Linear Regression Model for Predicting Dressing-Change Pain in Diabetic Foot Ulcers: Development, Internal Validation, and Personalized Analgesic Care

Summary

This study developed and validated a linear regression model to predict pain during diabetic foot ulcer dressing changes in 232 patients. Ulcer area, disease duration, and Self-Rating Anxiety Scale (SAS) were key predictors. A model-based personalized analgesic regimen reduced pain, improved compliance, and enhanced ulcer healing, supporting precision pain management strategies.

Abstract

This study included 232 patients with diabetic foot ulcers (DFUs) to construct and validate a linear regression model to predict pain during dressing changes and to evaluate the effectiveness of a model-based personalized analgesic care regimen. In stage 1,160 patients were divided into a training set (n = 113) and an internal validation set (n = 47); multivariable linear regression identified ulcer area, ulcer duration, duration of diabetes, and the Self-Rating Anxiety Scale (SAS) score as predictors of continuous dressing-change VAS. The model yielded an adjusted R² of 0.458 and an RMSE of 1.719 in the training set, while the validation-set RMSE, calibration intercept, calibration slope, and AUC for identifying high pain were 2.058, 1.414, 0.721, and 0.913, respectively. In stage 2 (72 cases), subjects were assigned to a control group and an intervention group. The intervention group received a personalized analgesic regimen based on model-derived risk stratification, while the control group received conventional analgesic care. After 4 weeks, the intervention group had a lower dressing-change VAS score than the control group (3.83 ± 1.61 vs 4.94 ± 1.91, P = 0.010), together with a higher overall adherence rate (86.11% vs 63.89%, P = 0.029), a higher ulcer area reduction rate (41.84% ± 11.01% vs 26.22% ± 7.55%, P < 0.001), and a shorter wound-healing time (32.0 [29.0–35.0] vs 38.0 [34.0–43.5] days, P < 0.001). The model showed useful discrimination for dressing-change pain, and the model-based personalized analgesic regimen improved short-term pain, adherence, and wound-related outcomes; however, external validation and further refinement of model calibration are required before routine clinical implementation.

Introduction

Diabetic foot ulcers (DFUs), one of the most serious chronic complications of diabetes mellitus, are associated with high morbidity, disability, and healthcare costs, substantially impairing patients' quality of life and increasing the burden on healthcare systems¹. Regular wound dressing changes are an essential component of DFU management because they facilitate wound cleansing, infection control, and tissue repair. However, dressing-change pain remains a frequent clinical challenge that may reduce treatment adherence, increase patient anxiety, and adversely affect wound healing outcomes^2,3.

Pain during dressing changes is commonly assessed using the Visual Analog Scale (VAS), which is a simple and widely accepted tool for measuring pain intensity. Nevertheless, VAS reflects the patient's pain after or during the procedure and cannot predict the expected level of pain before dressing changes. In addition, subjective pain assessment may be influenced by individual clinical and psychological characteristics, limiting its ability to support proactive analgesic planning^4,5.

Previous studies have demonstrated that dressing-change pain in patients with DFUs is associated with multiple clinical factors, including ulcer area, ulcer depth, infection status, duration of diabetes, and psychological characteristics such as anxiety. Although these investigations have improved understanding of pain-related factors, most have focused on individual predictors or exploratory analyses, and only limited attention has been given to developing and validating multivariable prediction models suitable for clinical application. Furthermore, personalized analgesic care strategies guided by prediction-model outputs have rarely been evaluated^6,7.

Prediction models based on routinely available clinical variables may assist clinicians in identifying patients at increased risk of dressing-change pain before wound care and facilitate individualized analgesic planning. However, models intended for clinical use require appropriate validation, and evidence from single-center studies with internal validation should be interpreted cautiously until external validation is performed.

Therefore, this study aimed to develop and internally validate a linear regression model for predicting dressing-change pain in patients with diabetic foot ulcers using routinely collected clinical variables. In addition, the study evaluated whether a model-guided personalized analgesic care regimen could improve short-term pain control, treatment adherence, and wound-healing outcomes (Table 1). By combining prediction model development with clinical evaluation of a personalized analgesic strategy, this study provides preliminary evidence supporting individualized pain management while acknowledging the need for further multicenter external validation before routine clinical implementation.

Protocol

This study was approved by the Ethics Committee of The Second People's Hospital of Liaocheng. Written informed consent was obtained from all participants before enrollment. The study was conducted at The Second People's Hospital of Liaocheng between December 2021 and October 2025. The reagents and the equipment used are listed in the Table of Materials.

Study subjects
Adult patients presenting with diabetic foot ulcers (DFUs) were screened according to the Practical Guidelines on the Prevention and Management of Diabetic Foot Ulcers. Participants were confirmed to be at least 18 years of age, to have a DFU requiring routine dressing changes at least twice weekly, to have had an ulcer for at least 1 week, and to be capable of understanding and completing pain assessments. Patients with severe organ failure, malignant disease, autoimmune disorders, psychiatric illness, cognitive impairment, allergy to the study analgesics, recent participation in another clinical trial, or severe infection requiring emergency surgery were excluded. The study procedures were explained to all eligible participants, and written informed consent was obtained before enrollment. Each participant was assigned a unique study identification number, and demographic and clinical information were recorded using a standardized case-report form.

Baseline clinical data collection
Baseline clinical data, including age, sex, body mass index, duration of diabetes, HbA1c, ulcer duration, ulcer area, ulcer depth, infection status, educational level, history of chronic pain, and Self-Rating Anxiety Scale (SAS) score, were recorded before the first dressing change. All collected data were independently verified by two trained investigators. Any discrepancies were resolved through source-document verification. The verified data were then entered into the electronic database using a double-entry verification process.

Standardized wound dressing procedure
All sterile dressing materials were prepared before the procedure was initiated. Participants were positioned comfortably with the affected foot fully exposed. The previous dressing was removed gently to minimize tissue trauma. The wound was thoroughly irrigated using sterile normal saline maintained at 22 °C ± 2 °C. Loose necrotic tissue was debrided using sterile instruments when clinically indicated. Antibacterial ointment was applied according to institutional wound-care guidelines. The wound was then covered with sterile gauze, and the dressing was secured. All dressing changes were performed using the same standardized technique by the same certified wound-care nurse whenever possible. Dressing changes were carried out twice weekly throughout the study period. Complete wound coverage and secure dressing placement were confirmed before the procedure was concluded.

Assessment of dressing-change pain
Immediately after each dressing change was completed, participants were instructed to rate their pain using the Visual Analog Scale (VAS). The VAS was explained, with a score of 0 representing no pain and a score of 10 representing the worst imaginable pain. Pain scores were recorded after three consecutive dressing changes, and the mean VAS score was calculated for each participant. The mean VAS score was used as the final dressing-change pain score. Participant understanding of the VAS scoring method was confirmed before the final score was recorded.

Wound photography
The wound was cleaned before image acquisition. Participants were positioned comfortably with the wound fully exposed. A sterile metric ruler was placed adjacent to the wound for image calibration. Wound photographs were captured using the same digital camera under identical lighting conditions. The camera was positioned perpendicular (90°) to the wound surface and maintained at a constant distance of approximately 30 cm from the wound. All photographs were saved in JPEG format without compression and labeled with the participant identification number and assessment date. The entire wound margin and the calibration ruler were confirmed to be clearly visible before each image was accepted.

Ulcer area measurement using ImageJ
Each wound photograph was opened using ImageJ software. The image scale was calibrated using the metric ruler included in the photograph by selecting Analyze → Set Scale. The entire wound margin was carefully traced using the Polygon Selection tool, and the ulcer area was measured by selecting Analyze → Measure. The ulcer area was recorded in cm². Each measurement was repeated twice, and the average value was used for statistical analysis. Image analysis was repeated whenever the wound margins could not be clearly identified.

Personalized analgesic care regimen
Participants predicted to have a Visual Analog Scale (VAS) score of <4 were classified as low risk and received reassurance, routine wound care, gentle dressing techniques, and music therapy during dressing changes. Routine prophylactic analgesics were not administered to this group.

Participants predicted to have a VAS score of 4–6 were classified as moderate risk. Oral ibuprofen sustained-release capsules (0.3 g) were administered 30 min before each dressing change. Guided abdominal breathing exercises were performed, and a periwound cold compress maintained at 4–8 °C was applied for 10 min before dressing changes.

Participants predicted to have a VAS score of ≥7 were classified as high risk. Oral ibuprofen sustained-release capsules (0.3 g) were administered 30 min before dressing changes, and cognitive-behavioral preparation and distraction techniques were provided. If the VAS score remained ≥7, intravenous flurbiprofen axetil (50 mg) was administered as rescue analgesia according to institutional guidelines. All assigned interventions were completed before every scheduled dressing change. Completion of the assigned analgesic intervention was confirmed before the dressing procedure was initiated.

Outcome assessment
Visual Analog Scale (VAS) scores were recorded at baseline and after 4 weeks. Ulcer area was measured at baseline and after 4 weeks using ImageJ. The percentage reduction in ulcer area was calculated using the following formula: Ulcer-area reduction (%) = [(Baseline ulcer area − Week 4 ulcer area) / Baseline ulcer area] × 100. Wound-healing time was recorded from the initiation of the intervention until complete epithelialization. Treatment adherence was monitored throughout the study, and adverse events, including gastrointestinal discomfort, dizziness, and somnolence, were recorded. Complete epithelialization was independently confirmed by two experienced wound-care clinicians before wound-healing time was documented.

Quality control
All investigators were trained before study initiation. Standardized operating procedures were followed throughout the study. Clinical data were verified using an independent double-entry verification process. The clinical assessment environment was maintained at 22 °C ± 2 °C whenever feasible. Wound photographs were reviewed before ImageJ analysis to ensure adequate image quality. ImageJ measurements were performed by trained investigators who were blinded to treatment allocation.

Statistical analysis
All verified data were entered into the study database after completion of double-entry verification. Statistical analyses were performed using IBM SPSS Statistics Version 26.0. Continuous variables were expressed as the mean ± standard deviation (SD) or median (interquartile range [IQR]), according to the data distribution, whereas categorical variables were expressed as frequencies and percentages. Data normality was assessed before statistical testing. Univariate linear regression analysis was performed to identify variables associated with dressing-change pain, and eligible variables were included in a multivariable linear regression model according to the predefined selection criteria. The final prediction model was constructed using the regression coefficients obtained from the multivariable analysis. Internal validation was performed using bootstrap resampling with 1,000 iterations. The required sample size was calculated using PASS Version 15.0. A two-sided P < 0.05 was considered statistically significant. Data completeness was verified, and all assumptions for linear regression were confirmed before the final prediction model was reported.

Results

Univariate analysis of pain-influencing factors
All 160 stage-1 patients and all 72 stage-2 patients were included in the respective analyses. No primary-outcome data were missing, and no imputation was performed. In the model-development cohort, the mean dressing-change VAS score among the 160 patients was 6.08 ± 2.40. Using the median value as the cutoff, patients were stratified into a high-pain group (VAS > 6, n = 64) and a low-pain group (VAS ≤ 6, n = 96). Univariate analysis revealed no stat...

Discussion

This protocol describes the development, internal validation, and preliminary clinical evaluation of a linear regression model for predicting dressing-change pain in patients with diabetic foot ulcers (DFUs). By integrating routinely collected clinical variables with a risk-stratified analgesic care regimen, the protocol provides a standardized workflow for identifying patients who may experience greater dressing-change pain and implementing individualized pain-management strategies. Effective pain control during dressin...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Digital Camera	Canon (China) Co., Ltd. (China)	EOS 850D	Used to capture high-definition wound images for ulcer area measurement using ImageJ software
Electronic Weight Scale	Omron Healthcare (China) Co., Ltd. (China)	HN-289	Used to measure patients' weight for BMI calculation, which was included as a potential influencing factor
Flurbiprofen Axetil Injection	Beijing Tide Pharmaceutical Co., Ltd. (China)	20220820	Specification: 50mg/vial; additional intravenous administration when VAS ≥ 7 during dressing change in the high-risk group
Glycated Hemoglobin Analyzer	Roche Diagnostics Products (Shanghai) Co., Ltd. (China)	Cobas c513	Used to detect patients' HbA1c levels, which were included in diabetes-related influencing factor analysis
Ibuprofen Sustained-Release Capsules	Fenbid Pharmaceutical Co., Ltd. (China)	20220612	Specification: 0.3g/capsule; used for oral analgesia in the moderate-risk group and control group, administered 30 minutes before dressing change
ImageJ Image Analysis Software	National Institutes of Health (NIH, USA)	1.53k	Used for accurate measurement of wound area from captured images and calculation of ulcer area reduction rate
Intramuscular Injection Syringe	Jiangsu Changfeng Medical Technology Co., Ltd. (China)	5ml/piece	Used for intramuscular injection of Tramadol Hydrochloride Injection in the control group when severe pain occurs
Intravenous Syringe	Shandong Weigao Group Medical Polymer Products Co., Ltd. (China)	10ml/piece	Used for additional intravenous injection of Flurbiprofen Axetil Injection in the high-risk group
Medical Cotton Swabs	Sinocare Inc. (China)	20221208	Used for auxiliary wound cleaning and drug application to ensure sterile operation
Medical Stadiometer	Shanghai Rongshun Medical Technology Co., Ltd. (China)	RS-200	Used to measure patients' height for BMI calculation
Normal Saline (0.9% Sodium Chloride Injection)	Sichuan Kelun Pharmaceutical Co., Ltd. (China)	20221122	Specification: 0.9%; used for wound irrigation and cleaning to remove necrotic tissue
Paracetamol and Oxycodone Hydrochloride Tablets	Pfizer Pharmaceuticals Ltd. (USA)	20220715	Composition: 325mg Paracetamol + 5mg Oxycodone/tablet; used for oral analgesia in the high-risk group, administered 30 minutes before dressing change
Paracetamol Tablets	Sino-US Tianjin SmithKline & French Pharmaceutical Co., Ltd. (China)	20220506	Specification: 0.5g/tablet; used for oral analgesia in the low-risk group, administered 30 minutes before dressing change
PASS Sample Size Calculation Software	NCSS LLC (USA)	15	Used to calculate sample size for model development and protocol validation phases based on primary outcome measure (VAS score)
Pressure Algometer	Beijing Jiehui Technology Co., Ltd. (China)	JHPT-100	Used to measure patients' pain threshold and provide quantitative data for pain influencing factor analysis
Self-Designed Treatment Adherence Scale	self made (Cronbach’s α=0.86)	20220220	0-100 point scale; used to evaluate patients' adherence to dressing change, medication, and intervention (≥80 points = complete adherence)
Self-Rating Anxiety Scale (SAS)	Standardized Psychological Assessment Scales (Chinese Association for Mental Health)	20220215	Used to assess patients' anxiety level; quantitative SAS scores were included as an influencing factor
Silver Ion Antibacterial Ointment	Hangzhou Yiqingchuang Biotechnology Co., Ltd. (China)	20221016	Used for antibacterial protection of wounds; applied to wounds before dressing in both groups
SPSS Statistical Software	IBM Corporation (USA)	26	Used for data entry and statistical analysis (t-test, ANOVA, logistic regression, etc.)
Sterile Gauze	Zhende Medical Products Co., Ltd. (China)	20221105	Used for wound covering and dressing; core consumable for dressing change
Sterile Gauze Pad	Jiangsu Yuyue Medical Equipment & Supply Co., Ltd. (China)	20221110	Used for wrapping sterile ice packs to avoid direct skin contact during cold compress
Sterile Ice Pack	Qingdao Hainuo Biotechnology Co., Ltd. (China)	20220418	Used for local cold compress on wounds in the moderate-risk group; applied with sterile gauze wrapping to prevent skin frostbite
Tramadol Hydrochloride Injection	Grünenthal GmbH (Germany)	20220910	Specification: 100mg/vial; used as rescue analgesia via intramuscular injection in the control group when severe pain occurs
Visual Analog Scale (VAS) for Pain Assessment	Self-prepared (based on international standard templates)	20220210	0-10 point scale; used for patients' subjective pain assessment during dressing change; recorded 3 consecutive times with average value taken