissions between January 2010 and December 2013 (inclusion criterion age >18 years old) were retrospectively collected. Patients with 1 or more PU/Is present on admission were excluded unless they acquired 1 or more PU/Is during their ICU stay. Patients without any information about their SOFA or mJ/C scores and patients whose ICU outcome could not be determined also were excluded.
Study variables. PU/I risk was assessed using the mJ/C risk scale.18,22,23 The first PU/I risk assessment was performed when the patient was admitted to the ICU; subsequent assessments were performed daily. If the mJ/C score was 29 points or less,18,23 PU/I risk was considered high or extremely high and patients were placed on an appropriate protective mattress if one had not already been provided. Otherwise, PU/I prevention care followed the NPUAP-EPUAP guidelines,21 and repositioning therapy was intensified as much as possible with consideration for the condition of the patient. SOFA scores24 were collected on admission and daily thereafter. Information on age, gender, LOS (noted in quartiles [Q]), and ICU transfer or discharge were extracted from the database.
mJ/C scale. The mJ/C scale consists of 12 main categories of specific patient variables, each graded from 1 (high risk) to 4 (low risk). These categories include age, weight/tissue viability, past medical history, general skin condition, mental condition, mobility, hemodynamics, respiration, oxygen requirements, nutrition, incontinence, and hygiene. Three (3) subcategories (transport to examinations or treatments, use of blood products, and hypothermia) can lead to the deduction of 1 point. Thus, the minimum score is 9 and the maximum score is 48, with a lower score indicating a higher risk for PU/I development.
SOFA. The SOFA scale includes 6 organ failure assessment categories: liver (serum bilirubin concentration), coagulation (platelet count), renal dysfunction (creatinine concentration), cardiovascular system (mean arterial pressure or need for vasopressors), presence of respiratory disorder (PaO2/FiO2) and nervous system status (Glasgow Coma Score [GCS]). The GCS assesses ocular, motor, and verbal responsiveness. Each category is scored from 0 (low risk) to 4 (high risk), so the SOFA score ranges from 0 to 24. Missing values in a single SOFA subgroup were given a value of zero (ie, normal), a situation mainly relevant to bilirubin concentration, which is not routinely collected on the day of admission. The SOFA score system has been validated; the higher the score, the more severe the patient’s condition and the higher the mortality.2
Outcome. The primary endpoint was the ICU outcome, which was defined as: 1) moved from ICU to a ward, recovering, 2) no response to ICU treatment/deceased, or 3) transferred elsewhere from the ICU and outcome could not be determined. The patients in group 3 were excluded from analysis. The overall hospital mortality rate of these ICU patients was collected from the database.
The primary variable of interest was PU/I development during the ICU stay and the overall ICU outcome, and the secondary variables were the SOFA and mJ/C scores, assessed separately or together.
Data collection. All data were retrospectively extracted by the database administrator of the ICU from the clinical documentation and information system. All patient, nurse, and physician identifiers were removed at this stage from the data sets. The data sets were transferred by the statistician to SAS 9.4 (SAS Institute Inc, Cary, North Carolina).
Data analysis. For data analysis, no distinction was made regarding those who developed 1 PU/I or several PU/Is during their ICU stay (ie, all were classified as having developed a PU/I during their ICU stay). LOS values were calculated from exact times. Deceased/transfers were recorded as such.
For initial assessment, the data were analyzed by first day mJ/C scores (<29 or >0) or SOFA scores (<6, 6–12, and ≥12), PU/I status (developed or not during ICU stay), and ICU outcome (alive or deceased). The χ2 test was used to compare categorical data in the different groups (see Table 1 and Table 2). Thereafter, the probability of the short-term outcome was determined using a logistic regression model25,26 that utilized the categorical values of both the mJ/C and the SOFA scores. For logistic regression, the response variable was binary and the relationship between response and explanatory variables was determined using the logit link function:
logit(π)=log(π/(1-π))= β0+β1x1+ ...+βkxk
where π is the probability of a positive outcome, x1,x2,…,xk are the values of the explanatory variables, and β0,β1,…,xk are the estimates of the predictability values of the different explanatory odds ratios (OR) used to summarize the results of the logistic regression modeling. The OR is closely connected to logistic regression, and it can be used to evaluate the effect of a single risk factor while other risk factors are fixed. The OR was used to examine the primary hypothesis H1, which denotes that PU/I(s) acquired during an ICU stay increase the risk of a negative, short-term outcome of ICU patients.
The primary hypothesis was evaluated using a logistic regression model with the following explanatory variables: admission SOFA score (groups <6, 6–11, ≥12), admission mJ/C score (≤29, ≥30), PU/I status (ICU acquired PU/Is, no PU/Is), and interactions between 1) SOFA score and PU status, between 2) SOFA score and mJ/C score, and between 3) PU/I status and mJ/C score.
logit(π)=log(π/(1-π))= β0+β1 *SOFAscore+
Logistic regression also was used to examine the correlation between mJ/C or SOFA scores and the negative short-term outcome of ICU patients.
In addition to the logistic regression model, descriptive contingency tables of different scenarios were used to further evaluate the relationships among different risk factors related to a negative outcome. The Wald χ2 test was used to assess the statistical significance of the contingency tables. A negative outcome was defined as no response to treatment or deceased.
Ethical approval. The study was approved by the Ethics Committee of Hospital District of Southwest Finland (T25/2011, 14.06.2011 §172).