One-Day Prevalence Study on Pressure Injuries in Intensive Care Units

Overview

The objective of this study is to provide an up-to-date, global picture of the extent and patterns of pressure injuries in ICUs. Point prevalence studies are only of value when performed on a vast scale. To sample a representative cohort, it is the intention to recruit about 1200 ICUs with all continents covered and as many countries as possible within each continent.

The objective is to provide an up-to-date, international "global" picture of the extent and patterns of pressure injuries in ICUs. Thereto , the plan is to perform a 1-day, prospective, multicenter point-prevalence study. The large scale of the project should allow thorough epidemiological analyses. More precisely the study will enable to identify: * major risk factors for pressure ulcer development; * preventive measures used in distinct ICU populations and countries; * shortages in the availability of evidence-based measures to prevent pressure sores; * malpractice pressure sore prevention in particular regions or countries; * occurrence rates of pressure ulcers with/without accurate adjustment for risk profile and preventive measures taken; * benchmarking between regions/countries; clinical outcomes associated with pressure ulcers (major organ derangements and mortality); * economic outcomes associated with pressure ulcers (length of ICU stay) and linking these outcomes with local practice regarding prevention measures applied/available. * country and regional differences in prevalence of pressure ulcers and outcome. Pressures ulcer stages will be graded following the classification system jointly developed by the National Pressure Ulcer Advisory Panel and European Pressure Ulcer Advisory Panel. Data to be recorded include patient demographics, data on severity of underlying disease and acute illness, organ failure, pressure ulcers, major risk factors for pressure ulcers, and measures taken to prevent pressure ulcers. Statistical Plan Power Calculation. For a risk factor with a prevalence in the study cohort of only 10% (for example patients with a BMI\<20) and an outcome difference of only 5% to be statistically significant(15% vs. 20% in decubitus occurrence rate), a sample size of 5255 patients is required (478 patients with the index risk factor and 4777 without) (alpha=0.05; Beta\>0.80). Data cleaning \&missing data. Exceptional values will be traced through distribution plotting. In case of uncertainties, the individual investigators will be contacted. Missing data will be handled by imputation (1, 2). Data quality checks will be performed, such as checks on completeness, consistency, correctness, and uniqueness. Descriptives. A single final analysis is planned at the end of the study; no interim analyses are planned. Socio-demographic study cohort characteristics will be described as proportions for categorical variables and for continuous variables as mean and standard deviation if normally distributed or median and inter-quartile range if not normally distributed (according to the Kolmogorov-Smirnov test for normality). The proportion of patients with decubitus (percentage, %, and their 95% confidence intervals) will be reported overall and according to geographic region (continent), country classification by income as defined by The World Bank (high-, upper-middle-, lower-middle-, and low-income countries), percentage of the gross domestic product spent on healthcare (obtained from the World Health Organization), and according to theEducation and Health Human Development Report of the United Nations Development Program. Subsequently, potentialdifferences in prevalence might offer the opportunity to evaluate variances in prevention measures on a large scale. Modelling. Covariates that will be evaluated on their relationship with the presence of decubitus encompass various organizational aspects of the ICU (e.g. nurse-to-patient ratio), decubitus prevention measures (e.g. type of matrasses used), and patient-specific characteristics (e.g. age, underlying conditions, severity of acute illness, body morphology, BMI,length of ICU stay etc.). Covariates will be considered for adjusted analysis when they have an association with pressure injuries at a statistical level \<0.25 in unadjusted (univariate) analysis or because of their clinical relevance. A stepwise approach willbe used to eliminate terms from the regression model; p\<0.15 or p\<0.10 will beset as the limit to keep covariates in the model. Relationships with binary outcome variables (e.g. decubitus, mortality) will be assessed by means of unadjusted statistical tests and multivariate logistic regression. The value of additional propensity score correction in the regression model will be assessed. Multinomial logistic regression will be performed to assess independent relationships with decubitus stages. Likewise, linear mixed-effect modelling will be used to assess unadjusted and adjusted relationships with continuous outcome variables (e.g. length of ICU stay and length of hospitalization). Results of logistic regression will be reported as adjusted odds ratios with 95% confidence intervals. To develop a decubitus prediction model for distinct ICU populations (e.g., trauma, surgical or medical patients) models will be build using machine learning techniques (e.g. Random Forest, Gradient Boosting Machine). In the process different techniques will be applied in order to receive the optimal accuracy. In order to gain insight in the correlation between predictors and variables, regression techniques will be applied, as state above. For validation of our models the study cohort will be split into a training, test and validation set. As such this gives a fair interpretation of the results. Alternatively ten-fold cross-validation can be applied to prevent overfitting. Benchmarking for individual centers based on decubitus will be performed by providing directly or indirectly standardized risks based on fixed center effects in a logistic regression model (3, 4). Besides the presence of the binary quality outcome (i.e. decubitus) and the center code, this model also includes patient-specific baseline co-variates to adjust for differential case-mix. The Firth correction will be applied to the logistic regression model to maintain convergence in the presence of very small centers (5). Statistical analysis will be performed using SPSS and R. The head investigator (SB) is responsible for all statistical analysis. Advanced statistical methods will be executed byMiekeDeschepper (Strategic Policy Cell at Ghent University Hospital).

Conditions

Eligibility

Locations (1)

Outcomes