A Dynamic Online Nomogram for Predicting Postoperative Acute Kidney Injury After Sleeve Gastrectomy

Introduction

Acute kidney injury (AKI) is a serious and increasingly recognized postoperative complication that contributes to higher mortality, longer hospitalization, and substantial healthcare costs following major noncardiac surgery [1]. Sleeve gastrectomy (SG) has become the predominant procedure for the treatment of morbid obesity worldwide. Although SG is generally safe, postoperative AKI remains underdiagnosed and clinically underestimated, with reported incidence rates of approximately 2% to 3% [2–4]. Given the rapid global expansion of SG and the long-term renal and cardiovascular consequences of AKI, early identification of high-risk patients is crucial for improving perioperative outcomes.

A growing body of evidence indicates that intraoperative hemodynamic instability is a key modifiable determinant of postoperative AKI [5]. Beyond absolute hypotension, several research groups have demonstrated that short-term blood pressure variability and hemodynamic complexity are independently linked to AKI. For example, Folks et al showed that higher short-term blood pressure variability quantified by sample entropy was associated with an increased risk of postoperative AKI in noncardiac surgery patients [5]. In patients with obesity, obesity-related alterations in renal autoregulation may further amplify vulnerability to perfusion instability [6,7]. Similarly, studies in pediatric cardiac surgery and major vascular surgery have reported that greater intraoperative blood pressure variability, captured by coefficients of variation and time-weighted deviations from target blood pressure ranges, is independently associated with postoperative AKI [8,9]. Moreover, a recent deep-learning model based on real-time intraoperative vital sign signals further underscored the central role of hemodynamic patterns in predicting postoperative AKI [10]. Taken together, these findings support the notion that dynamic blood pressure patterns, rather than isolated hypotensive episodes, are critical determinants of renal outcomes. They also provide a strong rationale for focusing on intraoperative systolic blood pressure variability as a novel predictor in the present study.

Recent AKI prediction studies have highlighted important limitations of conventional risk-stratification approaches, which rely predominantly on static preoperative variables and inadequately capture dynamic perioperative physiology. For example, Chen et al demonstrated that persistent postoperative AKI after cardiac surgery could be more accurately predicted by incorporating physiologic signal characteristics and applying rigorous model-development procedures [11]. Similar advances in real-time perioperative analytics show that physiologic complexity, rather than isolated hemodynamic values, provides superior discriminative ability for postoperative outcomes. Collectively, these findings reinforce the biologic plausibility and methodological relevance of evaluating intraoperative systolic blood pressure variability as a mechanistically informed predictor of AKI.

In addition to hemodynamic factors, SG itself can induce metabolic alterations that influence renal function. Because patients undergoing SG commonly exhibit obesity-related impairments in renal autoregulation and metabolic status, they may be particularly vulnerable to additional renal stressors during the perioperative period. Recent clinical data indicate that bariatric surgery, including SG, is associated with improvement in renal function (eGFR, albuminuria) and can even expand eligibility for kidney transplantation in patients with chronic kidney disease [12]. However, although SG improves metabolic parameters, it can also lead to postoperative nutritional and metabolic complications, such as elevated homocysteine levels and deficiencies in vitamin B12 and folate, which can contribute to renal or vascular dysfunction [13]. Therefore, understanding the potential renal implications of SG is clinically important to optimize perioperative management and prevent adverse metabolic sequelae.

Despite progress in identifying preoperative and demographic predictors of postoperative AKI [14,15], important knowledge gaps remain. Most existing models rely on static variables and fail to account for dynamic intraoperative physiology, particularly real-time SBP fluctuations that reflect ongoing changes in renal perfusion [16]. In addition, currently available AKI prediction tools are typically non-interactive, paper-based, or not optimized for perioperative clinical workflows, which limits their usefulness for timely decision-making. These limitations highlight the need for a dynamic, mechanism-informed, and clinically accessible prediction model specifically tailored to the SG population.

Nomograms offer an effective means of integrating multiple independent predictors into a user-friendly visual tool and are increasingly used in surgical and oncologic research to support individualized risk assessment [17–19]. When implemented as an online platform, they allow clinicians to obtain real-time, patient-specific risk estimates that can guide perioperative renal-protective interventions [20].

Growing evidence from critically ill and surgical populations suggests that hemodynamic coherence and multidimensional hemodynamic patterns, rather than single blood pressure thresholds, better reflect organ-specific perfusion and AKI risk [21]. However, no prior study has specifically evaluated intraoperative systolic blood pressure variability in the context of SG, where obesity-related alterations in renal autoregulation and metabolic status can further amplify susceptibility to perfusion instability.

Therefore, the objective of this study was to develop and internally validate a dynamic, web-based nomogram that integrates intraoperative systolic blood pressure variability and key perioperative predictors to estimate the risk of postoperative AKI in patients undergoing SG, thereby addressing key limitations of existing static prediction tools and supporting earlier, individualized perioperative risk stratification.

Material and Methods

STUDY DESIGN AND SETTING:

This study was designed as a retrospective observational cohort study conducted at the Second Affiliated Hospital of Anhui Medical University, a tertiary teaching hospital in Hefei, Anhui Province, China. Consecutive adult patients who underwent laparoscopic sleeve gastrectomy between January 1, 2016, and March 31, 2025, were eligible for screening. No data were prospectively collected for research purposes, and no variables were modified, reclassified, or filtered during data extraction. Any rescaling or selection of predictors was performed solely at the statistical modeling stage, as described below.

All perioperative data were retrospectively extracted from the hospital’s integrated electronic information systems, including the Electronic Medical Record (EMR), Anesthesia Information Management System (AIMS), Laboratory Information System (LIS), and the Integrated Perioperative Data Platform (IPDP). These systems routinely capture high-resolution perioperative physiological data, laboratory results, and clinical documentation as part of standard care. The retrospective design enabled analysis of a large real-world cohort without altering patient management.

A retrospective design was chosen to leverage this existing perioperative database, enabling inclusion of a large real-world cohort without the ethical and logistical challenges associated with prospective blood pressure manipulation. Formal a priori sample size or power calculations were not performed, as all eligible patients within the predefined time window were included. The final sample of 1386 patients provided adequate statistical power for multivariable logistic regression, exceeding the commonly recommended rule of at least 10 outcome events per predictor variable.

STUDY POPULATION:

Electronic medical records of patients who underwent laparoscopic sleeve gastrectomy between January 1, 2016, and March 31, 2025, were screened. A total of 1708 patients were initially identified.

Inclusion criteria were:

Exclusion criteria were:

After applying these criteria, 152 patients with missing intraoperative SBP data and 170 patients with >60% missing preoperative laboratory indicators were excluded, leaving 1386 patients for the final analysis. The inclusion criteria ensured that each patient had reliable baseline renal function data and complete intraoperative SBP recordings, which were essential for quantifying hemodynamic variability. Excluding patients with substantial missing laboratory or blood pressure data minimized measurement bias and improved the accuracy of risk estimates. Sensitivity checks showed similar baseline characteristics between included and excluded patients, suggesting minimal selection bias. The detailed screening and selection process is presented in Figure 1.

PERIOPERATIVE VARIABLES AND DATA SOURCES:

Information on potential covariates was extracted from the hospital’s Integrated Perioperative Data Platform (IPDP), which consolidates structured and unstructured data from the Electronic Medical Record (EMR), Anesthesia Information Management System (AIMS), Laboratory Information System (LIS), and perioperative monitoring systems. The IPDP is built on a standardized browser/server architecture, uses unified variable definitions, and applies consistent extraction rules across calendar years, ensuring stable data provenance and harmonization over the entire 2016–2025 study period.

Laboratory test data obtained within 30 days prior to surgery were used as baseline values for serum creatinine, eGFR, blood urea nitrogen (BUN), albumin, hemoglobin, glycated hemoglobin (HbA1c), fasting blood glucose, white blood cell count, neutrophil count, platelet count, C-reactive protein (CRP), urine specific gravity, urinary protein, and red blood cells in urine. All laboratory values were extracted directly from the LIS as the exact numeric outputs reported by analyzers. No averaging, truncation, rounding, normalization, smoothing, or range compression was applied during extraction or analysis. The IPDP performs only basic validation procedures – such as unit harmonization, timestamp checks, and plausibility screening – without modifying the underlying distributions of laboratory values. Clinically plausible borderline values are retained, and all values remain fully traceable to their original LIS entries.

Intraoperative data included anesthesia duration, use of vasopressors (phenylephrine, norepinephrine, dopamine, or dobutamine), blood loss, total infusion volume, volumes of crystalloid and plasma substitutes, and intraoperative urine output. At our center, insertion of a urinary catheter during laparoscopic sleeve gastrectomy is not routine and is reserved for selected patients. Consequently, most cases have no measurable intraoperative urine output. In the AIMS, when no catheter is placed, the urine-output field is automatically recorded as “0”, reflecting the absence of measurement rather than true physiologic anuria; low non-zero volumes are not rounded or truncated. Because non-zero urine-output values originate from a small subset of catheterized patients, the overall distribution is highly zero-inflated and does not represent a continuous physiological variable across the cohort. Therefore, intraoperative urine output was not included as a candidate predictor in the final multivariable model or nomogram.

Perioperative hemodynamic parameters – mean intraoperative SBP (mean_SBP), standard deviation of SBP (SD_SBP), and coefficient of variation of SBP (CV_SBP) – were also extracted and used as candidate predictors.

All covariates were selected based on clinical relevance and prior literature supporting their potential association with postoperative AKI.

OUTCOME DEFINITION (AKI):

The primary outcome was postoperative AKI, defined and staged according to the Kidney Disease: Improving Global Outcomes (KDIGO) serum creatinine criteria. AKI was diagnosed if either of the following occurred:

At our institution, postoperative serum creatinine is routinely measured on postoperative day (POD) 1 and POD 2 as part of the enhanced recovery after surgery (ERAS) pathway, with additional measurements obtained when clinically indicated. For AKI adjudication, the highest serum creatinine value recorded within the KDIGO-defined 48-h and 7-day windows was used. All patients in the final cohort had at least one postoperative creatinine measurement within 48 h; patients without postoperative creatinine testing were excluded according to prespecified criteria.

Baseline serum creatinine was defined as the most recent value measured within 30 days before surgery and was extracted directly from the LIS. This reflects standardized preoperative assessment practice at our hospital. All included patients had an available baseline value; therefore, no imputation or surrogate estimation of baseline creatinine was required.

Because urinary catheters are not routinely placed during sleeve gastrectomy, postoperative urine-output data were systematically unavailable. Consequently, only creatinine-based KDIGO criteria were used. As reported in prior noncardiac surgical cohorts, omission of urine-output criteria may slightly underestimate AKI incidence; this limitation is acknowledged in the Discussion section.

INTRAOPERATIVE BLOOD PRESSURE MONITORING AND VARIABILITY METRICS:

Invasive intraoperative SBP data were obtained from the AIMS, which automatically records arterial-line measurements at fixed 5-min intervals following standardized institutional calibration procedures. Before calculating variability metrics, we applied a predefined preprocessing pipeline:

Temporal consistency over the 2016–2025 period was verified by confirming that AIMS firmware updates did not alter sampling intervals, data structure, or metadata standards. These standardized procedures ensured high-quality, reproducible SBP time-series data for assessing intraoperative blood pressure variability.

Two variability metrics were calculated from all intraoperative SBP recordings for each patient:

CV_SBP was calculated as:

This standardized index has been widely applied in perioperative studies as a reliable measure of hemodynamic variability. SD_SBP and CV_SBP were initially considered as continuous candidate predictors and were entered into univariate and multivariable logistic regression analyses to evaluate their association with postoperative AKI. CV_SBP, prespecified as the primary hemodynamic marker of interest, was retained as a key predictor in the final model.

DATA QUALITY AND MISSING DATA:

Prior to analysis, the extent of missing data was assessed for all baseline and perioperative variables. Most demographic, comorbidity, and intraoperative variables had minimal missing data (<5%). Missingness primarily resulted from preoperative laboratory tests performed outside the institutional laboratory information system or outside the predefined extraction window.

Selected anthropometric and laboratory variables (eg, BMI and C-reactive protein) exhibited higher proportions of missing values. Variables with more than 60% missing data were excluded from baseline comparison and subsequent analyses, in accordance with institutional data quality recommendations for perioperative research. For the remaining variables, analyses were conducted using available-case data without imputation. Baseline characteristics are therefore reported based on the number of patients with available data for each variable, as detailed in Table 1.

DATA COLLECTION AND MANAGEMENT:

Data extraction was performed by trained members of the research team (L.C.X.) using standardized electronic extraction procedures implemented within the Integrated Perioperative Data Platform. All variables were retrieved directly from their source systems (EMR, AIMS, and LIS) using predefined variable definitions and uniform extraction rules. No manual case report forms were used, and no data were modified or reclassified during extraction.

Because this was a retrospective study based on routinely collected clinical data, investigators responsible for data extraction were not involved in outcome adjudication. Postoperative AKI was defined according to prespecified KDIGO criteria using laboratory values obtained from the LIS, ensuring objective outcome determination. These procedures minimized the potential for observer or information bias. Because outcome assessment was based solely on laboratory-defined KDIGO criteria automatically retrieved from the LIS, data collectors were effectively blinded to outcomes during extraction.

STATISTICAL ANALYSIS:

All statistical analyses were conducted using R version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria). The rms (version 6.2-0), DynNom (version 5.0), and pROC (version 1.18.0) packages in R were used for model construction, nomogram development, and performance evaluation. A two-sided P value <0.05 was considered statistically significant. Effect estimates are reported as odds ratios (ORs) with 95% confidence intervals (CIs).

Continuous variables were tested for normality using the Shapiro-Wilk test. Because most variables were non-normally distributed, continuous data are presented as medians and interquartile ranges (IQRs) and were compared between groups using the Mann-Whitney U test. Categorical variables are summarized as counts and percentages and were compared using the chi-square test or Fisher’s exact test, as appropriate. Standardized mean differences (SMDs) were additionally calculated to quantify between-group imbalance. All baseline comparisons presented in Table 1 were performed using available-case data.

RESCALING OF CONTINUOUS PREDICTORS:

To improve model interpretability and numerical stability, selected continuous predictors were rescaled prior to multivariable modeling. Estimated glomerular filtration rate (eGFR) was modeled per 10 mL/min/1.73 m² increase, hemoglobin per 10 g/L increase, intraoperative systolic blood pressure variability (CV_SBP) per 10% increase, and total intraoperative infusion volume per 100 mL increase. Accordingly, odds ratios represent the change in AKI risk associated with these clinically meaningful increments.

A nomogram was constructed based on the final multivariable logistic regression model to provide individualized estimates of postoperative AKI risk. Model discrimination was assessed using the area under the receiver operating characteristic curve (AUC). Calibration was evaluated using calibration plots and the Hosmer-Lemeshow goodness-of-fit test. Internal validation was performed using 1000 bootstrap resamples to obtain optimism-corrected performance estimates.

Clinical utility of the final prediction model was evaluated using decision curve analysis (DCA). Net benefit was calculated across a range of threshold probabilities and compared with default strategies of treating all patients or treating none. Because urine output was not included in AKI ascertainment due to incomplete perioperative recording, sensitivity analyses based on urine-output-related criteria were not applicable.

ETHICS APPROVAL AND CONSENT STATEMENT:

This study was designed as a retrospective cohort study and was conducted in accordance with the Declaration of Helsinki and relevant national and institutional regulations. The protocol was reviewed and approved by the Medical Ethics Committee of the Second Affiliated Hospital of Anhui Medical University (Approval ID: YX2025-199; date of approval: July 23, 2025). The approved protocol explicitly covered retrospective chart review and extraction of perioperative data for adult patients who underwent sleeve gastrectomy between January 2016 and March 2025.

In accordance with the institutional policy for minimal-risk retrospective research, the Ethics Committee confirmed that the use of de-identified clinical data collected before the approval date did not require written informed consent from individual patients. All data extraction, preprocessing, and statistical analyses were conducted after ethics approval had been granted, and no changes to patient management occurred as part of this research. All patient information was fully anonymized prior to analysis.

Results

PATIENT CHARACTERISTICS:

A total of 1386 patients were included in the final analysis, of whom 78 (5.6%) developed postoperative acute kidney injury (AKI). The median age of the cohort was 31 years (interquartile range [IQR], 26–36), and 72.6% of patients were female.

Baseline characteristics stratified by AKI status are summarized in Table 1. Compared with patients without AKI, those who developed AKI were more frequently male (38.5% vs 26.8%, P=0.034), had a higher prevalence of hypertension (19.2% vs 10.0%, P = 0.017), and were more often classified as American Society of Anesthesiologists (ASA) physical status III–IV (29.5% vs 17.9%, P=0.016). AKI patients also exhibited lower preoperative serum creatinine levels and higher baseline estimated glomerular filtration rate (eGFR).

Intraoperatively, patients who developed AKI received lower total infusion volumes and exhibited greater systolic blood pressure variability, including higher coefficients of variation of systolic blood pressure (CV_SBP). No significant between-group differences were observed for age, body mass index, diabetes mellitus, alcohol use, serum sodium, potassium, blood urea nitrogen, hemoglobin A1c, C-reactive protein, or most other routine laboratory parameters. Overall, baseline characteristics reflect a consecutive real-world cohort without matching or outcome-driven exclusion.

UNIVARIATE AND MULTIVARIATE ANALYSES:

In univariate logistic regression analyses, higher ASA grade, higher baseline eGFR, higher hemoglobin level, greater intraoperative systolic blood pressure variability (CV_SBP), and lower total intraoperative infusion volume were significantly associated with postoperative AKI. Variables demonstrating statistical significance in univariate analyses, together with clinically relevant predictors, were considered for multivariable modeling.

The final multivariable logistic regression model identified 5 independent predictors of postoperative AKI (Table 2). Higher ASA grade (III–IV vs I–II) was associated with an increased risk of AKI (OR=1.93, 95% CI: 1.13–3.22, P=0.014). Higher baseline eGFR (per 10 mL/min/1.73 m2 increase) was independently associated with AKI (OR=1.65, 95% CI: 1.33–2.04, P<0.001), as was higher hemoglobin level (per 10 g/L increase; OR=1.28, 95% CI: 1.09–1.50, P=0.002). Greater intraoperative systolic blood pressure variability, expressed as CV_SBP (per 10% increase), was also associated with increased AKI risk (OR=1.76, 95% CI: 1.20–2.47, P=0.002). In contrast, higher total intraoperative infusion volume (per 100 mL increase) was associated with a lower risk of postoperative AKI (OR=0.90, 95% CI: 0.84–0.96, P=0.003).

NOMOGRAM DEVELOPMENT AND MODEL PERFORMANCE:

A nomogram incorporating the 5 independent predictors identified in the final multivariable model was constructed to estimate individualized AKI risk (Figure 2). Each predictor contributed proportionally to the total risk score, which corresponded to an individualized predicted probability of postoperative AKI. A dynamic web-based version of the nomogram was developed to facilitate clinical use. This online tool is accessible at https://rexample-lcx.shinyapps.io/AKI_Predicted_Model3/ and may assist anesthesiologists and surgeons in perioperative decision-making and early risk stratification.

The model demonstrated good discriminatory performance, with an area under the receiver operating characteristic curve (AUC) of 0.716 (95% CI: 0.658–0.774) (Figure 3). Calibration analysis showed close agreement between predicted and observed risks, with a mean absolute error of 0.006 and a 90th percentile error of 0.009 (Figure 4). The Hosmer-Lemeshow goodness-of-fit test indicated no significant lack of fit (P=0.991).

Decision curve analysis demonstrated that the nomogram provided meaningful net clinical benefit across a range of threshold probabilities from 0.05 to 0.30, outperforming both “treat all” and “treat none” strategies (Figure 5).

Discussion

This study identified 5 independent predictors – ASA grade, baseline eGFR, hemoglobin level, intraoperative systolic blood pressure variability (CV_SBP), and total intraoperative infusion volume—that were significantly associated with the risk of postoperative AKI in patients undergoing sleeve gastrectomy. Notably, CV_SBP emerged as a novel intraoperative predictor, emphasizing the relevance of dynamic hemodynamic fluctuations rather than isolated blood pressure values. These findings are consistent with prior evidence suggesting the importance of maintaining intraoperative hemodynamic stability to preserve renal function in vulnerable populations [22]. The nomogram constructed from these variables demonstrated reasonable discriminative power (AUC=0.716), calibration, and clinical utility as validated by decision curve analysis, providing a practical and interpretable tool for individualized risk estimation. This dynamic, web-based model facilitates early identification of high-risk individuals, supports clinical decision-making, and has the potential to improve perioperative renal outcomes in bariatric surgical patients.

In our study, patients with higher ASA grades were more likely to develop postoperative AKI, in line with prior findings that comorbidities and impaired baseline functional status significantly influence renal outcomes [23,24]. Similarly, individuals classified as ASA grade III or higher were at increased risk, likely reflecting the cumulative burden of comorbidities and diminished physiological reserve [25,26]. Interestingly, higher baseline eGFR was independently associated with an increased risk of postoperative AKI in our cohort (odds ratio [OR] per 10 mL/min/1.73 m2 increase, 1.65; 95% confidence interval [CI], 1.33–2.04; P<0.001), which appears to contrast with conventional studies linking reduced renal function to AKI [27]. This discrepancy may partly reflect the use of creatinine-based KDIGO criteria, whereby patients with very low baseline creatinine are more likely to exceed diagnostic thresholds following small absolute postoperative increases. In addition, in obese patients undergoing sleeve gastrectomy, elevated eGFR may represent obesity-related glomerular hyperfiltration rather than superior renal reserve, potentially increasing vulnerability to intraoperative perfusion instability [28,29]. Importantly, the concurrent identification of CV_SBP as an independent predictor strengthens the interpretation that elevated eGFR in this cohort reflects increased susceptibility to perfusion instability rather than a statistical artifact alone. In contrast to studies in general surgical populations where anemia has been associated with AKI [30], higher preoperative hemoglobin was independently associated with increased AKI risk in our cohort (OR per 10 g/L, 1.28; 95% CI, 1.09–1.50; P=0.0023). This finding likely reflects population-specific mechanisms in obese patients undergoing sleeve gastrectomy, in whom higher hemoglobin may indicate hemoconcentration, increased blood viscosity, or relative intravascular volume depletion rather than enhanced oxygen-carrying capacity [31]. These factors may impair renal microcirculatory perfusion, particularly in the presence of intraoperative blood pressure variability, thereby increasing susceptibility to AKI [32]. In the context of sleeve gastrectomy, preoperative fasting, pneumoperitoneum, and reverse Trendelenburg positioning may further exacerbate the adverse microcirculatory effects of hemoconcentration. Higher intraoperative infusion volume was independently associated with a lower risk of postoperative AKI (OR per 100 mL, 0.90; 95% CI, 0.84–0.96; P=0.0028), suggesting that relative hypovolemia may be an important contributor to AKI in this sleeve gastrectomy population. This finding is clinically plausible given the perioperative propensity for dehydration and reduced effective circulating volume in bariatric patients, particularly under pneumoperitoneum and limited early oral intake [33]. Nevertheless, prior studies have reported both increased AKI with restrictive fluid strategies and increased AKI with excessive positive fluid balance, indicating a potential U-shaped relationship between fluid administration and renal outcomes [34]. Differences across studies may be driven by the exposure definition (total volume vs net balance), patient risk profiles, fluid regimens, and the distribution of administered volumes within each cohort. Our results reaffirmed intraoperative systolic blood pressure variability (CV_SBP) as an independent predictor of AKI. Greater intraoperative systolic blood pressure variability can impair renal autoregulation by inducing repeated episodes of perfusion instability [35–37]. This interpretation is supported by microcirculatory studies demonstrating that macro-hemodynamic parameters such as arterial pressure do not reliably reflect downstream organ perfusion, a phenomenon known as loss of hemodynamic coherence [38]. In major abdominal and high-risk surgery, transient dissociation between arterial pressure and microcirculatory flow has been strongly linked to postoperative organ injury, particularly AKI [32]. These findings suggest that dynamic blood pressure variability can serve as a pragmatic surrogate for unmeasured hemodynamic complexity: large beat-to-beat oscillations can indicate instability in real-time perfusion delivery even when mean arterial pressure remains within clinically acceptable ranges. The identification of CV_SBP as an independent predictor in our model therefore aligns with emerging physiological evidence emphasizing that renal vulnerability during surgery is driven not only by hypotension but also by loss of hemodynamic coherence and impaired microvascular autoregulation.

These associations underscore the importance of personalized perioperative strategies based on preoperative risk stratification and continuous intraoperative monitoring. Leveraging real-time hemodynamic assessment through anesthesia information management systems can enable timely adjustments in vasopressor use or fluid optimization, potentially mitigating renal injury. Practically, elevated intraoperative CV_SBP may prompt avoidance of rapid vasopressor titration, pursuit of smoother blood pressure control, and early goal-directed optimization of effective circulating volume rather than reactive correction of isolated hypotensive events. For patients identified as high-risk, targeted interventions including early nephrology consultation, renal-protective goal-directed fluid therapy, and minimization of nephrotoxic exposures may be beneficial.

Most existing studies on postoperative AKI have used static risk assessment approaches based on cross-sectional or single-timepoint indicators, limiting their ability to capture dynamic intraoperative physiology [2,4]. In contrast, our model incorporates intraoperative systolic blood pressure variability (CV_SBP) as a dynamic physiological variable, supporting real-time risk stratification and offering added value beyond traditional predictors. Importantly, the model is implemented as a web-based dynamic nomogram, enabling clinicians to obtain individualized AKI risk estimates instantly using routinely available perioperative data, with the potential to facilitate multidisciplinary communication and shared decision-making [39,40]. With further validation in larger and more diverse cohorts, this tool could be integrated into electronic health record systems and perioperative decision-support workflows, thereby bridging the gap between predictive modeling and bedside clinical practice and advancing precision perioperative care in bariatric surgical populations [41].

Several limitations merit consideration. First, this was a single-center study with relatively homogeneous perioperative pathways, which may limit generalizability to institutions with different patient populations or hemodynamic management strategies. Second, as detailed in the Methods section, intraoperative urine output was not used as a predictor because most patients were not catheterized and non-zero recordings were sparse, limiting its interpretability. Third, although arterial-line systolic blood pressure was acquired through standardized AIMS calibration procedures and signal-quality checks, retrospective extraction and artifact filtering can introduce minor measurement or documentation bias. Fourth, although several internal robustness checks were performed, external validation and additional sensitivity analyses in independent cohorts were not conducted and should be considered in future studies. Fifth, the study relied exclusively on creatinine-based KDIGO criteria and did not incorporate postoperative biomarkers or long-term renal outcomes, limiting insight into persistent or chronic kidney sequelae. Future research should include prospective multicenter validation, integration of biomarker and longitudinal renal data, and assessment of model performance within diverse perioperative decision-support workflows to strengthen clinical applicability and generalizability.

Conclusions

This study developed and validated a dynamic nomogram that effectively predicts the risk of postoperative AKI in patients undergoing sleeve gastrectomy. By integrating intraoperative systolic blood pressure variability as a key predictor, an element rarely included in previous AKI models, this work provides a novel, mechanism-informed approach to perioperative renal risk assessment. Given the strong associations observed with intraoperative systolic blood pressure variability and preoperative renal function, early identification of high-risk patients is essential to optimize perioperative management. This web-based tool offers a practical means to support individualized decision-making and may help guide renal-protective interventions during the perioperative period. The primary take-home message is that dynamic intraoperative physiology, particularly SBP variability, should be incorporated into perioperative risk stratification to improve early recognition of patients at risk for AKI. Integrating this model into clinical decision-support systems may further enhance real-time risk assessment and promote timely intervention. Future research should focus on multicenter prospective validation, assessment of long-term renal outcomes, and evaluation of the model’s performance within diverse clinical workflows. Such efforts will help determine the generalizability, scalability, and long-term impact of this prediction tool on perioperative renal outcomes.

Data Availability

The datasets generated and analyzed during the current study are not publicly available due to institutional privacy regulations but are available from the corresponding author on reasonable request.