Explainable SHAP-XGBoost models for pressure ulcers among patients requiring with mechanical ventilation in intensive care unit

Explainable SHAP-XGBoost models for pressure ulcers among patients requiring with mechanical ventilation in intensive care unit | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Explainable SHAP-XGBoost models for pressure ulcers among patients requiring with mechanical ventilation in intensive care unit Yu-juan Xue, Li Zheng, Zhen-nan Yuan, Xue-zhong Xing This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5322280/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 22 Mar, 2025 Read the published version in Scientific Reports → Version 1 posted 12 You are reading this latest preprint version Abstract Background Pressure ulcers are significant concern for ICU patients on mechanical ventilation. Early prediction is crucial for enhancing patient outcomes and reducing healthcare costs. This study aims to develop a predictive model using machine learning techniques, specifically XGBoost combined with SHAP, to identify key risk factors of pressure ulcers in this population. Methods Utilizing the MIMIC-IV 2.2 database, we included a cohort of 29,448 mechanically ventilated patients in ICU intensive unit. These patients were divided into a training set (20,614 patients, 70%) and an internal validation set (8,834 patients, 30%). Of these, 2,052 patients developed pressure ulcers. We applied the XGBoost algorithm to build the predictive model and used SHAP analysis to identify the top ten factors influencing pressure ulcer development: 'sepsis', 'age', 'the count of platelet', 'length of ICU stay', 'PaO2/FiO2 ratio', 'hemoglobin concentration', 'admission type', 'renal disease', 'albumin concentration', and 'ethnicity'. Results The predictive model achieved an area under the ROC curve (AUC) of 0.797 (95% CI: 0.786–0.808) in the training set and 0.739 (95% CI: 0.721–0.758) in the validation set. Calibration curves demonstrated good fit, and the decision curve analysis indicated clinical utility. Conclusion This study successfully developed a machine learning model that accurately predicts the risk of pressure ulcers in ICU patients with mechanical ventilation. This model could serve as a valuable tool for guiding early interventions, ultimately reducing the incidence of pressure ulcers in this vulnerable population. The integration of SHAP analysis offers insights into the most critical factors. Health sciences/Health care Health sciences/Risk factors pressure ulcer mechanical ventilation XGBoost Figures Figure 1 Figure 2 Figure 3 Figure 4 Background Pressure ulcers, also known as bedsores, represent a significant clinical challenge, particularly among critically ill patients in the intensive care unit (ICU) who are on mechanical ventilation. These patients are at heightened risk due to prolonged immobility, impaired skin integrity, and other complex comorbidities. Pressure ulcers not only lead to increased morbidity and prolonged hospital stays but also impose substantial economic burdens on healthcare systems[ 1 – 3 ]. Early identification of patients at risk for pressure ulcers is crucial to prevent their occurrence and improve patient outcomes. However, traditional risk assessment tools often lack predictive accuracy and fail to capture the multifactorial nature of pressure ulcer development in ICU patients. Machine learning is a type of artificial intelligence that can be used to build predictive models, but it is rarely used in research on pressure injuries[ 4 ]. The advent of machine learning, particularly models such as XGBoost, offers a promising avenue for developing more accurate predictive models. Furthermore, SHAP (SHapley Additive exPlanations) analysis enables the identification of key variables driving model predictions[ 5 ], offering deeper insights into the risk factors associated with pressure ulcer development. This study aims to leverage machine learning and SHAP analysis to build a robust predictive model for pressure ulcers in mechanically ventilated ICU patients, utilizing the large-scale MIMIC-IV database. Patients and Methods Database and Study Population A flowchart of the patient selection process is shown in Figure 1. This study was conducted using data from the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 2.2), a publicly available database containing detailed information on ICU admissions at the Beth Israel Deaconess Medical Center in Boston, MA, from 2008 to 2019. The dataset includes over 76,000 ICU admissions and provides comprehensive data, including demographics, laboratory results, vital signs, treatments, and outcomes. The study was exempt from institutional review board approval as it utilized a publicly available de-identified database, and access was granted after completing the Collaborative Institutional Training Initiative (CITI) program (Record ID: 36067767). For this study, we included ICU patients who were mechanically ventilated, as they represent a high-risk group for developing pressure ulcers. A total of 29,448 patients were selected, of whom 2,052 developed pressure ulcers. The data was randomly split into a training set (70%, n=20,614) and an internal validation set (30%, n=8,834). The selection criteria included adult patients (age ≥18 years) who had complete data for the variables of interest. Patients with more than 50% missing data or multiple ICU admissions (only the first admission was considered) were excluded. Data Collection Data were collected on various factors potentially related to the development of pressure ulcers, based on existing literature and clinical relevance. These factors included demographic information (age, gender), comorbidities (sepsis, renal disease, myocardial infarction, chronic pulmonary disease), ICU-specific metrics (length of ICU stay, admission type), vital signs, and laboratory values within the first 24 hours of ICU admission (e.g., PaO2/FiO2 ratio, hemoglobin, platelets, albumin). The primary outcome was the development of pressure ulcers during the ICU stay. Model Development and Validation The predictive model was developed using the XGBoost algorithm, a powerful machine learning technique known for its robustness in handling structured data. The model was trained on the training set, with hyperparameters optimized using the caret package in R. The key parameters included a learning rate (eta) of 0.1, a maximum depth of 2, and 100 boosting rounds. The model's performance was evaluated on the internal validation set using the area under the receiver operating characteristic (ROC) curve (AUC) to assess its discriminatory ability. SHAP Analysis To interpret the model and identify the most influential predictors of pressure ulcer development, SHAP (SHapley Additive exPlanations) values were calculated using the shapviz R package. SHAP values provide insights into how each feature contributes to the model's predictions. The top ten factors influencing pressure ulcer risk were identified, and SHAP summary plots were generated to visualize their impact. Statistical Analysis All statistical analyses were performed using R version 4.3.0. Continuous variables were summarized as mean ± standard deviation or median with interquartile range (IQR), and categorical variables were presented as frequencies and percentages. Comparisons between groups were made using the Chi-square test for categorical variables and independent samples t-test or nonparametric tests for continuous variables. Model performance was evaluated using ROC curves, calibration curves, and decision curve analysis (DCA) to assess the clinical utility of the model. Accuracy, sensitivity, specificity, and other metrics were reported for both the training and validation sets. Results Patient Characteristics The study cohort consisted of 29,448 mechanically ventilated ICU patients from the MIMIC-IV database, of whom 2,052 (6.97%) developed pressure ulcers. The training set included 20,614 patients, while the validation set included 8,834 patients. The baseline characteristics of patients in the training and validation sets were similar, ensuring a balanced distribution of demographic and clinical factors (Table 1). Table 1. Characteristics of study participants from training set and validation set. Factors Training sets N=20,614 Testing sets N=8,834 P Age (mean, y) 65.5±0.1 65.7±0.2 0.090 Female (N, %) 8,360(40.6) 3,588(40.5) 0.922 COPD (N, %) 5,006(24.3) 2,208(25.0) 0.194 Diabetes (N, %) 4,657(22.6) 1,969(22.3) 0.569 Myocardial infarct (N, %) 3,383(16.4) 1,452(16.4) 0.003 Congestive heart failure (N, %) 5,179(25.1) 2,212(25.0) 0.879 Peripheral vascular disease (N, %) 2,257(10.9) 985(11.2) 0.613 Cerebrovascular disease (N, %) 2,839(13.8) 1,211(13.7) 0.884 Dementia (N, %) 614 (3.0) 275 (3.1) 0.537 Renal disease (N, %) 3,735(18.1) 1,575(17.8) 0.553 Mild liver disease (N,%) 2,202(10.7) 905(10.2) 0.263 Paraplegia (N, %) 835(4.1) 370(4.2) 0.585 The presence of metastasis (N, %) 1,045(5.1) 476(5.4) 0.257 AIDS (N, %) 94(0.5) 48(0.5) 0.321 Cancer (N, %) 2,306 1,014 0.468 Respiration（PaO2/FIO2） 195.2±0.7 195.1±1.1 0.516 Glasgow (mean) 14.4±0.01 14.4±0.02 0.668 Hemoglobin (mean, g/L) 9.9±0.02 10.0±0.02 0.434 Platelets count(10^6/L) 183.1±0.7 182.5±1.1 0.674 WBC (10^9/L) 15.2±0.1 15.5±0.1 0.998 Albumin (mean, g/L) 3.2±0.0 3.2±0.0 0.809 Sepsis (N, %) 13,920(67.5) 5,948(67.3) 0.742 Shock (N, %) 2,274(11.0) 997 (11.3) 0.524 Lactate (mean, mmol/L) 2.9±0.0 3.0±0.0 0.115 Los of ICU (mean, day) 4.4±0.0 4.4±0.1 0.426 BMI (mean, kg/m^2) 23.5±0.0 23.0±0.1 0.806 Emergency admission (N, %) 13,515(65.6) 5,779(65.4) 0.608 Ethnicity (N, %) 0.836 Asian 541(2.6) 231(2.6) White 13,859(67.2) 5,867(66.4) Black 1,644(8.0) 682(77.2) Others 2,066(10.0) 897(10.2) Unknown 2,774(13.5) 1,157(13.1) Model Performance The XGBoost model demonstrated good performance in predicting the development of pressure ulcers, with an AUC of 0.797 (95% CI: 0.786-0.808) in the training set (Figure 2A) and 0.739 (95% CI: 0.721-0.758) in the validation set (Figure 2B). The calibration curves indicated that the model's predictions were well-calibrated, with predicted probabilities closely matching observed outcomes both in training set (Figure 2C) and testing set (Figure 2D). Clinical Utility Decision curve analysis (DCA) showed that the XGBoost model provided a net clinical benefit across a wide range of threshold probabilities, supporting its potential utility in clinical practice (Figure 2E). The model's performance suggests it could be effectively used to identify high-risk patients and guide early interventions to prevent pressure ulcers in mechanically ventilated ICU patients. SHAP Analysis SHAP analysis identified the ten most influential factors for predicting pressure ulcer development: 'sepsis', 'age', 'the count of platelet', 'length of ICU stay', 'PaO2/FiO2 ratio', 'hemoglobin concentration', 'admission type', 'renal disease', 'albumin concentration', and 'ethnicity'. These factors were ranked based on their average absolute SHAP values, highlighting their relative importance in the model (Figure 3A). To demonstrate how the XGBoost model evaluates the contributions of individual patient features, we utilize SHAP force plots to interpret individual predictions for two patients (Figure 3B, 3C). The color indicates the contribution of each feature, with red indicating that the feature has a negative effect on the prediction (arrow pointing left, SHAP value decreases) and yellow indicating that the feature has a positive effect on the prediction (arrow pointing right, SHAP value increases). The length of the color bar indicates the strength of the contribution, and E[f(x)] indicates the SHAP reference value, which is the mean predicted by the model. For a “true positive” group of patients, the XGBoost model predicted in-hospital mortality with a SHAP. Figure 4 illustrates the interaction summary plot for the top 10 most significant interacting features. The features have been sorted in descending order of their interaction importance: 'sepsis', 'age', 'the count of platelet', 'length of ICU stay', 'PaO2/FiO2 ratio', 'hemoglobin concentration', 'admission type', 'renal disease', 'albumin concentration', and 'ethnicity'. The interaction summary plot reveals that certain feature pairs have a significant combined impact on the model’s predictions. For instance, the interaction between sepsis and age, as well as sepsis and los of ICU, displays notable SHAP interaction values, suggesting that these pairs of features interact strongly to influence the model’s output. Discussion In this study, we developed and validated a predictive model for pressure ulcer development in mechanically ventilated ICU patients using the XGBoost machine learning algorithm. The model demonstrated strong predictive performance with an AUC of 0.797 in the training set and 0.739 in the internal validation set. Our findings underscore the utility of machine learning approaches in identifying high-risk patients and enhancing clinical decision-making processes. One way to consider our model’s performance is to place our results alongside the Braden Scale[ 6 ]. The Braden Scale is the most commonly used tool in North America for predicting risk for pressure injury and measures cumulative risk for pressure injuries via 7 categories: sensory perception, activity, mobility, moisture, nutrition, and friction/shear. Total scores range from 9 (very high risk) to 23 (very low risk)[ 6 ]. Our model’s relatively strong performance (area under the ROC curve = 0.79 vs 0.68 for the Braden Scale[ 7 ]) suggests the model would be a useful way to differentiate among critical care patients in order to apply preventive measures that are not feasible for every patient because of cost, such as specialty beds. Jenny[ 8 ] conducted a systematic review to identify risk factors independently predictive of pressure ulcer development among critical-care patients. They founded that age, mobility/activity, perfusion, and vasopressor infusion frequently emerged as important factors in pressure ulcer development. Sepsis emerged as the most significant predictor, consistent with its systemic impacts on the body. Studies have consistently shown that sepsis leads to widespread inflammation and coagulation abnormalities, which impair microcirculation and tissue perfusion, crucial factors in the pathogenesis of pressure ulcers[ 9 ]. The hyperinflammatory state associated with sepsis increases the risk of skin breakdown and inhibits wound healing processes, highlighting the need for vigilant monitoring and early intervention in septic patients. Age is a well-documented risk factor for pressure ulcer development. Older patients often have thinner, less elastic skin, and a decreased ability to redistribute pressure, increasing their vulnerability[ 10 ]. Findings for age is consistent with the results from a systematic review conducted by Coleman and colleagues in an acute, rehabilitative, long-term-care population[ 11 ]. Additionally, polypharmacy, comorbidities, and impaired mobility often present in older adults further exacerbate this risk. This emphasizes the importance of geriatric-specific preventive care strategies in the ICU. In this study, both high and low platelet counts contributed to the development of pressure ulcers. Based on the previous study[ 12 ], elevated platelet counts appear to be significantly associated with the development of pressure ulcers in bedridden elderly patients. The research demonstrated that patients with pressure ulcers had higher platelet counts and increased platelet aggregation compared to those without pressure ulcers (p < 0.03). This suggests that increased platelet activity could contribute to the pathogenesis of pressure ulcers by exacerbating local microcirculatory impairments and inflammatory responses, which in turn compromises tissue perfusion and integrity. These results corroborate the hypothesis that heightened platelet count and activity may be a critical factor in pressure ulcer formation. Conversely, low platelet counts can signify coagulopathy, impacting the body's ability to repair tissue damage effectively. Thus, platelet levels may serve as a surrogate marker for the inflammatory and hemostatic balance. Length of ICU Stay Longer ICU stays are inherently associated with increased pressure ulcer risk due to prolonged immobilization and exposure to a high-intervention environment. Extended bed rest and the critical nature of these patients often result in continuous pressure on bony prominences without sufficient relief or repositioning, underlining the necessity for proactive pressure ulcer prevention protocols in long-term ICU patients. In previous literature, long hospital stay, especially in ICU, is a risk factor for pressure ulcer[ 13 – 16 ]. PaO2/FiO2 Ratio The PaO2/FiO2 ratio, a measure of respiratory function and oxygenation, reflects the patient's ability to deliver oxygen to tissues. Lower ratios indicate inadequate oxygenation, leading to cellular hypoxia, which compromises tissue integrity and healing. Ensuring optimal respiratory function is thus crucial to prevent pressure ulcers in patients requiring mechanical ventilation. In this study, the status of each subgroup in sofa score was included in the prediction of pressure ulcer. Among the top ten important factors, respiratory factors, oxygenation index and renal function were more important in the prediction of pressure ulcer. In previous literature reports, Maarit[ 16 ] assessed the performance of the Sequential Organ Failure Assessment (SOFA) scale and its subcategories in predicting the development of PUs. They found that the Glasgow Coma score, renal and respiratory disorders, and hypotension were significantly (P < .0001) linked to PU development. First-day total SOFA score and its cardiovascular and respiratory subcategory scores were the most important predictors of PUs. Hemoglobin Levels Low hemoglobin levels, or anemia, impair oxygen transport capacity, leading to tissue hypoxia—a condition detrimental to skin integrity and repair[ 17 ]. Adequate hemoglobin levels are essential to ensure optimal tissue perfusion and the delivery of oxygen necessary for cellular metabolism and repair mechanisms. Monitoring and managing hemoglobin levels can therefore be a critical component of pressure ulcer prevention strategies. Admission Type The type of ICU admission often reflects the patient's baseline health status and urgency of care, which can inherently affect the risk for pressure ulcers. Emergency admissions are typically associated with acute illnesses and limited preparation time for preventive care, unlike elective admissions which allow for planned preoperative optimization. A pressure ulcer can develop in several hours, depending upon risk factors and use of pressure ulcer prevention activities[ 15 ]. This highlights the need for heightened vigilance and early intervention for patients admitted emergently. Renal Disease Chronic kidney disease and acute renal failure contribute to pressure ulcer risk through mechanisms like fluid overload, metabolic disturbances, and impaired nutritional status, which is in agreement with previous studies[ 18 ]. The accumulation of uremic toxins and altered protein metabolism can weaken skin and tissue health, making prevention and management particularly challenging in renal-compromised patients. Albumin Levels Hypoalbuminemia is a marker of malnutrition and an independent risk factor for pressure ulcer development, as low albumin levels can lead to decreased oncotic pressure, edema, and poor tissue repair. Chung[ 19 ]conducted a review about risk factors for pressure Injuries in Adult Patients. They found that hypoalbuminemia was identified as possible risk factor for pressure injury development. Ensuring adequate nutritional support and addressing hypoalbuminemia are therefore critical in preventing pressure ulcers in the ICU setting. Ethnicity While ethnicity as a risk factor can point to genetic predispositions for skin conditions or varying susceptibilities, it often reflects broader socio-economic disparities and differences in health care access or delivery[ 20 ]. Interventions should account for such disparities, ensuring equitable care and prevention measures across different ethnic groups. The study was carried out in Indonesia where the incidence of pressure ulcers has been reported to be as high as 33.3%[ 21 ]. The incidence of pressure ulcers is higher in Indonesia than in other Asian countries, where the incidence ranges from 2.1–31.3%[ 22 , 23 ]. Other international studies have reported incidence rates of 7–29% in an intensive care unit (ICU) or acute care setting[ 24 , 25 ]. Conclusion Our findings, rooted in both the predictive power of advanced machine learning techniques and supported by existing literature, provide valuable insights into the key factors influencing pressure ulcer development in ICU patients. Targeted interventions addressing these specific risk factors could significantly improve patient outcomes by preventing pressure ulcers among high-risk populations in critical care settings. Future studies should focus on external validation and real-world implementation of such predictive models to enhance preventive care strategies further. Limitations Despite the strengths of this study, including the large sample size and robust machine learning methodology, several limitations should be noted. First, the study was based on retrospective data from the MIMIC-IV database, which may limit the generalizability of the findings to other populations and healthcare settings. Although our model performed well in internal validation, it was not externally validated using data from a different hospital or patient population. External validation is crucial to confirm the model’s applicability in diverse clinical environments. Second, while the MIMIC-IV database provides a wealth of clinical data, the potential for missing or inaccurately recorded data remains a concern. We attempted to mitigate this by excluding patients with more than 50% missing data, but residual confounding may still exist. Additionally, the model did not include some potentially relevant variables, such as patient mobility, skin condition, and specific interventions for pressure ulcer prevention, which were not available in the MIMIC-IV database. Future studies should aim to incorporate these factors to refine the model further. Lastly, the model’s performance, while strong, suggests that there is room for improvement. The AUC values, particularly in the validation set, indicate that the model may not capture all the complexity associated with pressure ulcer development. Further refinement of the model, perhaps by incorporating additional data or using more advanced algorithms, could enhance its predictive accuracy. Conclusion This study demonstrates that machine learning, specifically the XGBoost algorithm combined with SHAP analysis, can effectively predict the risk of pressure ulcer development in mechanically ventilated ICU patients. The model's strong performance in both the training and validation sets highlights its potential utility in clinical practice. By identifying key risk factors and providing interpretable insights, this model could serve as a valuable tool for guiding early interventions, ultimately improving patient outcomes and reducing the incidence of pressure ulcers in this vulnerable population. Future work should aim to validate the model in diverse settings and explore the integration of additional variables to enhance predictive accuracy. Declarations Ethics approval and consent to participate: The data in this study were from two public de-identified databases. After completing Collaborative Institutional Training Initiative (CITI program), we got permission to access the database (Record ID: 36067767). This study involves human participants but the Ethics Committee(s) or Institutional Board(s) exempted this study. Acknowledgement : None. Consent for publication : Not applicable. Availability of data and materials : The datasets used and/or analysed during the current study are available from corresponding author upon reasonable request. Contributorship statement (I) Conception and design: ZN Yu, XZ Xing and YJ Xue; (II) Provision of study materials or patients: SN Qu, CL Huang, HJ Wang and H Wang; (III) Collection and assembly of data: YJ Xue and ZN Yuan; (IV) Data analysis and interpretation: ZN Yuan and H Zhang; (V) Final approval of manuscript: All authors. Competing interests The authors declare that they have no competing interests. Funding : None. References Spilsbury, K. et al. Pressure ulcers and their treatment and effects on quality of life: hospital inpatient perspectives. J. Adv. Nurs. 57 (5), 494–504 (2007). Frankel, H., Sperry, J. & Kaplan, L. Risk factors for pressure ulcer development in a best practice surgical intensive care unit. Am. Surg. 73 (12), 1215–1217 (2007). Slowikowski, G. C. & Funk, M. Factors associated with pressure ulcers in patients in a surgical intensive care unit. J. Wound Ostomy Cont. Nurs. 37 (6), 619–626 (2010). Raju, D., Su, X., Patrician, P. A., Loan, L. A. & McCarthy, M. S. Exploring factors associated with pressure ulcers: a data mining approach. Int. J. Nurs. Stud. 52 (1), 102–111 (2015). Ali, S. et al. The enlightening role of explainable artificial intelligence in medical & healthcare domains: A systematic literature review. Comput. Biol. Med. 166 , 107555 (2023). Braden, B. & Bergstrom, N. A conceptual schema for the study of the etiology of pressure sores. Rehabil Nurs. 12 (1), 8–12 (1987). Hyun, S. et al. Predictive validity of the Braden scale for patients in intensive care units. Am. J. Crit. Care . 22 (6), 514–520 (2013). Alderden, J., Rondinelli, J., Pepper, G., Cummins, M. & Whitney, J. Risk factors for pressure injuries among critical care patients: A systematic review. Int. J. Nurs. Stud. 71 , 97–114 (2017). Singer, M. et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA . 315 (8), 801–810 (2016). Baumgarten, M. et al. Extrinsic risk factors for pressure ulcers early in the hospital stay: a nested case-control study. J. Gerontol. Biol. Sci. Med. Sci. 63 (4), 408–413 (2008). Coleman, S. et al. A new pressure ulcer conceptual framework. J. Adv. Nurs. 70 (10), 2222–2234 (2014). Matsuyama, N. et al. The effect of anti-platelet aggregation to prevent pressure ulcer development: a retrospective study of 132 elderly patients. Gerontology . 46 (6), 311–317 (2000). Han, D. et al. Prolonged stay in the emergency department is an independent risk factor for hospital-acquired pressure ulcer. Int. Wound J. 17 (2), 259–267 (2020). Lima Serrano, M., Gonzalez Mendez, M. I., Carrasco Cebollero, F. M. & Lima Rodriguez, J. S. Risk factors for pressure ulcer development in Intensive Care Units: A systematic review. Med. Intensiva . 41 (6), 339–346 (2017). Naccarato, M. K. & Kelechi, T. Pressure ulcer prevention in the emergency department. Adv. Emerg. Nurs. J. 33 (2), 155–162 (2011). Ahtiala, M., Soppi, E. & Saari, T. I. Sequential Organ Failure Assessment (SOFA) to Predict Pressure Ulcer Risk in Intensive Care Patients: A Retrospective Cohort Study. Ostomy Wound Manage. 64 (10), 32–38 (2018). Alderden, J. et al. Predicting Pressure Injury in Critical Care Patients: A Machine-Learning Model. Am. J. Crit. Care . 27 (6), 461–468 (2018). Seo, Y., Oh, H., Na, Y., Kim, M. & Seo, W. A Prospective Study of Pressure Injury Healing Rate and Time and Influencing Factors in an Acute Care Setting. Adv. Skin. Wound Care . 35 (12), 1–9 (2022). Chung, M. L. et al. Risk Factors for Pressure Injuries in Adult Patients: A Narrative Synthesis. Int. J. Environ. Res. Public. Health 19 (2). (2022). Lyder, C. H., Shannon, R., Empleo-Frazier, O., McGeHee, D. & White, C. A comprehensive program to prevent pressure ulcers in long-term care: exploring costs and outcomes. Ostomy Wound Manage. 48 (4), 52–62 (2002). Suriadi, S. H. et al. A new instrument for predicting pressure ulcer risk in an intensive care unit. J. Tissue Viability . 16 (3), 21–26 (2006). Jun Seongsook, R. N., Jeong Ihnsook, R. N. & Lee Younghee, R. N. Validity of pressure ulcer risk assessment scales; Cubbin and Jackson, Braden, and Douglas scale. Int. J. Nurs. Stud. 41 (2), 199–204 (2004). Suriadi, S. H., Sugama, J., Thigpen, B. & Subuh, M. Development of a new risk assessment scale for predicting pressure ulcers in an intensive care unit. Nurs. Crit. Care . 13 (1), 34–43 (2008). Theaker, C., Kuper, M. & Soni, N. Pressure ulcer prevention in intensive care - a randomised control trial of two pressure-relieving devices. Anaesthesia . 60 (4), 395–399 (2005). Whittington, K. T. & Briones, R. National Prevalence and Incidence Study: 6-year sequential acute care data. Adv. Skin. Wound Care . 17 (9), 490–494 (2004). Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 22 Mar, 2025 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 28 Jan, 2025 Reviews received at journal 23 Jan, 2025 Reviewers agreed at journal 16 Jan, 2025 Reviews received at journal 26 Dec, 2024 Reviewers agreed at journal 18 Dec, 2024 Reviews received at journal 24 Nov, 2024 Reviewers agreed at journal 19 Nov, 2024 Reviewers invited by journal 19 Nov, 2024 Editor assigned by journal 19 Nov, 2024 Editor invited by journal 12 Nov, 2024 Submission checks completed at journal 11 Nov, 2024 First submitted to journal 23 Oct, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5322280","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":382013930,"identity":"aabff911-6f5e-408e-9f39-f80693c13c6c","order_by":0,"name":"Yu-juan Xue","email":"","orcid":"","institution":"Peking University People’s Hospital, Peking University","correspondingAuthor":false,"prefix":"","firstName":"Yu-juan","middleName":"","lastName":"Xue","suffix":""},{"id":382013931,"identity":"398e8b06-48ba-44c8-951b-d0d245aa3512","order_by":1,"name":"Li Zheng","email":"","orcid":"","institution":"National Cancer Center, National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences \u0026 Peking Union Medical College","correspondingAuthor":false,"prefix":"","firstName":"Li","middleName":"","lastName":"Zheng","suffix":""},{"id":382013932,"identity":"7e42b06f-f7e4-48a7-bb38-445dff13764c","order_by":2,"name":"Zhen-nan Yuan","email":"","orcid":"","institution":"National Cancer Center, National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences \u0026 Peking Union Medical College","correspondingAuthor":false,"prefix":"","firstName":"Zhen-nan","middleName":"","lastName":"Yuan","suffix":""},{"id":382013933,"identity":"c920d1aa-c441-4850-8d4f-43e7bea04efe","order_by":3,"name":"Xue-zhong Xing","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAApElEQVRIiWNgGAWjYPACGx5+/gbStKTJSM44QJqWwzYGDQlEqjW4kf7swY+a8zwGDAcYP3zMIUKLZM+BdMOeY7d5zJkbmCVnbiNCCz97wzFpxobbPJYNB9iYeYnRwsbM2AbUco7H4EACkVr42ZvZgFoOkKBFsucYGxAn80jOONhMnF9AISbxo8bOnp+/+eCHj8RoQQKMDaSpHwWjYBSMglGAGwAA8qsw72REIDIAAAAASUVORK5CYII=","orcid":"","institution":"National Cancer Center, National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences \u0026 Peking Union Medical College","correspondingAuthor":true,"prefix":"","firstName":"Xue-zhong","middleName":"","lastName":"Xing","suffix":""}],"badges":[],"createdAt":"2024-10-24 03:08:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5322280/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5322280/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-025-92848-2","type":"published","date":"2025-03-22T15:57:48+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":71603639,"identity":"26f5e6b7-5a73-4831-bc96-09f8c579b75a","added_by":"auto","created_at":"2024-12-17 05:59:43","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":42552,"visible":true,"origin":"","legend":"\u003cp\u003eFlow diagram of the patient selection in MIMIC IV 2.2. (MIMIC-IV, Medical Information Mart for Intensive Care).\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-5322280/v1/b5fe5705c86ce9818066fa94.png"},{"id":71603642,"identity":"fc605f97-814d-48d3-9683-775c51673a12","added_by":"auto","created_at":"2024-12-17 05:59:45","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":738339,"visible":true,"origin":"","legend":"\u003cp\u003eThe performance of the predicted model in the training set (A) and validation set (B). Calibration curves of the predicted model for predicting pressure ulcer both in the training set (C) and validation set (D). Decision-curve analysis of the predicted model in validation set (E).\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-5322280/v1/f550b590ebda8b83a49ecba0.png"},{"id":71603638,"identity":"9d300da3-bb56-48ce-9707-f6d26c3f66e2","added_by":"auto","created_at":"2024-12-17 05:59:42","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":787332,"visible":true,"origin":"","legend":"\u003cp\u003eA. Importance chart of SHAP variables, with the included features sorted by the average absolute value of SHAP from highest to lowest. B, C. SHAP force plot for two cases: Color indicates the contribution of each feature, purple indicates that the feature has a negative effect on the prediction (arrow to the left, SHAP value decreases), and yellow indicates that the feature has a positive effect on the prediction (arrow to the right, SHAP value increases). The length of the color bar indicates the strength of the contribution, and E[f(x)] indicates the SHAP reference value, which is the mean predicted by the model. f (x) represents the SHAP value of the individual.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-5322280/v1/da98363569f4e86166675f61.png"},{"id":71603643,"identity":"164c952c-227f-4b07-908a-2c3d75611f48","added_by":"auto","created_at":"2024-12-17 05:59:46","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":138713,"visible":true,"origin":"","legend":"\u003cp\u003eInteraction summary plot generated using SHAP values. This plot displays the top 10 most interacting features of the model. On the x-axis and y-axis, the features are listed according to their interaction importance, with the feature names ordered as follows: 'sepsis', 'age', 'the count of platelet', 'length of ICU stay', 'PaO2/FiO2 ratio', 'hemoglobin concentration', 'admission type', 'renal disease', 'albumin concentration', and 'ethnicity'. Each point on the plot represents the SHAP interaction value for a specific feature interaction, highlighting how pairs of features together impact the model’s predictions.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-5322280/v1/b44199a61ae43a74729cbee7.png"},{"id":79120449,"identity":"fd3acbfb-85c1-42f6-95db-b925abd533e0","added_by":"auto","created_at":"2025-03-24 16:08:24","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1459104,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5322280/v1/84dc91ca-0d63-4007-98fc-caa2f27a4777.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Explainable SHAP-XGBoost models for pressure ulcers among patients requiring with mechanical ventilation in intensive care unit","fulltext":[{"header":"Background","content":"\u003cp\u003e Pressure ulcers, also known as bedsores, represent a significant clinical challenge, particularly among critically ill patients in the intensive care unit (ICU) who are on mechanical ventilation. These patients are at heightened risk due to prolonged immobility, impaired skin integrity, and other complex comorbidities. Pressure ulcers not only lead to increased morbidity and prolonged hospital stays but also impose substantial economic burdens on healthcare systems[\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Early identification of patients at risk for pressure ulcers is crucial to prevent their occurrence and improve patient outcomes. However, traditional risk assessment tools often lack predictive accuracy and fail to capture the multifactorial nature of pressure ulcer development in ICU patients. Machine learning is a type of artificial intelligence that can be used to build predictive models, but it is rarely used in research on pressure injuries[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. The advent of machine learning, particularly models such as XGBoost, offers a promising avenue for developing more accurate predictive models. Furthermore, SHAP (SHapley Additive exPlanations) analysis enables the identification of key variables driving model predictions[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e], offering deeper insights into the risk factors associated with pressure ulcer development. This study aims to leverage machine learning and SHAP analysis to build a robust predictive model for pressure ulcers in mechanically ventilated ICU patients, utilizing the large-scale MIMIC-IV database.\u003c/p\u003e"},{"header":"Patients and Methods","content":"\u003cp\u003e\u003cstrong\u003eDatabase and Study Population\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA flowchart of the patient selection process is shown in Figure 1. This study was conducted using data from the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 2.2), a publicly available database containing detailed information on ICU admissions at the Beth Israel Deaconess Medical Center in Boston, MA, from 2008 to 2019. The dataset includes over 76,000 ICU admissions and provides comprehensive data, including demographics, laboratory results, vital signs, treatments, and outcomes. The study was exempt from institutional review board approval as it utilized a publicly available de-identified database, and access was granted after completing the Collaborative Institutional Training Initiative (CITI) program (Record ID: 36067767).\u003c/p\u003e\n\u003cp\u003eFor this study, we included ICU patients who were mechanically ventilated, as they represent a high-risk group for developing pressure ulcers. A total of 29,448 patients were selected, of whom 2,052 developed pressure ulcers. The data was randomly split into a training set (70%, n=20,614) and an internal validation set (30%, n=8,834). The selection criteria included adult patients (age \u0026ge;18 years) who had complete data for the variables of interest. Patients with more than 50% missing data or multiple ICU admissions (only the first admission was considered) were excluded.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Collection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData were collected on various factors potentially related to the development of pressure ulcers, based on existing literature and clinical relevance. These factors included demographic information (age, gender), comorbidities (sepsis, renal disease, myocardial infarction, chronic pulmonary disease), ICU-specific metrics (length of ICU stay, admission type), vital signs, and laboratory values within the first 24 hours of ICU admission (e.g., PaO2/FiO2 ratio, hemoglobin, platelets, albumin). The primary outcome was the development of pressure ulcers during the ICU stay.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eModel Development and Validation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe predictive model was developed using the XGBoost algorithm, a powerful machine learning technique known for its robustness in handling structured data. The model was trained on the training set, with hyperparameters optimized using the\u0026nbsp;caret\u0026nbsp;package in R. The key parameters included a learning rate (eta) of 0.1, a maximum depth of 2, and 100 boosting rounds. The model\u0026apos;s performance was evaluated on the internal validation set using the area under the receiver operating characteristic (ROC) curve (AUC) to assess its discriminatory ability.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSHAP Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo interpret the model and identify the most influential predictors of pressure ulcer development, SHAP (SHapley Additive exPlanations) values were calculated using the\u0026nbsp;shapviz\u0026nbsp;R package. SHAP values provide insights into how each feature contributes to the model\u0026apos;s predictions. The top ten factors influencing pressure ulcer risk were identified, and SHAP summary plots were generated to visualize their impact.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll statistical analyses were performed using R version 4.3.0. Continuous variables were summarized as mean \u0026plusmn; standard deviation or median with interquartile range (IQR), and categorical variables were presented as frequencies and percentages. Comparisons between groups were made using the Chi-square test for categorical variables and independent samples t-test or nonparametric tests for continuous variables. Model performance was evaluated using ROC curves, calibration curves, and decision curve analysis (DCA) to assess the clinical utility of the model. Accuracy, sensitivity, specificity, and other metrics were reported for both the training and validation sets.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003ePatient Characteristics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study cohort consisted of 29,448 mechanically ventilated ICU patients from the MIMIC-IV database, of whom 2,052 (6.97%) developed pressure ulcers. The training set included 20,614 patients, while the validation set included 8,834 patients. The baseline characteristics of patients in the training and validation sets were similar, ensuring a balanced distribution of demographic and clinical factors (Table 1).\u003c/p\u003e\n\u003cp\u003eTable 1. Characteristics of study participants from training set and validation set.\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"659\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eFactors\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eTraining sets\u003c/p\u003e\n \u003cp\u003eN=20,614\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eTesting sets\u003c/p\u003e\n \u003cp\u003eN=8,834\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eP\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eAge (mean, y)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e65.5\u0026plusmn;0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e65.7\u0026plusmn;0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.090\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eFemale (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e8,360(40.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e3,588(40.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.922\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eCOPD (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e5,006(24.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e2,208(25.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.194\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eDiabetes (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e4,657(22.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e1,969(22.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.569\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eMyocardial infarct (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e3,383(16.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e1,452(16.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.003\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eCongestive heart failure (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e5,179(25.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e2,212(25.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.879\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003ePeripheral vascular disease (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e2,257(10.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e985(11.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.613\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eCerebrovascular disease (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e2,839(13.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e1,211(13.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.884\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eDementia (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e614 (3.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e275 (3.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.537\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eRenal disease (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e3,735(18.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e1,575(17.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.553\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eMild liver disease (N,%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e2,202(10.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e905(10.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.263\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eParaplegia (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e835(4.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e370(4.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.585\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eThe presence of metastasis (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e1,045(5.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e476(5.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.257\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eAIDS (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e94(0.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e48(0.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.321\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eCancer (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e2,306\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e1,014\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.468\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eRespiration（PaO2/FIO2）\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e195.2\u0026plusmn;0.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e195.1\u0026plusmn;1.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.516\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eGlasgow (mean)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e14.4\u0026plusmn;0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e14.4\u0026plusmn;0.02\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.668\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eHemoglobin (mean, g/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e9.9\u0026plusmn;0.02\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e10.0\u0026plusmn;0.02\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.434\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003ePlatelets count(10^6/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e183.1\u0026plusmn;0.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e182.5\u0026plusmn;1.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.674\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eWBC (10^9/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e15.2\u0026plusmn;0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e15.5\u0026plusmn;0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.998\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eAlbumin (mean, g/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e3.2\u0026plusmn;0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e3.2\u0026plusmn;0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.809\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eSepsis (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e13,920(67.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e5,948(67.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.742\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eShock (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e2,274(11.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e997 (11.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.524\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eLactate (mean, mmol/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e2.9\u0026plusmn;0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e3.0\u0026plusmn;0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.115\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eLos of ICU (mean, day)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e4.4\u0026plusmn;0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e4.4\u0026plusmn;0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.426\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eBMI (mean, kg/m^2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e23.5\u0026plusmn;0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e23.0\u0026plusmn;0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.806\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eEmergency admission (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e13,515(65.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e5,779(65.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e0.608\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eEthnicity (N, %)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"6\" valign=\"top\" style=\"width: 66px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e0.836\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eAsian\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e541(2.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e231(2.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eWhite\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e13,859(67.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e5,867(66.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eBlack\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e1,644(8.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e682(77.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eOthers\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e2,066(10.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e897(10.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 291px;\"\u003e\n \u003cp\u003eUnknown\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\n \u003cp\u003e2,774(13.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003e1,157(13.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eModel Performance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe XGBoost model demonstrated good performance in predicting the development of pressure ulcers, with an AUC of 0.797 (95% CI: 0.786-0.808) in the training set (Figure 2A) and 0.739 (95% CI: 0.721-0.758) in the validation set (Figure 2B). The calibration curves indicated that the model\u0026apos;s predictions were well-calibrated, with predicted probabilities closely matching observed outcomes both in training set (Figure 2C) and testing set (Figure 2D).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical Utility\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDecision curve analysis (DCA) showed that the XGBoost model provided a net clinical benefit across a wide range of threshold probabilities, supporting its potential utility in clinical practice (Figure 2E). The model\u0026apos;s performance suggests it could be effectively used to identify high-risk patients and guide early interventions to prevent pressure ulcers in mechanically ventilated ICU patients.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSHAP Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSHAP analysis identified the ten most influential factors for predicting pressure ulcer development: \u0026apos;sepsis\u0026apos;, \u0026apos;age\u0026apos;, \u0026apos;the count of platelet\u0026apos;, \u0026apos;length of ICU stay\u0026apos;, \u0026apos;PaO2/FiO2 ratio\u0026apos;, \u0026apos;hemoglobin concentration\u0026apos;, \u0026apos;admission type\u0026apos;, \u0026apos;renal disease\u0026apos;, \u0026apos;albumin concentration\u0026apos;, and \u0026apos;ethnicity\u0026apos;. These factors were ranked based on their average absolute SHAP values, highlighting their relative importance in the model (Figure 3A). \u0026nbsp;To demonstrate how the XGBoost model evaluates the contributions of individual patient features, we utilize SHAP force plots to interpret individual predictions for two patients (Figure 3B, 3C). The color indicates the contribution of each feature, with \u0026nbsp;red indicating that the feature has a negative effect on the prediction \u0026nbsp;(arrow pointing left, SHAP value decreases) and yellow indicating that the \u0026nbsp;feature has a positive effect on the prediction (arrow pointing right,\u0026nbsp; SHAP value increases). The length of the color bar indicates the strength \u0026nbsp;of the contribution, and E[f(x)] indicates the SHAP reference value, \u0026nbsp;which is the mean predicted by the model. For a \u0026ldquo;true positive\u0026rdquo; group of \u0026nbsp;patients, the XGBoost model predicted in-hospital mortality with a SHAP. Figure 4 illustrates the interaction summary plot for the top 10 most significant interacting features. The features have been sorted in descending order of their interaction importance: \u0026apos;sepsis\u0026apos;, \u0026apos;age\u0026apos;, \u0026apos;the count of platelet\u0026apos;, \u0026apos;length of ICU stay\u0026apos;, \u0026apos;PaO2/FiO2 ratio\u0026apos;, \u0026apos;hemoglobin concentration\u0026apos;, \u0026apos;admission type\u0026apos;, \u0026apos;renal disease\u0026apos;, \u0026apos;albumin concentration\u0026apos;, and \u0026apos;ethnicity\u0026apos;. The interaction summary plot reveals that certain feature pairs have a significant combined impact on the model\u0026rsquo;s predictions. For instance, the interaction between sepsis and age, as well as sepsis and los of ICU, displays notable SHAP interaction values, suggesting that these pairs of features interact strongly to influence the model\u0026rsquo;s output.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we developed and validated a predictive model for pressure ulcer development in mechanically ventilated ICU patients using the XGBoost machine learning algorithm. The model demonstrated strong predictive performance with an AUC of 0.797 in the training set and 0.739 in the internal validation set. Our findings underscore the utility of machine learning approaches in identifying high-risk patients and enhancing clinical decision-making processes. One way to consider our model\u0026rsquo;s performance is to place our results alongside the Braden Scale[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. The Braden Scale is the most commonly used tool in North America for predicting risk for pressure injury and measures cumulative risk for pressure injuries via 7 categories: sensory perception, activity, mobility, moisture, nutrition, and friction/shear. Total scores range from 9 (very high risk) to 23 (very low risk)[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Our model\u0026rsquo;s relatively strong performance (area under the ROC curve\u0026thinsp;=\u0026thinsp;0.79 vs 0.68 for the Braden Scale[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]) suggests the model would be a useful way to differentiate among critical care patients in order to apply preventive measures that are not feasible for every patient because of cost, such as specialty beds. Jenny[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e] conducted a systematic review to identify risk factors independently predictive of pressure ulcer development among critical-care patients. They founded that age, mobility/activity, perfusion, and vasopressor infusion frequently emerged as important factors in pressure ulcer development.\u003c/p\u003e \u003cp\u003eSepsis emerged as the most significant predictor, consistent with its systemic impacts on the body. Studies have consistently shown that sepsis leads to widespread inflammation and coagulation abnormalities, which impair microcirculation and tissue perfusion, crucial factors in the pathogenesis of pressure ulcers[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. The hyperinflammatory state associated with sepsis increases the risk of skin breakdown and inhibits wound healing processes, highlighting the need for vigilant monitoring and early intervention in septic patients.\u003c/p\u003e \u003cp\u003eAge is a well-documented risk factor for pressure ulcer development. Older patients often have thinner, less elastic skin, and a decreased ability to redistribute pressure, increasing their vulnerability[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Findings for age is consistent with the results from a systematic review conducted by Coleman and colleagues in an acute, rehabilitative, long-term-care population[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Additionally, polypharmacy, comorbidities, and impaired mobility often present in older adults further exacerbate this risk. This emphasizes the importance of geriatric-specific preventive care strategies in the ICU.\u003c/p\u003e \u003cp\u003eIn this study, both high and low platelet counts contributed to the development of pressure ulcers. Based on the previous study[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e], elevated platelet counts appear to be significantly associated with the development of pressure ulcers in bedridden elderly patients. The research demonstrated that patients with pressure ulcers had higher platelet counts and increased platelet aggregation compared to those without pressure ulcers (p\u0026thinsp;\u0026lt;\u0026thinsp;0.03). This suggests that increased platelet activity could contribute to the pathogenesis of pressure ulcers by exacerbating local microcirculatory impairments and inflammatory responses, which in turn compromises tissue perfusion and integrity. These results corroborate the hypothesis that heightened platelet count and activity may be a critical factor in pressure ulcer formation. Conversely, low platelet counts can signify coagulopathy, impacting the body's ability to repair tissue damage effectively. Thus, platelet levels may serve as a surrogate marker for the inflammatory and hemostatic balance.\u003c/p\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eLength of ICU Stay\u003c/h2\u003e \u003cp\u003eLonger ICU stays are inherently associated with increased pressure ulcer risk due to prolonged immobilization and exposure to a high-intervention environment. Extended bed rest and the critical nature of these patients often result in continuous pressure on bony prominences without sufficient relief or repositioning, underlining the necessity for proactive pressure ulcer prevention protocols in long-term ICU patients. In previous literature, long hospital stay, especially in ICU, is a risk factor for pressure ulcer[\u003cspan additionalcitationids=\"CR14 CR15\" citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003ePaO2/FiO2 Ratio\u003c/h2\u003e \u003cp\u003eThe PaO2/FiO2 ratio, a measure of respiratory function and oxygenation, reflects the patient's ability to deliver oxygen to tissues. Lower ratios indicate inadequate oxygenation, leading to cellular hypoxia, which compromises tissue integrity and healing. Ensuring optimal respiratory function is thus crucial to prevent pressure ulcers in patients requiring mechanical ventilation. In this study, the status of each subgroup in sofa score was included in the prediction of pressure ulcer. Among the top ten important factors, respiratory factors, oxygenation index and renal function were more important in the prediction of pressure ulcer. In previous literature reports, Maarit[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e] assessed the performance of the Sequential Organ Failure Assessment (SOFA) scale and its subcategories in predicting the development of PUs. They found that the Glasgow Coma score, renal and respiratory disorders, and hypotension were significantly (P\u0026thinsp;\u0026lt;\u0026thinsp;.0001) linked to PU development. First-day total SOFA score and its cardiovascular and respiratory subcategory scores were the most important predictors of PUs.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eHemoglobin Levels\u003c/h2\u003e \u003cp\u003eLow hemoglobin levels, or anemia, impair oxygen transport capacity, leading to tissue hypoxia\u0026mdash;a condition detrimental to skin integrity and repair[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Adequate hemoglobin levels are essential to ensure optimal tissue perfusion and the delivery of oxygen necessary for cellular metabolism and repair mechanisms. Monitoring and managing hemoglobin levels can therefore be a critical component of pressure ulcer prevention strategies.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eAdmission Type\u003c/h2\u003e \u003cp\u003eThe type of ICU admission often reflects the patient's baseline health status and urgency of care, which can inherently affect the risk for pressure ulcers. Emergency admissions are typically associated with acute illnesses and limited preparation time for preventive care, unlike elective admissions which allow for planned preoperative optimization. A pressure ulcer can develop in several hours, depending upon risk factors and use of pressure ulcer prevention activities[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. This highlights the need for heightened vigilance and early intervention for patients admitted emergently.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eRenal Disease\u003c/h2\u003e \u003cp\u003eChronic kidney disease and acute renal failure contribute to pressure ulcer risk through mechanisms like fluid overload, metabolic disturbances, and impaired nutritional status, which is in agreement with previous studies[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. The accumulation of uremic toxins and altered protein metabolism can weaken skin and tissue health, making prevention and management particularly challenging in renal-compromised patients.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eAlbumin Levels\u003c/h2\u003e \u003cp\u003eHypoalbuminemia is a marker of malnutrition and an independent risk factor for pressure ulcer development, as low albumin levels can lead to decreased oncotic pressure, edema, and poor tissue repair. Chung[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]conducted a review about risk factors for pressure Injuries in Adult Patients. They found that hypoalbuminemia was identified as possible\u003c/p\u003e \u003cp\u003erisk factor for pressure injury development. Ensuring adequate nutritional support and addressing hypoalbuminemia are therefore critical in preventing pressure ulcers in the ICU setting.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eEthnicity\u003c/h2\u003e \u003cp\u003eWhile ethnicity as a risk factor can point to genetic predispositions for skin conditions or varying susceptibilities, it often reflects broader socio-economic disparities and differences in health care access or delivery[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. Interventions should account for such disparities, ensuring equitable care and prevention measures across different ethnic groups. The study was carried out in Indonesia where the incidence of pressure ulcers has been reported to be as high as 33.3%[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. The incidence of pressure ulcers is higher in Indonesia than in other Asian countries, where the incidence ranges from 2.1\u0026ndash;31.3%[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Other international studies have reported incidence rates of 7\u0026ndash;29% in an intensive care unit (ICU) or acute care setting[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eOur findings, rooted in both the predictive power of advanced machine learning techniques and supported by existing literature, provide valuable insights into the key factors influencing pressure ulcer development in ICU patients. Targeted interventions addressing these specific risk factors could significantly improve patient outcomes by preventing pressure ulcers among high-risk populations in critical care settings. Future studies should focus on external validation and real-world implementation of such predictive models to enhance preventive care strategies further.\u003c/p\u003e \u003cp\u003eLimitations\u003c/p\u003e \u003cp\u003eDespite the strengths of this study, including the large sample size and robust machine learning methodology, several limitations should be noted. First, the study was based on retrospective data from the MIMIC-IV database, which may limit the generalizability of the findings to other populations and healthcare settings. Although our model performed well in internal validation, it was not externally validated using data from a different hospital or patient population. External validation is crucial to confirm the model\u0026rsquo;s applicability in diverse clinical environments. Second, while the MIMIC-IV database provides a wealth of clinical data, the potential for missing or inaccurately recorded data remains a concern. We attempted to mitigate this by excluding patients with more than 50% missing data, but residual confounding may still exist. Additionally, the model did not include some potentially relevant variables, such as patient mobility, skin condition, and specific interventions for pressure ulcer prevention, which were not available in the MIMIC-IV database. Future studies should aim to incorporate these factors to refine the model further.\u003c/p\u003e \u003cp\u003eLastly, the model\u0026rsquo;s performance, while strong, suggests that there is room for improvement. The AUC values, particularly in the validation set, indicate that the model may not capture all the complexity associated with pressure ulcer development. Further refinement of the model, perhaps by incorporating additional data or using more advanced algorithms, could enhance its predictive accuracy.\u003c/p\u003e\n\u003ch3\u003eConclusion\u003c/h3\u003e\n\u003cp\u003eThis study demonstrates that machine learning, specifically the XGBoost algorithm combined with SHAP analysis, can effectively predict the risk of pressure ulcer development in mechanically ventilated ICU patients. The model's strong performance in both the training and validation sets highlights its potential utility in clinical practice. By identifying key risk factors and providing interpretable insights, this model could serve as a valuable tool for guiding early interventions, ultimately improving patient outcomes and reducing the incidence of pressure ulcers in this vulnerable population. Future work should aim to validate the model in diverse settings and explore the integration of additional variables to enhance predictive accuracy.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eEthics approval and consent to participate: The data in this study were from two public de-identified databases. After completing Collaborative Institutional Training Initiative (CITI program), we got permission to access the database (Record ID: 36067767). This study involves human participants but the Ethics Committee(s) or Institutional Board(s) exempted this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgement\u003c/strong\u003e: None.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e: Not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e: The datasets used and/or analysed during the current study are available from corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eContributorship statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(I) Conception and design: ZN Yu, XZ Xing and YJ Xue; (II) Provision of study materials or patients: SN Qu, CL Huang, HJ Wang and H Wang; (III) Collection and assembly of data: YJ Xue and ZN Yuan; (IV) Data analysis and interpretation: ZN Yuan and H Zhang; (V) Final approval of manuscript: All authors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e: None.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSpilsbury, K. et al. Pressure ulcers and their treatment and effects on quality of life: hospital inpatient perspectives. \u003cem\u003eJ. Adv. Nurs.\u003c/em\u003e \u003cb\u003e57\u003c/b\u003e (5), 494\u0026ndash;504 (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFrankel, H., Sperry, J. \u0026amp; Kaplan, L. Risk factors for pressure ulcer development in a best practice surgical intensive care unit. \u003cem\u003eAm. Surg.\u003c/em\u003e \u003cb\u003e73\u003c/b\u003e (12), 1215\u0026ndash;1217 (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSlowikowski, G. C. \u0026amp; Funk, M. Factors associated with pressure ulcers in patients in a surgical intensive care unit. \u003cem\u003eJ. Wound Ostomy Cont. Nurs.\u003c/em\u003e \u003cb\u003e37\u003c/b\u003e (6), 619\u0026ndash;626 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaju, D., Su, X., Patrician, P. A., Loan, L. A. \u0026amp; McCarthy, M. S. Exploring factors associated with pressure ulcers: a data mining approach. \u003cem\u003eInt. J. Nurs. Stud.\u003c/em\u003e \u003cb\u003e52\u003c/b\u003e (1), 102\u0026ndash;111 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAli, S. et al. The enlightening role of explainable artificial intelligence in medical \u0026amp; healthcare domains: A systematic literature review. \u003cem\u003eComput. Biol. Med.\u003c/em\u003e \u003cb\u003e166\u003c/b\u003e, 107555 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBraden, B. \u0026amp; Bergstrom, N. A conceptual schema for the study of the etiology of pressure sores. \u003cem\u003eRehabil Nurs.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e (1), 8\u0026ndash;12 (1987).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHyun, S. et al. Predictive validity of the Braden scale for patients in intensive care units. \u003cem\u003eAm. J. Crit. Care\u003c/em\u003e. \u003cb\u003e22\u003c/b\u003e (6), 514\u0026ndash;520 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlderden, J., Rondinelli, J., Pepper, G., Cummins, M. \u0026amp; Whitney, J. Risk factors for pressure injuries among critical care patients: A systematic review. \u003cem\u003eInt. J. Nurs. Stud.\u003c/em\u003e \u003cb\u003e71\u003c/b\u003e, 97\u0026ndash;114 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSinger, M. et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). \u003cem\u003eJAMA\u003c/em\u003e. \u003cb\u003e315\u003c/b\u003e (8), 801\u0026ndash;810 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaumgarten, M. et al. Extrinsic risk factors for pressure ulcers early in the hospital stay: a nested case-control study. \u003cem\u003eJ. Gerontol. Biol. Sci. Med. Sci.\u003c/em\u003e \u003cb\u003e63\u003c/b\u003e (4), 408\u0026ndash;413 (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eColeman, S. et al. A new pressure ulcer conceptual framework. \u003cem\u003eJ. Adv. Nurs.\u003c/em\u003e \u003cb\u003e70\u003c/b\u003e (10), 2222\u0026ndash;2234 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMatsuyama, N. et al. The effect of anti-platelet aggregation to prevent pressure ulcer development: a retrospective study of 132 elderly patients. \u003cem\u003eGerontology\u003c/em\u003e. \u003cb\u003e46\u003c/b\u003e (6), 311\u0026ndash;317 (2000).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHan, D. et al. Prolonged stay in the emergency department is an independent risk factor for hospital-acquired pressure ulcer. \u003cem\u003eInt. Wound J.\u003c/em\u003e \u003cb\u003e17\u003c/b\u003e (2), 259\u0026ndash;267 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLima Serrano, M., Gonzalez Mendez, M. I., Carrasco Cebollero, F. M. \u0026amp; Lima Rodriguez, J. S. Risk factors for pressure ulcer development in Intensive Care Units: A systematic review. \u003cem\u003eMed. Intensiva\u003c/em\u003e. \u003cb\u003e41\u003c/b\u003e (6), 339\u0026ndash;346 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNaccarato, M. K. \u0026amp; Kelechi, T. Pressure ulcer prevention in the emergency department. \u003cem\u003eAdv. Emerg. Nurs. J.\u003c/em\u003e \u003cb\u003e33\u003c/b\u003e (2), 155\u0026ndash;162 (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAhtiala, M., Soppi, E. \u0026amp; Saari, T. I. Sequential Organ Failure Assessment (SOFA) to Predict Pressure Ulcer Risk in Intensive Care Patients: A Retrospective Cohort Study. \u003cem\u003eOstomy Wound Manage.\u003c/em\u003e \u003cb\u003e64\u003c/b\u003e (10), 32\u0026ndash;38 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlderden, J. et al. Predicting Pressure Injury in Critical Care Patients: A Machine-Learning Model. \u003cem\u003eAm. J. Crit. Care\u003c/em\u003e. \u003cb\u003e27\u003c/b\u003e (6), 461\u0026ndash;468 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSeo, Y., Oh, H., Na, Y., Kim, M. \u0026amp; Seo, W. A Prospective Study of Pressure Injury Healing Rate and Time and Influencing Factors in an Acute Care Setting. \u003cem\u003eAdv. Skin. Wound Care\u003c/em\u003e. \u003cb\u003e35\u003c/b\u003e (12), 1\u0026ndash;9 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChung, M. L. et al. Risk Factors for Pressure Injuries in Adult Patients: A Narrative Synthesis. \u003cem\u003eInt. J. Environ. Res. Public. Health\u003c/em\u003e \u003cb\u003e19\u003c/b\u003e(2). (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLyder, C. H., Shannon, R., Empleo-Frazier, O., McGeHee, D. \u0026amp; White, C. A comprehensive program to prevent pressure ulcers in long-term care: exploring costs and outcomes. \u003cem\u003eOstomy Wound Manage.\u003c/em\u003e \u003cb\u003e48\u003c/b\u003e (4), 52\u0026ndash;62 (2002).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSuriadi, S. H. et al. A new instrument for predicting pressure ulcer risk in an intensive care unit. \u003cem\u003eJ. Tissue Viability\u003c/em\u003e. \u003cb\u003e16\u003c/b\u003e (3), 21\u0026ndash;26 (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJun Seongsook, R. N., Jeong Ihnsook, R. N. \u0026amp; Lee Younghee, R. N. Validity of pressure ulcer risk assessment scales; Cubbin and Jackson, Braden, and Douglas scale. \u003cem\u003eInt. J. Nurs. Stud.\u003c/em\u003e \u003cb\u003e41\u003c/b\u003e (2), 199\u0026ndash;204 (2004).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSuriadi, S. H., Sugama, J., Thigpen, B. \u0026amp; Subuh, M. Development of a new risk assessment scale for predicting pressure ulcers in an intensive care unit. \u003cem\u003eNurs. Crit. Care\u003c/em\u003e. \u003cb\u003e13\u003c/b\u003e (1), 34\u0026ndash;43 (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTheaker, C., Kuper, M. \u0026amp; Soni, N. Pressure ulcer prevention in intensive care - a randomised control trial of two pressure-relieving devices. \u003cem\u003eAnaesthesia\u003c/em\u003e. \u003cb\u003e60\u003c/b\u003e (4), 395\u0026ndash;399 (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWhittington, K. T. \u0026amp; Briones, R. National Prevalence and Incidence Study: 6-year sequential acute care data. \u003cem\u003eAdv. Skin. Wound Care\u003c/em\u003e. \u003cb\u003e17\u003c/b\u003e (9), 490\u0026ndash;494 (2004).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"pressure ulcer, mechanical ventilation, XGBoost","lastPublishedDoi":"10.21203/rs.3.rs-5322280/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5322280/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003ePressure ulcers are significant concern for ICU patients on mechanical ventilation. Early prediction is crucial for enhancing patient outcomes and reducing healthcare costs. This study aims to develop a predictive model using machine learning techniques, specifically XGBoost combined with SHAP, to identify key risk factors of pressure ulcers in this population.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eUtilizing the MIMIC-IV 2.2 database, we included a cohort of 29,448 mechanically ventilated patients in ICU intensive unit. These patients were divided into a training set (20,614 patients, 70%) and an internal validation set (8,834 patients, 30%). Of these, 2,052 patients developed pressure ulcers. We applied the XGBoost algorithm to build the predictive model and used SHAP analysis to identify the top ten factors influencing pressure ulcer development: 'sepsis', 'age', 'the count of platelet', 'length of ICU stay', 'PaO2/FiO2 ratio', 'hemoglobin concentration', 'admission type', 'renal disease', 'albumin concentration', and 'ethnicity'.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe predictive model achieved an area under the ROC curve (AUC) of 0.797 (95% CI: 0.786\u0026ndash;0.808) in the training set and 0.739 (95% CI: 0.721\u0026ndash;0.758) in the validation set. Calibration curves demonstrated good fit, and the decision curve analysis indicated clinical utility.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThis study successfully developed a machine learning model that accurately predicts the risk of pressure ulcers in ICU patients with mechanical ventilation. This model could serve as a valuable tool for guiding early interventions, ultimately reducing the incidence of pressure ulcers in this vulnerable population. The integration of SHAP analysis offers insights into the most critical factors.\u003c/p\u003e","manuscriptTitle":"Explainable SHAP-XGBoost models for pressure ulcers among patients requiring with mechanical ventilation in intensive care unit","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-12-17 05:59:30","doi":"10.21203/rs.3.rs-5322280/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-01-28T06:29:04+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-01-23T13:48:48+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"160747306736971524753501282608555706586","date":"2025-01-16T13:33:09+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-12-26T11:37:19+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"120807437612021911672867173279831738732","date":"2024-12-18T17:28:18+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-11-24T21:39:45+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"20036582552923623966191719530731313698","date":"2024-11-19T20:53:05+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-11-19T10:34:35+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-11-19T09:45:02+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2024-11-12T11:26:24+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-11-11T09:43:28+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2024-10-24T03:00:26+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"67a31559-c662-4ad5-b126-b456e97a892c","owner":[],"postedDate":"December 17th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":40681948,"name":"Health sciences/Health care"},{"id":40681949,"name":"Health sciences/Risk factors"}],"tags":[],"updatedAt":"2025-03-24T16:02:08+00:00","versionOfRecord":{"articleIdentity":"rs-5322280","link":"https://doi.org/10.1038/s41598-025-92848-2","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2025-03-22 15:57:48","publishedOnDateReadable":"March 22nd, 2025"},"versionCreatedAt":"2024-12-17 05:59:30","video":"","vorDoi":"10.1038/s41598-025-92848-2","vorDoiUrl":"https://doi.org/10.1038/s41598-025-92848-2","workflowStages":[]},"version":"v1","identity":"rs-5322280","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5322280","identity":"rs-5322280","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}