A retrospective analysis was conducted on patients who underwent hysteroscopic adhesiolysis at our hospital from October 2018 to September 2021. This study was approved by the Ethics Committee of Xuancheng Central Hospital. Inclusion criteria: (1) Patients diagnosed with uterine adhesions through hysteroscopy before surgery; (2) Patients with a history of sexual activity; (3) Patients without contraindications for hysteroscopic surgery; (4) The severity of adhesions was classified as Grade I–II in all cases (9). (5) Patients who signed an informed consent form. Exclusion criteria: (1) Patients with coagulation disorders; (2) Patients unable to undergo hysteroscopy due to excessive uterine curvature; (3) Patients with cervical adhesions; (4) Patients with consciousness disorders or mental illnesses; (5) Patients with tuberculous uterine adhesions.

All patients underwent hysteroscopic adhesiolysis. Preoperatively, vaginal cleansing was performed 3 days before surgery, and patients were instructed to fast with water restriction starting 6 h prior to surgery. Misoprostol 600 μg was inserted into the vagina 2 h before surgery for cervical softening. After anesthesia, the patient was positioned in the lithotomy position. The surgical area was disinfected, and the cervical canal was dilated to a size of 10–12. Physiological saline was injected into the uterus at a rate of 260–450 mL/min to expand it, while maintaining a pressure of 100–150 mmHg. Depending on the location, severity, and density of adhesions, separation was performed using hysteroscopic electrodes, micro-scissors, or direct-tip separation to avoid damaging the endometrium. Postoperative oral antibiotics were prescribed for 1 week to prevent infection. All procedures in this study were conducted by a single experienced surgeon at the same medical center.

This study obtained the required data through a retrospective review of the hospital’s electronic medical record system and medical documentation system. Data collection was conducted by three clinical nurses from the research team, all of whom possessed full-time bachelor’s degrees and had at least 5 years of experience in obstetrics and gynecology. To ensure consistency, the three data collectors independently extracted records for the first five subjects in both the recurrence and non-recurrence groups. Any discrepancies were then discussed, and the first author and corresponding author standardized the data collection criteria. Data collection was finalized once consensus was reached within the research team. Collected data included the number of preoperative intrauterine procedures (i.e., intrauterine surgery performed due to underlying disease), the number of pregnancies and deliveries, preoperative menstrual conditions, the number of pregnancy terminations, the duration of illness (based on a comprehensive assessment of medical history, symptoms, and auxiliary examinations), and age. The extent and nature of uterine adhesions were assessed by hysteroscopy, along with postoperative placement of an intrauterine device (not placed if postoperative recovery was good and no intrauterine infection occurred; otherwise, it was placed) and the duration of surgery.

Follow-up was conducted for 1 year post-surgery through phone calls, both outpatient and inpatient visits. The follow-up period for the modeling group ended in October 2021, and for the validation group in September 2022. The follow-up rate was 100%. Monthly follow-ups were conducted for three consecutive months, followed by quarterly follow-ups. Patients’ menstrual recovery and recurrence status (after postoperative menstrual improvement, hypomenorrhea or amenorrhea reappeared, with definite adhesion formation and AFS score ≥1) were assessed. Patients who experienced recurrence were instructed to seek timely medical consultation.

SPSS 25.0 software was used for the statistical analysis of the data in this study. Quantitative data were expressed as ( x ¯ ± s , mean ± standard deviation) and analyzed using the t-test; categorical data were expressed as cases (%) and analyzed using the Chi-square test (χ²). The Receiver Operating Characteristic (ROC) curve analysis was employed to assess the optimal cutoff value for disease duration. In the modeling group, multivariate logistic regression analysis was used to identify factors affecting postoperative recurrence. The prediction model, constructed using nomograms, was developed using R software. The model’s predictive validity was evaluated using the Hosmer–Lemeshow (H–L) goodness-of-fit test, and its discriminative ability was assessed using the ROC curve. The clinical utility of the model was evaluated using decision curve analysis (DCA). A p-value of <0.05 was considered statistically significant.

A retrospective review was conducted of 434 patients who underwent hysteroscopic adhesiolysis for intrauterine adhesions at our hospital between October 2018 and September 2021. Among these patients, 12 with coagulation disorders, 8 in whom the hysteroscope could not reach the uterine fundus due to excessive uterine curvature, 15 with cervical adhesions, 5 with impaired consciousness or psychiatric disorders, 7 with tuberculous intrauterine adhesions, and 3 who did not provide written informed consent were excluded. Ultimately, 384 patients were included in the present study. Patients were grouped according to the time of surgery: the modeling group comprised patients who underwent surgery between October 2018 and October 2020 (n = 256), and the validation group comprised patients treated between November 2020 and September 2021 (n = 128). There were no missing data in either group, and the follow-up rate was 100%. The postoperative recurrence rate of intrauterine adhesions after hysteroscopic adhesiolysis was 20.31% (52/256) in the modeling group and 23.44% (30/128) in the validation group. See Figure 1.

To ensure the validity of external validation, baseline characteristics of patients in the modeling and validation groups were compared (Table 1). The results showed no statistically significant differences between the two groups in disease duration, number of induced abortions, number of preoperative intrauterine procedures, extent of adhesions, or postoperative intrauterine device placement (all p > 0.05), indicating good homogeneity between the groups and suitability for external validation.

The postoperative recurrence rate of uterine adhesion separation in the modeling group was 20.31% (52/256). There were statistically significant differences between the two groups in terms of duration of illness, number of induced abortions, number of preoperative intrauterine procedures, extent of adhesions, and postoperative intrauterine device placement (p < 0.05). However, there were no significant differences in the number of pregnancies, number of deliveries, age, preoperative menstrual conditions, nature of adhesions, and operation time between the two groups (p > 0.05). See Table 2.

Variables with p < 0.05 in Table 2—including disease duration, number of induced abortions, number of preoperative intrauterine procedures, extent of adhesion, and postoperative intrauterine device (IUD) placement—were included in the multivariate logistic regression analysis using a stepwise forward method. The results showed that a disease duration >12 months and 12 days, more than two induced abortions, more than one preoperative intrauterine procedure, and an adhesion extent ≥1/2 were independent risk factors for recurrence after hysteroscopic adhesiolysis (p < 0.05). See Table 3.

A nomogram for postoperative recurrence of uterine adhesions was constructed using the results of the multivariate logistic regression analysis with R software. A duration of illness >12 months and 12 days scored 52.5 points, more than 2 induced abortions scored 100 points, more than 1 preoperative intrauterine operation scored 53 points, and an adhesion range ≥1/2 scored 62.5 points. See Figure 2.

The predictive validity of the model was assessed using the Hosmer–Lemeshow (H–L) goodness-of-fit test. The results showed good consistency between predicted and actual values (Figure 3A), with χ² = 6.427, p = 0.316. The discriminative ability of the predictive model was evaluated using the ROC curve, showing an area under the curve of 0.767 (95%CI: 0.689–0.845), with specificity and sensitivity of 77.45 and 65.38%, respectively (Figure 3B).

The postoperative recurrence rate of uterine adhesion separation in the validation group was 23.44% (30/128). There were statistically significant differences between the two groups in terms of duration of illness, number of induced abortions, number of preoperative intrauterine procedures, adhesion range, and whether insertion of an IUD after surgery (p < 0.05). There were no statistically significant differences between the two groups in terms of gravidity, parity, age, preoperative menstrual status, adhesion characteristics, and operation time (p > 0.05), as shown in Table 4. External validation results demonstrated good consistency between predicted and actual values in the H–L goodness-of-fit test, with χ² = 7.006, p = 0.352 (Figure 4A); the area under the ROC curve was 0.779 (95%CI: 0.706–0.852), with specificity and sensitivity of 77.50 and 65.40%, respectively (Figure 4B).

Decision curve analysis was performed with threshold probability on the x-axis and net benefit on the y-axis. In the modeling group, the nomogram demonstrated high clinical utility for predicting postoperative recurrence of intrauterine adhesions when the threshold probability ranged from 0.02 to 0.89 (Figure 5A). In the validation group, the nomogram similarly showed high clinical utility across threshold probabilities of 0.02–0.92 (Figure 5B).