In the experiment, a total of 883 women aged 26–62 years, with an average age of (39.87 $\pm$ 8.49) years, who had undergone physical examination from June 2021 to July 2021 in Meizhou people’s Hospital, were chosen as research objects. None of the included women were in menstruation or pregnancy, and all of them had no history of vaginal medication or antibacterial medication in the past 1 week, and had no sexual life in the past 3 days.

All participants in this study underwent vaginal examination by the same gynecologist. All included subjects had no history of vaginal irrigation within 24 hours, and underwent vaginal examination. Three cotton swabs were used to take the secretion of 1/3 section of vaginal wall for pH value detection, microscopic examination and amine test. Among them, one smear was added with normal saline to detect trichomonas, and one was sent to the laboratory for BV examination. BVBlue rapid detection method and Amsel gold standard method were used to detect polyamine. The other smear was heated, and then observed under a fixed mirror for clue cells.

According to the relevant standards of Obstetrics and Gynecology (9th Edition) [6]: ① increased vaginal secretion with thin character and fishy smell; ② positive amine test; ③ pH value of vaginal secretion above 4.5; ④ positive clue cells (prerequisite). BV was diagnosed when 3 of the above 4 items were positive.

A self-made questionnaire was used to investigate the information of patients, including age, education level, monthly income, abortion history, oral sex, sexual partner, whether to wash the vagina independently, sanitary pads used during menstruation, nursing pads used during non-menstruation, times of sexual activities per week, contraceptive methods, history of vaginitis, etc. All the patients agreed to participate voluntarily in the questionnaire survey. A total of 967 questionnaires were distributed, 883 of which were effectively recovered, with an effective recovery rate of 91.31%. The acquired data were checked and entered by two persons to ensure accuracy.

Data were processed by SPSS 22.0 software (IBM Corp., Armonk, NY, USA). Continuous variables conforming to normal distribution are expressed in the form of mean $\pm$ standard deviation ($\overline{x}$ $\pm$ S); classification variables are expressed by case number and rate (%), and $\chi$${}^2$ test was carried out. Taking the occurrence of BV as the dependent variable and the clinical data with statistical significance in univariate analysis as the independent variable, the risk factors affecting BV were screened by multivariate Logistic regression analysis with $\alpha$ = 0.05 as the test standard. Independent risk factors were introduced into R software (R3.6.3, R Foundation for Statistical Computing, Vienna, Austria), and rms package was applied to build a nomogram prediction model of BV risk. The Bootstrap method was used for internal verification by repeated sampling for 100 times. Receiver operating characteristic (ROC) curve and goodness-of-fit test were used to evaluate the discrimination and accuracy of the model.

There were 186 cases of BV among 883 women undergoing gynecological examination, and the incidence of BV was 21.06% (186/883). Among them, 17 cases were of grade I–II, accounting for 9.14%; 169 cases were above grade III, accounting for 90.86%; 156 cases were positive in amine test, accounting for 83.87%. There were 284 cases (32.16%) under 35 years old, 424 cases (48.02%) between 35 and 50 years old, and 175 cases (19.82%) over 50 years old. There were 346 cases with educational background of junior high school and below (39.18%), 305 cases (34.54%) with educational background of high school/technical secondary school, and 232 cases (26.27%) with educational background of junior college or above. There were 138 cases (15.63%) with monthly income less than ¥1000 (US $\$$138.66), 309 cases (34.99%) with monthly income between ¥1000 (US $\$$138.66) and ¥3000 (US $\$$415.99), and 436 cases (49.38%) with monthly income more than ¥3000 (US $\$$415.99). There were 93 cases unmarried (10.53%), 618 cases married (69.99%), and 172 cases with history of divorce (19.48%). There were 158 farmers (17.89%), 273 workers (30.92), 280 employees (31.71%) and 172 others (19.48%). There were 424 cases (48.02%) with the habit of wearing tights, and 232 cases (26.27%) who use nursing pads during non-menstrual period. There were 287 cases (32.50%) without abortion history, 380 cases (43.04%) with one abortion, and 216 cases (24.46%) with more than two abortions. There were 105 cases of oral sex (11.89%), 82 cases (9.29%) having no sexual partners, 692 cases (78.37%) having 1 partner, and 109 cases (12.34%) having 2 partners or more; there were 265 cases of self-vaginal irrigation (30.01%). There were 221 cases (25.03%) having sex 0–2 times per week, 483 cases (54.70%) having sex 3–4 times per week and 179 cases (20.27%) having sex more than 5 times per week. There were 392 cases (44.39%) of condom contraception, 164 cases (18.57%) of contraceptive contraception, 233 cases (26.39%) of intrauterine device (IUD) contraception and 94 cases (10.65%) of no contraception. There were 458 cases (51.87%) with history of vaginitis.

According to the occurrence of BV, 883 patients were divided into infected group (n = 186) and uninfected group (n = 697). Univariate analysis results showed that use of nursing pads during non-menstrual period, abortion history, oral sex, self-vaginal irrigation, frequency of sexual activity per week and contraceptive methods were the main factors affecting the occurrence of BV (p 0.05) as shown in Table 1.

Values were assigned to meaningful variables in single factor analysis, as shown in Table 2.

The significant variables of univariate analysis, including use of nursing pad during non-menstrual period, abortion history, oral sex, self-vaginal irrigation, frequency of sexual activity per week, contraceptive method, were taken as independent variables (see Table 2), and the index of BV occurrence was taken as a dependent variable. They were included in logistics regression equation for multivariate analysis. The results showed that use of nursing pad during non-menstrual period, abortion history $\ge$1 time, self-vaginal irrigation and frequency of sexual activity $\ge$5 times per week were the risk factors for BV occurrence (p 0.05) (see Table 3).

Based on the main influencing factors of BV obtained by multivariate logistic regression model, a nomogram model was established by using R software (Fig. 1), and the prediction probability associated with the sum of the index scores represents the predictive value for BV occurrence risk. To be specific, 35 points were assigned to the use of nursing pad during non-menstrual period, 39 points to abortion history once, 85 points to sexual activity frequency $\ge$2 times per week, 55 points to self-vaginal irrigation, 35 points to sex frequency of 3–4 times per week, 100 points to sex frequency $\ge$5 times per week, 34 points to contraceptive contraception, 47.5 points to intrauterine device contraception and 86 points to no contraception. If a patient uses nursing pad during non-menstrual period (35 points), has abortion history $\ge$2 times (100 points), does vaginal irrigation by himself/herself (55 points), has sexual activity frequency $\ge$5 times per week (100 points) and takes condom contraception (0 points), the total score would be 290 points, and the prediction probability is about 60%, and eventually this patient is evaluated as a high-risk patient.

Fig. 1.

Nomogram model of bacterial vaginosis (BV) incidence.

According to the prediction model formula, the risk prediction value of BV was calculated. The area under the ROC curve was used to evaluate the nomogram. The area under the curve was calculated to be 0.789 (95% confidence interval (CI): 0.751–0.827, Fig. 2A) in this work, indicating that the nomogram had good discrimination. The calibration curve of nomogram was plotted, and the slope of calibration curve was close to 1 (Fig. 2B). Hosmer-Lemeshow goodness-of-fit test showed = 6.408 and p = 0.602, indicating that the nomogram model had high degree of consistency in predicting BV infection risk. The decision curve analysis (DCA) curve shows that the nomogram model can make valuable judgments and has good clinical applicability (Fig. 2C).

Fig. 2.

Verification of nomogram model. (A) ROC curve analysis of nomogram model for predicting BV risk. (B) Calibration curve of nomogram model. (C) DCA curve of BV infection risk nomogram prediction model. ROC, receiver operating characteristic; BV, bacterial vaginosis; DCA, decision curve analysis