A total of 80 donors who underwent hand-assisted laparoscopic living donor kidney resection at Beijing Friendship Hospital, Capital Medical University from September 2022 to March 2024 were selected by a retrospective study (60 cases in the modeling group and 20 cases in the external verification group). This study complies with the Declaration of Helsinki and is approved by the Ethics Committee of our hospital. All subjects have signed informed consent, and the batch number is: BFHHZS20240073.

Inclusion criteria: (1) The operators in the modeling group and the external validation group are the same, but the operators between the groups are different. Exclusion criteria: (1) The operation time was affected by drug allergy or other conditions during the operation; (2) Lack of preoperative CT images, partial laboratory results or other examinations leads to incomplete medical records. (3) Preoperative lithiasis was found in the donor kidney.

We adopted an improved hand-assisted retrolaparoscopy technique consistent with previous studies [11]. The main improvements include: (1) The combination of blunt separation and sharp scissors separation was used for kidney removal, and the intermittent hemostasis was performed by an ultrasonic knife. (2) After the renal artery, vein and ureter were fully free, the surgeon made a parallel rectus abdominalis incision in the lateral abdomen and entered the retroperitoneal space to re-establish the pneumoperitoneum. (3) The surgeon moderately pulled the renal artery and vein with his left hand, and then used 2 Hem-o-lock clamps on the proximal end of the renal artery and vein successively, and directly removed the kidney donor after cutting the renal artery and vein with scissors.

The subjective difficulty score of each hand-assisted laparoscopic living donor nephrectomy was recorded, and the whole operation was divided into the following four main steps: Trocar placement, separating the renal hilum, separating the perinephric region and kidney removal, and the time consumption of different steps were recorded respectively. Demographic data and perioperative clinical data were also collected, including gender, age, BMI, underlying diseases, smoking history, history of abdominal surgery, left side of the surgery, anatomical anomaly, preoperative blood lipid level, ASA score, postoperative hospitalization days, drainage tube retention time, pain visual analogue scale(VAS), etc. Meanwhile, in this study, donor kidney length, width, thickness, and volume, the number of renal arteries, the number of donor renal veins, perirenal fat thickness [12, 13] (see Fig. 1 for measurement methods), subcutaneous fat thickness (see Fig. 2 for measurement methods), and Mayo adhesive probability score were also included [12], sagittal abdominal diameter (see Fig. 3 for measurement method), distance of the 12th rib - iliac crest (see Fig. 4 for measurement method), distance of the 12th rib - the 12th rib (see Fig. 4 for measurement methods), transverse pelvic diameter [14] (see Fig. 4 for measurement methods), Agatston score of renal artery calcification [15], etc. Among them, the subjective difficulty score was used to score the overall difficulty of surgery after surgery (1–3 points), and the higher the score, the more difficult the surgery was.

SPSS 25.0 software was used to analyze and process the research data. For comparison between groups of quantitative data, Bonferroni method was used for multiple comparisons if the data met the normal distribution, and Kruskal-Wallis multiple local rank sum test was used if the data did not meet the normal distribution. Chi-square test and Fisher’s exact probability method were used for the comparison of multi-group rates and composition ratios. α segmentation method was used for the multiple comparison of Chi-square test, that is, R*C cross table was divided into n 2*2 four-cell tables, and the test level α became α/n. In terms of influencing factors of surgical difficulty, quantitative data consistent with normal distribution were presented as mean ± standard deviation (x ± s), and Pearson correlation test was used for analysis. The classified data is expressed as an example (%). Quantitative data that did not conform to normal distribution were expressed as the median (interquartile distance) [M(Q1,Q3)], and Spearman correlation test was used for classified data or quantitative data that did not conform to normal distribution. With the difficulty of surgery as the dependent variable and the above influencing factors as the independent variables, bivariate correlation analysis was carried out, and all the variables with statistically significant differences were included in the multifactor linear regression analysis, so as to obtain the prediction equation of the difficulty of surgery.

Then, the model is verified internally and externally to evaluate its accuracy. In internal verification, 100 patients in the modeling group were selected by random number table method, and their predicted surgical difficulty was calculated by regression equation, and paired T-test was performed with the actual value. An additional 20 patients with laparoscopic donor nephrectomy assisted by another surgeon were collected as an external validation set. The prediction of surgical difficulty was calculated by regression equation, and paired T-test was performed with the actual value. With the exception of α segmentation, all test methods were considered statistically significant with P < 0.05.