Prognostic Models for Mortality in Patients with Diabetic Foot Ulcers: A Systematic Review

Abstract

Objective:

This study aims to systematically review and summarize existing prognostic models for all-cause mortality in patients with diabetic foot ulcers (DFUs). It further evaluates their characteristics, performance, and methodological quality to inform future model development and clinical application.

Approach:

We searched eight databases from their inception to February 24, 2025. Extracted data included general study characteristics and model-specific information. The prognostic models were classified using the transparent reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis guidelines and evaluated using the Prediction Model Risk of Bias Assessment Tool (PROBAST) to assess the risk of bias (ROB) and applicability.

Results:

Of the 1,580 retrieved studies, 19 were included, reporting a total of 22 prognostic models. Reported mortality rates among DFU patients ranged from 4.2% to 48.5%, with follow-up durations varying from 12 weeks to 22 years. All studies exhibited a high ROB due to issues with data sources, small sample sizes, inadequate data handling, and model validation flaws. Across the included prognostic models, the most frequently reported predictors were age, cardiovascular disease, renal disease, ulcer severity, ischemia, and serum albumin.

Innovation:

Prognostic assessment of mortality risk in patients with DFUs is critical for effective clinical management. This study systematically evaluated 22 published prognostic models using the PROBAST, identifying key methodological limitations and commonly used predictors. The findings provide a scientific basis for the future development of higher quality, clinically applicable prognostic models. In addition, this work offers valuable insights to inform clinical risk stratification and decision-making for DFU patients.

Conclusions:

All included prognostic models demonstrated a high ROB according to the PROBAST assessment. Future research should focus on using larger sample sizes, rigorous study designs, and multicenter external validation to develop more reliable prognostic models. Given the multifactorial nature of mortality risk in DFU patients, multidisciplinary collaboration is essential to improve clinical outcomes.

Keywords

diabetic foot ulcers mortality prognostic model systematic review

INTRODUCTION

In recent years, the global prevalence of diabetes has continued to rise. In 2021, an estimated 537 million people worldwide were living with diabetes. This number is projected to reach 643 million by 2030 and 783 million by 2045.¹ China has the highest number of diabetes cases globally, with approximately 140 million individuals aged 20–79 affected.² It is estimated that 19–34% of individuals with diabetes will develop diabetic foot ulcers (DFUs) during their lifetime.³ DFUs are common and serious complications of diabetes. They mainly result from lower extremity neuropathy or vascular disease, which can lead to local infection, ulceration, and even deep tissue necrosis.⁴ Globally, the prevalence of DFUs is approximately 6.3%.⁵

Patients with DFUs face high risks of recurrence and amputation. The 5-year mortality rate was as high as 50–70%,⁶ comparable with that of many cancers.⁷ Moreover, studies show that the mortality risk in DFU patients is 2.5 times higher than in diabetic patients without foot ulcers.⁸

Although DFUs primarily manifest as a localized condition, their clinical prognosis is significantly influenced by systemic complications.⁹ Brennan et al. found a strong association between the severity of DFUs and increased mortality risk.¹⁰ The presence of peripheral artery disease (PAD) further increases the mortality rate in this population.¹¹ A systematic review identified cardiovascular disease and infection as the leading causes of death among patients with DFUs.¹² In addition, poor nutritional status may further increase the risk of both mortality and amputation.¹³

Accurate prediction of mortality risk in patients with DFUs is essential for early intervention and personalized treatment. Clinical prognostic models estimate health outcome probabilities and support patient counseling and clinical decision-making.¹⁴ A clinical prognostic model is a statistical tool that combines two or more predictors—such as patient characteristics, clinical findings, or laboratory results—to estimate the probability of a specific outcome in an individual. Common forms include risk scores and nomograms. A risk score assigns numerical points to each predictor based on its weight in the model, and the total score corresponds to a predicted probability of the outcome.¹⁵ In contrast, a nomogram is a graphical tool that translates a multivariable regression model into a continuous scoring system, allowing clinicians to visually calculate the precise probability of an endpoint—such as disease progression or mortality—for an individual patient.¹⁶ Well-developed and validated prognostic models for DFU patients can help bridge information gaps, improve the accuracy of mortality estimates, and provide health care professionals with evidence-based decision support. This, in turn, facilitates the management of modifiable risk factors and enhances long-term patient outcomes.¹⁷ In recent years, research on DFU mortality prognostic models has grown; however, the quality of existing models varies significantly. Therefore, this study aims to systematically review and synthesize current DFU mortality prognostic models, assess their potential ROB, and provide guidance for selecting or developing high-performance models to support clinical prognosis management.

CLINICAL PROBLEM ADDRESSED

DFU is one of the most common and serious chronic complications of diabetes, associated with a high mortality rate. Although primarily localized, DFUs often coexist with systemic issues such as cardiovascular disease, PAD, and infections, which further elevate the risk of death. Therefore, accurately assessing the mortality risk in DFU patients is essential for early intervention and personalized treatment. However, the quality of existing mortality prognostic models varies widely. Thus, it is essential to conduct a systematic evaluation of these models to improve their scientific rigor and clinical utility. This will provide a solid foundation for health care professionals to select or develop high-performance prognostic tools and ultimately enhance the prognosis management of patients with DFUs.

MATERIALS AND METHODS

The study protocol was registered on PROSPERO (CRD420250652656, https://www.crd.york.ac.uk/PROSPERO/view/CRD420250652656), and no amendments were made after registration.

Terminology clarification

In this systematic review, we use the term “prognostic model” to refer to multivariable models developed to estimate the probability of future mortality in patients with DFUs. Although the term “predictive model” is sometimes used in the literature to describe similar models, we adopt “prognostic model” throughout this article for conceptual clarity and consistency. When citing external literature that uses different terminology, the original wording is retained.

Literature search strategy

To ensure a comprehensive and inclusive search, we conducted searches in both the Chinese and English databases. The databases searched included China National Knowledge Infrastructure, Wanfang Data, China Science and Technology Journal database, SinoMed, PubMed, Web of Science, The Cochrane Library, and Embase. All databases were searched from their inception to February 24, 2025. No additional sources, such as trial registers or conference proceedings, were searched. A combination of keywords was used, including: “diabetic foot,” “diabetic feet,” “diabetic foot ulcer,” “DF,” “DFU,” “prediction model,” “predictive model,” “risk prediction,” “risk factor*,” “risk assessment,” “prognostic model,” “risk score*,” “prediction tool,” “nomogram,” “predictive factor*,” “mortality,” “survival,” and “death.” The full electronic search strategies for each database, including all Boolean operators and field tags, are provided in Supplementary Data S1 to ensure reproducibility.

We utilized the PICOTS framework, recommended by the Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modeling Studies checklist.¹⁸ This system helps frame the review’s aim, search strategy, and study inclusion and exclusion criteria.¹⁹ The key items of our systematic review are described below:

P (Population): Patients diagnosed with DFUs.

I (Index model): Published prognostic models for mortality in DFU patients, with predictors ≥2.

C (Comparator): No competing model.

O (Outcome): All-cause mortality (rather than cause-specific mortality).

T (Timing): The outcome was predicted based on patient characteristics, clinical scoring scales, and laboratory indicators at admission or during follow-up.

S (Setting): The intended use of the prognostic model is to individualize mortality risk assessment in patients with DFUs, thereby aiding clinical decision-making and enabling early intervention to improve patient outcomes.

Inclusion and exclusion criteria

The inclusion criteria for this review were as follows: (1) studies involving patients diagnosed with DFUs as the primary population, regardless of whether these patients had comorbidities such as peripheral neuropathy or PAD; (2) observational study designs, including cohort, case–control, or cross-sectional studies; (3) development and/or reporting of a mortality risk prognostic model; and (4) the primary outcome of interest was all-cause mortality. Studies reporting only cause-specific mortality were excluded to maintain consistency in outcome definition and comparability across studies.

Studies were excluded if they met any of the following criteria: (1) no mortality risk prognostic model was developed or reported; (2) the study focused exclusively on specific causes of death rather than all-cause mortality; (3) the publication was not in English or Chinese; (4) the full text was unavailable despite attempts to contact the authors; (5) the study population consisted solely of DFU patients who had received specific interventions, such as prior lower limb amputation, vascular stenting, or revascularization surgery; (6) included mixed patient populations where DFU was not the primary study population; (7) follow-up limited to the hospitalization period only; or (8) a total sample size of fewer than 100 participants.

Study selection and screening

The study selection process was independently conducted by two reviewers (G.Y.H. and Z.J.Y.). Initially, duplicate records were removed using EndNote software. The remaining titles and abstracts were screened based on the predefined inclusion and exclusion criteria. Subsequently, full-text articles of potentially eligible studies were retrieved and reviewed. The reference lists of all included studies were also manually screened to identify additional relevant studies. Any disagreements during the selection process were resolved through discussion or consultation with a third reviewer (Z.H.R.).

Data extraction

Data extraction was independently performed by two reviewers. Any discrepancies were resolved through discussion, with a third reviewer consulted when consensus could not be reached. The primary outcome of interest was all-cause mortality in patients with DFUs, and all reported mortality outcomes during follow-up were extracted, excluding in-hospital mortality. When multiple follow-up time points were available, all relevant time points were recorded. If a study reported more than one prognostic model, all eligible models were included for analysis. Descriptive statistics were collected for each study.

The extracted data were categorized into two groups: (1) General study characteristics, including first author, publication year, country, study design, population characteristics, data source, follow-up duration, sample size, and mortality rate. (2) Model-related information, such as handling of missing data, modeling methods, predictor variables, model discrimination, calibration techniques, and validation approaches. As part of this process, the included studies were classified according to the transparent reporting of a multivariable prognostic model for Individual Prognosis Or Diagnosis guidelines to assess reporting transparency and completeness. Initial data extraction was conducted by one reviewer and subsequently cross-checked by a second reviewer to ensure accuracy and consistency. An electronic laboratory notebook was not used for data collection or management in this study.

All eligible studies were summarized descriptively due to heterogeneity in study design, predictors, outcome assessment, follow-up duration, and model development methods. Study characteristics and model features were tabulated for comparison, without performing meta-analysis, statistical conversions, or sensitivity analyses.

Quality assessment

ROB and applicability of the included studies were independently assessed by two reviewers using the Prediction Model Risk of Bias Assessment Tool (PROBAST), as recommended by Cochrane. PROBAST consists of 20 signaling questions across four domains: participant selection, predictors, outcomes, and analysis. Each question is rated as “Yes/Probably yes,” “No/Probably no,” or “No information.” For ROB, if all four domains were rated as low risk, the overall study risk was considered low; if any domain was rated as high risk, the overall study risk was considered high. Applicability was assessed in three domains: participant selection, predictors, and outcomes, with each rated as “low,” “high,” or “unclear” concern. Studies with all three domains rated as low concern were considered to have good applicability.

RESULTS

Study selection

Figure 2 illustrates the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 flowchart, detailing the comprehensive search process and screening results.

Figure 1.

Summary of graphic illustration.

Figure 2.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 flow diagram for the selection of studies.

A total of 1,580 articles were initially identified through database searches. After removing 438 duplicates using EndNote X9 reference management software, 1,142 articles remained. Following title and abstract screening, 1,030 articles were excluded. Subsequently, 112 reports were sought for retrieval, of which six could not be obtained, leaving 106 full-text articles assessed for eligibility. Based on the predefined inclusion and exclusion criteria, 19 studies were finally included in this systematic review.

Study characteristics

Table 1 summarizes the key characteristics of the 19 included studies in this systematic review, encompassing a total of 22 prognostic models for mortality in patients with DFUs. Among these studies, one reported three models,²⁴ one reported two models,³⁴ and the remaining studies each developed a single prognostic model. These studies were published between 2007 and 2024, with seven conducted in China,^{20,22,33,35–38} two in Italy,^23,31 two in Romania,^24,34 two in the United Kingdom,^27,32 and one each in South Korea,²¹ France,²⁵ Australia,²⁶ Saudi Arabia,²⁸ Tanzania,²⁹ and Spain, respectively.³⁰ Among the included studies, five were prospective cohort studies,^{25,29,30,32,33} and 14 were retrospective cohort studies.^{20–24,26–28,31,34–38} Two studies used data sourced from databases,^21,27 while the remaining 17 collected data from health care institutions,^{20,22–26,28–38} among which 13 were single-center studies,^{20,22–26,28,31,33,34,36–38} and four were multicenter studies.^29,30,32,35 The mean or median age of participants ranged from 52.4 to 72.2 years, with a predominance of males between 54.6% and 69.5%. Sample sizes varied widely, ranging from 103 to 71,000 participants. Follow-up durations ranged from 12 weeks to 22 years, and mortality rates varied between 4.2% and 48.5%, indicating substantial heterogeneity across studies.

Table 1.

Characteristics of the Included Studies

Author	Year	Country	Study Design	Number of Models	Participants	Data Source	Follow-Up Duration	Sample	Mortality	Mortality Rate
Li et al.²⁰	2011	China	Retrospective cohort study	1	Mean age: 64.8 ± 11.5 years; male: 66.9%	Department of Endocrinology, PLA General Hospital 306, China	3.71 ± 1.8 years	139	55	39.6%
Ha et al.²¹	2021	Korea	Retrospective cohort study	1	Mean age: 62.9 ± 12.2 years; male: 55.9％	Health Insurance Review and Assessment Service Database, South Korea	880 days	1289	158	12.30%
Liu et al.²²	2024	China	Retrospective cohort study	1	Male: 64.41%	Department of Endocrinology, The First Affiliated Hospital of Anhui Medical University, China	7–48 months	354	86	24.30%
Guarnotta et al.²³	2021	Italy	Retrospective cohort study	1	Mean age: 65.3 ± 12.1 years; male: 66.36%	Division of Endocrinology and Diabetology, AOUP Paolo Giaccone, University of Palermo, Italy	12 years	330	110	33.33%
Niță et al.²⁴	2023	Romania	Retrospective cohort study	3	Mean age: 61.34 ± 11.06 years; male: 66％	Department of Diabetes, Nutrition, and Metabolic Diseases, Emergency Clinical Hospital “Sf. Spiridon”, Romania	13 years	659	278	42.20%
Amadou et al.²⁵	2020	France	Prospective cohort study	1	New DFU; mean age: 65 ± 12 years; male: 68.3%	Department of Multidisciplinary Diabetic Foot Care, University Hospital, Paris, France	5 years	347	121	35%
Jeyaraman et al.²⁶	2019	Australia	Retrospective cohort study	1	Mean age: 59.9 ± 12.3 years; male: 62.8%	Multidisciplinary Foot Clinic, Royal Darwin Hospital, Northern Territory, Australia	Median follow-up: 5.8 years (IQR: 3.1–9.8)	513	199	5 years, 24.6%; 10 years: 45.4%
Holman et al.²⁷	2024	Britain	Retrospective cohort study	1	Median age: 71 years (IQR: 60–80); male: 67.7%	National Diabetes Foot Care Audit (NDFA), United Kingdom	12 weeks, 26 weeks, 52 weeks	71,000	12 weeks: 2,985; 26 weeks: 5,850; 52 weeks: 10,210	12 weeks: 4.2%; 26 weeks: 8.2%; 52 weeks: 14.4%;
Almohammadi et al.²⁸	2022	Saudi Arabia	Retrospective cohort study	1	Median age: 63.94 ± 13.97 years; male: 66.5%	Tertiary Educational Hospital, Western Region, Saudi Arabia	7 years	358	75	20.9%
Abbas et al.²⁹	2022	Tanzania	Prospective cohort study	1	Median age: 52.4 years; male: 58.3%	Multiple Clinics, Dar es Salaam, Tanzania	22 years	103	50	48.5%
Bundó et al.³⁰	2023	Spain	Prospective cohort study	1	Median age: 72.2 ± 12.7 years; male: 69.5%	Primary Care Healthcare Centers (n = 36), Catalonia, Spain	1 year	256	25	9.7%
Gazzaruso et al.³¹	2021	Italy	Retrospective cohort study	1	Median age: 71.1 ± 8.8 years; male: 55.9%	Outpatient Department, Diabetic Foot Unit, Clinical Institute Beato Matteo, Vigevano, Italy	42.8 ± 23.3 months	583	99	17.0%
Winkley et al.³²	2007	Britain	Prospective cohort study	1	New DFU; mean age: 62.0 ±13.9 years; male: 63.6%	Community and Hospital Podiatry Clinics (n = 5), South London, UK	18 months	253	40	15.8%
Yang et al.³³	2023	China	Prospective cohort study	1	Mean age: 67.29 ± 11.14 years; male: 68.6%	Department of Endocrinology, The First Affiliated Hospital of Chongqing Medical University, China	11–34 months	191	33	17.3%
Popa et al.³⁴	2023	Romania	Retrospective cohort study	2	Age range: 19–89 years; male: 69%	Emergency Clinical Hospital “Sf. Spiridon”, Iași, Romania	10 years	635	5 years: 221; 10 years: 263	5 years: 34.8%; 10 years: 41.4%
Yan et al.³⁵	2024	China	Retrospective cohort study	1	Median age: 64 years; male: 54.6％	Luhe Hospital and Anzhen Hospital, Capital Medical University, Beijing, China	5 years	229	49	21.40%
Huang et al.³⁶	2023	China	Retrospective cohort study	1	Perfusion, Extent, Depth, Infection, and Sensation grades 2–4	Diabetic Foot Center, People’s Hospital of Guangxi Zhuang Autonomous Region, China	Median follow-up: 60 months	172	34	19.80%
Zhu³⁷	2024	China	Retrospective cohort study	1	New DFU; mean age: 65.31 ± 12.03 years; male: 66.1%	The Third Affiliated Hospital of Nanchang University, China	Median follow-up: 30.45 months	283	50	17.70%
Wang³⁸	2018	China	Retrospective cohort study	1	Mean age: 67 ± 16,95 years; male: 64.9%	Shanghai Yuanyang Diabetic Foot Treatment Center, China	Median follow-up: 38 months	665	179	26.90%

Perfusion, Extent, Depth, Infection, and Sensation—a classification system for grading diabetic foot ulcers proposed by the International Working Group on the Diabetic Foot (IWGDF).

DFU, diabetic foot ulcer; IQR, interquartile range.

Table 2 summarizes the characteristics of the 22 prognostic models derived from the 19 included studies. Among the 22 models, 16 were developed using the Cox proportional hazards models, as reported in 14 studies.^{20–26,29,30,32,33,36–38} Two models used standard logistic regression models,^28,35 one model using a Poisson regression model,²⁷ one using a multivariable logistic regression model,³¹ and two using a multilayer perceptron approach within a single study.³⁴ Sixteen models did not report how missing data were handled,^20–32,38 including three models from a single study.²⁴ Only four models reported the area under the curve (AUC), with values ranging from 0.68 to 0.742, including two models from a single study.³⁴ The reported AUC values indicate moderate model discrimination (AUC >0.60).^22,34,36 Calibration was reported in four models, three of which were from a single study,²⁴ and one from another study.³¹ All models used the Hosmer–Lemeshow goodness-of-fit test. Internal validation of the prognostic model was performed in only one model.³⁴ None of the included models conducted external validation.

Table 2.

Characteristics of the Included Prognostic Models

Author	Missing Data Handling	Model Development Method	Predictor Factors	AUC (95% CI)	Calibration Method	Verification Method
Li et al.²⁰	—	Cox proportional hazards model	Age, smoking, hypertension, coronary artery disease, DKD	—	—	—
Ha et al.²¹	—	Cox proportional hazards model	Age, socioeconomic status, ophthalmopathy, circulatory complications, institution type	—	—	—
Liu et al.²²	—	Cox proportional hazards model	Age, SCr, Diabetes Complication Severity Index, C-reactive protein/albumin ratio, Alb,	0.686 (0.623–0.750)	—	—
Guarnotta et al.²³	—	Cox proportional hazards model	Age, eGFR, CKD	—	—	—
Niță et al.²⁴	—	Cox proportional hazards model	Model 1 (Texas): Cardiovascular disease, Hb, GFR, age	—	Hosmer–Lemeshow	—
	—	Cox proportional hazards model	Model 2 (Wagner–Meggitt): Cardiovascular disease, Hb, GFR, age	—	Hosmer–Lemeshow	—
	—	Cox proportional hazards model	Model 3 (SEWSS): Cardiovascular disease, Hb, GFR, age	—	Hosmer–Lemeshow	—
Amadou et al.²⁵	—	Cox proportional hazards model	Age, duration of diabetes, Pedis level, duration of ulcer in the first year	—	—	—
Jeyaraman et al.²⁶	—	Cox proportional hazards model	Age, CKD, Alb	—	—	—
Holman et al.²⁷	—	Poisson regression models	Age, ulcer characteristics (area, deep ulcers, hindfoot location), ischemia, various comorbidities (heart failure, myocardial infarction, stroke, renal replacement therapy, CKD)	—	—	—
Almohammadi et al.²⁸	—	Logistic regression	Age, ICU admission	—	—	—
Abbas et al.²⁹	—	Cox proportional hazards model	Neuropathy, PAD, previous amputation, retinopathy	—	—	—
Bundó et al.³⁰	—	Cox proportional hazards model	Caregiver presence, ischemia, infection	—	—	—
Gazzaruso et al.³¹	—	Multivariable logistic regression model	Age, body mass index, eGFR, previous amputation, persistent active ulcer	—	Hosmer–Lemeshow	—
Winkley et al.³²	—	Cox proportional hazards model	Age, HbA1c, ischemia, depression	—	—	—
Yang et al.³³	26 cases excluded	Cox proportional hazards model	Age, sarcopenia, SCr	—	—	—
Popa et al.³⁴	24 cases excluded	Multilayer perceptron	Model 1 (5-year mortality): Age, University of Texas Staging System for Diabetic Foot Ulcers score, Wagner–Meggitt classification, Saint Elian Wound Score System, GFR, topographic aspects and the depth of the lesion, ischemia, cardiovascular disease, DKD, hypertension, PAD, hyperuricemia	0.73 (0.4607–0.7184)	—	Internal validation (cross-validation)
Popa et al.³⁴	24 cases excluded	Multilayer perceptron	Model 2 (10-year mortality): Age, University of Texas Staging System for Diabetic Foot Ulcers score, Wagner–Meggitt classification, Saint Elian Wound Score System, GFR, topographic aspects and the depth of the lesion, ischemia, cardiovascular disease, DKD, hypertension	0.7339 (0.4779–0.7296)	—	Internal validation (cross-validation)
Yan et al.³⁵	28 cases excluded	Logistic regression	Age, Alb, cardiac dysfunction, revascularization	—	—	—
Huang et al.³⁶	407 cases excluded	Cox proportional hazards model	UA/HbA1c, UA	UA: 0.681 (0.574–0.788) UA/HbA1c: 0.742 (0.652–0.832)	—	—
Zhu³⁷	15 cases excluded	Cox proportional hazards model	Age, percentage of neutrophils	—	—	—
Wang³⁸	—	Cox proportional hazards model	Age, blood urea nitrogen, lactate dehydrogenase	—	—	—

Alb, serum albumin; AUC, area under the curve (a measure of model discrimination performance); CI, confidence interval (a range of values that likely contain the true population parameter with a given probability); CKD, chronic kidney disease; DKD, diabetic kidney disease; eGFR, estimated glomerular filtration rate; GFR, glomerular filtration rate; Hb, hemoglobin; HbA1c, hemoglobin A1c; PAD, peripheral arterial disease; SCr, serum creatinine; SEWSS: Site, Extent, Wound depth, Sepsis, and Ischemia score; Texas: ulcer depth (Grade 0–3) and stage according to infection and ischemia; UA, serum uric acid; Wagner–Meggitt: ulcer depth and extent of tissue involvement (Grade 0–5).

Quality assessment

Table 3 presents the results of ROB and applicability assessments of the included prognostic models using the PROBAST, together with brief rationales for each judgment. All models were rated as having a high ROB, indicating methodological limitations in either the development or validation of the prognostic models. In the “Participants” domain, bias was primarily related to data sources. According to the PROBAST criteria for low ROB in prognostic models, prospective cohort studies, registry data, or high-quality datasets are recommended as appropriate sources.³⁹ However, 13 models were rated as having a high ROB in this domain, primarily due to their use of retrospective cohort designs based on nondatabase sources, including three models from Niță et al.,²⁴ and two models from Popa et al.,³⁴ with the remaining models from other single-model studies.^{22,23,26,28,31,35–37} An additional three studies were rated as having unclear ROB,^20,33,38 mainly because the exclusion criteria were not explicitly stated, making it difficult to assess the representativeness and consistency of the study populations. In the “Predictors” domain, one model was rated as having unclear ROB,³⁷ primarily due to the lack of a clear definition or assessment method for the predictors. In the “Outcome” domain, all models were rated as having a low ROB. In the “Analysis” domain, 20 models were rated as having a high ROB.^{20–33,35–38} This was mainly attributed to several factors: failure to address missing data,^20–32,38 insufficient sample size to support model development,^20,29 lack of reporting on model calibration performance,^{20–23,25–30,32–38} and the absence of any internal or external model validation.^{20–33,35–38}

Table 3.

Domain-Level PROBAST Risk of Bias and Applicability Assessments with Brief Rationales

Author	Model	ROB				Applicability			Overall		Rationale
Author	Model	Participants	Predictors	Outcome	Analysis	Participants	Predictors	Outcome	ROB	Applicability	Rationale
Li et al.²⁰	—	?	+	+	−	+	+	+	−	+	Unclear exclusion criteria; missing data; insufficient sample size for model development; calibration performance not reported; no internal or external validation
Ha et al.²¹	—	+	+	+	−	+	−	+	−	−	Missing data; calibration performance not reported; no internal or external validation; narrow definition of socioeconomic predictors
Liu et al.²²	—	−	+	+	−	+	+	+	−	+	Retrospective design (nondatabase); missing data; calibration performance not reported; no internal or external validation
Guarnotta et al.²³	—	−	+	+	−	+	+	+	−	+	Retrospective design (nondatabase); missing data; calibration performance not reported; no internal or external validation
Niță et al.²⁴	Model 1	−	+	+	−	+	+	+	−	+	Retrospective design (nondatabase); missing data; no internal or external validation
	Model 2	−	+	+	−	+	+	+	−	+	Retrospective design (nondatabase); missing data; no internal or external validation
	Model 3	−	+	+	−	+	+	+	−	+	Retrospective design (nondatabase); missing data; no internal or external validation
Amadou et al.²⁵	—	+	+	+	−	−	+	+	−	−	Missing data; calibration performance not reported; no internal or external validation; limited population representativeness
Jeyaraman et al.²⁶	—	−	+	+	−	+	+	+	−	+	Retrospective design (nondatabase); missing data; calibration performance not reported; no internal or external validation
Holman et al.²⁷	—	+	+	+	−	+	+	+	−	+	Missing data; calibration performance not reported; no internal or external validation
Almohammadi et al.²⁸	—	−	+	+	−	+	+	+	−	+	Retrospective design (nondatabase); missing data; calibration performance not reported; no internal or external validation
Abbas et al.²⁹	—	+	+	+	−	+	+	+	−	+	Missing data; insufficient sample size for model development; Calibration performance not reported; no internal or external validation
Bundó et al.³⁰	—	+	+	+	−	+	+	+	−	+	Missing data; calibration performance not reported; no internal or external validation
Gazzaruso et al.³¹	—	−	+	+	−	+	+	+	−	+	Retrospective design (nondatabase); missing data; no internal or external validation
Winkley et al.³²	—	+	+	+	−	−	+	+	−	−	Missing data; calibration performance not reported; no internal or external validation; limited population representativeness
Yang et al.³³	—	?	+	+	−	+	+	+	−	+	Unclear exclusion criteria; calibration performance not reported; no internal or external validation
Popa et al.³⁴	Model 1	−	+	+	+	+	+	+	−	+	Retrospective design (nondatabase); calibration performance not reported
	Model 2	−	+	+	+	+	+	+	−	+	Retrospective design (nondatabase); calibration performance not reported
Yan et al.³⁵	—	−	+	+	−	+	+	+	−	+	Retrospective design (nondatabase); calibration performance not reported; no internal or external validation
Huang et al.³⁶	—	−	+	+	−	−	+	+	−	−	Retrospective design (nondatabase); calibration performance not reported; no internal or external validation; limited population representativeness
Zhu³⁷	—	−	?	+	−	−	+	+	−	−	Retrospective design (nondatabase); unclear predictor measurement methods; calibration performance not reported; no internal or external validation; limited population representativeness
Wang³⁸	—	?	+	+	−	+	+	+	−	+	Unclear exclusion criteria; missing data; calibration performance not reported; no internal or external validation

+ Indicates low ROB/low concern regarding applicability; − indicates high ROB/high concern regarding applicability; ? indicates unclear ROB/unclear concern regarding the applicability.

PROBAST, Prediction Model Risk of Bias Assessment Tool; ROB, risk of bias.

In terms of applicability concerns, five models were rated as having high concern regarding applicability,^{21,25,32,36,37} while the remaining 17 models were classified as having low concern.^{20,22–24,26–31,33–35,38} In the “Participants” domain, four models were considered to have high concern regarding applicability due to limited representativeness of the included populations^25,32,36,37 Specifically, some models restricted inclusion to newly diagnosed DFU patients,^25,32,37 or those with Perfusion, Extent, Depth, Infection, and Sensation grades 2–4.³⁶ These restrictive inclusion criteria may limit the generalizability of the findings to the broader DFU patient population. In the “Predictors” domain, one model was rated as having high concern regarding applicability due to a narrow definition of socioeconomic status.²¹ This model used only insurance type, income percentile, and residential area as indicators, omitting other key socioeconomic variables such as lifestyle and education level, which may compromise the external applicability of the prognostic model. In the “Outcome” domain, all models were rated as having low concern regarding applicability.

Prognostic factors in the included models

In this systematic review, a total of 19 studies were included, identifying 22 prognostic models for mortality risk in patients with DFUs. Across these models, 39 distinct predictors were identified. These predictors were categorized into five main domains: demographic characteristics, comorbidities and underlying conditions, laboratory and physiological indicators, foot ulcer-related factors, and hospitalization or treatment-related factors. For studies presenting multiple models,^24,34 only predictors consistently identified across models were extracted to prevent duplication. Among these, the most frequently reported predictors across the included models were age, cardiovascular disease, renal disease, ulcer severity, ischemia, and serum albumin (Alb), as summarized in Table 4. It should be noted that the frequency with which a predictor appears across models does not necessarily reflect the magnitude of its statistical association with mortality, as inclusion may depend on study design, sample characteristics, and modeling strategies. As this review focuses on prognostic modeling rather than causal inference, these predictors should be interpreted as contributors to risk estimation rather than independent causal determinants.

Table 4.

Prognostic Factors in the Included Models

Category	Predictive Factors	Number of Studies
Demographic characteristics	Age^{20–28,31–35,37,38}	16
	Smoking²⁰	1
	Socioeconomic status²¹	1
	Duration of diabetes²⁵	1
	Caregiver presence³⁰	1
	Body mass index³¹	1
	Sarcopenia³³	1
Comorbidities and underlying conditions	Cardiovascular disease (coronary artery disease, cardiac dysfunction, stroke)^{20,24,27,34,35}	5
	Renal disease (CKD, DKD, renal impairment)^{20,23,26,27,34}	5
	Hypertension^20,34	2
	Ophthalmic complications^21,29	2
	Diabetes Complication Severity Index²²	1
	Anemia²⁴	1
	Renal replacement therapy²⁷	1
	PAD²⁹	1
	Neuropathy²⁹	1
	Depression³²	1
Laboratory and physiological indicators	Alb^22,26,35	3
	SCr^22,33	2
	eGFR^23,31	2
	GFR^24,34	2
	C-reactive protein/albumin ratio²²	1
	Hb²⁴	1
	HbA1c³²	1
	UA³⁶	1
	UA/HbA1c³⁶	1
	Percentage of neutrophils³⁷	1
	Lactate dehydrogenase³⁸	1
	Blood urea nitrogen³⁸	1
Foot ulcer-related factors	Ulcer Severity^24,27,31,34	4
	Ischemia^{27,30–32,34}	4
	Ulcer classification^25,34	2
	Amputation-related factors^29,31	2
	Duration of ulcer in the first year²⁵	1
	Infection³⁰	1
Hospitalization and treatment factors	Institution type²¹	1
	Length of hospitalization²⁴	1
	ICU admission²⁸	1
	Revascularization³⁵	1

Alb, serum albumin; CKD, chronic kidney disease; DKD, diabetic kidney disease; eGFR, estimated glomerular filtration rate; GFR, glomerular filtration rate; Hb, hemoglobin; HbA1c, hemoglobin A1c; PAD, peripheral arterial disease; SCr, serum creatinine; UA, serum uric acid.

DISCUSSION

This systematic review identified and summarized 22 prognostic models for mortality risk in patients with DFUs, derived from 19 studies. The reported mortality rates varied widely, ranging from 4.2% to 48.5%. Several factors may explain this variation. One key factor is the length of follow-up. In general, longer follow-up periods result in more observed death events, leading to higher reported mortality rates. Geographic and population differences also play a role. Countries differ in health care quality, economic development, patient characteristics, health behaviors, and allocation of medical resources. Patients in resource-limited settings may face higher mortality risks due to delayed diagnosis and limited access to care. Previous research has shown a strong association between lower socioeconomic status and increased mortality risk in patients with DFUs.²¹ Most included studies were retrospective and conducted at single centers, which limits the generalizability and external validity of the findings. This may also affect the accuracy of mortality estimates. Future research should prioritize prospective, multicenter studies to improve data quality and representativeness. Moreover, greater attention should be given to the role of social determinants of health in DFU-related mortality, particularly in low- and middle-income countries and regions.

This systematic review analyzed the characteristics, methodological quality, and applicability of the included models. According to the PROBAST assessment, all models were rated as having a high ROB with the Analysis domain being the primary contributor. Key limitations included inappropriate data sources, small sample sizes, inadequate handling of missing data, and lack of model calibration and validation. These limitations explain why most existing DFU mortality prognostic models fail to meet contemporary methodological standards and have limited reliability for clinical decision-making. High ROB may introduce systematic errors that distort predictor selection and inflate effect estimates, reducing both reproducibility and generalizability. Regarding applicability, five studies were rated as having high concern. The main limiting factors included restrictive inclusion criteria and narrowly defined predictors, which may reduce the external validity of the models in clinical settings. Therefore, when applying these models in practice, both the ROB and applicability limitations should be considered, and the results interpreted with caution. To enhance the robustness, generalizability, and clinical utility, future models should consider adopting prospective study designs or utilizing high-quality databases to ensure data representativeness and reliability. Clearly defined inclusion and exclusion criteria are essential to minimize population heterogeneity. Sample size should meet the recommended threshold of at least 10 outcome events per predictor variable.⁴⁰ Blinded assessment of both outcomes and predictors is recommended to reduce measurement bias. Continuous variables should be retained during analysis whenever possible. If categorization is necessary, internal validation and coefficient shrinkage should be performed to mitigate overfitting.⁴⁰ Evidence suggests that machine learning approaches such as support vector machines and back propagation neural networks may offer superior prognostic performance compared with the traditional statistical methods.⁴¹ Therefore, future model development should incorporate a broader range of modeling techniques, including machine learning algorithms such as decision trees and random forests, to improve the identification of high-risk individuals and enhance model applicability. Importantly, both internal and external validations are essential to ensure prognostic performance and reduce bias risk. In summary, future DFU mortality risk models should follow PROBAST guidelines. Emphasis must be placed on sound methodology and thorough validation to produce reliable clinical tools.

This study highlights the considerable heterogeneity of predictors across mortality prognostic models for patients with DFUs. The variables included in the reviewed models varied substantially, with no two models sharing an identical set of predictors. The most frequently reported predictors were age, cardiovascular disease, renal disease, ulcer severity, ischemia, and Alb. Age is a nonmodifiable factor closely associated with chronic comorbidities and declining physiological reserve. It has been repeatedly included in various models, indicating its consistency and predictive value in mortality risk estimation. Cardiovascular and renal diseases represent the multisystem involvement commonly seen in DFU patients and are strong predictors of mortality. This multimorbidity substantially increases the risk of death. A large-scale study involving 66,323 U.S. veterans with DFUs found significantly higher mortality rates among those with cerebrovascular, cardiovascular, and renal comorbidities.¹⁰ A global systematic review and meta-analysis of DFU mortality reported that nearly 50% of deaths were attributable to cardiovascular causes.¹² Prior research also confirmed that active management of cardiovascular risk factors can reduce mortality in this population.⁴² Ulcer severity reflects wound depth, area, infection status, and degree of ischemia. More severe ulcers are harder to manage, increasing the risk of complications such as sepsis, which is responsible for approximately 14% of DFU-related deaths.⁴³ Furthermore, a systematic review demonstrated that DFU patients with PAD (a common manifestation of ischemia) had a higher mortality risk compared with those without PAD (HR = 1.88).¹² Alb levels reflect nutritional status. Malnutrition impairs wound healing and may exacerbate cardiovascular and cerebrovascular disease, potentially triggering acute adverse events.⁹ Certain nutritional factors have been shown to promote collagen deposition and support tissue repair, thereby facilitating wound healing.⁴⁴ From this perspective, nutritional status may represent a potentially modifiable determinant of prognosis in patients with DFUs. Therefore, a multidisciplinary and proactive approach is essential in DFU management. Optimal care should go beyond local wound treatment to comprehensively address comorbidities, vascular health, nutritional status, and systemic risk factors.

This review has several limitations, mainly related to the review process itself and the characteristics of the included studies. First, regarding the review process, this review may not have comprehensively captured all the existing mortality prognostic models for patients with DFUs. Unpublished models or those indexed in nonmainstream databases may have been missed, introducing selection bias. The literature search was restricted to English and Chinese publications, which may have excluded relevant evidence, particularly from low- and middle-income countries, thereby limiting the global applicability of the findings. In addition, to minimize confounding from specific interventions, studies enrolling only patients who had undergone amputation or revascularization were excluded; however, this restriction may have reduced the external generalizability of the results. Second, regarding the included studies, most prognostic models were developed using retrospective data and may therefore be subject to selection and information bias. When retrospective designs are used, propensity score methods can be considered to help reduce potential bias. Substantial heterogeneity existed across studies in model development methods, sample characteristics, follow-up durations, and validation strategies, which may lead to different accumulation of death events and thus complicate direct comparison and synthesis. Moreover, handling of confounding factors was inconsistent; for example, adjustment for key clinical variables such as ulcer and infection severity varied across studies, which may affect the stability and comparability of model performance. In addition, most studies did not stratify mortality by cause, and it was not possible to separate deaths directly attributable to DFU from deaths due to other comorbidities. Therefore, this review focused on all-cause mortality, which may limit the applicability of the prognostic models specifically to DFU-related deaths. These findings highlight the need for prospective, large-scale, multicenter studies with systematic adjustment for important confounders and a stronger external validation to improve the reliability and applicability of DFU mortality prognostic models.

KEY FINDINGS

All included models demonstrated high ROB, mainly due to issues such as inappropriate data sources, small sample sizes, inadequate handling of missing data, and insufficient model calibration and validation.

Age, cardiovascular disease, renal disease, ulcer severity, ischemia, and Alb were the most frequently included and clinically significant predictors across models.

This systematic review included 22 models. All models were rated as having a high ROB based on the PROBAST assessment, indicating that none met the current methodological quality standards. While the identified risk factors are not novel, our synthesis confirms their relevance and highlights consistent methodological limitations across studies. To improve the quality of future research, investigators should rigorously adhere to the PROBAST framework to improve future model development. Clinically, current DFU mortality prognostic models should be interpreted with caution due to their limited prognostic accuracy. Sole reliance on these models may result in misclassification of patient risk. This is particularly problematic when models are poorly calibrated, as risk estimates may systematically over- or underestimate true mortality probabilities in real-world settings. Until high-quality models are rigorously validated, clinicians should base decisions on comprehensive clinical assessments and prioritize consistently reported risk factors, including age, cardiovascular disease, renal disease, ulcer severity, ischemia, and Alb. Looking ahead, once rigorously validated, these models could be integrated into routine clinical workflows to enhance risk stratification and guide individualized management. For instance, embedding validated models into electronic health record systems could enable automated risk calculations at the point of care and trigger clinical alerts for high-risk patients, facilitating timely multidisciplinary interventions.

INNOVATION

DFUs are a serious complication of diabetes, often leading to increased mortality risk. Accurate prediction of mortality risk is essential for guiding timely clinical interventions and optimizing patient outcomes. This study systematically reviewed 22 published mortality prognostic models for DFU patients, derived from 19 studies. It identified key methodological limitations in the existing literature and synthesized high-frequency risk factors with stable predictive value to inform future model refinement and development. The findings provide useful guidance for improving risk stratification and support more personalized, evidence-based care for patients with DFUs.

AVAILABILITY OF DATA AND MATERIALS

All data collection forms, extracted data, and analysis materials used in this review are available from the corresponding author upon reasonable request. No public repository was used.

AUTHOR DISCLOSURE AND GHOSTWRITING

The authors declare no conflicts of financial or nonfinancial competing interest. This article was solely authored by the individuals listed, with no involvement of ghostwriters in its preparation.

Footnotes

FUNDING INFORMATION

The study was funded by the Health Science and Technology Project of the Pudong New Area Health Commission (grant/award number: PW2023A-09), the Academic Leaders Training Program of Pudong Health Bureau of Shanghai (grant/award number: PWRd2022-16), and the Tongji University “Third Year Action Plan for Discipline. Construction of School of Nursing” (Grant/Award Number: JS2210328).

ABOUT THE AUTHORS

Huiren Zhuang, PhD, is a nursing supervisor at Shanghai East Hospital, School of Medicine, Tongji University. Yihao Gu is an undergraduate student at Tongji University School of Medicine. Jingjing Guo is a master’s degree candidate at the School of Medicine, Tongji University. Weiying Zhang, PhD, is the nursing director at Shanghai East Hospital, School of Medicine, Tongji University. Xiaojing Li, MSc, is the director of the Nursing Department at the Punan Branch of Renji Hospital, Shanghai Jiao Tong University School of Medicine. Jinyi Zhu is an undergraduate student at Tongji University School of Medicine.

Supplemental Material

Abbreviations and Acronyms

References

IDF Diabetes Atlas IDF Diabetes Atlas 2022 Reports. Available from: https://diabetesatlas.org/2022-reports/ [Last accessed: 12 February, 2025].

Sun

, Saeedi

, Karuranga

, et al. IDF diabetes atlas: Global, regional and country‐level diabetes prevalence estimates for 2021 and projections for 2045. Diabetes Res Clin Pract, 2022; 183:109119; doi: 10.1016/j.diabres.2021.109119

Armstrong

, Boulton

AJM

, Bus

. Diabetic foot ulcers and their recurrence. N Engl J Med, 2017; 376(24):2367–2375; doi: 10.1056/NEJMra1615439

Huang

, Lu

, Yang

, et al. Microenvironment-Based diabetic foot ulcer nanomedicine. Adv Sci (Weinh), 2023; 10(2):e2203308; doi: 10.1002/advs.202203308

Zhang

, Lu

, Jing

, et al. Global epidemiology of diabetic foot ulceration: A systematic review and meta-analysis †. Ann Med, 2017; 49(2):106–116; doi: 10.1080/07853890.2016.1231932

McDermott

, Fang

, Boulton

AJM

, et al. Epidemiology, and Disparities in the Burden of Diabetic Foot Ulcers. Diabetes Care, 2023; 46(1):209–221; doi: 10.2337/dci22-0043

Armstrong

, Swerdlow

, Armstrong

, et al. Five year mortality and direct costs of care for people with diabetic foot complications are comparable to cancer. J Foot Ankle Res., 2020; 13(1):16; doi: 10.1186/s13047-020-00383-2

Walsh

, Hoffstad

, Sullivan

, et al. Association of diabetic foot ulcer and death in a population-based cohort from the United Kingdom. Diabet Med, 2016; 33(11):1493–1498; doi: 10.1111/dme.13054

Deng

, Xie

, Chen

, et al. Impact of acute hyperglycemic crisis episode on survival in individuals with diabetic foot ulcer using a machine learning approach. Front Endocrinol (Lausanne), 2022; 13:974063; doi: 10.3389/fendo.2022.974063

10.

Brennan

, Hess

, Bartle

, et al. Diabetic foot ulcer severity predicts mortality among veterans with type 2 diabetes. J Diabetes Complications, 2017; 31(3):556–561; doi: 10.1016/j.jdiacomp.2016.11.020

11.

Meloni

, Izzo

, Giurato

, et al. Prevalence, Clinical Aspects and Outcomes in a Large Cohort of Persons with Diabetic Foot Disease: Comparison between Neuropathic and Ischemic Ulcers. J Clin Med, 2020; 9(6):1780; doi: 10.3390/jcm9061780

12.

Chen

, Sun

, Gao

, et al. Global mortality of diabetic foot ulcer: A systematic review and meta-analysis of observational studies. Diabetes Obes Metab, 2023; 25(1):36–45; doi: 10.1111/dom.14840

13.

Lauwers

, Dirinck

, Van Bouwel

, et al. Malnutrition and its relation with diabetic foot ulcer severity and outcome: A review. Acta Clin Belg, 2022; 77(1):79–85; doi: 10.1080/17843286.2020.1800315

14.

Riley

, Collins

, Kirton

, et al. Uncertainty of risk estimates from clinical prediction models: Rationale, challenges, and approaches. BMJ, 2025; 388:e080749. Published 2025 Feb 13; doi: 10.1136/bmj-2024-080749

15.

Sullivan

, Massaro

, D’Agostino

RB.

Sr ., Presentation of multivariate data for clinical use: The Framingham Study risk score functions. Stat Med, 2004; 23(10):1631–1660; doi: 10.1002/sim.1742

16.

Pan

, You

, Bu

, et al. Development and validation of a nomogram for predicting cancer-specific survival in patients with Wilms’ tumor. J Cancer, 2019; 10(21):5299–5305; doi: 10.7150/jca.32741

17.

Monteiro-Soares

, Hamilton

, Russell

, et al. Guidelines on the classification of foot ulcers in people with diabetes (IWGDF 2023 update). Diabetes Metab Res Rev, 2024; 40(3):e3648; doi: 10.1002/dmrr.3648

18.

Moons

, de Groot

, Bouwmeester

, et al. Critical appraisal and data extraction for systematic reviews of prediction modelling studies: The CHARMS checklist. PLoS Med., 2014; 11(10):e1001744; doi: 10.1371/journal.pmed.1001744

19.

Debray

, Damen

, Snell

, et al. A guide to systematic review and meta-analysis of prediction model performance. BMJ, 2017; 356:i6460; doi: 10.1136/bmj.i6460

20.

, Xiao

, Wang

, et al. Relationship between long-term mortality and diabetic complications in 139 patients with diabetic foot ulcer. Chin J Endocrinol Metab, 2011; 27(2):128–132; doi: 10.3760/cma.j.issn.1000-6699.2011.02.009

21.

, Jin

, Park

. Association between socioeconomic position and diabetic foot ulcer outcomes: A population-based cohort study in South Korea. BMC Public Health, 2021; 21(1):1395; doi: 10.1186/s12889-021-11406-3

22.

Liu

, Li

, Qian

, et al. Influence and predictive value of DCSI and CRP/albumin ratio on all-cause mortality in patients with diabetic foot ulcer. Acta Universitatis Medicinalis Anhui, 2024; 59(12):2183–2189; doi: 10.19405/j.cnki.issn1000-1492.2024.12.017

23.

Guarnotta

, Radellini

, Vigneri

, et al. Diabetic foot ulcers: Retrospective comparative analysis from Sicily between two eras. PLoS One, 2021; 16(12):e0259405. Published 2021 Dec 7; doi: 10.1371/journal.pone.0259405

24.

Niță

, Arhire

, Mihalache

, et al. Evaluating classification systems of diabetic foot ulcer severity: A 12-year retrospective study on factors impacting survival. Healthcare (Basel), 2023; 11(14):2077; doi: 10.3390/healthcare11142077

25.

Amadou

, Carlier

, Amouyal

, et al. Five-year mortality in patients with diabetic foot ulcer during 2009-2010 was lower than expected. Diabetes Metab, 2020; 46(3):230–235; doi: 10.1016/j.diabet.2019.04.010

26.

Jeyaraman

, Berhane

, Hamilton

, et al. Mortality in patients with diabetic foot ulcer: A retrospective study of 513 cases from a single centre in the northern territory of australia. BMC Endocr Disord, 2019; 19(1):1; doi: 10.1186/s12902-018-0327-2

27.

Holman

, Yelland

, Young

, et al. Mortality rates in people presenting with a new diabetes-related foot ulcer: A cohort study with implications for management. Diabetologia, 2024; 67(12):2691–2701; doi: 10.1007/s00125-024-06262-w

28.

Almohammadi

, Alnashri

, Abdulrahman T Harun

, et al. Pattern and type of amputation and mortality rate associated with diabetic foot in Jeddah, Saudi Arabia: A retrospective Cohort Study. Ann Med Surg (Lond), 2022; 73:103174; doi: 10.1016/j.amsu.2021.103174

29.

Abbas

, Chockalingam

, Lutale

, et al. Predicting the risk of amputation and death in patients with diabetic foot ulcer. A long-term prospective cohort study of patients in Tanzania. Endocrinol Diabetes Metab, 2022; 5(3):e00336; doi: 10.1002/edm2.336

30.

Bundó

, Vlacho

, Llussà

, et al. Prediction of outcomes in subjects with type 2 diabetes and diabetic foot ulcers in Catalonian primary care centers: A multicenter observational study. J Foot Ankle Res, 2023; 16(1):8; doi: 10.1186/s13047-023-00602-6

31.

Gazzaruso

, Gallotti

, Pujia

, et al. Predictors of healing, ulcer recurrence and persistence, amputation and mortality in type 2 diabetic patients with diabetic foot: A 10-year retrospective cohort study. Endocrine, 2021; 71(1):59–68; doi: 10.1007/s12020-020-02431-0

32.

Winkley

, Stahl

, Chalder

, et al. Risk factors associated with adverse outcomes in a population-based prospective cohort study of people with their first diabetic foot ulcer. J Diabetes Complications, 2007; 21(6):341–349; doi: 10.1016/j.jdiacomp.2007.09.004

33.

Yang

, Ni

, Zhang

, et al. Sarcopenia is an independent risk factor for all-cause mortality rate in patients with diabetic foot ulcers. Front Nutr, 2023; 10:1097008; doi: 10.3389/fnut.2023.1097008

34.

Popa

, Gavril

, Popa

, et al. Survival prediction in diabetic foot ulcers: A machine learning approach. J Clin Med, 2023; 12(18):5816; doi: 10.3390/jcm12185816

35.

Yan

, Wang

, Yang

, et al. The efficacy of diabetic foot treatment in a “TOSF” Pattern: A five-year retrospective study. Diabetes Metab Syndr Obes, 2024; 17:1923–1939; doi: 10.2147/DMSO.S461112

36.

Huang

, Deng

, Huang

, et al. The ratio of serum uric acid to glycosylated haemoglobin as a predictor of all-mortality in elderly patients with diabetic foot ulcers: A longitudinal cohort study. Diabetes Metab Syndr Obes, 2023; 16:2779–2790; doi: 10.2147/DMSO.S423017

37.

Zhu

. Analysis of factors associated with long-term prognosis of diabetic foot ulcer [Master’s thesis]. Nanchang University Medical School; 2024.

38.

Wang

. Analysis of the relationship between prognosis and related indicators in patients with diabetic foot disease [Master’s thesis]. Shanghai Jiao Tong University; 2018; doi:10.27307/d.cnki.gsjtu.2018.002571

39.

Moons

KGM

, Wolff

, Riley

, et al. PROBAST: A Tool to Assess Risk of Bias and Applicability of Prediction Model Studies: Explanation and Elaboration. Ann Intern Med., 2019; 170(1):W1–W33; doi: 10.7326/M18-1377

40.

Moons

, Kengne

, Woodward

, et al. Risk prediction models: I. Development, internal validation, and assessing the incremental value of a new (bio)marker. Heart., 2012; 98(9):683–690; doi: 10.1136/heartjnl-2011-301246

41.

Luo

, Zhao

, Tu

. Clinical prediction models in the precision medicine era: Old and new algorithms. Ann Transl Med, 2020; 8(6):274; doi: 10.21037/atm.2020.02.63

42.

Cavallari

, Bhatt

, Steg

, et al. Causes and Risk Factors for Death in Diabetes: A Competing-Risk Analysis From the SAVOR-TIMI 53 Trial. J Am Coll Cardiol., 2021; 77(14):1837–1840; doi: 10.1016/j.jacc.2021.02.030

43.

Brownrigg

, Griffin

, Hughes

, et al. Influence of foot ulceration on cause-specific mortality in patients with diabetes mellitus. J Vasc Surg, 2014; 60(4):982–986.e3; doi: 10.1016/j.jvs.2014.04.052

44.

Cianciulli

, Delli Poggi

, Giannetta

, et al. Role of specialized nutrition for wound healing: A narrative review. Progr Nutr, 2019; 21(3):507–512; doi: 10.23751/pn.v21i3.7280

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.11 MB