Medical practice variation in Vietnam: a cross-sectional study Authors: TTDuong Doan, Van Phuoc Le, Cu Linh Le, Thi Thuy Dung Nguyen, Lam Tran Khanh Le, Steven Johnson, Othman Ouenes, John Peabody Published: 2026 ABSTRACT AbstractBackgroundVariation in clinical care is a recognized global issue and has been associated with poor health outcomes. This study describes the use of the E-Patient Quality Improvement and Standardization (EQIS) platform to assess provider clinical decision-making and to estimate financial waste from unnecessary laboratory and imaging tests in two clinical areas—obstetrics and cardiology—in Vietnam.MethodsThis cross-sectional study included maternal and cardiac health clinicians across all seven hospitals within a private health system in Vietnam. A total of 216 obstetric providers and 71 cardiac care providers participated. Each participant was randomly assigned three simulated e-patient cases relevant to their specialty on the EQIS platform. Overall scores were calculated as the percentage of correct responses for each participant. The number of unnecessary tests and their associated costs were estimated for each case.ResultsOverall, practice quality varied substantially, with mean scores of 73.5 (9.2), 95%CI: 72.3–74.7 for obstetric cases and 80.0 (7.5), 95% CI: 79.6–82.1 for cardiac cases. Across domains, diagnostic performance had the lowest mean score (< 70%) and the greatest variability (SD: 19.0–19.9). Mean scores also varied by provider type, educational level, department, average daily patient volume, hospital, age, and years of experience. Relative to evidence-based recommendations, providers ordered 2.4–2.9 unnecessary tests per case. These patterns were associated with estimated excess costs of 330 to 495 million VND per month for the hospital system.ConclusionsClinical quality varied substantially, and unnecessary testing was common in both obstetric and cardiac care across the seven private hospitals included in this study in Vietnam. The EQIS platform may be a useful tool for assessing clinical practice patterns and estimating unnecessary costs.Keywords: Obstetrics, Cardiology, Clinical practice, Clinical variation, EQIS platform, Simulated e-patients Background Medical practice variation refers to differences in clinical decision-making, including the overuse and underuse of services [1]. These variations are associated with adverse patient outcomes and increased burden on healthcare systems [2]. Contributing factors are classified into three categories: patient characteristics, healthcare system attributes, and provider-related factors [3]. Traditional approaches to assessing healthcare worker competency are increasingly constrained. Advances in medical technology have made hands-on training with real patients more ethically challenging [4]. Simulation-based education provides an effective approach for competency assessment and active learning [5]. The Patient Quality Improvement and Standardization (EQIS) approach, for example, has been used to measure variation across diverse providers and clinical disciplines [6–10]. Healthcare quality in Vietnam remains suboptimal [11]. A 2015 study in two public hospitals reported high medical error rates which approached 40% [12]. Prior research has focused mainly on healthcare utilization and patient satisfaction, particularly in primary care and resource-limited settings [13–15]. Assessments of clinical competency are limited [16]. This study evaluated competencies and practice variation among obstetric and cardiac providers in Vietnam using simulated e-patient cases via the EQIS platform. It also quantifies financial waste from unnecessary laboratories and imaging tests to inform hospital administrators. Methods Study setting The study was conducted in a private hospital system in Vietnam in June 2024. At the time of data collection, the system comprised seven hospitals located in major cities across the country (Fig. 1). Across these hospitals, the number of monthly births ranged from 50 to 200, with annual cesarean section rates of 45%–70%. The hospitals also provided cardiovascular care, with approximately 50–700 outpatient visits and 10–90 inpatient admissions per month. Seven private hospitals of an enterprise across Vietnam in 2024 Study participant selection This study focused on two clinical areas – (1) obstetric care and (2) cardiac care—selected by the hospital system. In total, 216 eligible obstetric providers and 71 cardiac physicians participated. The sample was not intended to be representative of other private hospitals in Vietnam. Obstetric providers included all frontline physicians, nurses, and midwives involved in antenatal care, normal deliveries, and obstetric emergencies. Providers in infertility units, diagnostic technicians (e.g., ultrasound or laboratory staff), and those on maternity leave during the assessment period were excluded. For cardiac care, all physicians from the Departments of Cardiology (invasive and non-invasive units), Internal Medicine, Emergency, and Resuscitation were included. Development of simulated e-patient cases Three expert physicians developed 18 e-patient cases for each clinical area. All members had completed residency training in their specialties and had 13–25 years of clinical experience. The team underwent three training sessions on e-patient case development, covering literature review of relevant epidemiology, definition of assessment objectives for each subclinical area, and translation of objectives into simulated cases. Each clinical area was divided into three subclinical areas. Assessment objectives were reviewed for consistency with updated international and national evidence-based guidelines, including those of the Vietnamese Ministry of Health [17–19], the World Health Organization [20], the American College of Obstetricians and Gynecologists, and the European Society of Cardiology [21]. Each case was structured around six domains of clinical practice: medical history; physical examination; laboratory tests and imaging work-up; diagnosis; management; and follow-up strategies. The case development process for these clinical areas is described below and has been previously applied in other settings [6–10]. Case-writing experts developed case summaries based on predefined assessment objectives. The summaries followed six clinical domains and included all relevant response options, including appropriate and inappropriate choices and necessary and unnecessary tests. The experts uploaded the objectives and summaries to the EQIS platform, which was developed by an IT team and incorporated artificial intelligence (ChatGPT 3.5). The platform generated initial case drafts, with 22–25 structured questions per case depending on clinical updates. AI-generated cases were reviewed in multiple rounds by case-writing experts and a clinical committee to ensure alignment with clinical practice (Fig. 2). Cases were developed in English and translated into Vietnamese for assessment. The development and quality assurance process of simulated e-patient cases Data collection Before data collection, participants received online or in-person training on simulated e-patient cases and the EQIS platform. Written informed consent was obtained during this session. Cases were randomly assigned using the platform’s built-in randomization algorithm, ensuring an equal distribution of three cases across participants. Each case required approximately 20–25 min, with a total completion time of about one hour. The platform provided real-time feedback followed by an overall score and summary feedback after completion of each case. Participants who did not complete the assessment received up to three email reminders over two weeks. All eligible providers were enrolled and completed the assessment (completion rate 100%), likely influenced by institutional participation requirements and CME incentives. This study was approved by the ethics review committees of the participating hospitals and VinUniversity (Decision No 108/2023-QD-VMEC). Variables The primary outcome was the EQIS case score. Each case was scored using a predefined weighting system based on domain importance. Diagnosis and management were weighted 4 and 6 points, respectively, while the remaining four domains were weighted 1–2 points each. Weights were defined by the case-writing team and the clinical committee. Domain and overall scores were standardized to a 0–100%. Higher scores indicate greater adherence to evidence-based guidelines. Independent variables included sex (male, female), provider type (physician, midwife, nurse), department (cardiology, emergency), educational level (college/bachelor ≤ 4 years; physician or CKI/MSc 5–7 years; CKII/PhD ≥ 7 years of training), and average daily patient volume (obstetric care: ≤1, 1.1–5, ≥ 5; cardiac care: 1, 2, ≥ 3). Additional variables included completion of continuing medical education (CME) and hospital location (Hanoi, Quang Ninh, Hai Phong, Da Nang, Nha Trang, Ho Chi Minh City, Phu Quoc). CKI and CKII refer to Vietnam’s clinical specialist training programs (levels 1 and 2), undertaken by licensed physicians working in hospitals, typically requiring 1–3 years to complete. Data analysis Summary statistics were computed for all variables. Continuous variables are presented as means ± standard deviations (SDs) or medians with interquartile ranges, as appropriate. One-way analysis of variance (ANOVA) was used to compare scores across clinical areas, domains, and categorical independent variables. Pearson’s correlation was used to assess associations between age, years of experience, and overall scores. A p-value < 0.05 was considered statistically significant. Analyses were performed using IBM SPSS Statistics 27 (IBM Corp. Released 2020. IBM SPSS Statistics for Windows, Version 27.0. Armonk, NY: IBM Corp). The cost of unnecessary laboratory tests and imaging workups was estimated using the maximum prices specified in Circular No. 13/2023/TT-BYT (29/06/2023) of the Vietnam Ministry of Health. This circular defines price ceilings and valuation methods for on-demand medical examination and treatment services in public healthcare facilities [22]. We estimated the total healthcare cost as follows: (1) obtained monthly hospital records to estimate the number of patients presenting with moderate or severe maternal and cardiac conditions; (2) calculated the average number and cost of unnecessary tests per case to derive the cost per patient; and (3) multiplied the number of patients by the per-patient cost of unnecessary testing. Results Characteristics of participants Among the 216 obstetric healthcare providers, 42 were physicians (19.4%), 144 were midwives (66.7%), and 30 were nurses (13.9%). Of the 71 physicians in the cardiac cohort, 30 (42.3%) worked in Cardiology and 41 (57.7%) in the Emergency Department. The mean age was 36.4 (8.4) years for obstetric providers and 41.0 (7.2) years for cardiac providers. The mean years of experience were 13.0 (7.9) for obstetric providers and 13.9 (6.4) for cardiac providers. Educational attainment varied across groups. Among obstetric providers, 76.9% held college or bachelor’s degrees, 17.1% had 5–7 years of training, and 6.0% had ≥ 7 years of training. Among cardiac providers, 70.4% had 5–7 years of training, while the remainder had ≥ 7 years of training (Table 1). Table 1Characteristics of study participantsObstetric CareCardiac Care Characteristics N = 216 Characteristics N = 71Age: year, mean (SD)36.4 (8.4)Age: year, mean (SD)41.0 (7.2)Female, n (%)202 (93.5)Female, n (%)13 (18.3)Provider type, n (%)Department Physician42 (19.4)Cardiology30 (42.3) Midwife144 (66.7)Emergency/Resuscitation41 (57.7) Nurse30 (13.9)Educational level, n (%)Educational level, n (%)Bachelor/College166 (76.9)Physician, CKI specialist, MSc50 (70.4)Physician, CKI specialist, MSc37 (17.1)CKII specialist, PhD21 (29.6)CKII specialist, PhD13 (6.0)Experience: year, mean (SD)13.0 (7.9)Experience: year, mean (SD)13.9 (6.4)Average daily patient volume, n (%)Average daily patient volume, n (%) <=189 (41.2)<=127 (38.0) 1.1–585 (39.4)217 (23.9) >=542 (19.4)327 (38.0)Completed CME, n (%)182 (84.3)Completed CME, n (%)50 (70.4)Hospital, n (%)Hospital, n (%) Ha Noi87 (40.3)Ha Noi24 (33.8) Quang Ninh14 (6.5)Quang Ninh4 (5.6) Hai Phong23 (10.6)Hai Phong6 (8.5) Da Nang30 (13.9)Da Nang7 (9.9) Nha Trang25 (11.6)Nha Trang8 (11.3) Ho Chi Minh City29 (13.4)Ho Chi Minh City19 (26.8) Phu Quoc8 (3.7)Phu Quoc3 (4.2)Abbreviations. SD: standard deviation; MSc: Master of Science; PhD: Doctor of Philosophy; CME: continuing medical education; CKI or CKII: specialist clinical training programs (level I or II) in Vietnam for medical graduates working in hospitals, typically requiring one to three years to complete Characteristics of study participants Abbreviations. SD: standard deviation; MSc: Master of Science; PhD: Doctor of Philosophy; CME: continuing medical education; CKI or CKII: specialist clinical training programs (level I or II) in Vietnam for medical graduates working in hospitals, typically requiring one to three years to complete Clinical competence and variation Overall, practice quality varied widely, with scores ranging from 24% to 93%, a mean of 73.5 (9.2), 95%CI: 72.3–74.7 for obstetric cases and ranging from 59% to 95.4%, with a mean of 80.0 (7.5), 95%CI: 79.6–82.1 for cardiac cases (Table 2). Table 2Mean score, standard deviation (SD), and 95% confidence intervals (CIs) of health care providers by domainDomainObstetric care (n = 216)Cardiac care (n = 71)Mean (SD)95%CIMean (SD)95% CIOverall73.5 (9.2)72.3–74.780.0 (7.5)79.6–82.1History82.7 (12.1)81.1–84.392.8 (6.6)91.2–94.4Physical examination84.6 (13.1)82.9–86.489.7 (8.7)87.6–91.7Work-up80.5 (14.1)78.6–82.489.2 (8.5)87.2–91.2Diagnosis65.2 (19.0)63.6–67.769.0 (19.9)64.3–73.7Management74.3 (11.3)72.8–75.876.9 (8.9)74.8–79.0Follow-up care66.6 (16.0)64.5–68.879.4 (12.7)76.4–82.4 Mean score, standard deviation (SD), and 95% confidence intervals (CIs) of health care providers by domain In both clinical areas, the diagnostic domain had the lowest mean scores and greatest variability (obstetric: 65.2 (19.0), 95% CI: 63.6–67.7; cardiac: 69.0 (19.9), 95% CI: 64.3–73.7). The second lowest scores were observed in follow-up care for obstetric cases (66.6 (16.0), 95% CI: 64.5–68.8) and in management for cardiac cases (76.9 (8.9), 95% CI: 74.8–79.0). We observed significant differences in scores across provider type, educational level, department, average daily patient volume, hospital, age, and years of experience (p < 0.05) (Tables 3 and 4). Table 3Clinical competency and variation by characteristics of obstetric providersObstetric careCharacteristicsMean (SD) of score p Sex 0.847 Female73.4 (7.4) Male73.4 (9.4) Provider type 0.008 Physician76.7 (7.1) Nurse70.0 (8.6) Midwife73.3 (9.6) Educational level 0.030 Bachelor/College72.6 (9.6) Physician, CKI specialist, MSc76.1 (7.3) CKII specialist, PhD77.6 (7.7) Average daily patient volume 0.442 <=172.7 (8.1) 1–573.7 (10.6) >=574.8 (8.5) Completion of CME 0.685 Yes73.6 (9.5) No72.9 (7.4) Hospital 0.214 Ha Noi71.9 (10.3) Quang Ninh78.2 (7.3) Hai Phong73.2 (6.5) Da Nang75.8 (9.7) Nha Trang73.0 (7.6) Ho Chi Minh City74.1 (8.9) Phu Quoc73.7 (9.1) Age (year) 0.18 0.01* Experience (year) 0.15 0.027* Abbreviations. SD: standard deviation; MSc: Master of Social Science; PhD: Doctor of Philosophy; CME: continuing medical education; *Pearson’s correlation coefficient and corresponding p value; CKI or CKII: specialist clinical training programs (level I or II) in Vietnam for medical graduates working in hospitals, typically requiring one to three years to complete Clinical competency and variation by characteristics of obstetric providers Abbreviations. SD: standard deviation; MSc: Master of Social Science; PhD: Doctor of Philosophy; CME: continuing medical education; *Pearson’s correlation coefficient and corresponding p value; CKI or CKII: specialist clinical training programs (level I or II) in Vietnam for medical graduates working in hospitals, typically requiring one to three years to complete Table 4Clinical competence and variation by characteristics of cardiac care providersCardiac care Characteristics Mean (SD) of scorep Sex 0.068 Female83.8 (6.6) Male79.6 (7.5) Department 0.004 Cardiology83.3 (6.4) Emergency/Resuscitation78.2 (7.5) Educational level 0.162 Physician, CKI specialist, MSc79.5 (6.8) CKII specialist, PhD82.3 (8.6) Average daily patient volume 0.004 176.7 (7.1) 281.7 (7.6) >=383.1 (6.4) Completion of CME 0.708 Yes80.5 (7.7) No79.8 (6.8) Hospital 0.040 Ha Noi78.7 (8.1) Quang Ninh79.3 (8.3) Hai Phong76.6 (2.1) Da Nang76.9 (4.9) Nha Trang79.6 (9.9) Ho Chi Minh City84.1 (5.7) Phu Quoc88.5 (2.7) Age (year) -0.25 0.035* Experience (year) -0.24 0.043* Abbreviations. SD: standard deviation; MSc: Master of Social Science; PhD: Doctor of Philosophy; CME: continuing medical education; *Pearson’s correlation coefficient and corresponding p value; CKI or CKII: specialist clinical training programs (level I or II) in Vietnam for medical graduates working in hospitals, typically requiring one to three years to complete Clinical competence and variation by characteristics of cardiac care providers Abbreviations. SD: standard deviation; MSc: Master of Social Science; PhD: Doctor of Philosophy; CME: continuing medical education; *Pearson’s correlation coefficient and corresponding p value; CKI or CKII: specialist clinical training programs (level I or II) in Vietnam for medical graduates working in hospitals, typically requiring one to three years to complete Among obstetric providers (Table 3), physicians had the highest mean score (76.7 (7.1)), followed by midwives (73.3 (9.6)) and nurses (70.0 (8.6)), p = 0.008. Providers with ≥ 5 years of training achieved higher scores (76.5 (7.2)) than those with bachelor’s (73.1 (9.4)) or college-level training (71.9 (9.9)), p = 0.03. Age was positively correlated with mean score (r = 0.18, p = 0.01), as were years of experience (r = 0.15, p = 0.027). No significant differences were observed by hospital, sex, average daily patient volume, or possession of continuing medical education certificates. Among cardiac care providers (Table 4), physicians working in the cardiology department had a higher mean score (83.3 (6.4)) than those in the emergency department (78.2 (7.5), p = 0.004. Providers who saw at least three patients per day achieved the highest mean score (83.1 (6.4)), whereas those who saw fewer than one patient per day had the lowest mean score (76.7 (7.1), p = 0.004. There was a significant negative correlation between age and mean score (r = − 0.25; p = 0.035), as well as between years of experience and mean score (r = − 0.24; p = 0.043). No significant differences were observed with respect to sex, educational level, or completion of CME. Unnecessary testing and cost among simulated e-patient cases Numerous unnecessary tests were ordered, each incurring additional cost (Table 5). Among the 648 obstetric care cases, a total of 1,888 unnecessary tests were identified, corresponding to a cost of 890,000,000 VND (~ 35,000 USD). On average, obstetric providers ordered 2.9 unnecessary tests per case, resulting in an additional cost of approximately 1.4 million VND (~ 54 USD) per case. As shown in Table 6, the most frequently ordered unnecessary test was the liver function test (24.8% of cases), followed by C-reactive protein (20.8%), bilirubin (13.4%), and rapid urinalysis tests (13.0%). Table 5Unnecessary testing and cost by clinical area among simulated e-patient casesObstetric care (n = 648)Cardiac care (n = 213)Total of unnecessary tests1888515Average unnecessary tests per case2.92.4Total cost of unnecessary tests890,000,000 VND* (~ 35,000 USD)560,000,000 VND*(~ 22,000 USD)Average cost of unnecessary tests per case1,400,000 VND(~ 54 USD)2,600,000 VND(~ 110 USD)* The cost of unnecessary laboratory tests or imaging workups was calculated based on the price levels stipulated in Circular No. 13/2023/TT-BYT dated 29/06/2023 of the Vietnam Ministry of Health Abbreviations. VND: Vietnam Dong, USD: United States Dollar Unnecessary testing and cost by clinical area among simulated e-patient cases 560,000,000 VND* (~ 22,000 USD) 1,400,000 VND (~ 54 USD) 2,600,000 VND (~ 110 USD) * The cost of unnecessary laboratory tests or imaging workups was calculated based on the price levels stipulated in Circular No. 13/2023/TT-BYT dated 29/06/2023 of the Vietnam Ministry of Health Abbreviations. VND: Vietnam Dong, USD: United States Dollar Table 6Top 10 unnecessary tests and associated costs by clinical area among simulated e-patient casesObstetric care (n = 648)Cardiac care (n = 213)No Testsn (% of cases)Testsn (% of cases)1Liver function (LFTs)161 (24.8)Chest X-ray36 (16.9)2C-reactive protein (CRP)135 (20.8)Abdominal ultrasound33 (15.5)3Bilirubin87 (13.4)Carotid ultrasound31 (14.6)4Thyroid function84 (13.0)Blood gas26 (12.2)5Quick urinalysis test80 (12.3)Uric acid24 (11.3)6Echocardiogram76 (11.7)Liver function (LFTs)15 (7.0)7Ionogram66 (10.2)Ultrasound of lower limb blood vessels14 (6.6)8Vaginal swab for Group B Streptococcus (GBS)49 (7.6)Flu testing13 (6.1)9Fetal echocardiography34 (5.2)Coagulation Profile12 (5.6)10Blood glucose levels33 (5.1)Follow-up NT-proBNP levels11 (5.2) Top 10 unnecessary tests and associated costs by clinical area among simulated e-patient cases Cardiac care providers ordered slightly fewer unnecessary tests (2.4 per case) than obstetric providers; however, the associated cost per case was substantially higher (2.6 million VND ~ 100 USD). The most ordered unnecessary tests in cardiology were chest X-ray (16.9%), abdominal ultrasound (15.5%), carotid ultrasound (14.6%), blood gas analysis (12.2%), and uric acid testing (11.3%) Based on an estimated monthly volume of 6,000 cardiac patients and 2,400 obstetric patients across seven hospitals, the total unnecessary cost to the health care system was approximately 330 million VND per month. This included 270 million VND for cardiac care (~ 10,500 USD) and 58 million VND for obstetric care (~ 2,300 USD). The largest contributors to unnecessary costs were coronary CT angiography in cardiac patients (38.5%) and abdominal CT in obstetric patients (11.6%). The cost per test for these procedures was 1.5 and 1.2 times higher, respectively, in the private hospital system. If private-sector pricing were applied, the total estimated unnecessary costs would range from 396 to 495 million VND per month Discussion Simulated e-patient approaches have previously been developed and applied to evaluate and improve quality of care [6–10]. However, each new clinical area, country context, and implementation model requires careful localization, field testing, and validation In this study, we developed an AI-supported platform to create simulated e-patient cases. The use of AI in clinical education and case development offers potential advantages in efficiency, scalability, and expansion of educational resources [23]. A key limitation of this approach is that AI models may not reflect local clinical norms and can produce biased, variable, or nonreproducible outputs, including hallucinations and other “black box” challenges [24–26]. To help mitigate these limitations, AI was used only to generate the initial structured drafts of cases based on predefined learning objectives and case summaries. All AI-generated content underwent multiple rounds of review and refinement by case-writing experts, clinical reviewers, and case-structure specialists. This process helped identify and correct inaccuracies, omissions, and inconsistencies within the cases We found that AI was useful for generating explanations and reference ranges for laboratory tests, describing imaging workups, and developing clinically relevant scenarios. However, AI systems also produced inaccurate, incomplete, or internally inconsistent information, particularly in complex clinical scenarios requiring longitudinal reasoning and integration of medical history, laboratory findings, and management decisions. Although these issues were mitigated through extensive manual review and expert oversight, we acknowledge that such processes may not fully overcome the limitations of AI-generated content. Future studies should evaluate the reliability, reproducibility, and educational outcomes of AI-supported case development using standardized prompts, external validation methods, and comparisons with traditionally developed clinical cases. While our findings suggest that this approach has potential for scalable formative assessment, further work is needed to establish psychometric validity, conduct multi-institutional validation, reduce hallucinations and bias, and evaluate learner-centered outcomes Using simulated e-patient cases to evaluate participant competencies, we found that clinical care scores varied across both obstetric and cardiac services. The lowest scores and greatest variability were observed in the diagnosis and treatment domains. These domains are directly related to care delivery and patient outcomes and coincide with higher levels of unnecessary testing observed in this study [27]. Our findings are broadly consistent with previous studies on the quality of care in Vietnam, which have also reported substantial variation in clinical practice. For example, a more limited study found variability in physicians’ ability to identify the underlying etiology of diabetes [9]. Consistent with our findings, prior studies internationally have reported variation in care across provider characteristics, including provider type, educational level, department, average patient volume, hospital, age, and years of experience We found no significant differences in performance by sex in either specialty. In obstetrics, age and years of experience were positively correlated with clinical scores. In cardiology, years of experience showed a negative correlation with performance. These patterns may reflect differences in training exposure or familiarity with current clinical guidelines across provider groups, particularly in specialties such as cardiology where evidence and technique involve rapidly. Our findings indicate substantial variation in care quality and frequent provision of unnecessary services. These findings suggest opportunities for improvement when care is measured and standardized. In a previous study, standardization of cardiology practice was associated with reductions in readmissions and direct costs [28]. In another study, reduced variation in the management of chronic obstructive pulmonary disease was associated with 233 fewer inpatient days per year, 371 fewer revisit days, and inpatient cost savings of nearly $1 million across 15 hospitals in a large U.S. health care system [29]. Future studies in Vietnam and other settings may help inform targeted training programs that build on existing strengths while addressing identified gaps to improve clinical competencies A notable finding in this study is the consistently high number of unnecessary tests (2.4–2.9 per case) and the corresponding use of potentially avoidable resources. The estimated cost of these tests was 1.4–2.6 million VND (~ 54–110 USD) per case. Cost estimates were based on prices from state-owned health facilities rather than private hospitals, which may underestimate actual costs and potential savings. Even using these conservative estimates, the cost of unnecessary care represents approximately 40%–70% of the average monthly income per person (~ 6 million VND). In addition, one analysis reported that at least 85% of per-person health expenditure in an urban area in 2022 was attributable to routine checks (approximately 2.8 million VND) [30]. Our findings are consistent with prior research suggesting a high level of medical intervention in obstetric care in Vietnam. Previous studies have reported that approximately 30% of women had more than 11 antenatal visits, with an average of 10 ultrasound examinations during pregnancy. In addition, around 40% of deliveries were performed by cesarean section, including one third of cases without clear medical indications [31, 32]. When considered alongside estimates of unnecessary testing and overprovision from other recent studies, our findings suggest substantial potential for cost savings in Vietnam This study has several limitations. Participating providers were drawn from higher-performing private hospital groups and may not be representative of clinicians in public-sector settings. State-owned health facilities were not included, and comparisons between private and public providers were beyond the scope of this study. The analysis was based on e-patient simulation cases. Further research is needed to examine how measured competencies relate to actual clinical practice and patient outcomes, particularly for maternal care in health facilities. In addition, the cost-per-case analysis did not include other potential economic components, such as unnecessary medications, costs associated with misdiagnoses, or downstream follow-up care. Previous studies have reported a high correlation between simulated and actual patient care [33]. In addition, improvements in e-patient scores have been associated with better health outcomes and lower system costs [33, 34]. Within this context, our findings may have practical relevance as Vietnam’s health care system seeks to standardize care processes, including through strengthened continuing medical education and the introduction of a new medical practice licensing framework. The use of the EQIS platform to measure clinical competency represents an innovative approach that may help address traditional challenges related to cost and scalability Conclusions We used the EQIS platform to assess the clinical decision-making of 287 health care providers in a private hospital system in Vietnam. The analysis showed substantial variation in clinical care across both obstetric and cardiac services. Gaps were observed between evidence-based standards and measured care, particularly in diagnostic accuracy and clinical management of simulated e-patients. Notable findings included lower diagnostis scores and considerable use of unnecessary tests and imaging. These patterns were associated with estimated excess costs of 330 to 495 million VND per month for the hospital system. Overall, simulated e-patient cases appear to be a feasible and scalable approach for assessing variation in medical practice and patterns of unnecessary resource use. Acknowledgements The authors would like to thank the participants who voluntarily participated in this study, the writers and research assistants who developed simulation cases, and the staff of the hospitals who willingly assisted in data collection. Abbreviations Artificial intelligence E-Patient Quality and Improvement and Standardization Continuing Medical Education SD United States Dollar Vietnam Dong Author contributions Conceptualization, J.P., L.V.P, L.C.L; methodology, J.P., L.V.P., L.C.L., L.T.K.L, N.N.H, N.T.T.D and D.T.T.D; analysis, T.T.D.D.; writing—original draft preparation, T.T.D.D, L.T.K.L; review and editing J.P., S.J., O.O, T.T.D.D.; supervision, J.P., L.V.P, L.C.L; project coordination, N.T.T.D. All the authors have reviewed and agreed to the published version of the manuscript. Funding This research was funded by Peabody Health Philanthropies. Data availability The database used for this study is part of an ongoing project in Hanoi, Vietnam. The data can be assessed by contacting Duong Doan ([doan.ttd@vinuni.edu.vn](mailto: doan.ttd@vinuni.edu.vn)), the corresponding author. Declarations Ethics approval and consent to participate The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Institutional Ethical Review Board for Biomedical Research Vinmec International General Hospital JSC- VinUniversity (Decision No 108/2023-QD-VMEC dated November 11th, 2023). All subjects gave both verbal and written informed consent for inclusion before they participated in the study. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The database used for this study is part of an ongoing project in Hanoi, Vietnam. The data can be assessed by contacting Duong Doan ([doan.ttd@vinuni.edu.vn](mailto: doan.ttd@vinuni.edu.vn)), the corresponding author.