2.1 Patients selection and follow-up
In 2018, we conducted a prospective study to evaluate the development of peristomal dermatitis in patients who underwent ileostomy within one month (during hospitalization or at the outpatient department). The criteria for the diagnosis of peristomal dermatitis were developing symptoms of erythema, edema, possible vesicles, maceration, and loss of skin integrity due to contact with several substances including urine, feces, medicaments, and ostomy bag systems. The information included in this study was prospectively collected, while their surgical data was retrospectively extracted from the hospital information system (HIS) before this study was conducted. We screened 459 patients who underwent ileostomy for various primary diseases. Patients with missing values and inadequate information were excluded (Fig. 1), resulting in a total of 441 patients included in the model construction. With this sample size, the study achieved approximately 85% power to detect a difference of 0.100 in the area under the curve (AUC) between 0.600 and 0.700, using a two-sided significance level of 0.05. We included 18 independent variables for the construction of the nomogram, which are described as follows.
Diabetes
This variable indicates whether the patient has diabetes, regardless of medication use and disease management. It's a dichotomous variable.
Ileostomy positioning
This variable refers to the pre-ostomy positioning. The standardized principles of positioning were in accordance with the protocol [1]. Briefly, the ileostomy site should be flat, visible to the patient, and easily managed. Patients who received pre-ostomy positioning would have their ileostomy site marked.
Stoma care lecture
This variable indicates whether the patient received a lecture on how to care for their stoma during the hospitalization. The values of this variable include “Yes” and “No”.
Residence
This variable indicates whether the patient lives in Guangdong Province or not.
Depressed skin
This variable refers to whether the skin adjacent to the stoma is lower than the surrounding skin. The values of this variable include “Yes” and “No”.
Sex
This variable indicates the sex of the patients. The values of this variable include “Female” and “Male”.
Surgery data
We recorded the operation time as months. Due to the uneven distribution of the number of cases per month, we did not set dummy variables for this variable.
Job
This variable indicates whether the patient had a job before coming to the hospital.
Body mass index
Body Mass Index (BMI) is calculated as the weight in kilograms divided by the square of the height in meters.
Loop ileostomy
This variable indicates the type of ileostomy. If it is a loop ileostomy, it is labeled as “Yes”, otherwise it is labeled as “No”.
Stoma flow
This variable represents the amount of fluid that comes out of the stoma when the flow is relatively stable. If it is over 800ml per day, it is labeled as “Yes”, otherwise it is labeled as “NO”.
Stoma height
This variable refers to the height of the stoma relative to the surrounding skin observed on the first postoperative day. If is over 1cm, it is labeled as “Yes”, otherwise, “No”.
Primary surgery
This variable indicates the primary operation of the ostomy. The types of primary surgery are summarized in the footnote of Table 1. Due to the uneven distribution of t cases per surgery type, we did not set dummy variables.
Table 1
Demographic features of the patients
Peristomal dermatitis | No | Yes | P-value |
| N = 359 | N = 82 | |
Age | 58.0 [48.0;66.0] | 57.0 [50.0;65.0] | 0.805 |
Body mass index | 22.0 (3.54) | 22.6 (3.16) | 0.148 |
Sex: | | | 0.056 |
Male | 252 (70.2%) | 48 (58.5%) | |
Female | 107 (29.8%) | 34 (41.5%) | |
Primary surgery: | | | 0.440 |
Rectal cancer resection | 242 (67.4%) | 54 (65.9%) | |
Sigmoid cancer resection | 27 (7.52%) | 8 (9.76%) | |
Colon cancer resection | 23 (6.41%) | 2 (2.44%) | |
Others* | 67 (18.7%) | 18 (22.0%) | |
Ileostomy positioning: | | | 0.009 |
No | 38 (10.6%) | 18 (22.0%) | |
Yes | 321 (89.4%) | 64 (78.0%) | |
Loop ileostomy: | | | 0.452 |
No | 168 (46.8%) | 34 (41.5%) | |
Yes | 191 (53.2%) | 48 (58.5%) | |
Stoma flow: | | | 0.600 |
<800mL/day | 333 (92.8%) | 78 (95.1%) | |
≥800mL/day | 26 (7.24%) | 4 (4.88%) | |
Stoma care lecture: | | | 0.022 |
No | 235 (65.5%) | 65 (79.3%) | |
Yes | 124 (34.5%) | 17 (20.7%) | |
Stoma height; | | | 0.558 |
<1cm | 293 (81.6%) | 64 (78.0%) | |
≥1cm | 66 (18.4%) | 18 (22.0%) | |
Diabetes: | | | <0.001 |
No | 335 (93.3%) | 65 (79.3%) | |
Yes | 24 (6.69%) | 17 (20.7%) | |
Neoadjuvant chemotherapy: | | | 1.000 |
No | 233 (64.9%) | 53 (64.6%) | |
Yes | 126 (35.1%) | 29 (35.4%) | |
Neoadjuvant radiotherapy: | | | 1.000 |
No | 337 (93.9%) | 77 (93.9%) | |
Yes | 22 (6.13%) | 5 (6.10%) | |
Job: | | | 0.197 |
No | 83 (23.1%) | 13 (15.9%) | |
Yes | 276 (76.9%) | 69 (84.1%) | |
Residence: | | | 0.047 |
In Guangdong province | 273 (76.0%) | 53 (64.6%) | |
Outside Guangdong province | 86 (24.0%) | 29 (35.4%) | |
Education level: | | | 1.000 |
College degree or below | 309 (86.1%) | 71 (86.6%) | |
College degree and above | 50 (13.9%) | 11 (13.4%) | |
Marriage: | | | 0.776 |
Married | 340 (95.5%) | 78 (95.1%) | |
Unmarried | 16 (4.49%) | 4 (4.88%) | |
Depressed skin: | | | 0.046 |
No | 225 (62.7%) | 41 (50.0%) | |
Yes | 134 (37.3%) | 41 (50.0%) | |
Others included: 5 ulcerative colitis, 4 bowel obstruction caused by peritoneal cancer, 1 total colectomy by familial adenomatous polyposis, 1 subtotal colectomy by congenital megacolon, 1 sigmoidoscopy by sigmoid perforation, 1 Crohn’s disease, 1 radiation enteritis, 2 proctectomy by rectostenosis, 1 rectosigmoid junction cancer resection, 1 sigmoid lymphoma resection in the NO group; 34 bowel obstruction caused by peritoneal cancer, 10 total colectomy by familial adenomatous polyposis, 5 ulcerative colitis, 4 Crohn’s disease, 2 rectosigmoid junction cancer resection, 2 complicated anal fistula, 1 total colorectal resection by congenital megacolon, 1 sigmoid mesenchymoma resection, 1 rectectomy desmoid fibromatosis, 1 rectal stromal tumor resection, 1 solitary fibrous tumor resection, 1 peritonitis, 1 anterior sacral cyst surgery, 1 abdominal trauma, 1 intestinal allergic purpura, 1 perforation of intestinal tuberculosis in the YES group. |
Age
This variable indicates the age of the patients.
Marital status
This variable refers to whether the patients are married. It is labeled as “Married” or “Unmarried” according to their marital status.
Education level
This variable refers to the level of patients’ education. It is labeled as “above college degree” or “below college degree”.
Neoadjuvant chemotherapy
This variable indicates whether patients had received neoadjuvant chemotherapy. The values of this variable include “Yes” and “No”.
Neoadjuvant radiotherapy
This variable indicates whether patients had received neoadjuvant radiotherapy. The values of this variable include “Yes” and “No”.
The study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of the Sixth Affiliated Hospital of Sun Yat-sen University (NO.2020ZSLYEC-040). Informed consent was waived for this current study.
2.2 Statistical analysis
Continuous variables with a normal distribution ware presented with mean ± standard deviation, and Welch’s t-test was used to test their means between dermatitis and non-dermatitis group. For variables with skewed distribution, median and 1st, 3rd quartile was utilized to describe their statistical features and Mann-Whitney U test was used to test their means between dermatitis and non-dermatitis group. A two-sided significance level with P ≤ 0.05 was set for all analysis.
To explore the association between the odds ratio (OR) and body mass index (BMI), we used restricted cubic splines (RCS) and examined whether it was necessary to convert the variable type from continuous to categorical. These splines were knotted by 3 knots (33rd, 66th, and 99th percentiles). We selected 19 kg/m2 and 65 years as the reference (i.e., OR = 1), respectively. Statistical significance of the overall association and for the nonlinearity of the splines were evaluated with likelihood ratio tests [11]. The DeLong’ test was used to assess whether the AUCs of the two models are significantly different[12].
To improve the robustness and validity of the models, we randomly divided the 441 patients into a training set and a validation set with a ratio of 7:3 without replacement. Based on the training set, we preformed univariate and multivariate logistics regression to identify candidate factors with a P-value ≤ 0.05 for inclusion in the nomogram. According to Occam’s Law of Razor, the best model to achieve optimal results are models with fewer variables[13]. Therefore, we further screened the optimal nomogram model based on the area under the curve (AUC) of the receiver operating characteristic (ROC) analysis. In addition, the discrimination and calibration ability of the final nomogram model was depicted with AUC and calibration curve in both the training and validation set. Besides, the clinical utility of the nomogram model was assessed by the decision curve analysis (DCA). All data analysis were performed using R-studio (version 4.1.0 for windows).