Construction and validation of a predictive model for hypothermia complication during endoscopic thyroidectomy for thyroid cancer
Hui Ye, Lu Xia, Tian Zhan, Haiwei Zhang
TL;DR
This study creates a model to predict hypothermia risk during endoscopic thyroid cancer surgery, helping nurses identify high-risk patients.
Contribution
A novel predictive model for intraoperative hypothermia in endoscopic thyroidectomy, validated for clinical utility.
Findings
Infusion volume and irrigation fluid temperature are independent risk factors for intraoperative hypothermia.
BMI is an independent protective factor against intraoperative hypothermia.
The model showed excellent predictive accuracy (AUC = 0.945) and maintained strong performance in external validation (AUC = 0.831).
Abstract
Intraoperative hypothermia frequently occurs during surgery and can negatively impact patient outcomes. The study focuses on establishing a clinical prediction model to identify the risk of intraoperative hypothermia in patients undergoing endoscopic thyroidectomy for thyroid cancer. Univariate analysis was performed to identify potential indicators associated with intraoperative hypothermia. Multivariable logistic regression analysis was employed to select the independent predictors for model construction. The predictive performance and clinical utility of the model were assessed using receiver operating characteristic (ROC) curve, calibration plots and decision curve analysis (DCA). External validation was conducted to evaluate its generalizability. Univariate analysis revealed that age, body mass index (BMI), anesthesia duration, duration of surgery, infusion volume, intraoperative…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
FIGURE 1
FIGURE 2
FIGURE 3
FIGURE 4| Variables | Assignment |
|---|---|
| Intraoperative temperature | 0 indicates intraoperative temperature ≥36.0 °C, 1 indicates <36.0 °C |
| Gender | 0 indicates female, 1 indicates male |
| Age (years) | 0 indicates <30, 1 indicates 30–60, and 2 indicates >60 |
| BMI (kg/m2) | 0 indicates <18.5, 1 indicates 18.5–23.9, and 2 indicates >23.9 |
| History of underlying diseases (diabetes or hypertensive) | 0 indicates no, 1 indicates yes |
| HR (beats/min) | 0 indicates <60, 1 indicates 60–100, and 2 indicates >100 |
| Preoperative blood pressure (mmHg) | 0 indicates low, values < 90/60, 1 indicates normal, values = 90–140/60–90, and 2 indicates high, values >140/90 |
| Basal body temperature | 0 indicates <36.5 °C, 1 indicates 36.5–37 °C, and 2 indicates >37.0 °C |
| Operating room temperature | 0 indicates ≥23.0 °C, 1 indicates <23.0 °C |
| ASA | 0 indicates grade 1, 1 indicates grade 2, and 2 indicates grade 3 |
| Anesthesia duration (min) | 0 indicates <120, 1 indicates 120–180, and 2 indicates >180 |
| Duration of surgery (min) | 0 indicates <120, 1 indicates 120–180, and 2 indicates >180 |
| Intraoperative infusion volume (mL) | 0 indicates <1,000, 1 indicates 1,000–2000, and 2 indicates >2000 |
| Intraoperative irrigation volume (mL) | 0 indicates ≥2000, 1 indicates <2000 |
| Irrigation fluid temperature | 0 indicates room temperature, 1 indicates constant temperature (37 °C) |
| Intraoperative blood loss (mL) | 0 indicates ≥100, 1 indicates <100 |
| Blood transfusion | 0 indicates no, 1 indicates yes |
| Demographic characteristics [n (%)] | Non-hypothermia group (n = 144) | Hypothermia group (n = 109) | Statistical value |
|
|---|---|---|---|---|
| Gender | 0.310 | 0.577 | ||
| Female | 83 (57.64) | 59 (54.13) | ||
| Male | 61 (42.36) | 50 (45.87) | ||
| Age (years) | 12.851 |
| ||
| <30 | 18 (12.50) | 6 (5.50) | ||
| 30–60 | 52 (45.61) | 23 (21.10) | ||
| >60 | 74 (51.39) | 80 (73.39) | ||
| BMI (kg/m2) | 23.267 |
| ||
| <18.5 | 25 (17.36) | 49 (44.95) | ||
| 18.5–23.9 | 73 (50.69) | 40 (36.70) | ||
| >23.9 | 46 (31.94) | 20 (18.35) | ||
| History of underlying diseases | 1.435 | 0.231 | ||
| No | 54 (37.50) | 33 (30.28) | ||
| Yes | 90 (62.50) | 76 (69.72) | ||
| Heart rate (beats/min) | 1.393 | 0.498 | ||
| <60 | 7 (4.86) | 3 (2.75) | ||
| 60–100 | 132 (91.67) | 104 (95.41) | ||
| >100 | 5 (3.47) | 2 (1.83) | ||
| Preoperative blood pressure (mmHg) | 0.754 | 0.686 | ||
| Low | 12 (8.33) | 6 (5.50) | ||
| Normal | 122 (84.72) | 95 (87.16) | ||
| High | 10 (6.94) | 8 (7.34) | ||
| Basal body temperature | 2.671 | 0.263 | ||
| <36.5 °C | 84 (58.33) | 57 (52.29) | ||
| 36.5–37 °C | 49 (34.03) | 47 (43.12) | ||
| >37.0 °C | 11 (7.64) | 5 (4.59) | ||
| Operating room temperature | 0.001 | 0.976 | ||
| ≥23.0 °C | 127 (88.19) | 96 (88.07) | ||
| <23.0 °C | 17 (11.81) | 13 (11.93) | ||
| ASA | 0.957 | 0.620 | ||
| Grade 1 | 5 (3.47) | 6 (5.50) | ||
| Grade 2 | 98 (68.06) | 69 (63.30) | ||
| Grade 3 | 41 (28.47) | 34 (31.19) | ||
| Anesthesia duration (min) | 21.919 |
| ||
| <120 | 37 (25.69) | 22 (20.18) | ||
| 120–180 | 82 (56.94) | 39 (35.78) | ||
| >180 | 25 (17.36) | 48 (44.04) | ||
| Duration of surgery (min) | 18.757 |
| ||
| <120 | 38 (26.39) | 29 (26.61) | ||
| 120–180 | 87 (60.42) | 42 (38.53) | ||
| >180 | 19 (13.19) | 38 (34.86) | ||
| Infusion volume (mL) | 59.275 |
| ||
| <1,000 | 42 (29.17) | 6 (5.50) | ||
| 1,000–2000 | 79 (54.86) | 37 (33.94) | ||
| >2000 | 23 (15.97) | 66 (60.55) | ||
| Intraoperative irrigation volume (mL) | 5.116 |
| ||
| ≥2000 | 25 (17.36) | 32 (29.36) | ||
| <2000 | 119 (82.64) | 77 (70.64) | ||
| Irrigation fluid temperature | 21.001 |
| ||
| Room temperature | 35 (24.31) | 57 (52.29) | ||
| Constant temperature (37 °C) | 109 (75.69) | 52 (47.71) | ||
| Intraoperative blood loss (mL) | 6.461 |
| ||
| ≥100 | 76 (52.78) | 40 (36.70) | ||
| <100 | 68 (47.22) | 69 (63.30) | ||
| Blood transfusion | 0.161 | 0.688 | ||
| No | 140 (97.22) | 105 (96.33) | ||
| Yes | 4 (2.78) | 4 (3.67) |
| Variables |
| Standard error | Wald value |
| 95% |
|
|---|---|---|---|---|---|---|
| Age | 0.036 | 0.019 | 3.447 | 1.036 | 0.998–1.076 | 0.063 |
| BMI | −1.053 | 0.222 | 22.399 | 0.349 | 0.226–0.540 |
|
| Anesthesia duration | 0.011 | 0.035 | 0.097 | 1.011 | 0.944–1.083 | 0.755 |
| Duration of surgery | −0.058 | 0.034 | 2.795 | 0.944 | 0.882–1.010 | 0.095 |
| Infusion volume | 7.917 | 2.026 | 15.276 | 8.115 | 3.474–21.182 |
|
| Intraoperative irrigation volume | 0.805 | 1.033 | 0.608 | 2.236 | 0.296–16.925 | 0.436 |
| Irrigation fluid temperature | 3.417 | 0.823 | 17.224 | 30.486 | 6.070–153.100 |
|
| Intraoperative blood loss | 0.872 | 1.074 | 0.633 | 1.089 | 0.988–1.190 | 0.412 |
| Demographic characteristics [n (%)] | Non-hypothermia group (n = 62) | Hypothermia group (n = 38) | Statistical value |
|
|---|---|---|---|---|
| Gender | 6.629 |
| ||
| Female | 35 (56.45) | 31 (81.58) | ||
| Male | 27 (43.55) | 7 (18.42) | ||
| Age (years) | 15.382 |
| ||
| <30 | 11 (17.74) | 0 (0.00) | ||
| 30–60 | 16 (25.81) | 23 (60.53) | ||
| >60 | 35 (56.45) | 15 (39.47) | ||
| BMI (kg/m2) | 7.678 |
| ||
| <18.5 | 17 (27.42) | 7 (18.42) | ||
| 18.5–23.9 | 32 (51.61) | 13 (34.21) | ||
| >23.9 | 13 (20.97) | 18 (47.37) | ||
| History of underlying diseases | 1.181 | 0.277 | ||
| No | 21 (33.87) | 17 (44.74) | ||
| Yes | 41 (66.13) | 21 (55.26) | ||
| Heart rate (beats/min) | 3.879 | 0.144 | ||
| <60 | 6 (9.68) | 1 (2.63) | ||
| 60–100 | 53 (85.48) | 37 (97.37) | ||
| >100 | 3 (4.84) | 0 (0.00) | ||
| Preoperative blood pressure (mmHg) | 1.040 | 0.594 | ||
| Low | 9 (14.52) | 3 (7.89) | ||
| Normal | 46 (74.19) | 31 (81.58) | ||
| High | 7 (11.29) | 4 (10.53) | ||
| Basal body temperature | 19.106 |
| ||
| <36.5 °C | 20 (32.26) | 29 (76.32) | ||
| 36.5–37 °C | 35 (56.45) | 6 (15.79) | ||
| >37.0 °C | 7 (11.29) | 3 (7.89) | ||
| Operating room temperature | 0.614 | 0.433 | ||
| ≥23.0 °C | 52 (83.87) | 34 (89.47) | ||
| <23.0 °C | 10 (16.13) | 4 (10.53) | ||
| ASA | 1.213 | 0.545 | ||
| Grade 1 | 18 (29.03) | 8 (21.05) | ||
| Grade 2 | 41 (66.13) | 29 (76.32) | ||
| Grade 3 | 3 (4.84) | 1 (2.63) | ||
| Anesthesia duration (min) | 21.294 |
| ||
| <120 | 23 (37.10) | 5 (13.16) | ||
| 120–180 | 30 (48.39) | 11 (28.95) | ||
| >180 | 9 (14.52) | 22 (57.89) | ||
| Duration of surgery (min) | 9.012 |
| ||
| <120 | 23 (37.10) | 7 (18.42) | ||
| 120–180 | 31 (50.00) | 17 (44.74) | ||
| >180 | 8 (12.90) | 14 (36.84) | ||
| Infusion volume (mL) | 19.458 |
| ||
| <1,000 | 24 (38.71) | 2 (5.26) | ||
| 1,000–2000 | 28 (45.16) | 17 (44.74) | ||
| >2000 | 10 (16.13) | 19 (50.00) | ||
| Intraoperative irrigation volume (mL) | 9.564 |
| ||
| ≥2000 | 9 (14.52) | 16 (42.11) | ||
| <2000 | 53 (85.48) | 22 (57.89) | ||
| Irrigation fluid temperature | 16.352 |
| ||
| Room temperature | 6 (9.68) | 17 (44.74) | ||
| Constant temperature (37 °C) | 56 (90.32) | 21 (55.26) | ||
| Intraoperative blood loss (mL) | 0.075 | 0.784 | ||
| ≥100 | 36 (58.06) | 21 (55.26) | ||
| <100 | 26 (41.94) | 17 (44.74) | ||
| Blood transfusion | 0.255 | 0.614 | ||
| No | 60 (96.77) | 36 (94.74) | ||
| Yes | 2 (3.23) | 2 (5.26) |
| Demographic characteristics [n (%)] | Training cohort (n = 253) | Validation cohort (n = 100) | Statistical value |
|
|---|---|---|---|---|
| Status | 0.762 | 0.383 | ||
| Hypothermia | 109 (43.08) | 38 (38.00) | ||
| Non-hypothermia | 144 (56.92) | 62 (62.00) | ||
| Gender | 2.887 | 0.089 | ||
| Female | 142 (56.13) | 66 (66.00) | ||
| Male | 111 (43.87) | 34 (34.00) | ||
| Age (years) | 3.573 | 0.168 | ||
| <30 | 24 (9.49) | 11 (11.00) | ||
| 30–60 | 75 (29.64) | 39 (39.00) | ||
| >60 | 154 (60.87) | 50 (50.00) | ||
| BMI (kg/m2) | 1.343 | 0.511 | ||
| <18.5 | 74 (29.25) | 24 (24.00) | ||
| 18.5–23.9 | 113 (44.66) | 45 (45.00) | ||
| >23.9 | 66 (26.09) | 31 (31.00) | ||
| History of underlying diseases | 0.409 | 0.522 | ||
| No | 87 (34.39) | 38 (38.00) | ||
| Yes | 166 (65.61) | 62 (62.00) | ||
| Heart rate (beats/min) | 1.479 | 0.477 | ||
| <60 | 10 (3.95) | 7 (7.00) | ||
| 60–100 | 236 (93.28) | 90 (90.00) | ||
| >100 | 7 (2.77) | 3 (3.00) | ||
| Preoperative blood pressure (mmHg) | 3.992 | 0.136 | ||
| Low | 18 (7.11) | 12 (12.00) | ||
| Normal | 217 (85.77) | 77 (77.00) | ||
| High | 18 (7.11) | 11 (11.00) | ||
| Basal body temperature | 2.091 | 0.352 | ||
| <36.5 °C | 141 (55.73) | 49 (49.00) | ||
| 36.5–37 °C | 96 (37.94) | 41 (41.00) | ||
| >37.0 °C | 16 (6.32) | 10 (10.00) | ||
| Operating room temperature | 0.301 | 0.583 | ||
| ≥23.0 °C | 30 (11.86) | 14 (14.00) | ||
| <23.0 °C | 223 (88.14) | 86 (86.00) | ||
| ASA | 53.288 |
| ||
| Grade 1 | 11 (4.35) | 26 (26.00) | ||
| Grade 2 | 167 (66.01) | 70 (70.00) | ||
| Grade 3 | 75 (29.64) | 4 (4.00) | ||
| Anesthesia duration (min) | 1.477 | 0.478 | ||
| <120 | 59 (23.32) | 28 (28.00) | ||
| 120–180 | 121 (47.83) | 41 (41.00) | ||
| >180 | 73 (28.85) | 31 (31.00) | ||
| Duration of surgery (min) | 0.459 | 0.795 | ||
| <120 | 67 (26.48) | 30 (30.00) | ||
| 120–180 | 129 (50.99) | 48 (48.00) | ||
| >180 | 57 (22.53) | 22 (22.00) | ||
| Infusion volume (mL) | 2.518 | 0.284 | ||
| <1,000 | 48 (18.97) | 26 (26.00) | ||
| 1,000–2000 | 89 (35.18) | 45 (45.00) | ||
| >2000 | 116 (45.85) | 29 (29.00) | ||
| Intraoperative irrigation volume (mL) | 0.245 | 0.620 | ||
| ≥2000 | 196 (77.47) | 75 (75.00) | ||
| <2000 | 57 (22.53) | 25 (25.00) | ||
| Irrigation fluid temperature | 5.827 |
| ||
| Room temperature | 92 (36.36) | 23 (23.00) | ||
| Constant temperature (37 °C) | 161 (63.64) | 77 (77.00) | ||
| Intraoperative blood loss (mL) | 3.566 | 0.059 | ||
| ≥100 | 137 (54.15) | 57 (57.00) | ||
| <100 | 116 (45.85) | 43 (43.00) | ||
| Blood transfusion | 0.153 | 0.695 | ||
| No | 245 (96.84) | 96 (96.00) | ||
| Yes | 8 (3.16) | 4 (4.00) |
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsThermal Regulation in Medicine · Thyroid and Parathyroid Surgery · Ultrasound and Hyperthermia Applications
Introduction
1
Thyroid cancer is one of the most common endocrine malignancies. Papillary thyroid carcinoma and follicular thyroid carcinoma together constitute over 90% of all thyroid carcinoma cases (Prete et al., 2020). Most patients with these subtypes have an excellent prognosis, with more than 90% achieving disease-specific survival beyond 10 years (Jemal et al., 2009). The incidence of thyroid carcinoma has increased significantly over the past 30 years (Kim et al., 2020). Young women represent a high-risk population for thyroid cancer, accounting for approximately 7.5%–10% of all cases (Bleyer et al., 2006). Therefore, during surgical management, clinicians should aim not only for complete disease eradication but also for optimal postoperative cosmetic outcomes to address patients’ dual concerns regarding functional integrity and aesthetic appearance (Sephton, 2019). In recent years, with advances in technology, endoscopic radical resection for thyroid cancer has been increasingly adopted due to its advantages, including improved surgical visualization, minimal incisions, and aesthetically concealed incision sites (Li et al., 2023). However, because this procedure is technically complex and typically involves prolonged operative duration, 50%–70% of patients experience intraoperative hypothermia (Zou et al., 2020).
Intraoperative hypothermia refers to a decrease in the patient’s core body temperature below 36 °C for any reason during surgery, and it represents a common perioperative complication (Wang et al., 2022). Intraoperative hypothermia is correlated with multiple adverse consequences, such as cardiovascular complications following surgery, perioperative hemorrhage, disrupted drug metabolism, surgical wound infections, heightened risk of postoperative delirium, and a greater probability of deep vein thrombosis occurrence (Frank et al., 1997; Ju et al., 2023; Stöckler et al., 1989; Li et al., 2022). Furthermore, intraoperative hypothermia may also delay emergence from anesthesia (Pan et al., 2025), diminish thermal comfort and patient satisfaction, and lead to elevated healthcare spending (Soilly et al., 2023). Therefore, maintaining stable intraoperative body temperature is a critical intervention for reducing the risk of perioperative and anesthesia-related complications.
Numerous studies have demonstrated that intraoperative hypothermia is associated with a range of contributing factors. Anesthetic agents play a key role, particularly general anesthetics, which impair central thermoregulation, induce peripheral vasodilation, and consequently enhance heat loss (Cho et al., 2024). Environmental and procedural elements further contribute to hypothermia risk, including low ambient operating room temperatures, prolonged patient exposure to cold environments, and the administration of unwarmed irrigation or intravenous fluids (Campbell et al., 2015). Moreover, longer surgical duration and greater exposure of body cavities increase the surface area for heat dissipation, thereby elevating the likelihood of significant thermal loss (Zhao et al., 2023). Furthermore, the incidence of hypothermia varies across surgical types, being notably higher in major abdominal procedures, trauma surgeries, and extended laparoscopic operations (Shen and He, 2024). Therefore, systematic identification of these risk factors and the development of tailored perioperative temperature management strategies are essential for reducing the occurrence of intraoperative hypothermia and its associated complications.
In recent years, the rapid advancement of artificial intelligence has driven a growing number of studies on intraoperative hypothermia prediction models, offering robust support for early and accurate risk stratification of intraoperative hypothermia (Jiang et al., 2024; Huang et al., 2025). However, research on predictive models for intraoperative hypothermia in patients undergoing endoscopic radical thyroidectomy remains limited, and the development and validation of these models necessitate systematic investigation and rigorous clinical evaluation. Here, we identified key risk factors associated with hypothermia during endoscopic radical thyroidectomy and developed a predictive model based on these variables. The model demonstrates strong discriminatory capacity in identifying high-risk patients, thereby offering a reliable and evidence-based foundation for timely intraoperative hypothermia prevention and management.
Materials and methods
2
Study design
2.1
The study is a retrospective study and follows TRIPOD statement. A total of 280 patients who underwent endoscopic radical thyroidectomy at Jiangsu Provincial People’s Hospital between January 2022 and January 2023 were included in the training cohort. An independent external validation cohort comprising 120 patients who underwent the same procedure at Jiangsu Cancer Hospital during the same period was established, thereby enhancing the model’s generalizability and clinical applicability. Inclusion criteria were as follows: patients aged ≥18 years who had complete clinical characteristic data. The nasopharynx was used as the site for core body temperature monitoring (Bindu et al., 2017). Body temperature data were continuously collected at each time point from the induction of anesthesia until patient transfer out of the operating room using an anesthesia monitor. A patient was classified as having intraoperative hypothermia if body temperature dropped to <36 °C at any time during this period.
Sample
2.2
Based on a comprehensive literature review and expert consensus (Cho et al., 2022; Tavares Mendonça et al., 2021), this study incorporated 16 potential risk factors. It has been widely reported that a sample size of 10–15 times the number of independent variables is necessary to ensure reliable parameter estimation. Considering an expected sample loss rate of 10%–20%, the required sample size ranges from 176 to 288 cases. Therefore, the 253 cases in the training cohort used for modeling satisfy the sample size requirements.
Clinical data collection
2.3
All data were extracted from patients’ medical records, nursing records, and anesthesia record sheets using a self-designed data collection form. The organized data has been encoded according to the preset rules as shown in Table 1.
Model construction and validation
2.4
This study first employed univariate analysis to conduct a preliminary screening of the association between independent variables and intraoperative hypothermia, in order to identify potential influencing factors with statistical significance. Subsequently, significant variables identified through the logistic regression analysis were incorporated as predictors into the predictive model. R language was used to build a postoperative hypothermia prediction model. Model performance was assessed using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA), with external validation performed on the validation set. Modeling is achieved through the use of “readxl”, “pROC”, “rms”, “dcurves” and “dplyr” R packages.
Statistical analysis
2.5
All data were analyzed using SPSS (version 28.0) and R software (version 3.6.1). Categorical variables are reported as frequency and percentage (%), and group comparisons were conducted using the chi-square (χ^2^) test. The p value <0.05 was considered statistically significant.
Results
3
Study participant selection and flow diagram
3.1
A total of 400 patients who met the inclusion criteria were enrolled in this study. Of these, 280 patients were recruited from Jiangsu Provincial People’s Hospital and constituted the training cohort, while 120 patients were recruited from Jiangsu Cancer Hospital and formed the external validation cohort. In the training cohort, 16 patients with incomplete temperature monitoring data and 11 with incomplete baseline data were excluded, resulting in a final inclusion of 253 patients. In the validation cohort, 15 patients with incomplete temperature monitoring data and 5 with incomplete baseline data were excluded, yielding a final sample of 100 patients. The detailed patient selection process is illustrated in Figure 1.
Flowchart of the study.
Baseline characteristics of the study population
3.2
A total of 253 patients were included in the training cohort, of whom 109 experienced intraoperative hypothermia and 144 maintained normal body temperature during surgery. The overall incidence of intraoperative hypothermia was 43.08%. More sample characteristics are shown in Table 2. Univariate analysis revealed that age, BMI, anesthesia duration, duration of surgery, infusion volume, intraoperative irrigation volume, irrigation fluid temperature and intraoperative blood loss were significantly associated with the occurrence of intraoperative hypothermia (Table 2).
Feature selection and model construction
3.3
To identify the independent predictors, the eight variables significantly associated with intraoperative hypothermia were entered into a multivariate logistic regression model. As shown in Table 3, logistic regression analysis showed that infusion volume and irrigation fluid temperature were independent risk factors for intraoperative hypothermia in patients undergoing endoscopic radical thyroidectomy for thyroid cancer, whereas BMI was an independent protective factor (P < 0.05).
Subsequently, independent risk factors associated with intraoperative hypothermia in patients undergoing endoscopic radical thyroidectomy were systematically integrated into R software to develop a nomogram-based predictive model, facilitating individualized risk assessment and clinical decision-making. This nomogram model is shown in Figure 2.
Developed intraoperative hypothermia nomogram. The intraoperative hypothermia nomogram was developed in the training cohort, with irrigation fluid temperature, infusion volume and BMI. BMI: body mass index.
Evaluation of the predictive value of the model
3.4
Internal validation of the nomogram model was performed using the Bootstrap method with 1,000 resampling iterations to assess its stability (Xia et al., 2024). The predictive accuracy of the model for intraoperative hypothermia in patients with thyroid cancer undergoing endoscopic surgery was assessed using ROC curve analysis. The area under the curve (AUC) was 0.945 [95% CI (0.916–0.974)], with a sensitivity of 0.936 and a specificity of 0.819, indicating that the risk prediction model exhibited good discriminatory performance and could effectively stratify the risk of intraoperative hypothermia in patients undergoing endoscopic thyroidectomy (Figure 3A). The calibration curve demonstrates good agreement between the observed and predicted probabilities, further confirming the accuracy of the model (Figure 3B).
Predictive nomogram model evaluation. (A) ROC curve, (B) calibration plots and (C) decision curve analysis of the intraoperative hypothermia nomogram prediction in the training cohort. ROC, Receiver operating characteristic.
In addition, to assess the clinical value of the nomogram model, the DCA method was used. DCA evaluates the clinical utility of predictive models by quantifying the net benefit across a range of threshold probabilities in a patient population (Mei et al., 2022). As shown in Figure 3C, this model demonstrates a higher net benefit across a range of threshold probabilities, indicating significant clinical advantages.
External validation
3.5
A total of 100 patients were included in the validation cohort, of whom 38 experienced intraoperative hypothermia and 62 maintained normal body temperature during surgery. The overall incidence of intraoperative hypothermia was 38%. In contrast to the training cohort, the validation cohort results demonstrated that gender and basal body temperature were significantly associated with the occurrence of intraoperative hypothermia, whereas intraoperative blood loss showed a weaker association (Table 4).
Subsequently, a statistical analysis was performed on the 16 clinical characteristics in both the training and validation sets. As shown in Table 5, apart from american society of anesthesiologists (ASA) classification and irrigation fluid temperature, all other variables showed no statistically significant differences. It is possible that the difference of irrigation fluid temperature contributes to the slightly lower incidence of intraoperative hypothermia observed in the validation cohort compared to the training cohort.
To evaluate the stability and applicability of the constructed predictive model, we conducted an external validation of it. The ROC curve yielded an AUC of 0.831, with a sensitivity of 0.842 and a specificity of 0.774 (Figure 4A). The calibration curves showed good agreement between predicted and observed probabilities, with both internally and externally corrected curves closely aligning with the ideal diagonal line (Figure 4B). The results of DCA also indicated that this model has high clinical application value (Figure 4C).
Model validation by external data. (A) ROC curve, (B) calibration plots and (C) decision curve analysis of the intraoperative hypothermia nomogram prediction in the validation cohort. ROC, Receiver operating characteristic.
Discussion
4
The incidence of thyroid cancer is increasing significantly, particularly among female populations (Wiltshire et al., 2016; Chen et al., 2018). Surgery remains one of the primary treatment modalities for thyroid cancer (Kaliszewski et al., 2020). Endoscopic radical thyroidectomy, as a representative minimally invasive technique, has been widely adopted in clinical practice due to its favorable safety profile, rapid postoperative recovery, and reduced pain intensity (Bian et al., 2018). However, this procedure requires general anesthesia, which may impair the patient’s thermoregulatory function (Daniel, 2008). Furthermore, intraoperative administration of large volumes of unwarmed intravenous fluids and prolonged exposure of the patient to the cool operating room environment collectively contribute to an increased risk of intraoperative hypothermia (Leslie and Sessler, 2003). Therefore, identifying high-risk patients for intraoperative hypothermia among individuals with thyroid cancer undergoing endoscopic radical surgery, and proactively implementing evidence-based thermal protection nursing interventions, can enhance surgical outcomes and facilitate postoperative recovery.
Currently, nomograms have been widely applied in the fields of oncology and internal medicine due to their user-friendly digital interfaces, higher predictive accuracy, and intuitive expression of prognosis information, becoming an important tool for assisting clinical decision-making. Dong et al. (2022) developed a nomogram to predict the risk of postoperative delirium in patients undergoing laparoscopic surgery for gynecologic cancers. The model exhibited robust predictive accuracy, with an area under the receiver operating characteristic curve of 0.833. Additionally, previous studies have demonstrated the application of nomogram models in predicting the risk of intraoperative hypothermia across various surgical procedures, including laparoscopic radical resection for colorectal cancer (Yan et al., 2023), video-assisted thoracoscopic lobectomy (Xia et al., 2025), and lung transplantation (Huang et al., 2024), among others. However, research on predictive models for intraoperative hypothermia in patients undergoing endoscopic radical thyroidectomy remains limited. This study represents the first application of nomograms in endoscopic thyroid surgery for predicting the risk of intraoperative hypothermia, offering a novel quantitative approach to perioperative temperature management.
In the study, age, BMI, anesthesia duration, duration of surgery, infusion volume, intraoperative irrigation volume, irrigation fluid temperature and intraoperative blood loss were found to be closely related to the occurrence of intraoperative hypothermia in training cohort. The results exhibit a slight discrepancy compared to the validation cohort, which may be attributed to differences in data sources. Further research indicates that fluid infusion volume and irrigation fluid temperature are independent risk factors for intraoperative hypothermia, whereas BMI serves as an independent protective factor.
Due to preoperative fasting and substantial intraoperative fluid loss, large volumes of room-temperature or refrigerated intravenous fluids and blood products are commonly administered to restore circulating blood volume, maintain internal homeostasis, and ensure hemodynamic stability. However, the infusion of large quantities of such fluids not only increases metabolic heat demand but also results in direct thermal loss due to their subphysiological temperatures, thereby contributing to the development of intraoperative hypothermia (Okada et al., 2020). Similar to previous studies, the passage of irrigation fluid through the patient’s body contributes to increased heat loss, thereby predisposing patients to shivering and intraoperative hypothermia at room temperature (Liu et al., 2025). Moreover, the present study demonstrates that a higher BMI is linked to a protective effect against hypothermia in patients, consistent with evidence from prior research (Yi et al., 2017). According to the gradient theory, heat produced by metabolic activity is transferred from the core region to peripheral tissues and subsequently dissipated into the external environment (Emmert et al., 2018). Therefore, a higher BMI may confer protective benefits, as adipose tissue serves as an effective thermal insulator, attenuating heat loss during the conduction process (Groene et al., 2019).
Based on the above three important factors, the study developed a highly accurate prediction model for intraoperative hypothermia in patients undergoing endoscopic thyroidectomy for cancer. External validation demonstrated strong discriminatory ability and excellent calibration, indicating that the nomogram has robust applicability and high precision in similar surgical settings. For patients at high risk of intraoperative hypothermia, the implementation of comprehensive, individualized warming interventions is recommended to effectively mitigate the likelihood of temperature decline during surgery.
It should be pointed out that this study has certain limitations. Firstly, this study is retrospective in design, and the analysis of selected risk factors did not encompass all variables potentially influencing intraoperative hypothermia. Certain plausible factors (such as psychological stress or preoperative anxiety) were not captured during data collection. Future efforts will focus on expanding data acquisition to include these and other overlooked variables, enabling iterative refinement and enhanced generalizability of the predictive model. Secondly, the sample size was limited, lacking large-scale data to further verify the stability and generalization ability of the prediction model. Subsequent multi-center large data validation is needed. Thirdly, the study is limited by a narrow methodological approach. Future work will incorporate advanced machine learning techniques to enable a more comprehensive evaluation of the model’s performance and predictive robustness. Finally, unlike previous research results, this study did not list age and comorbidities as risk factors, and this difference needs to be confirmed in future studies (Ozaki et al., 1997). Therefore, it is still necessary to conduct well-designed, multi-center, large-sample studies to further improve the accuracy and clinical applicability of the risk prediction model.
Conclusion
5
In conclusion, this study developed a nomogram model with excellent predictive performance, which can assist nursing staff in assessing the risk of intraoperative hypothermia among patients undergoing endoscopic thyroid surgery. This capability enables timely implementation of proactive warming interventions to maintain stable intraoperative body temperature.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Bian C. Liu H. Yao X.-Y. Wu S.-P. Wu Y. Liu C. (2018). Complete endoscopic radical resection of thyroid cancer via an oral vestibule approach. Oncol. Lett. 16 (5), 5599–5606. 10.3892/ol.2018.9369 30344714 PMC 6176244 · doi ↗ · pubmed ↗
- 2Bindu B. Bindra A. Rath G. (2017). Temperature management under general anesthesia: compulsion or option. J. Anaesthesiol. Clin. Pharmacol. 33 (3), 306–316. 10.4103/joacp.JOACP_334_16 29109627 PMC 5672515 · doi ↗ · pubmed ↗
- 3Bleyer A. Viny A. Barr R. (2006). Cancer in 15- to 29-year-olds by primary site. Oncologist 11(6):590–601. 10.1634/theoncologist.11-6-590 16794238 · doi ↗ · pubmed ↗
- 4Campbell G. Alderson P. Smith A. F. Warttig S. (2015). Warming of intravenous and irrigation fluids for preventing inadvertent perioperative hypothermia. Cochrane Database Syst. Rev. 2015 (4), CD 009891. 10.1002/14651858.CD 009891.pub 2 25866139 PMC 6769178 · doi ↗ · pubmed ↗
- 5Chen C. Huang S. Huang A. Jia Y. Wang J. Mao M. (2018). Total endoscopic thyroidectomy versus conventional open thyroidectomy in thyroid cancer: a systematic review and meta-analysis. Ther. Clin. Risk Manag. 14, 2349–2361. 10.2147/TCRM.S 183612 30584310 PMC 6287425 · doi ↗ · pubmed ↗
- 6Cho S.-A. Lee S.-J. Yoon S. Sung T.-Y. (2022). Risk factors for postoperative hypothermia in patients undergoing robot-assisted gynecological surgery: a retrospective cohort study. Int. J. Med. Sci. 19 (7), 1147–1154. 10.7150/ijms.73225 35919810 PMC 9339412 · doi ↗ · pubmed ↗
- 7Cho S.-A. Lee S.-J. Kwon W. Jung J.-Y. You H.-J. Yoon S.-E. (2024). Effect of remimazolam on the incidence of intraoperative hypothermia compared with inhalation anesthetics in patients undergoing endoscopic nasal surgery: a prospective randomized controlled trial. Int. J. Med. Sci. 21 (13), 2510–2517. 10.7150/ijms.100262 39439451 PMC 11492874 · doi ↗ · pubmed ↗
- 8Daniel I. (2008). Sessler. Temperature monitoring and perioperative thermoregulation. Anesthesiology 109 (2), 318–338. 10.1097/ALN.0b 013e 31817 f 6d 76 18648241 PMC 2614355 · doi ↗ · pubmed ↗