We conducted a retrospective analysis of the data from patients with EP treated at our hospital between July 2022 and June 2024. The inclusion criteria were: ① diagnosis of EP according to relevant diagnostic criteria [7]; ② indications for hysteroscopic resection; ③ clear consciousness and no cognitive abnormalities; ④ complete clinical data available. The exclusion criteria were: ① presence of other gynecological diseases such as uterine fibroids; ② inability to tolerate hysteroscopic surgery due to severe systemic diseases; ③ abnormal immune system or associated infections; ④ hormonal therapy within the last 3 months prior to surgery; ⑤ women who are pregnant or breastfeeding. This study adhered to ethical medical standards.

The hysteroscopic resection of EP was performed by the same surgical team. The surgery was scheduled 3 to 5 days after the end of the menstrual period. The procedure was as follows: the patient was positioned in the lithotomy position, and routine disinfection of the surgical area was performed. Anesthesia was administered, and once it took effect, a speculum was inserted, followed by dilation of the cervix to 9–10 mm. A hysteroscope was then inserted to examine the uterine cavity, assessing the location, number, and base width of the polyps. The polyp tissue was excised using an electrosurgical knife along the vascular base of the polyp, ensuring the surrounding normal endometrium was preserved. Finally, the uterine cavity was irrigated. Postoperatively, prophylactic antibiotics were administered for 1 day.

A hysteroscopic examination was performed on all patients 6 months after their operation, to assess the presence of IUA. The criteria for determining IUA were as follows [8]: ① symptoms such as reduced menstrual flow or even amenorrhea, and periodic abdominal pain; ② hysteroscopic examination revealing central adhesions (anteroposterior wall adhesions with widened ends, with the occlusion of the uterine horn or loss of uterine cavity symmetry), peripheral adhesions (appearing as semi-curtain or crescent-shaped, with the occlusion of the uterine horn or loss of uterine cavity symmetry), and mixed adhesions (small obstructed cavities). Based on the presence or absence of IUA, the patients were divided into the IUA group and the non-IUA group.

Data were collected from our hospital’s electronic medical records system, including age, disease duration, body mass index (BMI), comorbidities, number of pregnancies, fertility desires, use of intrauterine devices (IUDs), history of pelvic inflammatory disease, history of cesarean section, history of dilation and curettage (D&C), history of pregnancy termination, history of polyp disease, endometrial hyperplasia status, polyp diameter, number of polyps, location of polyps, surgery duration, and intraoperative blood loss.

Data were analyzed using SPSS 25.0 software (IBM Corp., Armonk, NY, USA). Continuous variables that followed a normal distribution were described as mean $\pm$ standard deviation (SD) ($\overline{x}$ $\mathrm{\pm }$ s) and analyzed using t-test. Categorical variables were described as [n (%)] and analyzed using $\chi$² test. Logistic regression analysis was performed to identify factors influencing the occurrence of IUA in EP patients postoperatively. Variable selection was carried out using the stepwise forward selection method, with variables which had a p-value 0.100 in univariate analysis being included in the multivariate model. A nomogram model for predicting IUA occurrence was constructed using the rms package in R software version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria). Receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) was calculated to evaluate the model’s discrimination ability. The calibration curve was used to assess the model’s consistency, and decision curve analysis was performed to evaluate its clinical applicability. A p-value of 0.050 was considered statistically significant.

In this study, a total of 409 EP patients were initially collected from the electronic medical record system. After excluding 41 cases lost to follow-up, 46 cases were further excluded based on the exclusion criteria (34 cases with other gynecological diseases and 12 cases who had received hormonal treatment within the 3 months before their surgery). Ultimately, 322 EP patients were included in the study. These patients were randomly divided into a modeling group (n = 248) and a validation group (n = 74). A flowchart of the participants is shown in Fig. 1. In the modeling group of 248 patients, 51 developed IUA postoperatively, with an incidence rate of 20.56%. In the validation group of 74 patients, 14 developed IUA postoperatively, with an incidence rate of 18.92%.

Fig. 1.

Flow chart of subjects.

There were no significant differences in the data between the modeling group and the validation group (p $>$ 0.050, as shown in Table 1.

In the IUA group, there were significant differences compared to the non-IUA group in terms of the use of IUDs, history of pelvic inflammatory disease, history of D&C, history of pregnancy termination, and surgical duration (p 0.050), as shown in Table 2.

Univariate analysis showed that factors potentially associated with the occurrence of postoperative IUA in EP patients included the use of intrauterine devices, history of pelvic inflammatory disease, history of D&C, history of pregnancy termination, and surgical duration (p 0.050), as shown in Table 3.

Using the occurrence of IUA (no = 0, yes = 1) in EP patients after hysteroscopic resection as the dependent variable and the five factors identified as potentially significant in the univariate analysis (Table 3) as independent variables (assignments shown in Table 4), a multivariate logistic regression analysis was conducted. Variable screening was performed using the stepwise forward method. It was found that the use of intrauterine devices (OR = 5.316), history of pelvic inflammatory disease (OR = 4.059), history of D&C (OR = 5.314), history of pregnancy termination (OR = 3.514), and surgical duration (OR = 1.149) were influencing factors for the occurrence of IUA in EP patients postoperatively (p 0.050), as shown in Table 5.

A nomogram model was developed based on five key factors to predict the occurrence of IUA in patients with EP following hysteroscopic resection. The following points were assigned for each factor: 60 points for the use of intrauterine devices, 50.5 points for history of pelvic inflammatory disease, 59 points for history of D&C, 45 points for history of pregnancy termination, and for surgical duration, every additional 2 minutes beyond the baseline surgical time of 24 minutes, 10 points were assigned. When the score reached 44 points, the maximum score (100 points) was achieved. The total score corresponds to the predicted probability of IUA occurrence in EP patients after surgery, as shown in Fig. 2.

Fig. 2.

Construction of nomogram model for IUA occurrence after hysteroscopic resection in EP patients.

Internal validation showed that the model’s predicted AUC was 0.815 (95% CI: 0.753–0.876), with a sensitivity and specificity of 80.40% and 70.10%, respectively (see Fig. 3A). The calibration curve demonstrated good fit (see Fig. 3B), and the Hosmer-Lemeshow (HL) test yielded $\chi$² = 9.598, p = 0.294, suggesting excellent predictive performance of the model. The decision curve analysis showed that the net benefit was highest when the model’s threshold probability ranged from 6% to 78% (see Fig. 3C).

Fig. 3.

Internal validation of the nomogram model. (A) Receiver operating characteristic (ROC) curve for internal validation of the model. (B) Calibration curve for internal validation of the model. (C) Decision curve of the modeling group.

External validation showed that the model’s predicted AUC was 0.808 (95% CI: 0.747–0.870), with a sensitivity and specificity of 86.30% and 65.00%, respectively (see Fig. 4A). The calibration curve demonstrated good fit (see Fig. 4B), and the Hosmer-Lemeshow (HL) test yielded $\chi$² = 9.701, p = 0.282, indicating excellent predictive performance of the model. The decision curve showed that the net benefit was higher when the threshold probability of the model was between 8% and 77% (see Fig. 4C).

Fig. 4.

External validation of the model. (A) ROC curve for external validation of the model. (B) Calibration curve for external validation of the model. (C) Decision curve of the verification group.