Structured Education Improves Radiographic Detection and Referral in Charcot Foot: A Quasi-experimental Study

Abstract

Background:

Charcot foot, characterized by progressive osseous and articular destruction, necessitates early diagnosis and a multidisciplinary approach to prevent severe complications. This study assessed the efficacy of an educational intervention in improving radiographic detection and referral for Charcot foot among health care professionals.

Methods:

Thirty health care professionals, including foot and ankle surgeons (FAs), orthopaedic/trauma specialists (OTs), and family physicians (FDs), participated in this quasi-experimental, pre-post study. Participants completed a baseline assessment and received an educational intervention (face-to-face/webinar or self-directed PDF), followed by a reassessment. Primary outcomes included changes in diagnostic and referral accuracy, evaluated through score improvements and Fleiss κ for interobserver agreement. Reliability in the Meary angle measurement was also assessed.

Results:

FA specialists demonstrated consistently high baseline scores. Following the intervention, OT and FD groups exhibited significant improvements in both diagnostic and referral scores. Fleiss κ increased from 0.463 to 0.700 for OT diagnosis, and from 0.439 to 0.723 for referral. Fleiss κ values for FD improved from 0.352 to 0.546 for diagnosis and from 0.071 to 0.608 for referral. Face-to-face/webinar training outperformed the PDF format, yielding higher scores and agreement rates. Interobserver reliability for the Meary angle was excellent (intraclass correlation coefficient = 0.988).

Conclusion:

Structured educational interventions significantly enhanced clinicians’ ability to diagnose and appropriately refer patients with Charcot foot. Face-to-face instruction proved more effective than self-directed formats.

Level of Evidence:

Level III, quasi-experimental pre-post study.

Graphical Abstract.

Introduction

Charcot neuroarthropathy is a chronic, progressive condition characterized by joint deformity and osseous degeneration, frequently resulting in foot dislocation and bone destruction, which substantially increases the risk of ulceration, infection, and lower-limb amputation. ³ Charcot foot requires the presence of peripheral neuropathy, most often due to diabetes mellitus (approximately 75%), whereas the remaining 25% involve other etiologies such as alcohol abuse, spinal cord injury, or leprosy.^5,24,26 Among diabetic patients with peripheral neuropathy, the prevalence of Charcot foot ranges from 0.08% to 13%, with an annual ulceration risk of 2.5% and amputation rate of 4.1 per 100 person-years.^4,19

Early diagnosis of Charcot foot is essential to prevent severe complications. In the acute phase, clinical signs such as warmth, erythema, and edema are predominant. An important stage in Charcot pathophysiology is the prodromal or stage 0 phase, characterized by typical clinical signs but without radiographic changes. Magnetic resonance imaging is required to detect early bone marrow edema. Early diagnosis during this phase can prevent progression and collapse.^{2,9,11,12,14,15,16,27} In subacute and chronic phases, deformities become evident, particularly on radiographs, including alterations in the Meary angle—a key radiologic parameter correlating with disease progression and ulceration risk.^1,7,18,25

Prior studies have demonstrated the positive impact of targeted educational interventions in enhancing diagnostic accuracy and interobserver reliability in various radiologic contexts. Notably, training programs on COVID-19 diagnostic criteria and such as the Breast Imaging Reporting and Data System (BI-RADS) used in mammography have shown significant improvements in diagnostic agreement among health care professionals.^17,20

Despite advances in Charcot foot management, early detection—especially in primary care—remains a challenge. Although early MRI detection is crucial for identifying prodromal Charcot foot, our study focused on cases with radiographic findings. Prodromal (stage 0) Charcot cases were excluded. This study evaluates whether a structured educational intervention can enhance clinicians’ radiographic recognition and referral decisions of Charcot foot. Although family physicians typically depend on radiology reports, in many clinical settings they are required to interpret plain radiographs as part of initial assessment. Despite prior validation, a secondary objective was to assess inter- and intraobserver variability in the Meary angle measurement. We reassessed Meary angle because of its relevance in training nonspecialist clinicians and its role in predicting plantar ulceration.

Materials and Methods

Study Design

A quasi-experimental, pre-post study was conducted to evaluate the effectiveness of a structured educational intervention in improving radiographic recognition corresponding to Eichenholtz stage 1 or more advanced Charcot foot and referral for Charcot neuroarthropathy. Referral accuracy was assessed by comparing participant decisions to a gold standard consensus established by 3 foot and ankle surgeons based on radiographic criteria.

Participants

A total of 30 licensed health care professionals were enrolled using convenience sampling, divided in 3 groups: 10 foot and ankle surgeons (FA), 10 orthopaedic/trauma specialists without foot and ankle subspecialization (OT), and 10 family physicians (FD) (Figure 1). OT and FD participants were affiliated with the host institution, whereas 3 FA were internal and 7 were invited from other hospitals via email. All participants provided informed consent and met the eligibility criteria.

Figure 1.

Flowchart of participant inclusion and pre- to post-intervention allocation. Source: author’s elaboration.

Inclusion and Exclusion Criteria

Eligible participants were actively practicing clinicians within the categories of FA, OT, or FD and required a basic proficiency in radiographic interpretation. Basic proficiency was defined as self-reported routine interpretation of plain radiographs in clinical practice. Exclusion criteria included (1) formal training in Charcot foot diagnosis within the preceding 12 months, (2) current non-practicing status, or (3) failure to complete all phases of the study.

Sample Size Calculation

Sample size estimation was based on previous studies of educational interventions targeting diagnostic agreement. Assuming a mean improvement of 1 point on a 0 to 10 scale (SD 1.5), a sample of 18 participants was needed to achieve 80% power with a 5% significance level. To strengthen subgroup comparisons, the sample size was increased to 30.

Variables

• - Radiographic identification score: Continuous (0-10); number of correctly classified radiographs as Charcot or non-Charcot.

• - Referral decision score: Continuous (0-10); number of correct decisions on specialist referral.

• - Interobserver agreement: Fleiss κ, used to assess diagnostic and referral consistency across participants. κ is a statistical measure used to evaluate the degree of agreement among multiple raters when assigning categorical ratings. Values range from 0 (no agreement) to 1 (perfect agreement).

• - The Meary angle reliability: Intraobserver agreement assessed with intraclass correlation coefficient (ICC).

• - Educational modality: Categorized as face-to-face/webinar or self-directed PDF.

• - Professional group: FA, OT, or FD, used to compare baseline and post-intervention performance.

Study Procedure

Phase 1: Baseline assessment

Conducted in March-April 2023, participants completed an online assessment comprising 11 anonymized weight-bearing lateral foot radiographs (2 normal, 2 flatfoot, 4 Charcot, 2 cavus, and 1 duplicate flatfoot for intraobserver reliability) (Figure 2). Participants were instructed to (1) determine the presence of Charcot foot (yes/no), (2) measure the Meary angle using an online protractor (with a video tutorial provided), and (3) assess ulceration risk and referral necessity (yes/no).

Figure 2.

Weight-bearing lateral foot radiographs in survey assessment. Source: De-identified radiographic images, author’s elaboration.

Phase 2: Educational intervention and reassessment

In May-June 2023, OT and FD participants received a standardized 10-minute group educational training session on Charcot foot recognition and management, conducted by the investigators at the host institution. Participants were randomly assigned in-person or via a digital PDF format. Training was conducted in small groups, either in person or via interactive webinar. FA participants were excluded from this phase because of prior expertise. A second assessment using the same radiographs (excluding the duplicate) was completed post-intervention. No incentives were provided.

Phase 3: Scoring

Scores were assigned as follows: 1 point per correct identification of Charcot status and 1 point per appropriate referral decision, for a maximum of 10 points in each domain.

To evaluate the impact of an educational intervention on diagnostic and referral consistency for Charcot foot in lateral foot radiographs, we compared Fleiss κ coefficient values before and after training among OT and FD.

To evaluate the impact of different educational delivery methods on diagnostic consistency for Charcot foot in lateral radiographs, we compared Fleiss κ values before and after an intervention among 2 groups: one receiving face-to-face education, and the other, a self-administered PDF. To preserve participant anonymity, collective scores from the OT and FD groups were assessed.

Statistical Analysis

Data was stored in Microsoft Excel and analyzed using SPSS version 22.0 (IBM Corp). All responses were anonymized to minimize observer bias. Descriptive statistics (mean ± SD) were calculated for continuous variables. Paired t tests evaluated pre- and post-intervention differences; 1-way analysis of variance assessed intergroup comparisons (FA, OT, FD). κ was used to evaluate interobserver agreement in identification and referral tasks, whereas ICC assessed intraobserver reliability in the Meary angle measurement. CIs (95%) were reported for all reliability measures. Significance was set at P <.05. No adjustments were made for multiple comparisons.

Results

Participant Characteristics

All 30 participants (FA, OT, and FD) completed the baseline phase of the study (Table 1). The same participants complete both pre- and post-intervention phases.

Table 1.

Baseline Scores and Fleiss κ Coefficients for Specialists in Diagnostic and Referral.

	Score				κ a
Specialty	Mean	SD	95% CI	ANOVA	κ	SE	95% CI	P Value
Diagnostic
Foot and ankle	9.9	0.3	9.7-10.1	<.005	0.958	0.01	0.94-0.98	<.001
Ortho/trauma	8.4	2.1	6.9-9.9	<.005	0.463	0.035	0.39-0.53	<.001
Family doctors	7.5	1.5	6.4-8.6	<.005	0.352	0.038	0.28-0.43	<.001
Referral
Foot and ankle	9.8	0.4	9.5-10.1	<.001	0.918	0.013	0.89-0.94	<.001
Ortho/trauma	7.8	1.1	6.9-8.6	<.001	0.439	0.035	0.37-0.51	<.001
Family doctors	6.7	1.7	5.5-7.9	<.001	0.071	0.045	−0.02-0.16	.119

Abbreviation: ANOVA, analysis of variance.

^aFleiss κ coefficient: 0.8-1.0: almost perfect; 0.6-0.8, considerable; 0.4-0.6, moderate; 0.2-0.4, acceptable; 0.01-0.2, mild; 0.00, poor.

Baseline Performance

Fleiss κ coefficient

Diagnostic and referral agreement among FA was excellent (κ = 0.958, SE = 0.010) (κ = 0.918, SE = 0.013). OT participants exhibited moderate agreement for diagnosis (κ = 0.463, SE = 0.035) and referral (κ = 0.439, SE = 0.035), while FD participants demonstrated fair agreement for diagnosis (κ = 0.352, SE = 0.038) and poor agreement for referral (κ = 0.071, SE = 0.045).

Mean scores

FA specialists achieved mean scores of 9.9 (SD = 0.3) for radiologic identification and 9.8 (SD = 0.4) for referral. OT participants scored 8.4 (SD = 2.1) for identification and 7.8 (SD = 1.1) for referral. FD participants scored 7.5 (SD = 1.5) for identification and 6.7 (SD = 1.7) for referral (Table 2, Figure 3).

Table 2.

Comparative Analysis of Fleiss κ Coefficients and Scores Pre- and Post-intervention for Diagnostic and Referral Accuracy for Ortho/Trauma and Family Doctors.

	Pre-Int. Score		Post-Int. Score			Pre-Int. κ a		Post-Int. κ a
Specialty	Mean	95% CI	Mean	95% CI	t test	Mean	95% CI	Mean	95% CI	z	z test
Diagnostic
OT	8.4	7.10-9.70	9.2	8.64-9.76	.278	0.463	0.394-0.532	0.700	0.641-0.759	5.55	<.001
FD	7.5	6.57-8.43	8.8	8.06-9.54	.49	0.352	0.278-0.426	0.546	0.481-0.611	3.93	<.001
Referral
OT	7.8	7.12-8.48	9.4	8.90-9.90	.002	0.439	0.370-0.508	0.723	0.668-0.778	6.64	<.001
FD	6.7	5.56-7.75	8.9	8.03-9.77	.006	0.071	−0.017-0.159	0.608	0.547-0.669	11.16	<.001

Abbreviations: FD, family doctors; OT, Ortho/trauma; Post-Int., post-intervention; Pre-Int., pre-intervention.

^aFleiss κ coefficient: 0.8-1.0: almost perfect; 0.6-0.8, considerable; 0.4-0.6, moderate; 0.2-0.4, acceptable; 0.01-0.2, mild; 0.00, poor.

Figure 3.

Comparative analysis of radiological identification and specialized referral scores. Source: Author’s elaboratio.

Impact of the Educational Intervention

Mean score analysis

OT participants improved their diagnostic scores from 8.4 (SD = 2.1) to 9.2 (SD = 0.9), although not significantly (P = .278). Referral scores significantly increased from 7.8 (SD = 1.1) to 9.4 (SD = 0.8) (P = .002). FD participants’ diagnostic scores improved from 7.5 (SD = 1.5) to 8.8 (SD = 1.2) (P = .490) and referral scores from 6.7 (SD = 1.7) to 8.9 (SD = 1.4) (P = .006).

Fleiss κ coefficient

Following the intervention, OT participants showed a significant improvement in diagnostic (κ = 0.700, SE = 0.030; z = 5.39, P < .001) and referral agreement (κ = 0.723, SE = 0.028; z = 6.01, P < .001). FD participants agreement also improved for diagnostic (κ = 0.546, SE = 0.033; z = 3.60, P = .00032), and referral rose markedly from κ = 0.071 to κ = 0.610 (SE = 0.031; z = 10.76, P < .001) (Table 3, Figure 4).

Table 3.

Comparative Analysis of Fleiss κ Coefficients and Scores Across Intervention Formats for Diagnostic and Referral Accuracy. ^a

	Pre-Int. Score		Post-Int. Score			Pre-Int. κ b		Post-Int. κ b
Format	Mean	95% CI	Mean	95% CI	t test	Mean	95% CI	Mean	95% CI	z test
Diagnostic
F2F/W	7.9	7.11-8.69	9.8	9.63-9.98	<.001	0.446	0.399-0.493	0.704	0.612-0.796	<.001
PDF			8.5	7.89-9.11	.406			0.567	0.475-0.659	.022
Referral
F2F/W	7.3	6.64-7.96	9.8	9.63-9.98	<.001	0.179	0.132-0.226	0.905	0.813-0.997	<.001
PDF			8.5	7.89-9.11	.036			0.533	0.441-0.625	<.001

Abbreviations: F2F/W, face-to-face/webinar; PDF, self-directed PDF; Post-Int., post-intervention; Pre-Int., pre-intervention.

^aOutcomes from this table stem from the analysis of the combined OT and FD groups.

^bFleiss κ coefficient: 0.8-1.0: almost perfect; 0.6-0.8, considerable; 0.4-0.6, moderate; 0.2-0.4, acceptable; 0.01-0.2, mild; 0.00, poor.

Figure 4.

Comparison of scores by educational intervention format. Source: author’s elaboration.

Effect of Training Modality

Fleiss κ coefficient

Face-to-face/webinar training led to greater improvements in diagnostic agreement (κ = 0.046 to κ = 0.704, SE = 0.047; P < .001) and referral agreement (κ = 0.179 to κ = 0.905, SE = 0.047; P < .001). The PDF group also improved: diagnostic agreement rose from κ = 0.046 to κ = 0.567 (SE = 0.047; P = .022), and referral agreement increased from κ = 0.179 to κ = 0.533 (SE = 0.047; P < .001).

Mean score analysis

Face-to-face training resulted in significant increases in diagnostic scores (7.9 [SD = 1.8] to 9.8 [SD = 0.4], P < .001) and referral scores (7.3 [SD = 1.5] to 9.8 [SD = 0.4], P < .001). PDF format also showed improvements in diagnostic scores (7.9 to 8.5, P = .022) and referral scores (7.3 to 8.5, P < .001).

Meary Angle Measurement

Interobserver reliability among all participants was excellent (ICC = 0.988). Intraobserver reliability was substantial for FA (ICC = 0.975) and FD (ICC = 0.965), and moderate for OT (ICC = 0.934) (Table 4).

Table 4.

ICCs for Meary Angle Measurements for Groups. ^a

	Intraobserver		Interobserver
Speciality	ICC	95% CI	ICC	95% CI	P Value
Foot and ankle	0.991	0.98-1.00	0.988	0.97-1.00	<.001
Ortho/trauma	0.991	0.98-1.00
Family doctors	0.982	0.96-1.00

Abbreviation: ICC, intraclass correlation coefficient.

^aICC: >0.99, almost perfect; 0.95-0.99, substantial; 0.90-0.95, moderate; <0.90, poor.

Discussion

This study demonstrates that a structured educational intervention significantly improves radiologic identification and specialized referral of Charcot neuroarthropathy, among OT and FD. Post-intervention, participants in both groups showed increased diagnostic accuracy, with more pronounced improvement in the specialized referral task.

These findings are consistent with previous research highlighting the effectiveness of targeted educational programs in enhancing diagnostic skills and reducing interobserver variability.^{6,8,9,10,17,19,22,23} Specifically, Sharma et al ¹⁷ demonstrated improved interobserver agreement following an online course focused on radiologic diagnosis of COVID-19 lesions in chest radiographs, whereas Timmers et al ¹⁷ reported enhanced consistency among radiologists following specialized training in the Breast Imaging Reporting and Data System.

In our study, Meary angle was chosen because of its superior reproducibility and its sensitivity to early sagittal plane deformity, particularly medial column collapse, which represents a key radiographic hallmark of early Charcot neuroarthropathy. Meary angle demonstrated excellent interobserver and intraobserver reliability across all groups, supporting its utility as a reproducible and objective measure in the radiographic assessment of Charcot foot. These findings are consistent with those of Sheth et al, ¹⁸ who reported near-perfect agreement in cuboid height measurements across health care professionals. Bevan and Tomlinson ¹ and Wukich et al ²⁵ further identified that a Meary angle greater than −27 degrees was a significant predictor of ulceration risk in diabetic Charcot neuroarthropathy, emphasizing its role in clinical monitoring and risk stratification. In contrast, Hastings et al ⁷ reported that Meary angle had the lowest measurement precision among foot alignment metrics in patients with diabetic neuropathy, largely because of challenges in visualizing anatomical landmarks in cases of severe deformity. Nevertheless, the angle remains clinically relevant for assessing deformity progression in Charcot neuroarthropathy, supporting its utility as one of the reproducible and objective measures in the radiographic assessment of Charcot foot.

Moreover, although both educational modalities—face-to-face/webinar and self-directed PDF—resulted in significant improvements, our findings suggest that interactive, instructor-led formats may be more effective in enhancing clinical decision making, particularly with respect to specialist referral decisions among family physicians. This aligns with the findings of Plowright et al, ¹³ who emphasized the positive impact of interactive educational approaches on health care outcomes. Similarly, Waheed et al ²¹ demonstrated that targeted educational interventions substantially improved the knowledge and clinical practices of primary care physicians in the management of diabetic foot ulcers, highlighting the essential role of continuing medical education in optimizing professional performance and patient care delivery.

However, several limitations must be acknowledged. First, the questionnaire was exclusively based on radiographic images and did not incorporate clinical photographs, physical examination findings or the prodromal phase (stage 0), where early intervention may prevent foot collapse and which are essential for the comprehensive diagnosis of Charcot foot. Second, the use of convenience sampling may introduce selection bias, limiting the generalizability; reusing the same radiographs pre- and post-intervention also raises a potential learning/recall effect. Third, the lack of information regarding years in practice of participants limits our ability to correlate clinical experience with performance outcomes, future studies should account for clinician seniority. Finally, the study did not include a follow-up period to evaluate long-term retention of knowledge or changes in clinical practice. Nevertheless, key strengths of this study are its clinical relevance, the rigorous methodologic approach enabling objective assessment of educational outcomes, the use of robust statistical measures to evaluate interobserver agreement, and the practical insights provided by comparing different educational delivery formats.

Future research should include prodromal Charcot detection using magnetic resonance image and clinical criteria to promote earlier diagnosis and intervention. Also, longitudinal assessments, randomized participant allocation, and multimodal clinical scenarios to strengthen external validity. Additionally, increasing sample sizes and involving a broader spectrum of health care professionals—such as podiatrists, endocrinologists, and diabetes educators—would enhance the evaluation of the scalability and effectiveness of educational interventions aimed at improving early detection and management of Charcot foot. As an exploratory study, these findings provide a basis for future, more robust research designs

Conclusions

A structured educational intervention significantly improved radiographic identification and specialist referral for Charcot foot (excluding the prodromal stage 0), particularly among family physicians. Face-to-face/webinar formats yielded superior outcomes compared with self-directed materials. The Meary angle exhibited excellent interobserver and intraobserver reliability, supporting its utility as one reproducible and objective measure in the radiographic assessment of Charcot foot. Our study underscores the importance of educational interventions in managing Charcot foot, proposing practical changes that can be implemented in primary care to improve patient prognosis.