PREDICTIVE MODELS FOR INJURY RISK FACTORS IN AN INTENSIVE CARE UNITABSTRACT
Objective: to develop predictive models integrating clinical, environmental and demographic variables associated with the occurrence of Pressure Ulcer, Incontinence-Associated Dermatitis, and Medical Device-Related Pressure Ulcers.
Method: longitudinal study, carried out in a tertiary hospital in Fortaleza, Brazil. Data collected from January to October 2024 through patient record data and physical skin examination. Descriptive and bivariate statistical analyses, logistic regression, and predictive models were performed.
Results: the absence of skin protection increased the risk of Pressure Ulcers by 84 % (p=0.000, OR=1.841) and doubled the likelihood of Medical Device-Related Ulcers (p=0.008, OR=2.009). Elevated humidity levels showed a strong association with pressure ulcers (p=0.000, Cramér's V=0.299) and Incontinence-Associated Dermatitis (p=0.000, Cramér's V=0.348). Patients with very poor nutritional status had a higher risk of PU (p=0.000, Cramér's V=0.207). In the prediction, the variable "humidity" was the most relevant for dermatitis (R²=0.149), while spinal cord trauma increased the risk of medical device ulcers by more than six times (p=0.000, OR=6.727).
Conclusion: predictive models have been developed effectively by identifying risk factors, such as high humidity, inadequate nutrition, and lack of skin protection, associated with injuries in intensive care units. These findings reinforce the need for targeted preventive interventions to minimize risks in patients.
DESCRIPTORS:
Pressure ulcer dermatitis; Intensive care units; Nursing; Risk factors; Statistical models
RESUMO
Objetivo: desenvolver modelos preditivos que integrem variáveis clínicas, ambientais e demográficas associados à ocorrência de Lesão por Pressão, Dermatite Associada à Incontinência e Lesões por Pressão Relacionadas a Dispositivos Médicos.
Método: estudo longitudinal, realizado em hospital terciário em Fortaleza, Brasil. Dados coletados de janeiro a outubro de 2024 por meio de dados dos prontuários dos pacientes e exame físico da pele. Foram realizadas análises estatísticas descritivas, bivariadas, regressão logística e modelos preditivos.
Resultados: a ausência de proteção da pele aumentou em 84 % o risco de Lesão por Pressão (p=0,000, OR=1,841) e dobrou a probabilidade de Lesões Relacionadas a Dispositivos Médicos (p=0,008, OR=2,009). Níveis elevados de umidade mostraram forte associação com lesões por pressão (p=0,000, V de Cramer=0,299) e Dermatite Associada à Incontinência (p=0,000, V de Cramer=0,348). Pacientes com estado nutricional muito pobre apresentaram maior risco de LP (p=0,000, V de Cramer=0,207). Na predição, a variável "umidade" foi a mais relevante para dermatite (R²=0,149), enquanto o trauma raquimedular aumentou em mais de seis vezes o risco de lesões por dispositivo médico (p=0,000, OR=6,727).
Conclusão: os modelos preditivos desenvolveram-se eficazmente ao identificar fatores de risco, como umidade elevada, nutrição inadequada e ausência de proteção da pele, associados a lesões em unidades de terapia intensiva. Esses achados reforçam a necessidade de intervenções preventivas direcionadas para minimizar riscos em pacientes.
DESCRITORES:
Lesão por pressão; Dermatite; Unidades de terapia intensiva; Enfermagem; Fatores de risco; Modelos estatísticos
RESUMEN
Objetivo: desarrollar modelos predictivos que integren variables clínicas, ambientales y demográficas asociadas a la ocurrencia de Lesiones por Presión, Dermatitis Asociada a Incontinencia y Lesiones por Presión Relacionadas con Dispositivos Médicos.
Método: estudio longitudinal, realizado en un hospital terciario de Fortaleza, Brasil. Datos recopilados de enero a octubre de 2024 a través de datos de registros de pacientes y examen físico de la piel. Se realizaron análisis estadísticos descriptivos y bivariados, regresión logística y modelos predictivos.
Resultados: la ausencia de protección de la piel aumentó el riesgo de lesión por presión en un 84 % (p=0,000, OR=1,841) y duplicó la probabilidad de lesiones relacionadas con dispositivos médicos (p=0,008, OR=2,009). Los niveles elevados de humedad mostraron una fuerte asociación con las lesiones por presión (p = 0,000, V de Cramér = 0,299) y la dermatitis asociada a la incontinencia (p = 0,000, V de Cramér = 0,348). Los pacientes con un estado nutricional muy pobre presentaron mayor riesgo de LP (p=0,000, V de Cramér=0,207). En la predicción, la variable “humedad” fue la más relevante para la dermatitis (R²=0,149), mientras que el traumatismo medular aumentó el riesgo de lesiones por dispositivos médicos en más de seis veces (p=0,000, OR=6,727).
Conclusión: se han desarrollado modelos predictivos eficaces identificando factores de riesgo, como alta humedad, nutrición inadecuada y falta de protección de la piel, asociados con lesiones en unidades de cuidados intensivos. Estos hallazgos refuerzan la necesidad de intervenciones preventivas específicas para minimizar los riesgos en los pacientes.
DESCRIPTORES:
Úlcera por presión; Dermatitis; Unidades de cuidados intensivos; Enfermería; Factores de riesgo; Modelos estadísticos
INTRODUCTION
Pressure ulcers (PUs) are a relevant public health problem, especially in patients admitted to intensive care units (ICUs), with a significant impact on morbidity and mortality1. Although largely preventable, their prevention requires an integrated approach focused on skin protection and constant monitoring of medical devices, which frequently cause medical device-related pressure ulcers (MDRPUs). The presence of these lesions, associated with incontinence-associated dermatitis (IAD), demands preventive strategies, continuous monitoring, and specific interventions from health professionals and represents a significant challenge in intensive care units (ICUs), contributing to increased morbidity, mortality, and hospital costs2. These demands increase the consumption of ICU resources, while the need for specialized care grows, potentially compromising the efficiency and quality of care.
Critically ill patients with long hospital stays are at increased risk of developing pressure ulcers due to limited mobility, prolonged exposure to humidity, and constant use of medical devices3 (NPUAP, 2014). Data indicate that, under these conditions, prevalence may exceed 40 %. During the COVID-19 pandemic, prolonged use of devices and difficulties in repositioning patients in respiratory distress aggravated the occurrence of MDRPU in ICUs4. Furthermore, the intensive care environment is directly associated with a high prevalence of IAD, a condition that can progress to complex injuries5. These conditions increase the risk of serious infections such as sepsis, overburden healthcare teams, prolong hospital stays, and increase hospital costs6.
A predictive model using machine learning techniques was developed to identify key risk factors for PUs in mechanically ventilated ICU patients. The main factors identified included sepsis, advanced age, platelet count, length of ICU stay, PaO2/FiO2 ratio, hemoglobin concentration, type of admission, renal disease, albumin concentration, and ethnicity7.
Complications related to IAD and MDRPU are also little explored in the literature, with a lack of studies on factors such as humidity control and prolonged use of medical devices8. Furthermore, the lack of technical knowledge among health professionals reinforces the need for evidence-based preventive protocols6.
PUs, IAD, and MDRPU were selected for this study due to their high prevalence in ICU patients and their significant impacts on patient morbidity, mortality, and quality of life. These injuries share common risk factors, such as prolonged immobility, continuous exposure to moisture, inadequate nutrition, and lack of skin protection, requiring specific and ongoing preventive strategies from healthcare teams. The selection of these conditions is also justified by the scarcity of integrated predictive studies that simultaneously consider clinical, environmental, and demographic variables to identify risks early and guide more effective preventive interventions. Therefore, the focus of the present study is to develop predictive models that expand the risk identification capacity for these three categories of injuries in critical care settings.
In view of the above, the question is: “which clinical and environmental factors are associated with the occurrence of PUs, IAD, and MDRPU in patients admitted to the ICU?”. The objective of the study is to develop predictive models that integrate clinical, environmental, and demographic variables associated with the occurrence of PUs, IAD, and MDRPU. The justification for this study lies in the need to develop predictive models capable of early identification of risk factors associated with PU, IAD, and MDRPU, guiding clinical decision-making for the implementation of more effective and targeted preventive interventions. In this sense, this study proposes the construction of statistical models that consider, in an integrated manner, multiple risk factors, aiming to contribute to the improvement of care provided in critical environments.
METHOD
This study is longitudinal, carried out in a tertiary hospital in Fortaleza, Ceará, from January to October 2024. Data were collected at two different times: initially, through information recorded in the patients' medical records during admission to the ICU; and, subsequently, through direct skin inspection performed by the researcher on a single day during the patients' bath. This longitudinal design allowed the assessment of skin conditions at two points in time, allowing the identification of changes related to the risk factors analyzed.
The population of this study consisted of all patients admitted to the six ICUs of a tertiary hospital in Fortaleza - Ceará, from January to October 2024. Each ICU has 10 beds. All patients admitted during the collection period who were admitted without skin lesions were included in the study. Exclusion criteria included patients in critical or severe condition, who could not be lateralized for inspection of the back, as well as those under sedation. Data collection was performed using a form with personal and clinical information about each patient. The capture of recorded information was carried out through data in the patients' medical records and complemented by inspection of the skin during bathing, ensuring direct analysis of the skin conditions of hospitalized patients. The total population was 803 patients hospitalized during the collection period. After applying the exclusion criteria, the final sample consisted of 601 patients.
The ICUs included in the study followed a standardized protocol for preventing PU, IAD, and MDRPU. This protocol included evidence-based practices such as changing position every 2 hours, using support surfaces for pressure relief, inspecting skin daily for early identification of injuries and risk assessment, keeping skin clean and dry, using protective moisture barriers in high-risk areas, and moisturizing the skin. Medical devices are inspected regularly to prevent friction and prolonged pressure. Checklists are used to ensure compliance with established measures.
The variables studied included clinical, environmental, and demographic factors associated with the development of PU, IAD, and MDRPU in ICU patients. The exposure variables considered were: age, sex, reason for hospitalization (such as traumatic brain injury, burns, spinal cord injury, among others), nutritional status, humidity level, mobility, use of adequate support surfaces, skin protection, patient repositioning, prolonged use of medical devices, and humidity control. The outcomes analyzed were the occurrence of PU, IAD and MDRPU.
The Braden Scale was used to assess the risk of PU, considering criteria such as sensory perception, degree of humidity, activity, mobility, nutritional status and friction/shear. The analyses involved predictive models to identify the variables with the greatest impact on outcomes, revealing that factors such as skin protection, moisture control, and inadequate nutrition were significantly associated with an increased risk of injury.
The statistical analyses of this study were performed using the statistical package Statistical Package for the Social Sciences (SPSS), version 23.0, to explore the associations between exposure variables (age, sex, nutritional status, humidity level, use of medical devices, among others) and outcomes (occurrence of PU, IAD, and MDRPU). Initially, descriptive statistics were applied to characterize the sample, using means, medians, standard deviations and relative frequencies. To identify significant associations between categorical variables and outcomes, the chi-square test of independence, Fisher's exact test, and Cramér's V for effect size were used. Bivariate analysis was conducted to explore the associations between exposure variables and outcomes, considering a significant p-value <0.05. For inclusion in subsequent analyses (regression), p<0.20 was considered.
Subsequently, multivariate logistic regression models were applied to identify the independent risk factors associated with the outcomes, calculating the odds ratios (OR) with 95 % confidence intervals (95 % CI). The Omnibus test was used to verify the overall significance of the models, while the Hosmer-Lemeshow test evaluated the models adjustment to the observed data. Furthermore, multicollinearity tests were performed to ensure the independence of the variables included in the models.
The predictive capacity of the models was analyzed using Nagelkerke's R-squared, indicating the proportion of variability explained by the factors included. These statistical methods allowed the identification of variables such as humidity control, very poor nutritional status, and skin protection as significant predictors for the development of PU, IAD, and MDRPU, reinforcing the importance of specific preventive measures to minimize these risks in critically ill patients admitted to ICUs. Machine learning algorithms were applied in R statistical software to explore the predictive capacity of these variables and identify more complex patterns that can guide preventive and personalized interventions. The results were presented through tables and graphs.
All ethical precepts established for conducting research with human beings were strictly followed throughout this study. The project was submitted to and approved by the institution's Research Ethics Committee, ensuring that all ethical standards and guidelines were met. Moreover, all participants received clear information about the objectives of the study and signed the informed consent form, ensuring voluntary participation and respect for the privacy and confidentiality of the data collected.
RESULTS
The prevalence analysis revealed that, in a sample of 601 patients, 59.23 % had PU, while 20.47 % were diagnosed with IAD, and 14.14 % with MDRPU. Patients’ mean age was 44.77 years (standard deviation = 17.02), with values ranging from 1 to 85 years and a median of 43 years.
The distribution by sex indicated that 77.04 % of patients were women. Regarding the Braden scale, the mean scores at the time of ICU admission and in the last 24 hours were, respectively, 10.44 (standard deviation = 1.70) and 9.55 (standard deviation = 2.38), suggesting a high risk for the development of PU. The most frequent reasons for hospitalization included Traumatic Brain Injury (TBI) with 241 cases, followed by burns (39 cases), and suicide attempts (16 cases), highlighting the diversity of clinical causes that impact the management of ICU care.
Table 1 presents the association between PU, IAD, and MDRPU and the variables sex, reasons for hospitalization, use of adequate support surface, skin protection, repositioning, nutritional control, and humidity.
The absence of skin protection showed a strong association with the occurrence of PU (p=0.000, OR=1.841 [95 %CI 1.316 - 2.576]), indicating that patients without skin protection measures have an 84 % higher risk of developing PU compared to those who received these measures. This variable was also significantly associated with MDRPU (p=0.008, OR=2.009 [95 %CI 1.190 - 3.391]), demonstrating that the absence of protection doubles the probability of this type of injury occurring.
Another crucial factor was the humidity level, which showed high statistical significance in relation to the three outcomes. For PU, the p-value=0.000 (Cramér's V of 0.299), indicating a moderate to strong association between high humidity levels and the development of PU. For IAD, the effect was even more significant, with p=0.000 (Cramér's V of 0.348), confirming that patients in conditions of high humidity or completely wet have a much higher probability of developing IAD. Humidity was also significantly associated with MDRPU (p=0.000, Cramér's V=0.170).
Nutrition also stood out as a determining variable for the outcome of PU. Patients with very poor nutritional status showed a significant association (p=0.000, Cramér's V=0.207), indicating that inadequate nutrition dramatically increases the risk of developing PU.
Among the reasons for hospitalization, spinal cord injury (SCI) was the variable with the greatest impact on the occurrence of MDRPU (p=0.000, OR=6.727 [95 %CI 1.394 =6.242]). This datum reveals that patients with SCI have more than six times the risk of developing medical devices-related injuries compared to patients hospitalized for other reasons. Finally, patient repositioning was shown to be a significant factor for MDRPU (p=0.000, Cramér's V=0.187).
When variables with p<0.20 are entered into a regression model for the presence of PU, there is general statistical significance by the Omnibus test (p<0.001), indicating that the predictor variables significantly improve the prediction of PU compared to the null model. The Hosmer and Lemeshow test (p=0.556) indicates that the model is well-adjusted with the observed data (Table 2)
The inclusion of variables in the model increased the overall predictive capacity to 63.6 %, with emphasis on the significance of the variables "Undergoes Skin Protection" (p=0.008, Exp(B)=0.618), which reduces the risk of PU, and "Very Poor Nutritional Control" (p=0.002, Exp(B)=3.497), which increases the risk of developing PU. The reason for hospitalization due to Spinal Cord Trauma was also significant (p=0.040, Exp(B)=1.446).
In the model for the presence of IAD, the Omnibus test confirmed the general significance of the regression model (p=0.000) for the presence of IAD, and the Hosmer and Lemeshow statistic (p=0.000) indicated an adequate adjustment. Nagelkerke's R-squared was 0.113, indicating that the variables explain 11.3 % of the variability. Regression results are presented in Table 3.
The results of the logistic regression indicated that Humidity Control was significantly associated with the presence of IAD (B=1.084, SE=0.175, Wald=38.576, p=0.000, Exp(B)=2.957), indicating that poor humidity management increases the chances of developing IAD by almost three times. The reason for hospitalization related to burns was also significant (B=1.385, EP=0.513, Wald=7.297, p=0.007, Exp(B)=3.995), with an almost fourfold increase in the risk of IAD in patients hospitalized for this reason.
The overall classification of the regression model for the outcome presence of MDRPU (Table 4) was 87.5 %, with an excellent capacity to correctly identify negative cases (99 % accuracy), although with limitations in predicting positive cases (17.6 %). The Hosmer and Lemeshow test (p=0.556) indicated good adjustment of the model to the data.
The analysis revealed that some variables are significantly associated with the presence of Medical Device-Related Pressure Ulcers (MDRPUs). Skin protection demonstrated a significant effect (p=0.043, Exp(B)=0.557, 95 %CI=[0.317, 0.981]), indicating that its absence increases the risk of MDRPU. Repositioning was also identified as a significant protective factor (p=0.047, Exp(B)=0.345, 95 %CI=[0.121, 0.984]), reducing the risk of injury by approximately 65 %. Furthermore, hospitalization due to spinal cord trauma showed a relevant association (p=0.000, Exp(B)=1.207, 95 %CI=[1.089-1.477]), showing that patients with this condition have a greater chance of developing MDRPU.
The regression analyses performed highlighted significant associations between clinical and environmental variables and the outcomes of PU, IAD, and MDRPU. Based on these results, machine learning algorithms were applied to explore the predictive capacity of these variables and identify more complex patterns that can guide preventive and personalized interventions. In the Figure 1, we present the results of the predictive analyses, highlighting the most relevant variables and the performance of the models.
Predictive capability for protection against pressure injuries, incontinence-associated dermatitis and medical device injuries. Fortaleza, CE, Brazil, 2024.
Predictive analysis for PU (Image 1) revealed that the model explained 5.8 % of the variance (R²=0.058), with a mean square error (MSE) of 0.244 and a RMSE of 0.494. The most important variables for prediction were Age (impact of 9.391 on node purity), followed by Braden scale value at ICU admission (impact of 4.213) and Humidity (impact of 3.797).
For IAD, the model presented a moderate performance, with R²=0.149, MSE=0.149 and RMSE=0.386. “Humidity” was identified as the most relevant predictor (significant impact on node purity), followed by “age” and Braden scale scores performed in the last 24 hours and scores assessed at ICU admission. These findings highlight the central role of humidity in the development of IAD.
In the prediction of MDRPU, the model achieved R²=0.123, MSE=0.080, and RMSE=0.283. The most important variables were “Age”, with the greatest impact on node purity (5.473), followed by “Humidity” (impact of 1.630), and “Repositioning” (impact of 1.333). “Spinal Cord Injury” was strongly associated with the risk of MDRPU (OR=6.727), reflecting the vulnerability of this group.
DISCUSSION
PU risk management is often accomplished with scales that track vulnerable patients and assist nurses in identifying and making decisions9. Therefore, they should be used for skin care management.
The literature consistently indicates that advanced age is a significant predictor for the development of pressure injuries, especially in intensive care settings (ICU), where older patients have greater skin fragility, reduced tissue regeneration capacity, and increased comorbidity. A recent retrospective study of 1,158 ICU patients found a significant association between the occurrence of pressure ulcers and age, demonstrating that for each year's increase in age, the risk of developing these lesions increased by 1 % (p<0.001)10.
However, when analyzing the present study, it is observed that the average age of the sample was considerably lower than that found in other studies on pressure ulcers in ICUs. This datum may be directly related to the profile of patients treated at the reference hospital, which focuses on trauma care, involving younger individuals. This context may have influenced the average age downwards, differing from samples from studies in general ICUs that include a higher proportion of older patients. This age discrepancy may have contributed to an underestimation of the incidence of pressure injuries in predictive models, since younger patients are less susceptible to these conditions.
Therefore, the inclusion of older patients in future samples may increase the predictive capacity of the developed models, providing a more robust estimate of the associated risk factors. Therefore, it is recognized that the reduced age range of the participants constitutes a relevant aspect to be considered in the interpretation of the findings.
PU and IAD share several risk factors, such as reduced mobility, compromised tissue layers due to poor skin conditions, insufficient oxygenation, and long hospital stays. These factors contribute to increased risk scores in predictive tools such as the Braden Scale11.
Impaired mobility is an important factor in increasing the risk of IAD, especially in older people. Studies show that the risk of IAD is 2.4 times higher in patients with limited mobility, according to the Katz index. Furthermore, a European study found that greater dependence is related to the emergence of IAD. Patients with frequent incontinence, prolonged use of diapers, poor skin condition, reduced mobility, cognitive impairment, inability to maintain personal hygiene, and factors such as pain, fever, medication use, malnutrition, and critical illness are more predisposed to the condition12.
One of the triggering factors for incontinence-associated dermatitis is prolonged contact of the skin with moisture from effluents, urine, and perspiration caused by the increase in the local microclimate. The interaction of irritating agents from eliminations, such as lipases, proteases and urease, enhances the deterioration of the stratum corneum, making it susceptible to damage in perineal areas, increasing permeability to pathogens13-14.
Excessive moisture in the stratum corneum is a factor that worsens the skin's susceptibility to injury, leading to cellular hyperhydration. This process compromises the structural cohesion of the tissue, making the skin more fragile and vulnerable to damage caused by friction and shear15. Excess moisture, from incontinence, open wounds and perspiration, is aggravated by the local microclimate. This is common in bedridden or incontinent patients, especially with improper diaper use. To prevent injuries, it is essential to control moisture, changing diapers regularly, using protective products such as sprays and barrier creams, and diapers that allow ventilation3,16.
People who are bedridden, incontinent, or wearing inadequate diapers are especially exposed. Excessive moisture worsens the skin's susceptibility to injury, resulting in cellular hyperhydration that weakens the tissue, making it vulnerable to friction and shear15.
Patients who remained exposed to high levels of humidity had a substantially higher risk of skin lesions. This finding reinforces the need for standardized clinical practices to control the skin microclimate in ICUs, including frequent diaper changes, use of protective barriers, and keeping critical areas dry12. Implementing these preventive measures can reduce the burden of dermatological complications in critically ill patients.
The effectiveness of preventive measures is presented, such as the application of protocols based on the Braden Scale, which results in a reduction in injuries and length of hospital stay12. Among the factors evaluated, nutrition stands out. Nutrition is crucial in preventing PU, as inadequate nutrition impairs the maintenance of skin integrity and wound healing. Patients with nutritional deficiencies have a higher risk of developing pressure injuries due to reduced collagen synthesis, skin fragility, and decreased immune response. Nutritional interventions, such as protein supplementation and monitoring by nutritionists, help reduce this risk17.
Spinal cord injury patients face elevated risks of developing MDRPUs. Prolonged immobility and loss of sensation hinder protective responses, such as the perception of discomfort and position change, increasing the risk of injury by more than six times. To mitigate this risk, the adoption of specific interventions is recommended, such as continuous skin monitoring, frequent adjustment of medical devices, and the use of support surfaces that reduce pressure18.
The use of medical devices increases the risk of MDRPUs19. Patients with paralysis or spinal cord injuries are more vulnerable because they do not perceive or react to discomfort, hindering positioning change. The healthcare team should perform regular inspections, repositioning, adjustments to the devices and, when possible, their temporary removal20. Regular repositioning of patients is an effective and low-cost strategy for preventing PU, in addition to allowing continuous assessment of skin integrity. This practice is crucial in ICUs, where patients' skin is constantly exposed to risks from devices such as catheters, cervical collars, and pulse oximeters19.
Among the main measures for preventing medical devices-related ulcers, besides repositioning, several specific strategies stand out. These include a checklist of the devices used by the patient and the creation of a skin protection routine for each one. Additionally, the application of protective dressings to areas of friction is recommended, such as foams, hydrocolloids, and transparent films, especially after bathing, in addition to the rotation of application and fixation of the devices every eight hours, ensuring their correct installation and minimizing friction with the skin3.
Medical devices are essential in diagnostic, preventive, and therapeutic practices. However, their improper or prolonged use can unintentionally cause MDRPU21. Predisposing factors include material rigidity, inappropriate size selection, inappropriate fixation methods, and prolonged contact with the skin.
Interventions such as regular position changes, skin hydration, humidity control, daily skin inspections, and use of support surfaces have been shown to be effective in significantly reducing the incidence of injuries. In addition to these measures, skin protection in critically ill patients involves the use of physical barriers to reduce direct contact with moisture and friction. Among the materials used are multilayer silicone dressings, transparent protective films, barrier creams and specific sprays, which help preserve skin integrity and prevent injuries3. The 2025 Protocol on Prevention and Management of Pressure Injuries/Ulcers recommends the use of multilayer silicone adhesive dressings with foam and skin barrier protective products such as sprays or creams. These dressings complement standard prevention guidelines by redistributing pressure through forces dispersion over a larger area. They also minimize external shear forces and maintain a microclimate favorable for skin integrity3.
Implementing standardized protocols in ICUs can prevent the formation of these injuries and substantially improve clinical outcomes for patients. The 2025 international guideline, drawn up by the National Pressure Injury Advisory Panel (NPIAP), European Pressure Ulcer Advisory Panel (EPUAP) and Pan Pacific Pressure Injury Alliance (PPPIA), highlights that prevention must be a priority at all stages of intensive care, including early risk assessment and ongoing interventions to minimize predisposing factors3.
The predictive models developed in the study showed a moderate explanatory capacity for the outcomes of PU, IAD, and MDRPU. Although they were statistically significant, they only explained part of the variance in the results. This may indicate the need to integrate more clinical and environmental variables, as well as other longitudinal data, to refine predictive algorithms. Implementations that incorporate machine learning, combined with constant updates to monitored variables, can also improve the accuracy of these models1.
Furthermore, the continuous training of the multidisciplinary team, promoting collaborative education, the use of scales for skin risk assessment, device positioning, the standardization of procedures and records, as well as the development of specific guidelines, are crucial aspects highlighted in a systematic review on interventions in the prevention of MDRPU. These efforts are essential to overcome the challenges posed by the diversity and specificity of medical devices, as well as the wide range of interventions available for the prevention of MDRPUs22-24.
The findings of this study underscore the importance of preventive protocols based on identified risk factors, such as strict humidity control, adequate skin protection, and nutritional monitoring in critically ill patients in ICUs. The application of predictive models can improve clinical decision-making, allowing early interventions to minimize the occurrence of pressure injuries, incontinence-associated dermatitis, and medical device-related injuries, increasing the quality of care and reducing complications.
Although the developed predictive algorithms have demonstrated the ability to identify risk factors, the lack of longitudinal data and the complexity of clinical scenarios in ICUs may limit their applicability. Supporting AI algorithms requires continuous integration into care systems and training teams to interpret the data generated25-27.
This study has some limitations that should be considered. First, data collection was carried out in a single hospital center, which may limit the generalization of findings to other institutional realities and distinct regional contexts. Furthermore, the analysis was based on information from medical records and skin inspection at specific times, which may be subject to variations in completion and possible information biases. Another limitation refers to the lack of control over external variables, such as interventions carried out by different health teams and variations in care practices between shifts, which may influence the observed outcomes. Exclusion of critically ill patients who could not be lateralized for skin inspection may limit the applicability of predictive algorithms to those at highest risk. Patients who cannot be mobilized have a high risk of injury, requiring specific protocols. Finally, although the developed predictive models have demonstrated the ability to identify risk factors for skin lesions, external validation in other clinical scenarios is required to consolidate their applicability and robustness in different ICU contexts.
CONCLUSION
The study showed that variables such as inadequate moisture control, very poor nutritional status, lack of skin protection, and insufficient repositioning are significantly associated with an increased risk of outcomes (PU, IAD, and MDRPU). The application of predictive models has proven effective in identifying the main risk factors, providing support for the implementation of targeted preventive measures. Thus, the findings reinforce the importance of evidence-based protocols for injury prevention in ICUs, highlighting the need for integrated interventions that address multiple aspects of critical patient care.
REFERENCES
-
1. Silva TF, Tristão FS, Echeverria-Guanilo ME, Zillmer JGV, Oliveira CM, Blair IJQ. Care for prevention of pressure injuries performed by nurses in a teaching hospital. Rev Urug Enferm [Internet]. 2023 [cited 2025 Feb 10];18(2):e2023v18n2a8. Available from: https://doi.org/10.33517/rue2023v18n2a8
» https://doi.org/10.33517/rue2023v18n2a8 -
2. Andrade VSN, Sebold LF. Bundle for prevention of pressure injury associated with medical devices in obese patients. Cogitare Enferm [Internet]. 2023 [cited 2025 Feb 09];28:e89516. Available from: https://doi.org/10.1590/ce.v28i0.92827
» https://doi.org/10.1590/ce.v28i0.92827 -
3. National Pressure Injury Advisory Panel (NPIAP), European Pressure Ulcer Advisory Panel (EPUAP), Pan Pacific Pressure Injury Alliance (PPPIA). Prevention and Treatment of Pressure Ulcers/Injuries: Clinical Practice Guideline [Internet]. 2025 [cited 2025 Feb 09]. Available from: https://npiap.com/page/InternationalGuidelines
» https://npiap.com/page/InternationalGuidelines -
4. Melo CM, Bueno ALG, Rosseto TL, Almeida BG, Moreto MBC. Perssure injury in intensive care unit: Prevalence and associated factors in COVID-19 patientes. Rev Gaúcha Nurs [Internet]. 2023 [cited 2025 Feb 02];44:e20210345. Available from: https://doi.org/10.1590/1983-1447.2023.20210345.p
» https://doi.org/10.1590/1983-1447.2023.20210345.p -
5. Ramalho AO, Pacheco AO, Freitas PSS, Brandão ACMAG, Silva RA. Prevalence of incontinence-associated dermatitis and its related factors in intensive care patients. Braz J Enterostomal Ther [Internet]. 2024 [cited 2025 Mar 03];22:e1472. Available from: https://doi.org/10.30886/estima.v22.1472_PT
» https://doi.org/10.30886/estima.v22.1472_PT -
6. Nogueira DNG, Candido EF, Fuganti CCT, Fracarolli IFL, Pontel MC, Boll KM. Estimativa dos custos com antimicrobianos em casos de infecção de sítio cirúrgico. Rev SOBECC [Internet]. 2024 [cited 2025 Mar 09];29:E2429942. Available from: https://doi.org/10.5327/Z1414-4425202429942
» https://doi.org/10.5327/Z1414-4425202429942 -
7. Zhao Y, Zhang X, Li J, Wang H, Liu Y. Development of a predictive model for pressure injury risk in ICU patients using XGBoost and SHAP analysis. Sci Rep [Internet]. 2025 [cited 2025 Feb 09];15(1):92848. Available from: https://doi.org/10.1038/s41598-025-92848-2
» https://doi.org/10.1038/s41598-025-92848-2 -
8. Silva NR, Holanda ATP, Araújo MFN, Barbosa IKS, Santana GIS, Vieira TPLG, et al. Nursing management of patients with pressure injury in the intensive care unit. Rev Fisioter Terap [Internet]. 2024 [cited 2025 Mar 09];28:138. Available from: https://revistaft.com.br/manejo-da-enfermagem-ao-paciente-com-lesao-por-pressao-na-unidade-de-terapia-intensiva/
» https://revistaft.com.br/manejo-da-enfermagem-ao-paciente-com-lesao-por-pressao-na-unidade-de-terapia-intensiva/ -
9. Raimundo LS, Vieira LAS, Gálvez CL, Chen YW, Arcanjo AS. Prevention of pressure injuries in intensive care units: Understanding nurses' perceptions of pressure injuries related to medical devices. Rev REMECS [Internet]. 2024 [cited 2025 Mar 09];4(3):24. Available from: https://revistaremecs.com.br/index.php/remecs/article/download/1704/1739/2054
» https://revistaremecs.com.br/index.php/remecs/article/download/1704/1739/2054 -
10. Teixeira AO, Brinati LM, Toledo LV, Silva Neto JF, Teixeira DLP, Januário CF, et al. Factors associated with the incidence of pressure injury in critically ill patients: Cohort study. Rev Bras Nurse [Internet]. 2022 [cited 2025 Jan 03];75:6. Available from: https://doi.org/10.1590/0034-7167-2021-0267pt
» https://doi.org/10.1590/0034-7167-2021-0267pt -
11. Noie A, Jackson AC, Taheri M, Sayadi L, Bahramnezhad F. Determining the frequency of pressure ulcers incidence and associated risk factors in critical care patients: A 3-year retrospective study. Int Wound J [Internet]. 2024 [cited 2025 Feb 09];21(3):70120. Available from: https://pubmed.ncbi.nlm.nih.gov/39567442/
» https://pubmed.ncbi.nlm.nih.gov/39567442/ -
12. Souza MAP, Pessoa TM, Almeida LFS, Moura AF, Reinaldo AMC, Damião DBS, et al. Paciente crítico: utilização da escala de Braden na prevenção de lesão por pressão em pacientes de UTI. Rev Feridas [Internet]. 2022 [cited 2025 Mar 09];10(52):1867-76. Available from: https://doi.org/10.36489/feridas.2022v10i52p1867-1876
» https://doi.org/10.36489/feridas.2022v10i52p1867-1876 -
13. Ferreira M, Abbade L, Boccgi SCM, Miot HA, Boas PV, Guimaraes HQP. Incontinence-associated dermatitis in elderly patients: Prevalence and risk factors Dermatitis relacionada a la incontinencia en ancianos hospitalizados. Rev Bras Enferm [Internet]. 2020 [cited 2025 Feb 05];73 Suppl 3:e20180475. Available from: https://doi.org/10.1590/0034-7167-2018-0475
» https://doi.org/10.1590/0034-7167-2018-0475 -
14. Owen EJ, Heylen RA, Stewart K, Winyard PG, Jenkins ATA. The multi-factorial modes of action of urease in the pathogenesis of incontinence-associated dermatitis. Skin Health Dis [Internet]. 2024 [cited 2025 Mar 02];4(3):e349. Available from: https://doi.org/10.1002/ski2.349
» https://doi.org/10.1002/ski2.349 -
15. Deprez J, Ohde N, Wallin AE, Bååth C, Hommel A, Hultin L, et al. Prognostic factors for the development of incontinence-associated dermatitis (IAD): A systematic review. Int Wound J [Internet]. 2024 [cited 2025 Mar 01];21(7):e14962. Available from: https://doi.org/10.1111/iwj.14962
» https://doi.org/10.1111/iwj.14962 -
16. Monteiro DS, Borges EL, Spira JAO, Garcia TF, Matos SS. Incidence of skin lesions, risk, and clinical characteristics of critically ill patients. Texto Contexto Enferm [Internet]. 2021 [cited 2025 Jan 02];30:e20200125. Available from: https://doi.org/10.1590/1980-265X-TCE-2020-0125
» https://doi.org/10.1590/1980-265X-TCE-2020-0125 -
17. Silva A, Costa AQ, Rocha DM, Borges JWP, Andrade EMLR, Galiza FT, et al. Best practices of nurses in managing fecal incontinence in a hospital setting: An integrative review. Texto Contexto Enferm [Internet]. 2022 [cited 2025 Jan 02];31:e20220026. Available from: https://doi.org/10.1590/1980-265X-TCE-2022-0026pt
» https://doi.org/10.1590/1980-265X-TCE-2022-0026pt -
18. Jansen RC, Silva KBA, Moura MES. A escala de Braden na avaliação do risco para lesão por pressão . Rev Bras Enferm [Internet]. 2020. [cited 2025 Mar 09];73(6):e20190413. Available from: https://doi.org/10.1590/0034-7167-2019-0413
» https://doi.org/10.1590/0034-7167-2019-0413 -
19. Vecin NM, Gater DR. Pressure Injuries and Management after Spinal Cord Injury. J Pers Med [Internet]. 2022 [cited 2025 Jan 15];12(7):1130. Available from: https://doi.org/10.3390/jpm12071130
» https://doi.org/10.3390/jpm12071130 -
20. Soldera D, Girondi JBR, Hammerschidt KSA, Ouriques Neta EL, Amarante LN. Medical device-related pressure injuries in nurses' clinical practice. Enferm Foco [Internet]. 2021 [cited 2025 Mar 03];12(2):209-22. Available from: https://doi.org/10.21675/2357-707X.2021.v12.n2.3427
» https://doi.org/10.21675/2357-707X.2021.v12.n2.3427 -
21. Abud ACF, Santos GKBB, Santana LBF, Barreiro MSC, Cunha MOG. Factors associated with the occurrence of Medical-Device-Related Pressure Injury (MDRPI). Res Soc Dev [Internet]. 2024 [cited 2025 Mar 09];13(7):e4113746240. Available from: https://doi.org/10.33448/rsd-v13i7.46240
» https://doi.org/10.33448/rsd-v13i7.46240 -
22. Lyu Y, Huang YL, Li ZY, Lin F. Interventions and strategies to prevent medical device-related pressure injuries in adult patients: A systematic review. J Clin Nurs [Internet]. 2023 [cited 2025 Mar 09];32:6863-78. Available from: https://doi.org/10.1111/jocn.16790
» https://doi.org/10.1111/jocn.16790 -
23. Paz L, Gomes MMS, Ribeiro MC, Capellari C. Prevention of pressure injury: An integrative literature review. Rev Enferm Atual Derme [Internet]. 2023 [cited 2024 Feb 15];97(4):e023191. Available from: https://doi.org/10.31011/reaid-2023-v.97-n.4-art.160
» https://doi.org/10.31011/reaid-2023-v.97-n.4-art.160 -
24. Tayyib N, Asiri MY, Danic S, Sahi SL, Lasafin J, Generale LF, et al. The effectiveness of the SKINCARE bundle in preventing medical-device related pressure injuries in critical care units: A clinical trial. Adv Skin Wound Care [Internet]. 2021 [cited 2025 Jan 10];34(2):75-80. Available from: https://doi.org/10.1097/01.ASW.0000733062.34156.89
» https://doi.org/10.1097/01.ASW.0000733062.34156.89 -
25. Lin F, Wu Z, Song B, Coyer F, Chaboyer W. The effectiveness of multicomponent pressure injury prevention programs in adult intensive care patients: A systematic review. Int J Nurs Stud [Internet]. 2020 [cited 2025 Mar 05];102:103483. Available from: https://doi.org/10.1016/j.ijnurstu.2019.103483
» https://doi.org/10.1016/j.ijnurstu.2019.103483 -
26. Topaz M, Pruinelli L. Big data and nursing: Implications for the future. Stud Health Technol Inform [Internet]. 2017 [cited 2025 Jan 10];232:165-71. Available from: https://https://ebooks.iospress.nl/publication/46077
» https://https://ebooks.iospress.nl/publication/46077 -
27. Sanchez-Pinto LN, Luo Y, Churpek MM. Big data and data science in critical care. Chest [Internet]. 2018 [cited 2025 Jan 10];154(5):1239-48. Available from: https://doi.org/10.1016/j.chest.2018.04.037
» https://doi.org/10.1016/j.chest.2018.04.037
NOTES
-
APPROVAL OF ETHICS COMMITTEE IN RESEARCH
Approved by the Research Ethics Committee of the Instituto Dr. José Frotas, opinion no. 5,675,703/2022, Certificate of Presentation for Ethical Assessment 61145922.3.0000.5047.
-
TRANSLATED BY
Denise Costa Rodrigues.
-
DATA AVAILABILITY
The entire dataset supporting the study can be accessed at: https://doi.org/10.17605/OSF.IO/8Z7RM
Edited by
The entire dataset supporting the study can be accessed at: https://doi.org/10.17605/OSF.IO/8Z7RM
