A nomogram to predict hypoxemia in the post-anesthesia care unit after laparoscopic bariatric surgery

Abstract

Background

Hypoxemia in the post-anesthesia care unit (PACU) is a common and distressing complication following laparoscopic bariatric surgery (LBS). Despite its clinical significance, there remains a lack of effective, integrated predictive models to guide the prevention and management of hypoxemia in the PACU for this patient population. 

Methods

 In this retrospective single-center study, a total of 1,220 cases were included. The database was randomly divided into training and validation cohorts at a 7:3 ratio. Using least absolute shrinkage and selection operator (LASSO) regression combined with multivariable logistic regression, we identified independent predictors and constructed a nomogram for hypoxemia prediction. The nomogram’s performance was evaluated and validated through assessment of the area under the receiver operating characteristic curve (AUC), calculation of the concordance index (C-index), generation of calibration plots, and performance of decision curve analysis (DCA).

 Results

The study cohort of 1,220 patients was randomly divided into training (n = 854) and validation (n = 366) cohorts. A nomogram prediction model was constructed using five independent predictors: age, BMI, presence of two or more comorbidities, intraoperative opioid use ≥ 60 mg, and use of a neuromuscular blocking (NMB) reversal agent. In the training cohort, the AUC and the optimism-corrected C-index were 0.738 (95% CI 0.701–0.775) and 0.731, while in the validation cohort they were 0.735 (95% CI 0.682–0.789) and 0.715, respectively. Calibration plots demonstrated good agreement between predicted probabilities and actual observations. The DCA results confirmed the clinical utility of the nomogram across a range of threshold probabilities.

Conclusions

We developed and validated a nomogram-based prediction model for postoperative hypoxemia in the PACU among patients undergoing elective LBS. This model integrates five key clinical variables and demonstrates acceptable discriminative ability and good calibration, offering a valuable tool for individualized risk assessment and perioperative management.

Introduction

Obesity has emerged as one of the most pressing global health challenges, with its prevalence escalating dramatically over the past few decades [1]. According to recent estimates, over 890 million adults worldwide are classified as obese, making it a leading risk factor for a multitude of chronic diseases [2]. Laparoscopic bariatric surgery (LBS) has become a safe and effective treatment for obesity, but the incidence of pulmonary complications after surgery is as high as 20% [3]– [4]. Patients with obesity, particularly those with comorbidities such as obstructive sleep apnea (OSA), are at a heightened risk of postoperative hypoxemia, especially in the post-anesthesia care unit (PACU) after LBS [5]. Hypoxemia in the PACU is associated with adverse outcomes, including prolonged hospital stays, increased morbidity, and even mortality.

Despite the growing incidence of postoperative hypoxemia in obese patients, there is a lack of effective predictive tools tailored to this specific population. Existing models for predicting postoperative respiratory complications often have limitations, such as insufficient validation, limited applicability, or failure to account for key variables related to dynamic perioperative variables. For instance, the PRODIGY score, which predicts opioid-induced respiratory depression, does not fully address the unique challenges faced by bariatric surgery patients [6].

To address this gap, the development of a nomogram, a visual and user-friendly predictive tool, offers a promising solution. Nomograms integrate multiple risk factors into a single model, providing clinicians with a practical means to identify high-risk patients early and implement targeted interventions. This study aims to develop and validate a nomogram specifically designed to predict hypoxemia in the PACU following LBS, thereby enhancing patient safety and postoperative care outcomes.

Methods

Study design and participants

This single-center, retrospective study was conducted at an accredited bariatric center within a tertiary hospital in China. The study protocol received ethical approval from the Institutional Review Board, with a waiver of informed consent (approval No. 2023-SR-986). The study cohort comprised patients aged 18–60 years who underwent LBS under general anesthesia between January 2024 and December 2024 and were subsequently transferred to the PACU for anesthesia recovery. The study flow diagram is presented in Fig. 1.

The inclusion criteria were: (1) patients aged 18–60 years undergoing LBS under general anesthesia, and (2) patients admitted to the PACU for postoperative anesthesia recovery. The following exclusion criteria were applied: (1) unplanned conversion to laparotomy during surgery, (2) concurrent procedures involving other anatomical regions, (3) unplanned admission to the intensive care unit (ICU) postoperatively, (4) incomplete or missing data related to surgery or anesthesia.

Definition of hypoxemia

In this study, SpO₂ ≤ 90% was defined as the standard for hypoxemia for statistical data analysis [7]– [8]. During anesthesia recovery in the PACU, patients were categorized into the hypoxemia group if they exhibited SpO₂ ≤ 90% for a continuous duration of ≥ 5 min while receiving standard nasal catheter oxygen supplementation (flow rate of 3–4 L/min).

Anesthesia management

For anesthesia induction, patients received a combination of propofol, sufentanil, midazolam, and either rocuronium or cisatracurium. Anesthesia was maintained using a balanced approach with propofol, sevoflurane, remifentanil, and neuromuscular blocking agents (rocuronium or cisatracurium), with intermittent supplementation of sufentanil titrated to clinical requirements. Dexmedetomidine was administered as needed for adjunctive sedation.

Postoperatively, patients were transferred directly to the PACU with endotracheal intubation maintained until recovery criteria were met. Postoperative monitoring followed similar principles as intraoperative monitoring, with the continuous assessment by the same anesthesiology team. Vital parameters monitored included oxygenation, level of consciousness, respiratory function, hemodynamic stability, and pain intensity. To optimize recovery, rescue analgesics, propofol, remifentanil infusions, vasoactive agents, and reversal agents (flumazenil, neostigmine, or sugammadex) were administered as clinically indicated. After extubation, patients received supplemental oxygen via nasal cannula at a flow rate of 3–4 L/min. SpO₂ was continuously monitored using a pulse oximeter (probe type: Nellcor DS100A, Shanghai, China; monitor type: Philips IntelliVue MP70, Boeblingen, Germany) with automated 5-minute interval data logging and real-time integration into the anesthesia information management system (AIMS). Recovery quality was evaluated using the modified Aldrete scoring system. Patients achieving a score ≥ 9 were deemed fit for transfer to a general ward for continued observation and postoperative care.

Data collection

Patient data were systematically collected through the hospital electronic medical record (EMR) system and the AIMS. The EMR provided comprehensive baseline demographics, including age, sex, height, weight, and body mass index (BMI), as well as preoperative variables such as American Society of Anesthesiologists (ASA) physical status classification, comorbidities (e.g., hypertension, diabetes, coronary artery disease), and smoking history. The AIMS captured detailed intraoperative and postoperative data, including anesthetic techniques, medication administration, and physiological parameters. Intraoperative variables recorded included the use of opioids and dexmedetomidine, fluid management (type and volume), surgical duration, and surgical types. Postoperative data encompassed the administration of neuromuscular blocking (NMB) reversal agent (neostigmine or sugammadex), extubation timing, requirement for mechanical ventilation, incidence of hypoxemia, rescue analgesia requirements, and use of vasoactive drugs in the PACU. To standardize opioid exposure assessment across cases, all intraoperative and postoperative opioid doses were converted to intravenous morphine milligram equivalents (MME) using validated equianalgesic conversion factors. The following conversion ratios were applied: intravenous morphine 1 mg = 1 MME; intravenous sufentanil 10 µg = 10 MME; intravenous fentanyl 0.1 mg = 10 MME [9]. Remifentanil was excluded from the total opioids calculation due to its esterase metabolism and consequent ultrashort half-life, which results in a rapid offset of action.

Statistical analysis

To address missing data, the mean imputation method was adopted. The Shapiro-Wilk test was used to assess the distribution of quantitative data. Continuous variables following a normal distribution were presented as mean ± standard deviation (SD), whereas those with a skewed distribution were reported as median (interquartile range [IQR]). Comparisons of continuous data between groups were performed using Student’s t test for normally distributed data and the Mann-Whitney U test for non-normally distributed data. Categorical variables were expressed as frequencies (percentages), and their significance was evaluated using the chi-squared test or Fisher’s exact test, as appropriate.

The database was randomly divided into training and validation cohorts at a 7:3 ratio. Least absolute shrinkage and selection operator (LASSO) regression was adopted to select potential predictors in the training cohort. The optimal tuning parameter lambda (λ) was determined through cross-validation, ensuring that the coefficients of non-significant variables were shrunk to zero. Subsequently, the non-zero factors identified by LASSO regression were entered into a multivariable logistic regression analysis to identify independent predictors associated with hypoxemia. Based on the results of the multivariable logistic regression, a nomogram was constructed in the training cohort. Coefficients and odds ratios (OR) with 95% confidence intervals (CI) were calculated.

The predictive performance of the model was systematically evaluated through discrimination assessment, calibration analysis, and clinical utility validation. Discrimination was assessed by constructing a receiver operating characteristic (ROC) curve, with the model’s effectiveness quantified using the concordance index (C-index) and the area under the ROC curve (AUC). Additionally, the net reclassification index (NRI) and integrated discrimination improvement (IDI) were employed to evaluate the improvement in predictive accuracy of the nomogram compared to the baseline model comprising age and BMI. This baseline was chosen because both age and BMI represent well-established, clinically fundamental risk factors for postoperative hypoxemia. By comparing the new model against this parsimonious reference, the NRI and IDI effectively quantify the incremental predictive value contributed by additional variables. For calibration assessment, a calibration plot was generated to visually demonstrate the alignment between predicted probabilities and actual outcomes. Furthermore, decision curve analysis (DCA) was employed to evaluate the clinical utility of the model by quantifying net benefits across a range of threshold probabilities. All statistical analyses were performed using R software (version 4.2.0, R Foundation for Statistical Computing, Vienna, Austria) and SPSS statistical software (version 21.0, SPSS Inc. Chicago, Illinois, United States), with two-tailed P-values < 0.05 considered statistically significant.

Results

The flowchart of the study population with inclusions and exclusions was illustrated in Fig. 1. A total of 1220 patients were included in the study and randomly allocated into training (n = 854) and validation cohorts (n = 366). Baseline characteristics of the patients are summarized in Table 1. The overall prevalence of hypoxemia was 33.0%, with specific rates of 32.1% and 35.2% in the training and validation cohorts, respectively. The median age in the training cohort was 32 (IQR 27–37) years, and 71.0% of patients were female. In the validation cohort, the median age was 32 (IQR 27–38) years, with 74.0% female patients. All baseline characteristics were comparable between the two cohorts.

Fig. 1.

Flow chart of enrolling patients used for the development of nomogram

Table 1.

Baseline characteristics of patients included in the study

Characteristic	Training cohort (n = 854)	Validation cohort (n = 366)	P value
Hypoxemia, n(%)	274(32.1)	129(35.2)	0.269
Age, years	32(27–37)	32(27–38)	0.916
Male/Female	248/606	95/271	0.272
Height, cm	168 ± 8	167 ± 8	0.616
Weight, kg	108 ± 23	107 ± 25	0.957
BMI, kg/m2	37(33–42)	36(33–41)	0.329
Morbid obesitya, n(%)	278(32.6)	117(32.0)	0.841
ASA physical status, n(%)			0.846
Ⅱ	521(61.0)	217(59.3)
Ⅲ	328(38.4)	147(40.2)
Ⅳ	5(0.6)	2(0.5)
Type of comorbidities, n(%)
Diabetes	196(23.0)	77(21.0)	0.463
Hypertension	150(17.6)	64(17.5)	0.975
Coronary artery disease	78(9.1)	31(8.5)	0.710
Arrhythmia	70(8.2)	27(7.4)	0.628
Othersb	54(6.3)	16(4.4)	0.179
≥ 2 comorbidities, n(%)	136(15.9)	54(14.8)	0.605
Smoking, n(%)	147(17.2)	52(14.2)	0.193
Intraoperative opioids, mg	58.9 ± 9.7	59.5 ± 9.7	0.269
Intraoperative opioids ≥ 60 mg, n(%)	280(32.8)	138(37.7)	0.097
Dexmedetomidine, n(%)	646(75.6)	282(77.0)	0.598
Fluid infusion , ml	1500(1000–1500)	1500(1100–1525)	0.137
Duration of the anesthesia, min	92(76–104)	90(76–105)	0.693
Duration of the operation, min	74(61–88)	75(62–89)	0.670
Duration from stopping muscle relaxants until the end of surgery, min	48(36–68)	48(36–66)	0.374
Types of surgery, n(%)			0.200
LSG	555(65.0)	248(67.8)
LSG-DJB	33(3.9)	15(4.1)
LSG-JJB	197(23.0)	86(23.5)
OAGB	69(8.1)	17(4.6)
Duration of mechanical ventilation, min	119(104–134)	118(103–136)	0.936
Duration of extubation, min	28(18–38)	26(18–38)	0.794
Neostigmine use, n(%)	415(48.6)	177(48.4)	0.910
Sugammadex use, n(%)	372(43.6)	152(41.5)	0.512
Rescue analgesia, n(%)	66(7.7)	24(6.6)	0.473
Use of vasoactive drugs in the PACU, n(%)	81(9.5)	36(9.8)	0.849

Abbreviations: BMI body mass index, LSG laparoscopic sleeve gastrectomy, LSG–DJB laparoscopic sleeve gastrectomy plus duodenojejunal bypass, LSG–JJB laparoscopic sleeve gastrectomy plus jejunojejunal bypass, OAGB one anastomosis gastric bypass

^aMorbid obesity: BMI ≥ 40 kg/m², ^bOthers: chronic glomerulonephritis, chronic cardiac insufficiency, Cushing’s syndrome, hyperthyroidism, hypothyroidism, asthma, chronic pneumonia, reflux esophagitis, nephrotic syndrome, vasculopathy, autoimmune disease, pituitary tumour, meningioma, epilepsy, teratoma, anxiety, depression, bipolar disorder

A total of 28 clinical variables were incorporated as candidate predictors in the LASSO regression analysis (Table 1). The model was determined to be optimal at a lambda value of 0.017589. Through LASSO regression selection, 11 non-zero characteristic variables were identified as significant predictors (Fig. 2A, B). These selected predictive variables included age, gender, height, BMI, ASA (ASA_Ⅳ), ≥ 2 comorbidities, rescue analgesia, types of surgery (LSG-JJB), morbid obesity, intraoperative opioids ≥ 60 mg, and NMB reversal agent (neostigmine). To account for potential confounding factors, these variables identified through LASSO regression were included in subsequent multivariable logistic regression models. The results of logistic regression are presented in Table 2, where age, BMI, ≥ 2 comorbidities, intraoperative opioids ≥ 60 mg and NMB reversal agent were identified as significant independent variables.

Fig. 2.

Predictors selection using least absolute shrinkage and selection operator (LASSO) regression analysis and nomogram prediction model. A LASSO coefficient profiles of all the variables, a coefficient profile plot was provided against the log (lambda) sequence. B Identifcation of the optimal penalization coefficient lambda in the LASSO model. C Nomogram to estimate the risk of hypoxemia in the PACU after laparoscopic bariatric surgery

Table 2.

Multivariable logistic regression analysis of variables predicting hypoxemia in training cohort

Characteristics	Coefficients	OR (95% CI)a	P value
Age	0.029	1.030 (1.009–1.050)	0.004
BMI	0.095	1.100 (1.052–1.152)	< 0.001
≥ 2 comorbidities	0.672	1.958 (1.265–3.026)	0.003
Intraoperative opioids ≥ 60 mg	0.624	1.867 (1.310–2.657)	< 0.001
NMB reversal agent
None	Ref
Neostigmine	0.780	2.180 (1.102–4.571)	0.003
Sugammadex	0.229	1.258 (0.626–2.625)	0.529

OR odds ratios, CI confdence intervals

^aCoefficients and odds ratios with 95% confidence intervals were calculated

Using the five independent predictors, we developed a nomogram to illustrate the risk factors for postoperative hypoxemia in the PACU following elective LBS (Fig. 2C). This nomogram provides a quantitative assessment of each predictor’s contribution to the model and estimates the probability of hypoxemia in the PACU following LBS. The AUC was 0.738 (95% CI 0.701–0.775) (Fig. 3A), indicating acceptable discriminative ability. Internal validation was performed using bootstrap resampling with 1,000 iterations, yielding an optimism-corrected C-index of 0.731, confirming the reliability of the model. The calibration plot demonstrated good agreement between the nomogram-predicted probabilities and actual clinical outcomes (Fig. 3C). In the validation cohort, the nomogram exhibited an AUC of 0.735 (95% CI 0.682–0.789) and a bootstrap-corrected C-index of 0.715, further supporting its validity (Fig. 3B). Additionally, the calibration curve for the validation cohort showed strong alignment between predicted and observed risks (Fig. 3D). To evaluate the clinical utility of the model, DCA was performed for both the training (Fig. 3E) and validation cohorts (Fig. 3F). These analyses demonstrated that the nomogram provided a distinct net benefit across a wide range of threshold probabilities, particularly between 20% and 60%.

Fig. 3.

Validation of the nomogram prediction model in both the training and validation cohorts. Receiver operating characteristic curve to assess discrimination performance in training (A) and validation cohort (B). Calibration plots for the prediction model in training (C) and validation cohort (D). Decision curve analysis for the prediction model in training (E) and validation cohort (F)

The changes in AUC, NRI, and IDI were used to compare the accuracy between the nomogram and the baseline model comprising age and BMI. In the training cohort, the nomogram demonstrated significantly superior discriminative ability [AUC = 0.738 (95% CI 0.701–0.775)] compared to the baseline model [AUC = 0.658 (95% CI 0.619–0.698)] (P = 0.004). Furthermore, the NRI was 0.228 (95% CI 0.162–0.295, P < 0.001) and the IDI was 0.233 (95% CI 0.220–0.246, P < 0.001). These results were consistently validated in the validation cohort (Table 3), confirming that the nomogram more accurately predicts postoperative hypoxemia in the PACU following LBS than the baseline model based solely on age and BMI.

Table 3.

Predictive power of the nomogram compared with the baseline model in the training and validation cohorts

Index	Training cohort			Validation cohort
	Baseline model	Nomogram	P value	Baseline model	Nomogram	P value
AUCa	0.658(0.619–0.698)	0.738(0.701–0.775)	0.004	0.666(0.609–0.724)	0.735(0.682–0.789)	0.036
NRI	Ref	0.228(0.162–0.295)	< 0.001	Ref	0.208(0.108–0.304)	< 0.001
IDI	Ref	0.233(0.220–0.246)	< 0.001	Ref	0.247(0.224–0.271)	< 0.001

AUC area under the receiver operating characteristic curve, IDI integrated discrimination improvement, NRI net reclassification index

^aThe DeLong’s test for statistical significance was used to test differences between AUC curves

Discussion

In this study, we analyzed 1220 patients and found that the prevalence of hypoxemia in the PACU following LBS was 33.0%. Our results indicate that advanced age, elevated BMI, the presence of two or more comorbidities, intraoperative opioid use ≥ 60 mg (equivalent to intravenous morphine), and the use of NMB reversal agent are independent risk factors for PACU hypoxemia among patients after elective LBS. Both internal and external validation demonstrated that the nomogram developed in this study exhibits high predictive accuracy and robust clinical applicability.

The implementation of this model is expected to refine risk assessment and perioperative management for bariatric surgery patients, which may ultimately lead to a lower rate of postoperative hypoxemia in the PACU and better clinical outcomes. The DCA further provides a tangible framework for implementation. The “clinical decision window” (between 20% and 60% probability) can be used to triage patients after LBS at PACU admission. For example, those with a predicted risk score exceeding 30% may be selected for intensified monitoring or prophylactic interventions—such as high-flow nasal oxygen—thus enabling personalized management and efficient resource allocation.

In the current literature, postoperative hypoxemia is defined using various criteria. These include a peripheral capillary oxygen saturation of less than 93% on room air or a ratio of arterial partial pressure of oxygen to fractional inspired oxygen of less than 300 mmHg [10]. Another study defined it as a peripheral oxygen saturation of less than 90% for at least two consecutive minutes [11], while Templeton et al. defined it as continuous oxygen desaturation below 90% for at least three minutes during oxygen therapy [12]. The incidence of postoperative hypoxemia varies widely, ranging from 3% to 65%, and is influenced by the definition used, anesthetic techniques, surgical type, patient selection, and monitoring methods [13]– [14]. In our study, we found that among 1220 patients who underwent LBS, 33.0% experienced hypoxemia in the PACU. This finding underscores the importance of a standardized definition and highlights the need for effective predictive models to address this common postoperative complication.

Multiple studies have established a link between advanced age and the development of postoperative hypoxemia across various surgical contexts, a finding that aligns with our study’s results. Szeles et al.‘s multivariate analysis identified age as an independent predictor of severe hypoxemia following myocardial revascularization [15]. Previous research has suggested that age-related vascular degeneration, characterized by reduced reactivity of small vessels and compromised perfusion, contributes to hypoxemia [16]. Schouten’s work further implicated age-related declines in the balance of the pulmonary renin-angiotensin system, which may exacerbate inflammation and pulmonary damage [17]. Additionally, the natural aging process is associated with degenerative changes in organs, diminishing patients’ tolerance for anesthesia and extensive surgical procedures [18]. Notably, despite the median age of patients in our study being 32 years—lower than in some previous studies—the incidence of PACU hypoxemia still exhibited a general upward trend with increasing age, underscoring the consistent impact of age on postoperative oxygenation across diverse patient populations.

BMI has been widely recognized as a significant independent risk factor for hypoxemia in numerous studies. Kendale et al. demonstrated that the incidence of perioperative hypoxemia was 16% in patients with a normal BMI, 28% in those with a BMI >30 kg/m², and 35% in those with a BMI >40 kg/m² [19]. Several mechanisms underlie this association. Obesity leads to reduced thoracic cavity volume due to diaphragm elevation, increased airway resistance, decreased lung compliance, higher tissue and organ oxygen consumption, and a higher prevalence of comorbid chronic diseases [20]. Consequently, increased BMI is closely associated with the incidence, severity, and duration of hypoxemia in the PACU. In obese patients, adipose tissue accumulation in the posterior pharyngeal wall and around the trachea can cause pharyngeal stenosis and airway compression, increasing the likelihood of acute upper airway obstruction and subsequent hypoxemia and hypercapnia post-extubation. Additionally, the prevalence of obstructive sleep apnea syndrome (OSAS) rises with increasing obesity, affecting approximately 10–20% of morbidly obese individuals, often remaining undiagnosed [21]. Studies indicate that obese patients with preoperative OSAS exhibit heightened sensitivity to opioids, predisposing them to early postoperative hypoxemia.

The presence of two or more comorbidities has been identified as a significant risk factor for hypoxemia in the PACU following LBS. Common comorbidities associated with obesity, such as hypertension, diabetes, and coronary artery disease, contribute to respiratory complications [3]. These conditions can impair pulmonary function, increase the workload on the respiratory system, and reduce the body’s ability to maintain adequate oxygenation. For instance, hypertension may lead to pulmonary hypertension and reduced lung perfusion, while diabetes can cause microvascular damage and impair oxygen delivery [22]– [23]. Coronary artery disease may further compromise cardiac output, exacerbating hypoxemia. The presence of multiple comorbidities complicates anesthesia management and postoperative recovery, necessitating closer monitoring and more intensive care to mitigate the risk of hypoxemia and ensure optimal patient outcomes. High doses of opioids can lead to respiratory depression by suppressing the central nervous system’s respiratory drive, reducing tidal volume and respiratory rate [24]. This effect is particularly pronounced in the PACU, where patients are recovering from anesthesia and may exhibit increased sensitivity to opioids. Studies have shown that excessive intraoperative opioid administration is associated with a higher incidence of postoperative hypoxemia [25–27]. Additionally, the use of opioids can cause other respiratory complications, such as airway obstruction and reduced lung compliance, further exacerbating the risk of hypoxemia [28]. Therefore, careful titration of opioid doses during surgery and close monitoring in the PACU are essential to mitigate this risk and improve patient outcomes.

Studies have shown that residual neuromuscular block (RNMB) is associated with an increased risk of postoperative pulmonary complications, including hypoxemia [29–31]. The use of NMB reversal agents, such as neostigmine and sugammadex, is intended to reverse the effects of NMB agents and reduce the risk of RNMB. However, the effectiveness of these agents in preventing hypoxemia remains controversial. Some studies suggest that the use of NMB reversal agents may not significantly decrease the incidence of hypoxemia [32]– [33], while others indicate that specific agents like sugammadex may be more effective in reducing the risk of hypoxemia compared to traditional agents like neostigmine [34]. The choice of NMB reversal agent and the timing of its administration can influence postoperative outcomes, and further research is needed to determine the optimal strategies for minimizing the risk of hypoxemia in the PACU.

This retrospective study has inherent limitations that warrant consideration. One limitation pertains to the extraction of clinical information during the recovery phase. Data were collected at 5-minute intervals, which may potentially miss transient or rapidly changing physiological parameters critical for accurate assessment. Furthermore, the inclusion criteria for comorbidities were not exhaustive. Notably, a significant proportion of patients lacked preoperative polysomnography results, leading to an underdiagnosis of OSAS. Consequently, key sleep-related breathing parameters, such as the longest apnea duration and apnea–hypopnea index, were unavailable for analysis. Additionally, we acknowledge that prediction models alone cannot fully capture the multifactorial and dynamic nature of respiratory depression. Their true clinical utility lies in enabling early identification of high-risk patients who may benefit from targeted monitoring and preventive interventions. Future prospective studies should incorporate high-resolution continuous monitoring and systematic screening for sleep-disordered breathing to improve predictive accuracy and, most importantly, test whether the application of this model leads to meaningful improvements in patient outcomes.

Conclusions

We developed and externally validated a nomogram prediction model for hypoxemia in the PACU among patients undergoing elective LBS. The model incorporates five key clinical variables: age, BMI, presence of two or more comorbidities, intraoperative opioid use ≥ 60 mg (equivalent to intravenous morphine), and use of an NMB reversal agent. This nomogram may provide valuable guidance for individualized treatment decisions and enhance perioperative care by identifying patients at high risk of PACU hypoxemia.

Abbreviations

PACU

Post-anesthesia care unit

LBS

Laparoscopic bariatric surgery

LASSO

Least absolute shrinkage and selection operator

AUC

Area under curve

C-index

Concordance index

DCA

Decision curve analysis

NMB

Neuromuscular blocking

MME

Morphine milligram equivalents

OSA

Obstructive sleep apnea

ICU

Intensive care unit

EMR

Electronic medical record

AIMS

Anesthesia information management system

BMI

Body mass index

ASA

American Society of Anesthesiologists

SD

Standard deviation

IQR

Interquartile range

OR

Odds ratios

CI

Confidence intervals

ROC

Receiver operating characteristic

IDI

Integrated discrimination improvement

NRI

Net reclassification improvement

OSAS

Obstructive sleep apnea syndrome

LSG–DJB

Laparoscopic sleeve gastrectomy plus duodenojejunal bypass

LSG–JJB

Laparoscopic sleeve gastrectomy plus jejunojejunal bypass

OAGB

One anastomosis gastric bypass

RNMB

Residual neuromuscular block

Authors’ contributions

Jingjing Fan, Xiaodong Chen and Min Wang designed the study. Min wang, Yifan Xu and Xiaodong Chen collected data. Yiting Li, Qingyun Gong and Min Wang analysed data. Min Wang and Yiting Li contributed to manuscript writing. Xiaodong Chen and Jingjing Fan contributed to manuscript revision. All authors read and approved the final version of the manuscript.

Funding

Not applicable.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This study received approval from the Ethics Committee of the First Affiliated Hospital of Nanjing Medical University. Additionally, due to the retrospective nature of the study, the requirement for informed consent was waived by the same Ethics Committee. All methods were conducted in accordance with the relevant guidelines and regulations.

Consent to publication

Not applicable.

Competing interests

The authors declare no competing interests.