Assessing the Efficacy of POCUS Training in The Republic of Somaliland: A Retrospective Analysis of 500 Scans Conducted by General Practitioners

ABSTRACT

Objective

Evaluate the effectiveness of a point‐of‐care ultrasound (POCUS) training programme for public sector general practitioners in The Republic of Somaliland.

Methods

This retrospective, observational study analysed 500 ultrasound scans of adult patients from four public hospitals. Twelve general practitioners with no prior non‐obstetric ultrasound experience completed a 12‐day, multi‐organ training curriculum, then performed scans in the emergency department, ICU and outpatient clinics. Participants submitted their scans with a questionnaire providing their answer to one or more (out of 21 total) specific diagnostic questions, termed ‘studies’, answered by ultrasound. Blinded experts assessed image quality and accuracy. Concordance between learner and expert was calculated for each organ system alongside interrater reliability for each of the individual studies.

Results

Overall concordance between expert and learner was 90% (n = 500). Concordance was highest for renal (95%, n = 242), lung (91%, n = 365) and hepatobiliary (91%, n = 217) imaging, whereas cardiac (87%, n = 400) and DVT (83%, n = 42) concordance was lower. ICU studies had the highest concordance (91%), followed by those conducted in the clinic (89%) and emergency department (87%). Cohen's kappa coefficient between learner and expert interpretation was > 0.8 (p value < 0.05) for 18 out of 18 studies, indicating at least substantial agreement.

Conclusion

This POCUS training programme in The Republic of Somaliland resulted in high concordance between learner and expert interpretations, indicating effective training across multiple organ systems. Several areas for improvement were identified. These findings underscore POCUS's potential to enhance diagnostic capabilities in resource‐limited settings.

Abbreviations

CT

computed tomography

DVT

deep venous thrombosis

ED

emergency department

ICU

intensive care unit

IVC

inferior vena cava

LV

left ventricle

MoHD

Ministry of Health Development

MRI

magnetic resonance imaging

NCD

non‐communicable disease clinic

POCUS

point‐of‐care ultrasound

RV

right ventricle

Summary.

• This study explored how a short training program in point‐of‐care ultrasound (POCUS) helped general practitioners in The Republic of Somaliland improve their diagnostic skills.

• Researchers reviewed 500 scans from doctors who had completed a 12‐day ultrasound course, examining organs like the kidney, heart, lungs, and liver, as well as detecting blood clots.

• Learners and experts agreed on imaging findings in approximately 9 out of 10 cases, reflecting the effectiveness of the curriculum across multiple organ systems.

• These results underscore POCUS's potential to strengthen healthcare in underserved regions, while also pinpointing areas for further training to achieve even more accurate patient evaluations

1. Introduction

Point‐of‐care ultrasound (POCUS), due to its cost‐effectiveness, portability, speed, versatility and safety, is rapidly expanding as a diagnostic modality, particularly in low‐resource settings [1]. Although POCUS technology has become more affordable and portable, other imaging techniques, such as computed tomography (CT) and magnetic resonance imaging (MRI), remain prohibitively expensive. In many developing countries, accessibility of both CT and MRI is further limited by the low overall number of machines, geographic restriction to urban medical centres predominantly within the private sector, and the need for highly trained radiologists, radiology technicians, repairmen and support staff [2, 3]. In contrast, POCUS machines can easily be transported to rural areas, and the acquisition and interpretation of ultrasound imaging can effectively be taught to providers in a variety of clinical settings for a wide range of clinical purposes [4, 5].

Sub‐Saharan Africa remains one of the most resource‐limited regions globally, particularly when it comes to healthcare technology, such as medical imaging [2, 3]. Several prior studies have examined POCUS utilisation in Sub‐Saharan Africa across a variety of clinical settings. For example, studies in rural Rwanda, Liberia and Ghana have indicated that POCUS is a cost‐effective way to improve diagnostic speed and accuracy [6, 7, 8, 9, 10]. In one study analysing the impact of the introduction of ultrasound services in rural Rwanda, ultrasound changed management in 43% of total patients scanned and 96% concordance rates were described between experts and learners [6]. Another study analysed the impact of ultrasound on dyspneic patients in an urban Rwandan emergency department, and ultrasound changed the leading diagnosis in 66% of cases [7]. In Ghana, focused cardiac and pulmonary POCUS training for emergency medicine residents improved initial diagnostic accuracy of critically ill patients from 57.1% to 71.9% [8]. Another study in Ghana found that POCUS improved diagnostic accuracy in critically ill patients [9]. In Liberia, an analysis concluded that POCUS was primarily used for obstetric studies and changed management in 62% of cases [10]. To date, however, no prior studies have evaluated POCUS initiatives in The Republic of Somaliland.

The Republic of Somaliland, located in the northern Horn of Africa, is home to 5 million Somali people with a federal budget of 400 million dollars, of which only 6% is spent on healthcare [11]. Due to this dearth of public resources, access to medical imaging in the public sector is limited. Many regional hospitals lack any advanced imaging modalities, and the few public hospitals that have CT scanners or ultrasound capacity charge patients high fees for these services, approximately 120 USD and 30 USD, respectively. Inspired by prior efforts to promote POCUS in sub‐Saharan Africa [6, 7, 8, 9, 10], the Ministry of Health Development (MoHD) started initiatives to develop local ultrasound expertise and infrastructure. In this study, we sought to assess the efficacy of ultrasound training that has taken place thus far and identify areas for improvement prior to national scale‐up.

2. Materials and Methods

This is a retrospective, observational study of a convenience sample of 500 ultrasound scans of adult patients who presented to one of four public hospitals: Hargeisa Group Hospital (the national referral hospital), Gabiley Hospital, Arabsiyo Hospital and Kalabaydh Hospital (regional hospitals). Twelve general practitioners with no prior non‐obstetric ultrasound experience were selected for participation in POCUS training (several practitioners had prior ultrasound experience in obstetrics, but not in other POCUS applications). Ultrasound training was conducted by physicians with POCUS expertise and board‐certified by either the American Board of Internal Medicine or the American Board of Family Medicine. The ultrasound training curriculum spanned 24 half‐day sessions that took place over the course of 12 days, including instruction with lecture and image review for 2–3 h in the morning, followed by real‐world scanning sessions in the afternoon. The first series of lectures covered the basics of ultrasonography, how to operate the Butterfly IQ+ (Butterfly Network Inc., Guilford, CT) and Sonosite Micromaxx (Fujifilm Sonosite Inc., Bothell, WA) ultrasound probes and upload images to the cloud for review, training expectations, ultrasound limitations and reporting form instructions. The second series of lectures and exams covered cardiac ultrasound [12]. The third series of lectures focused on lung ultrasound of the dyspneic patient and the adapted bedside lung ultrasound in emergency (BLUE) protocol [13]. The fourth series of lectures focused on abdominal ultrasound of visceral structures, including kidneys, liver, gallbladder and small bowel. The fifth series of lectures concluded the course with lower extremity ultrasound for deep venous thrombosis and procedural applications of ultrasound. One month prior to the ultrasound training, Somali general practitioners were also provided with access to online courses, quizzes and educational materials (Global Ultrasound Institute, Walnut Creek, CA).

Following the completion of the course, general practitioners used the provided probes in their clinical work and uploaded scans (in the form of six‐second, MP4 videos) directly from the tablets to Butterfly Cloud (Butterfly Network Inc., Guilford, CT) as they were conducted. Participants were instructed to fill out a standardised reporting form with their interpretations of each scan (Figure A1), covering at least one of 21 sonographic variables (Table 1). Scans were acquired in three separate practice settings: the emergency department (ED), intensive care unit (ICU) and outpatient clinics. Quality and accuracy of the scans were evaluated through blinded image review conducted by two physician ultrasound experts. Feedback on ultrasound images with educational findings was provided to general practitioners using chat messaging forums after the scans were uploaded. Data from the first 500 forms were entered into a spreadsheet for analysis. Of note, each form represents a separate clinical encounter, but not necessarily a unique patient (as the same patient may have been assessed on separate occasions throughout their hospital stay). Concordance was calculated as the percentage of agreement between learner and expert. Abnormal concordance was calculated as the percentage of agreement between learner and expert for abnormal findings only (studies in which the expert identified an abnormal finding). For cardiac and pulmonary scans, experts also classified the image quality as either ‘fail’ or ‘pass’ based on gestalt (taking into account resolution, depth, gain, axis, etc.) as to whether or not the uploaded study could be used to make a conclusion on relevant sonographic variables. Interrater agreement and reliability were calculated for individual sonographic variables using kappa statistics. Cohen's kappa values ≤ 0 suggest no agreement, 0.01–0.20 none to slight, 0.21–0.40 fair, 0.41–0.60 moderate, 0.61–0.80 substantial and 0.81–1.00 almost perfect agreement. p < 0.05 was used to determine statistical significance. R statistical software (version 4.3.1) was used for analyses.

TABLE 1.

Frequency of learner and expert interpretations for each ultrasound variable across organ systems.

Organ system	Variable	Possible responses	Learner responses	Expert responses
Cardiac	Image quality	Pass	NA	330
		Fail	NA	40
	LV function	Preserved	197	199
		Moderately reduced	96	88
		Severely reduced	67	64
		Indeterminate	1	10
	RV dilation	Absent	169	165
		Present	74	65
		Indeterminate	22	35
	Effusion	Absent	266	265
		Present	55	47
		Indeterminate	5	14
	IVC	Dilated	52	48
		Mid‐range	77	76
		Small/collapsing	66	52
		Indeterminate	18	37
Pulmonary	Image quality	Pass	NA	306
		Fail	NA	13
	Lung profile	A‐line predominant	128	133
		B‐line predominant	65	63
		AB‐mixed profile	111	104
		Indeterminate	1	1
	Effusion	Absent	165	152
		Present	103	106
		Indeterminate	4	14
	Consolidation	Absent	129	117
		Present	74	72
		Indeterminate	11	25
	Sliding	Present	236	238
		Absent	2	2
		Indeterminate	3	1
Renal	Size	Normal	81	88
		Small	12	8
		Enlarged	6	3
		Indeterminate	1	1
	Hydronephrosis	Absent	143	147
		Present	22	17
		Indeterminate	1	2
	Cortex differentiation	Present	156	164
		Absent	23	13
		Indeterminate	6	8
	Perinephric fluid	Absent	38	38
		Present	22	21
Indeterminate		1	2
Hepatobiliary	Evidence of cirrhosis	Absent	111	110
		Present	28	26
		Indeterminate	2	5
	Perihepatic fluid	Absent	46	43
		Present	38	41
		Indeterminate	1	1
	Gallbladder wall thickening	Absent	64	62
		Present	23	17
		Indeterminate	7	15
	Gallstones	Absent	81	73
		Present	27	22
		Indeterminate	2	15
	Biliary dilation	Absent	31	31
		Present	4	2
		Indeterminate	9	11
DVT	Image quality	Pass	NA	36
		Fail	NA	5
	Evidence of DVT	Absent	27	24
		Present	15	14
		Indeterminate	0	4
Skin and soft tissue	Cellulitis/edema	Absent	0	0
		Present	2	2
		Indeterminate	0	0
	Abscess	Absent	1	1
		Present	0	0
		Indeterminate	0	0
Small bowel	Dilation	Absent	0	0
		Present	0	0
		Indeterminate	0	0

Abbreviations: DVT, deep vein thrombosis; IVC, inferior vena cava; LV, left ventricle; NA, not applicable; RV, right ventricle.

2.1. Ethics Statement

Ethical approval was obtained in writing from the Research Committee of the Somaliland Ministry of Health Development (MoHD) on June 10, 2022 (Committee Chair, Director General of MoHD, Mohamed Abdi Hargeye, MD). Informed consent for participants was waived by the Research Committee of the MoHD.

3. Results

Our analysis included 500 scans (see Tables 1, 2, 3 for results). There was overall high concordance (90%, N = 500) between learner and expert interpretation of ultrasound studies acquired in the clinic (N = 106), emergency department (N = 253) and intensive care unit (N = 141). Concordance remained high when assessing abnormal findings only (83%). Concordance was highest for renal imaging (95%, N = 242), lung imaging (91%, N = 365) and hepatobiliary imaging (91%, N = 217) while concordance was lower for cardiac imaging (87%, N = 400) and DVT imaging (83%, N = 42). Imaging quality was higher for lung studies (96% pass rate, N = 365) than for cardiac studies (89% pass rate, N = 400). Concordance was also compared between practice settings. The ICU studies had the highest overall concordance (91%), followed by the clinic studies (89%), whereas studies conducted in the ED had the lowest overall concordance (87%). Cohen's kappa coefficient between learner and expert interpretation was statistically significant and greater than 0.8 for 18 out of 18 studies (evaluation for small bowel dilation, soft tissue abscess and soft tissue cellulitis/edema were excluded from analysis due to small sample size), indicating substantial agreement in all studies. The studies with the highest kappa coefficient were the evaluation for lung sliding (0.98, N = 241), hydronephrosis (0.97, N = 166), cirrhosis (0.97, N = 141), perinephric fluid (0.95, N = 61) and perihepatic fluid (0.95, N = 85). The studies with the lowest kappa coefficient were the evaluation of lung consolidation (0.82, N = 214), IVC (0.83, N = 213), DVT (0.83, N = 42), biliary dilation (0.84, N = 44) and gallbladder wall thickening (0.85, N = 94).

TABLE 2.

Summary of results. (a) Pass rate and concordance by organ system. (b) Overall pass rate and concordance across organ systems. (c) Number of studies by practice setting.

(a)
Variable	Cardiac	Lung	Renal	Hepatobiliary	DVT	S&ST	Small bowel
Number of studies	400	365	242	217	42	2	0
Percent of total studies	0.32	0.29	0.19	0.17	0.03	0.00	0.00
Image pass rate	0.89	0.96	NA	NA	0.88	NA	NA
Concordance	0.87	0.91	0.95	0.91	0.83	1.00	NA
Abnormal concordance a	0.81	0.88	0.90	0.80	0.72	1.00	NA

(b)
Combined results across organ systems
Pass rate	0.92
Concordance	0.90
Abnormal concordance	0.83

(c)
Number of studies by practice setting
Clinic	106
ICU	141
ED	253
Total	500

Abbreviations: DVT, deep vein thrombosis; ED, emergency department; ICU, intensive care unit; NA, not applicable; S&ST, skin and soft tissue.

^a The percent of agreement between learner and expert for studies in which the expert identified an abnormal finding only.

TABLE 3.

Interrater agreement and reliability.

Indicators	Observations, n	Percentage agreement, %	Cohen's kappa (95% CI) a , b
Cardiovascular
LV function	361	88.9	0.82 (0.77, 0.86)
RV dilation	265	87.9	0.77 (0.72, 0.82)
Pericardial effusion	326	89.9	0.68 (0.63, 0.73)
IVC	213	83.1	0.77 (0.71, 0.83)
DVT	42	83.3	0.68 (0.54, 0.82)
Pulmonary
Profile	305	90.5	0.85 (0.81, 0.89)
Pleural effusion	272	91.9	0.84 (0.79, 0.88)
Consolidation	214	82.2	0.68 (0.62, 0.74)
Lung sliding	241	98.3	0.50 (0.44, 0.56)
Renal
Size	65	93.8	0.76 (0.65, 0.86)
Hydronephrosis	166	97	0.87 (0.82, 0.92)
Cortex differentiation	185	94.6	0.78 (0.72, 0.84)
Perinephric fluid	61	95.1	0.90 (0.83, 0.98)
GI
Cirrhosis	141	96.5	0.90 (0.85, 0.95)
Perihepatic fluid	85	95.3	0.91 (0.85, 0.97)
GB wall thickening	94	85.1	0.70 (0.61, 0.79)
Gallstones	110	86.4	0.70 (0.61, 0.79)
Biliary dilation	44	84.1	0.65 (0.51, 0.79)

Abbreviations: DVT, deep vein thrombosis; GI, gastrointestinal; IVC, inferior vena cava; LV, left ventricle; RV, right ventricle.

^a Cohen's kappa values ≤ 0 suggest no agreement; 0.01–0.20 none to slight, 0.21–0.40 fair, 0.41–0.60 moderate, 0.61–0.80 substantial and 0.81–1.00 almost perfect agreement.

^b p values < 0.05 for K for all the indicators.

4. Conclusion and Discussion

The above results indicate high concordance between learner and expert assessment of POCUS studies in a variety of clinical settings. Concordance was highest in lung, renal and hepatobiliary POCUS, and relatively lower for cardiac and DVT assessments. Of note, there were no small bowel studies and only two skin and soft tissue studies uploaded, and so neither of these evaluations were included in the analysis. This suggests that either the participants did not encounter scenarios where they felt POCUS would be useful for bowel and skin assessment or that learners may not have been comfortable with these assessments. The individual studies with the lowest kappa coefficients, such as the evaluation for pulmonary consolidation and assessing for biliary dilation, also merit more attention in training going forward. A closer inspection of the data identifies other areas for improvement. For example, identification of a reduced ejection fraction should often prompt assessment of the IVC and lung parenchyma to evaluate for evidence of volume overload. However, of the 164 studies where a reduced EF was identified by the learner, the IVC and lung profile were not assessed in about a third of cases. It is unclear if these assessments were not completed, performed but not uploaded, or deemed unnecessary in patients who clinically appeared well compensated from a cardiopulmonary perspective.

This study contributes to the limited but growing body of literature on ultrasound education in sub‐Saharan Africa. There are three prior studies in the region that examine the utilisation and impact of ultrasound following an educational intervention: Shah et al., Henwood et al. and Kotlyar et al. [6, 7, 10]. In Rwanda, Shah et al. [6] reported a similarly high overall concordance rate between experts and learners (96% vs. 90% in our study). A subsequent study in Rwanda by Henwood et al. [7] further supports the effectiveness of ultrasound training, reporting a sensitivity of 94% and a specificity of 98%. While both Shah et al. [6] and Henwood et al. found a change in management rate of around 40% and Kotlyar et al. in Liberia reported a rate of 62%; our study observed a change in management rate of 85%. This higher rate is likely influenced by selection bias, given the relatively low response rate for this question. Another notable difference in our study compared to these previous studies was the distribution of scan types. Obstetric scans were the most common application in Shah et al. [6] and Kotlyar et al., abdominal ultrasound for free fluid was the most common in Henwood et al., and cardiac scans were the most common application in this study. These differences are likely due to variations in patient populations and training.

There are several strengths of this study. First, the study evaluates effectiveness based on pragmatic, real‐world applications of ultrasound imaging. This is in contrast to several previous studies in the ultrasound training literature, which often assess training modalities with online quizzes or simulations [14, 15, 16]. While such studies benefit from internal validity, the practical implications and external validity of the findings are uncertain. Second, this is the first study to implement and evaluate ultrasound training in The Republic of Somaliland and, more broadly, within the Horn of Africa. While there have been several prior studies focusing on ultrasound training in other low‐resource settings, findings cannot necessarily be extrapolated. The Republic of Somaliland has unique public health structural challenges and opportunities, and therefore merits targeted inquiry. Third, because trainees maintain access to training equipment and educational resources after the didactic period is completed, and can continue to upload studies and receive feedback through ongoing collaboration with ultrasound experts, this initiative is likely to result in sustainable ultrasound infrastructure. This sustainability is further supported by strong, ongoing collaboration and partnership between the MoHD, regional hospitals, physician learners and physician educators in both the United States and The Republic of Somaliland.

This study also has several limitations. First, because there is no pre‐intervention assessment, the degree of improvement in ultrasound skills cannot be precisely measured. This limits the internal validity of the study. However, trainees involved had very limited exposure to ultrasound prior to enrolling in the study. It can therefore be assumed that most sonography skills exhibited after the didactic portion of the study are attributed to learning that took place during the training. Regardless, acquiring baseline data would strengthen evidence in future studies. Validity could further be enhanced by using several experimental arms. For example, learners could be randomised to purely online learning versus in‐person didactic learning. This would not only help assess the effectiveness of the programme overall, but also help identify the most time and cost‐effective methods of ultrasound training to guide future investment in the programme. Second, there are several areas where selection bias could be present in this study. Although learners are encouraged to upload all images acquired during the study period to the cloud, it is probable that not all studies that were conducted were ultimately uploaded for analysis and feedback. For example, if a learner was having difficulty acquiring an adequate cardiac window, or was able to get an acceptable window but had trouble interpreting the findings, then they may be inclined not to upload the study for evaluation. Therefore, the image pass rate may be overestimated in this study. Conversely, if the learner found a clear case of abnormal pathology or obtained particularly clear images, they may be more likely to upload a study for evaluation. Further, because data was not collected on which individual learners uploaded which scans, it is possible that a small number of learners uploaded a disproportionate number of scans. Learners less confident in their POCUS skills may have been less inclined to upload scans. Another area for potential selection bias is in the selection of physician learners. These learners volunteered for the training programme through the MoHD and may have a higher than average interest in POCUS or predisposition for learning new skills. Third, there are inherent limitations in retrospective image review. While experts made their best attempt to assess the quality and interpretation of uploaded studies, it is not possible to make a complete assessment of skill and quality without hands‐on instruction. For example, for a study assessing the IVC, it is possible that the uploaded scan did not include the largest diameter cross‐section or that too much pressure was applied to the abdomen, resulting in a confounded study. Fourth, the study did not account for changes in scan quality or accuracy of image interpretation over time. Future studies could collect data on individual participants and evaluate if further experience improves skills or, alternatively, if skill deteriorates further from the training period.

In conclusion, this study demonstrates that, when supported by online POCUS resources, a concise didactic training period, and sustained remote feedback, general practitioners with limited prior experience can achieve substantial concordance with expert assessments in the interpretation of early clinical scans. Areas for improvement going forward include further emphasis on cardiac, bowel, skin and soft tissue and DVT imaging, improving internal validity by acquiring baseline data and developing an experimental design, reducing selection bias by ensuring more complete data acquisition, and enhancing generalisability by including a more diverse cohort of physicians in different geographic areas.

Author Contributions

Jack Inglis: primary and corresponding author, drafted the work, contributed to the acquisition, analysis, and interpretation of the work. Karly Boll: drafted the work, contributed to the acquisition, analysis, and interpretation of the work. Abdihakim M. Mohamed: contributed to the conception and design of the work, contributed to the acquisition, analysis, and interpretation of the work. Max Fraden: drafted the work, contributed to the conception and design of the work, contributed to the acquisition, analysis, and interpretation of the work. Abdirahman Madar: drafted the work, contributed to the conception and design of the work, contributed to the acquisition, analysis, and interpretation of the work. Stephen Merjavy: drafted the work, contributed to the conception and design of the work, contributed to the acquisition, analysis, and interpretation of the work. Ahmed Abdi Hirsi: contributed to the conception and design of the work, contributed to the acquisition, analysis, and interpretation of the work. Matthew S. Haldeman: contributed to the conception and design of the work, contributed to the acquisition, analysis, and interpretation of the work. Celestin Hategeka: principal investigator, drafted the work, contributed to the conception and design of the work, contributed to the acquisition, analysis, and interpretation of the work.

Conflicts of Interest

The authors declare no conflicts of interest.

Appendix A.

FIGURE A1.

A screenshot from Butterfly Cloud depicting a sample study (left) and participant worksheet (right). Participants uploaded acquired images in video loop format (left), with each clip lasting up to 6 s, along with their interpretation of the findings (right). Experts graded these interpretations in the notes section of the worksheet.

Calculating Concordance

Concordance was calculated as the percentage of agreement between learner and expert. A score of 1 was assigned to individual studies in which there was agreement between learner and expert, and a score of 0 in which there was disagreement between learner and expert. These scores were aggregated and reported for each organ system that the learner assessed. For example, if a learner uploaded a cardiac study in which they assessed LV function, RV dilation, and pericardial effusion, and the expert agreed with all of their findings, this would be reported as 100% concordance—the expert and learner agreed on all of the findings. If the expert agreed with the learner's interpretation of LV function and RV dilation, but not on the interpretation of pericardial effusion, this would be reported as 66.7% concordance by averaging the score assigned to the three studies: (1 + 1 + 0)/3 = 66.67%. If a learner and expert both evaluated a study as indeterminate, this would be considered a concordant study (and assigned a score of 1). If the learner felt a study was indeterminate but the expert felt that a more definitive conclusion could be drawn from the scan, or vice versa, this was classified as a discordant study (and assigned a score of 0).

Calculating Abnormal Concordance

Abnormal concordance was calculated as the percentage of agreement between learner and expert for abnormal studies only (those in which the expert identified an abnormality). Abnormal concordance was calculated due to the theoretical concern that learners would be less able to identify abnormal findings than to confirm normal ones, due to the relative infrequency of encountering abnormal findings. For example, if a learner uploaded a study in which they assessed LV function and RV dilation, and the expert interpretation was that LV function is normal and that RV dilation is present, the abnormal concordance would be 100% if the learner also identified RV dilation, and would be 0% if the learner determined there was no RV dilation. LV function would not be included in this calculation because the expert determined the LV function was normal. Thus, averaging abnormal concordance for each organ system provides a more accurate assessment of a learner's ability to identify relevant abnormalities compared to total concordance.

Inglis J., Boll K., Mohamed A. M., et al., “Assessing the Efficacy of POCUS Training in The Republic of Somaliland: A Retrospective Analysis of 500 Scans Conducted by General Practitioners,” Australasian Journal of Ultrasound in Medicine 28, no. 3 (2025): e70022, 10.1002/ajum.70022.

Data Availability Statement

All data were entered and stored using Google Sheets and can be accessed using the following URL: https://docs.google.com/spreadsheets/d/16OkX1nFflz63lrSTm‐_tiIwMF4InJtRoDWGnF39cr7w/edit?gid=0#gid=0.