Governing AI-Assisted Acute Kidney Injury Recognition in Viet Nam: AKI Criteria Verification, Medication Safety Review and Post-AKI Follow-Up

Governing AI-Assisted Acute Kidney Injury Recognition in Viet Nam: AKI Criteria Verification, Medication Safety Review and Post-AKI Follow-Up

A patient is admitted to the hospital with pneumonia and dehydration. On admission, the serum creatinine concentration is 0.9 mg/dL. Within 48 hours, it rises to 1.3 mg/dL. The patient is receiving several medicines, urine output has declined, and the clinical team is focused primarily on treating the infection.

An electronic kidney surveillance system identifies the change and generates an alert suggesting possible acute kidney injury (AKI). The hospital uses an AI-assisted platform, although the particular alert may have been generated by a conventional rule comparing creatinine measurements.

The alert draws attention to a potentially important problem, but it does not explain why kidney function has deteriorated or which treatment would be safest. The clinician still needs to verify the diagnostic criteria, assess the patient, investigate possible causes and determine whether urgent intervention is necessary.

Medication review may be important, but stopping treatment indiscriminately could cause harm. Similarly, administering intravenous fluids without assessing volume status may worsen the condition of someone with cardiac dysfunction or fluid overload.

Even if kidney function begins to improve before discharge, the episode may require further attention. The clinical team needs to determine whether follow-up testing is indicated, which medicines should be reassessed and who will evaluate kidney recovery.

These responsibilities illustrate why recognizing AKI is only one part of protecting patients.

AI may help healthcare organizations predict kidney injury, recognize clinically important changes or identify patients who need further assessment. Its contribution becomes meaningful when those capabilities support timely clinical decisions, appropriate treatment and continuity of care after hospitalization.

For hospitals in Viet Nam, the challenge is to incorporate AI into a reliable kidney safety process without confusing automated detection with diagnosis, allowing alerts to replace professional judgment or generating unnecessary interventions.

Acute Kidney Injury Is More Than an Abnormal Laboratory Result

Acute kidney injury is a clinical syndrome characterized by an abrupt deterioration in kidney function. It may develop during severe illness, surgery, exposure to certain medicines or other circumstances affecting kidney perfusion, structure or function.

The underlying causes are varied and frequently coexist. Reduced circulating volume, sepsis, impaired cardiac function, venous congestion, urinary obstruction and intrinsic kidney disease can all contribute. A patient may have several mechanisms operating simultaneously, making clinical assessment considerably more complex than interpreting a laboratory result.

AKI is associated with electrolyte disturbances, fluid imbalance, prolonged hospitalization, the need for kidney replacement therapy and increased risks of chronic kidney disease, cardiovascular complications and mortality.

These associations require careful interpretation. AKI may contribute directly to adverse outcomes, but it can also reflect the severity of another illness. A patient developing AKI during septic shock, for example, faces several interacting risks that cannot all be attributed to kidney injury itself.

Nevertheless, timely recognition matters because some contributing factors are reversible. Delays may result in missed opportunities to correct hypovolemia, identify urinary obstruction, address medication-related problems or prevent further deterioration.

AKI can also develop without obvious kidney-related symptoms. Some patients continue producing urine despite substantial kidney dysfunction, while others develop reduced urine output before serum creatinine has risen sufficiently to meet a diagnostic threshold.

Serum creatinine remains an important marker, but its concentration may increase after kidney function has already deteriorated. Fluid balance, muscle mass, creatinine generation and certain medicines can influence the result.

These limitations mean that a single laboratory measurement may not fully represent the patient's current kidney condition.

Electronic surveillance can help healthcare professionals recognize changes that might otherwise receive insufficient attention. However, its clinical usefulness depends on reliable information, appropriate interpretation and the actions taken after a concerning result is identified.

Verifying AKI Against Established Clinical Criteria

Hospitals need a consistent definition of AKI before evaluating an automated detection or prediction system.

The Kidney Disease: Improving Global Outcomes (KDIGO) 2012 Clinical Practice Guideline defines AKI when at least one of three criteria is met.

The first is an increase in serum creatinine of at least 0.3 mg/dL (26.5 µmol/L) within 48 hours. The second is an increase in serum creatinine to at least 1.5 times baseline, known or presumed to have occurred within the previous seven days. The third is urine output below 0.5 mL/kg/hour for at least six hours.

These criteria focus on changes in kidney function rather than relying exclusively on whether a laboratory result exceeds the normal reference range.

In the opening scenario, serum creatinine increased from 0.9 to 1.3 mg/dL within 48 hours. The absolute increase of 0.4 mg/dL meets the KDIGO creatinine-change criterion for AKI.

The relative increase is approximately 1.44 times the initial value, which is below the separate 1.5-fold threshold. If the measurements are valid and no more severe criterion is present, the creatinine change is consistent with KDIGO stage 1 AKI.

However, the reported decline in urine output also requires assessment. Without reliable information about urine volume, patient weight and the duration of oliguria, the complete AKI stage cannot be established from the scenario alone.

Conversely, a creatinine concentration of 1.8 mg/dL does not independently establish AKI if previous kidney function and the timing of any change are unknown.

KDIGO classifies AKI into three stages according to the severity of creatinine changes and urine output abnormalities. Additional criteria, including initiation of kidney replacement therapy, are relevant to stage 3.

Staging helps communicate severity and guide the intensity of clinical assessment, but it does not replace evaluation of the patient's condition or underlying cause.

A patient with chronic kidney disease may develop AKI superimposed on existing impairment. Another may arrive with community-acquired AKI already present, making the first measurement obtained during hospitalization an unsuitable baseline.

Urine output may identify AKI before the creatinine criteria are met, but automated assessment depends on accurate measurement and documentation. Incomplete fluid charts, unreliable collection and differences in nursing documentation can affect classification.

A safe system should distinguish between an electronic rule being satisfied, a clinical diagnosis requiring confirmation and situations where the available information remains insufficient.

The clinical team must determine whether the evidence supports AKI and whether additional investigation or immediate intervention is needed.

Baseline Creatinine and Laboratory Timing Matter

Selecting an appropriate baseline serum creatinine value is one of the most difficult aspects of electronic AKI recognition.

A hospital information system may contain results from the current admission, previous outpatient consultations or earlier hospitalizations. These measurements can provide useful information, but they are not necessarily interchangeable.

A result obtained during a previous episode of AKI may not represent stable baseline function. A measurement from several months earlier may be informative but may no longer accurately reflect the patient's condition immediately before the current illness.

Some electronic detection systems use the lowest creatinine measurement within a defined period as an operational baseline. This approach makes automated surveillance practical, but it can generate incorrect classifications when the selected value does not represent genuine baseline kidney function.

Systems relying only on the current admission may also miss community-acquired AKI that was already present when the patient arrived.

These limitations should be examined during local validation rather than treated as minor technical exceptions.

Estimated glomerular filtration rate (eGFR) introduces another challenge. Conventional creatinine-based eGFR equations assume that serum creatinine is relatively stable. During rapidly changing kidney function, the calculated eGFR may not accurately represent the patient's current filtration rate.

A sudden change in eGFR should therefore not be interpreted in isolation as a substitute for established AKI criteria.

Laboratory timing also matters. A specimen may be collected at one time, analyzed later and entered into the electronic record after an additional delay.

If an algorithm calculates the interval between measurements using the time results were entered instead of the relevant specimen collection times, it may classify the change incorrectly.

Different reporting units, changes in laboratory methods, duplicate results and patient-matching errors can create further risks.

These are patient-safety concerns because an algorithm may produce an apparently precise conclusion from information that is incomplete or incorrectly interpreted.

Healthcare professionals reviewing an alert should be able to identify the relevant laboratory values, comparison period, baseline assumptions and important limitations.

AI Prediction and Criteria-Based AKI Alerts Are Different

Not every automated AKI alert uses artificial intelligence.

A conventional electronic detection system may apply predefined rules to laboratory results, comparing serum creatinine changes against established KDIGO thresholds.

Such a system can provide useful clinical decision support without using machine learning.

An AI prediction model performs a different task. It may analyze several clinical variables to estimate the probability that a patient will develop AKI during a future period.

Depending on the application, these variables may include previous kidney function, laboratory trends, vital signs, medication exposure, clinical diagnoses and other information available in electronic health records.

The two approaches answer different questions.

A criteria-based detection system identifies whether available data suggest that AKI criteria have already been met. A prediction model estimates the likelihood of a future event.

A patient identified as high-risk by a prediction model may not currently have AKI. Conversely, a patient who already meets AKI criteria may receive a relatively low prediction score from a system designed to forecast future deterioration.

These results should not be combined into an undifferentiated category of AI-detected kidney injury.

Some applications also support medication review, prioritize patients for clinical assessment or track whether relevant actions have been completed. Each function creates different expectations about performance and professional oversight.

A system validated for predicting severe AKI should not automatically be presented as a diagnostic tool for established AKI. Similarly, a conventional alert that accurately reproduces creatinine-based criteria has not necessarily demonstrated that it improves clinical outcomes.

This distinction matters during procurement and implementation. Hospitals should know whether they are purchasing a rule-based surveillance system, a predictive AI model, a clinical decision-support application or a combination of these functions.

The evidence required to justify each function should reflect its intended clinical purpose.

The Status of the KDIGO 2026 AKI/AKD Guideline Update

Kidney disease guidance is evolving, and healthcare organizations need to distinguish established recommendations from proposed changes.

KDIGO released a draft of its updated Clinical Practice Guideline for Acute Kidney Injury and Acute Kidney Disease (AKI/AKD) for public review in March 2026. The consultation period ended in April.

As of November 1, 2026, the official KDIGO guideline page continued to describe the document as being prepared for publication following public consultation. A final updated clinical practice guideline had not yet been identified there as published.

The proposed update reflects important advances in understanding the continuum between acute kidney injury, acute kidney disease, kidney recovery and chronic kidney disease.

It discusses functional and structural measures of kidney injury, biomarkers, risk prediction, electronic alerts, fluid management, medication stewardship, kidney replacement therapy and follow-up after acute kidney disorders.

These developments are relevant to digital kidney care, particularly as AI systems become more capable of integrating laboratory results and other clinical information.

However, recommendations in a public consultation draft should not be presented as final clinical guidance or as having replaced the established KDIGO 2012 criteria.

For example, the draft discusses additional diagnostic approaches involving cystatin C and biomarkers of kidney damage. Their proposed inclusion does not mean that every available biomarker can independently establish AKI or that an AI system can determine its clinical significance without appropriate validation.

The intended clinical use, supporting evidence, regulatory status and applicable professional guidance remain important.

Hospitals implementing AI should therefore monitor authoritative guideline developments and assess whether changes affect their local clinical pathways.

Any modification to an algorithm's diagnostic criteria, prediction logic or recommended actions should undergo appropriate review before routine implementation.

A software update incorporating proposed guidance should not be mistaken for a validated change in clinical standards.

Clinical Assessment Must Address the Underlying Cause

Recognizing possible AKI is not enough. The clinical team must determine what is happening to the patient.

A rising creatinine concentration may occur in someone with sepsis, hypovolemia, heart failure, urinary obstruction or intrinsic kidney disease. These conditions can require substantially different responses.

A patient with genuine volume depletion may benefit from carefully assessed fluid replacement. Another patient may have venous congestion or fluid overload, making indiscriminate intravenous fluid administration harmful.

A patient with urinary obstruction may require prompt investigation and intervention to restore drainage.

An electronic alert should not automatically trigger the same treatment in each situation.

The National Institute for Health and Care Excellence (NICE) guideline NG148, Acute Kidney Injury: Prevention, Detection and Management, last updated in October 2024, emphasizes timely assessment, identification of underlying causes and attention to potentially reversible conditions.

The guidance addresses clinical examination, urine testing, investigation of urinary obstruction and circumstances requiring specialist involvement.

The urgency of care depends on the patient's condition.

Severe hyperkalemia, significant metabolic acidosis, pulmonary edema, concerning uremic complications or rapidly worsening kidney function may require urgent medical evaluation and specialist involvement.

An ordinary electronic review queue must not delay care when the patient has an immediately threatening complication.

The same principle applies when AKI develops during sepsis or another time-critical illness.

AI may help identify relevant information or bring a patient to clinical attention, but it cannot determine the underlying cause or the safest intervention from a risk score alone.

The appropriate response may involve hemodynamic assessment, review of fluid balance, medication evaluation, urine examination, relevant imaging or nephrology consultation.

These actions should be selected according to the patient's clinical condition rather than automatically imposed whenever an alert appears.

Medication Safety Requires More Than Stopping Nephrotoxic Drugs

Medication review is an important component of caring for patients with suspected or confirmed AKI.

Many hospitalized patients receive medicines that require reassessment when kidney function deteriorates.

Some medicines may contribute directly to kidney injury. Others may accumulate because renal clearance has declined, increasing the risk of toxicity even when the medicine did not cause the AKI.

Other treatments affect blood pressure, renal hemodynamics or fluid balance.

These mechanisms should not be treated as equivalent.

Nonsteroidal anti-inflammatory drugs, certain antimicrobial agents and selected other treatments may create kidney-related risks under particular clinical circumstances.

Medicines such as aminoglycosides and vancomycin require appropriate attention to indication, exposure, dosing and monitoring.

Other renally cleared medicines may require dose adjustment or changes in administration intervals as kidney function changes.

Because standard creatinine-based estimates may be unreliable during rapidly changing kidney function, dosing decisions may also require consideration of clinical trends, drug concentrations where appropriate and other patient-specific information.

However, medication safety review should not become a universal instruction to discontinue every medicine associated with kidney-related risk.

Angiotensin-converting enzyme inhibitors, angiotensin receptor blockers and diuretics may require temporary adjustment during selected acute illnesses, particularly when hypotension or volume depletion is present.

At the same time, these treatments may provide important benefits for patients with heart failure, proteinuric kidney disease or other cardiovascular conditions.

The decision to continue, adjust or temporarily withhold treatment depends on blood pressure, volume status, potassium concentration, underlying conditions and current clinical priorities.

Similarly, a patient with a serious infection may require continued antimicrobial treatment despite potential kidney-related adverse effects.

The appropriate response may involve dose optimization, therapeutic drug monitoring or selection of a suitable alternative rather than indiscriminate discontinuation.

A poorly designed AI application could cause harm if it encouraged clinicians to stop necessary treatments without considering these circumstances.

Medication stewardship should therefore distinguish avoiding unnecessary exposure, adjusting treatment to current kidney function and maintaining therapies whose benefits remain important.

Clinical pharmacists can contribute by reviewing medication indications, doses, interactions, monitoring requirements and potential alternatives. Prescribing decisions remain the responsibility of appropriately authorized healthcare professionals.

Medication changes also need to remain visible after the acute episode.

A medicine temporarily withheld during AKI should not automatically remain discontinued indefinitely. Conversely, a treatment stopped because of a serious adverse reaction may require a different long-term plan.

Reliable documentation should explain why treatment changed, whether reassessment is needed and who is responsible for the next decision.

Contrast Imaging Requires Balanced Clinical Judgment

Iodinated contrast media are another area where simplified electronic warnings can create unintended consequences.

Contrast-associated kidney concerns may be relevant when evaluating patients with AKI or substantially impaired kidney function. However, potential kidney risk should not automatically prevent a clinically necessary investigation.

NICE updated its guidance on iodinated contrast media in October 2024, including recommendations concerning kidney risk and the consequences of delaying urgent imaging.

The guidance emphasizes that emergency use of iodinated contrast media should not be delayed when the clinical risk of postponing the investigation is likely to be significant.

This distinction is important for AI-assisted clinical decision support.

A patient with suspected pulmonary embolism, major vascular injury or another life-threatening condition may need urgent contrast-enhanced imaging.

An AI system that blocks the investigation solely because the patient has an elevated AKI risk score could delay diagnosis and expose the patient to serious harm.

The appropriate approach may involve professional review, preventive measures when clinically indicated and subsequent monitoring rather than automatic cancellation.

Hospitals should also distinguish contrast-associated AKI from contrast-induced AKI.

An increase in serum creatinine after contrast administration does not necessarily establish that the contrast medium caused the kidney injury.

AI-assisted systems should support a balanced assessment of diagnostic benefit and kidney-related risk rather than replacing that assessment with rigid restrictions.

Connecting an AKI Alert to Clinical Action

A reliable AKI recognition pathway must connect important information with an identified professional responsibility.

When a system generates an alert, the organization should establish who is expected to review it, how urgency is determined and what constitutes an appropriate clinical response.

Depending on the setting, relevant professionals may include the treating physician, nursing team, clinical pharmacist, nephrologist or other specialists.

Responsibilities should reflect the patient's condition rather than assuming that every AKI alert requires identical action.

A mild creatinine increase in a clinically stable patient may prompt timely assessment and monitoring. A patient with rapidly rising creatinine, oliguria, severe electrolyte disturbances or hemodynamic instability may require immediate evaluation and escalation.

The alerting process should help distinguish these circumstances.

An electronic acknowledgment alone does not demonstrate that AKI has been appropriately assessed or managed.

The clinical record should make it possible to determine whether relevant information was reviewed, whether the patient was assessed, what decisions were made and whether necessary actions were implemented.

These actions may include reassessing volume status, investigating possible causes, reviewing medicines, repeating clinically indicated measurements or obtaining specialist advice.

Their appropriateness matters as much as their timeliness.

A system that increases unnecessary laboratory testing, inappropriate fluid administration or avoidable specialist referrals may add workload and expose patients to interventions that offer little benefit.

Hospitals should therefore evaluate whether alerts produce meaningful clinical review rather than merely increasing the number of recorded actions.

What the 2026 Evidence Shows About AI-Supported AKI Care

Recent clinical research demonstrates why identifying patients at risk and improving patient outcomes must be evaluated separately.

A randomized clinical trial by Churpek and colleagues, published in JAMA Network Open on July 10, 2026, investigated whether structured early nephrology consultation triggered by a machine-learning AKI risk score could improve outcomes among hospitalized patients at increased risk.

The trial enrolled 180 patients who did not have serum creatinine-defined AKI at enrollment.

Eighty-nine patients were assigned to early nephrology consultation, while 91 received usual care.

The intervention involved an in-person nephrology assessment and recommendations concerning volume status, kidney perfusion, medication selection and dosing, electrolyte disturbances, nutrition and additional testing.

The primary outcome was the peak change in serum creatinine during the seven days following enrollment.

Early consultation did not produce a statistically significant improvement in the primary outcome compared with usual care.

The investigators also found no statistically significant differences in major secondary outcomes, including AKI development, kidney replacement therapy and mortality.

An additional finding was that consultation recommendations were less likely to be followed in the early consultation group than in the usual-care group.

This matters because even appropriate recommendations may have limited effect when they are not consistently implemented.

The trial was relatively small and evaluated a particular intervention within a defined patient population. Its findings should not be interpreted as proof that all AI-based AKI prediction is ineffective or that early nephrology involvement has no value in other circumstances.

Nevertheless, the study provides an important lesson for healthcare organizations.

Identifying patients at increased risk does not guarantee better outcomes. The benefit depends on whether the subsequent intervention addresses modifiable risks, reaches the responsible healthcare professionals and results in appropriate action.

An accompanying editorial by Thongprayoon and colleagues emphasized the need to translate risk prediction into targeted and reliable care.

For hospitals considering AI procurement, this evidence argues against treating predictive accuracy alone as proof of clinical effectiveness.

Lessons From Electronic AKI Alert Studies

Studies of conventional electronic AKI alerts reinforce the need to evaluate what happens after a warning is generated.

A randomized clinical trial by Li and colleagues, published in JAMA Network Open in January 2024, evaluated electronic AKI alerts combined with a care bundle among 2,208 hospitalized adults in China.

The intervention changed several aspects of clinical management.

Patients receiving alerts were more likely to receive intravenous fluids, less likely to be exposed to nonsteroidal anti-inflammatory drugs and more likely to have AKI documented at discharge.

However, these changes did not produce statistically significant improvements in the primary kidney function outcome or the patient-centered secondary outcomes assessed.

The findings demonstrate that increased clinical activity does not necessarily mean better care.

A separate study by Su and Hsu, published in the Journal of Medical Systems in June 2026, evaluated an automated AKI alert system integrated with medication-focused clinical decision support at a large tertiary hospital in Taiwan.

The system used KDIGO serum creatinine criteria to identify hospital-acquired stage 2–3 AKI.

Alerts were generated through scheduled processing four times daily and accompanied by medication-related guidance. The system did not automatically place treatment orders.

The investigators reported sensitivity of 90.94% and accuracy of 99.65% when comparing the deployed system with a retrospective computerized reference algorithm designed to reproduce its detection logic.

These findings support the technical feasibility of operating an electronic AKI detection system within hospital information infrastructure.

However, the reference process was based on computerized criteria rather than independent clinical adjudication of every case.

The reported results should therefore not be interpreted as proof that the system correctly diagnosed all clinically relevant AKI episodes or improved kidney outcomes.

The application was also primarily a criteria-based alert system integrated with medication decision support, rather than an independent machine-learning prediction model.

Together, these studies suggest that electronic systems can support recognition and influence clinical behavior, but the clinical value of implementation depends on whether appropriate care follows.

This distinction is particularly important when healthcare organizations evaluate commercial claims about AI-based AKI detection.

Kidney Recovery May Remain Incomplete at Discharge

A clinically important gap in AKI care can occur after the immediate hospitalization ends.

Serum creatinine may begin to improve before discharge, leading the clinical team to consider the acute kidney problem resolved.

However, improvement does not necessarily mean that kidney function has returned to its previous baseline or that longer-term risks have disappeared.

Some patients experience persistent kidney dysfunction. Others develop recurrent AKI or are subsequently diagnosed with chronic kidney disease.

The KDIGO 2012 guideline recommends evaluating patients three months after AKI to assess resolution, development of CKD or worsening of pre-existing kidney disease.

NICE Quality Standard QS76 also recommends clinical review within three months after hospital discharge, or sooner for patients at increased risk of poor outcomes.

The appropriate timing depends on individual circumstances.

A patient who has experienced mild AKI with apparently complete recovery may need a different follow-up approach from someone with severe AKI, incomplete recovery, underlying CKD or significant cardiovascular disease.

Some patients require substantially earlier reassessment.

Follow-up may include serum creatinine measurement, evaluation of kidney function, urinary albumin or protein testing where appropriate, blood pressure assessment and review of medication changes.

Medication reassessment is particularly important.

Treatments temporarily withheld during acute illness may need to be restarted, adjusted or replaced when the patient's condition becomes stable. Failure to reconsider these decisions may leave patients without medicines that provide important cardiovascular or kidney benefits.

A review by Selby and colleagues, published in the Clinical Kidney Journal in March 2026, emphasizes practical post-AKI care, including clear discharge communication, risk-based follow-up, urinary albumin testing, medication optimization and coordination between healthcare services.

The review also highlights cardiovascular risk management, patient education and the broader physical and psychological consequences that may follow serious illness.

The authors recognize that evidence remains incomplete regarding which follow-up interventions improve outcomes for particular patient groups.

This uncertainty supports individualized care rather than applying an identical follow-up schedule to every patient.

Patients and caregivers should understand that AKI occurred, why further assessment may be needed and who is responsible for arranging it.

A discharge summary stating only that kidney function has improved may not provide enough information for safe continuing care.

The receiving clinician may need to know the likely cause of AKI, its severity, kidney function at discharge, important medication decisions and the recommended timeframe for reassessment.

Evidence of Gaps in Post-AKI Care

A population-based cohort study by Mok and colleagues, published in BMC Medicine in August 2026, provides important evidence about what happens after AKI-related hospitalization.

The researchers examined linked hospital and primary care records involving 209,222 adults in England, representing 279,187 inpatient AKI episodes.

They assessed adherence to 14 indicators of recommended post-AKI care, covering documentation, clinical contact, kidney monitoring and medication management.

Only 19.5% of episodes had an AKI diagnosis recorded in primary care within 30 days of discharge.

Within the study's assessment window of 90 ± 30 days after discharge, serum creatinine was measured after 34.2% of episodes, while urine albumin-to-creatinine ratio testing was recorded after only 4.2%.

The researchers also identified variation in follow-up according to patient characteristics and opportunities to improve medication management.

These findings illustrate how an AKI episode may become insufficiently documented or incompletely followed after a patient leaves the hospital.

However, the study was observational and based on healthcare records from England.

It does not establish corresponding rates in Viet Nam or prove that every missing investigation represented inappropriate care.

The clinical need for a particular test depends on the patient's circumstances and the purpose of follow-up. Recorded information may also be affected by the completeness of data capture.

The broader lesson concerns continuity of care.

Hospitals need reliable arrangements for communicating the AKI episode, identifying follow-up needs and establishing who is responsible for reassessment.

AI-assisted tracking may help identify patients with outstanding follow-up tasks, but sending a reminder does not demonstrate that care has been completed.

A useful system should distinguish between recommendations awaiting action, investigations that have been arranged, completed assessments and cases requiring further review.

Where care is transferred to another healthcare organization, the follow-up process should also consider whether relevant information and responsibility have been appropriately handed over.

Understanding the Transition From AKI to AKD and CKD

The period following AKI requires attention to how kidney function evolves over time.

Some patients recover rapidly, while others experience abnormalities that persist beyond the initial episode.

Acute kidney disease (AKD) describes acute or subacute kidney abnormalities occurring over a period of less than three months, including abnormalities that may persist after AKI.

Chronic kidney disease (CKD), by contrast, involves abnormalities of kidney structure or function that persist for at least three months and have implications for health.

These concepts are related but should not be treated as interchangeable.

A patient with persistent kidney dysfunction two weeks after AKI should not automatically receive a new CKD diagnosis solely because recovery remains incomplete.

At the same time, persistent abnormalities warrant appropriate evaluation and monitoring, particularly when kidney function is not improving as expected.

Patients with established CKD may also experience superimposed AKI, making it necessary to distinguish pre-existing abnormalities from new changes.

The KDIGO 2026 consultation draft places greater emphasis on the continuum linking AKI, AKD and CKD.

For healthcare organizations, the practical implication is that kidney care should not be managed as a series of disconnected encounters.

Relevant information about the acute event, subsequent recovery and remaining clinical risks should remain available to the professionals responsible for continuing care.

AI may support longitudinal tracking, but interpretation of kidney function and decisions about chronic diagnoses remain professional clinical responsibilities.

This also distinguishes post-AKI follow-up from routine CKD case finding.

CKD case finding seeks to identify people who may have an unrecognized chronic condition.

Post-AKI care begins with a known or suspected acute event and focuses on recovery, recurrent risk, medication reassessment and possible development or worsening of persistent kidney disease.

The associated clinical questions, responsibilities and timeframes are different.

Measuring Whether AI-Assisted AKI Recognition Improves Care

The effectiveness of AI-assisted AKI recognition should not be assessed solely by the number of alerts generated.

A meaningful evaluation begins with the reliability of case identification.

For criteria-based alerts, hospitals should examine how accurately the system recognizes relevant creatinine or urine output changes using an appropriate reference process.

This includes reviewing baseline selection, laboratory timing, missing information and potentially incorrect classifications.

For predictive AI, evaluation should address discrimination, calibration, clinically relevant thresholds and performance across the intended patient population.

However, technical performance is only the first part of the assessment.

Hospitals should also examine whether important alerts receive timely professional review, whether AKI criteria are appropriately verified and whether relevant causes are investigated.

Medication-related measures may include appropriate review of kidney-related risks, correction of unsafe dosing and documentation of decisions to continue, adjust, withhold or resume treatment.

An alert acknowledged electronically should not automatically count as a completed clinical intervention.

Outcome monitoring may include AKI progression, severe AKI, kidney replacement therapy, electrolyte or fluid complications, length of stay and mortality.

These outcomes require careful interpretation because underlying illness severity and patient characteristics strongly influence them.

Post-discharge measures are equally important.

Hospitals can examine whether patients have documented recovery plans, whether recommended follow-up occurs within clinically appropriate timeframes and whether persistent kidney dysfunction or unresolved medication issues receive further attention.

Balancing measures should also be included.

Excessive alerts, unnecessary laboratory testing, inappropriate fluid administration, unjustified discontinuation of essential medicines and delays in urgent imaging may represent unintended harms.

An increase in documented AKI after implementing electronic surveillance may reflect better recognition rather than a true increase in kidney injury.

Conversely, fewer recorded cases may indicate underdetection if monitoring practices or data quality have changed.

Evaluation should therefore examine the quality of the underlying information and the complete clinical process, not merely changes in reported AKI rates.

The central question is whether the intervention improves care compared with the organization's previous practice.

Local Validation in Vietnamese Healthcare Settings

AI systems developed in other healthcare environments may not perform reliably when introduced into Vietnamese hospitals.

Patient populations, clinical documentation, laboratory reporting practices and information-system integration can differ substantially.

In some facilities, serum creatinine results may be available electronically while urine output is recorded separately in nursing documentation.

Medication orders may be structured, but information about actual administration may be incomplete.

Previous kidney function results may be held by another healthcare organization and therefore unavailable to the local system.

These gaps affect both predictive models and criteria-based recognition.

A model developed using comprehensive longitudinal records may perform differently when historical information is incomplete.

Before implementation, hospitals should define the intended population, clinical function and information required for reliable operation.

If a proposed application concerns adult medical and surgical inpatients, local validation should include representative patients from those settings.

The system should not automatically be extended to pediatric, neonatal, transplant or specialized critical care populations without appropriate evidence.

Hospitals should also examine clinically important exceptions, including patients with known CKD, incomplete baseline information, recent kidney replacement therapy and rapidly changing fluid balance.

Cases not identified by the system must be included in the evaluation.

Reviewing only positive alerts may conceal important false-negative results and create a misleading impression of safety.

Workflow reliability also requires assessment.

An accurate alert may have little practical value if it reaches the wrong professional, arrives too late or does not lead to appropriate clinical assessment.

A prediction system that creates more work than the clinical team can reasonably manage may reduce attention to other important patient needs.

The organization should therefore assess alert volume, staffing capacity, professional responsibilities and access to downstream services before routine deployment.

Substantial changes in laboratory systems, clinical documentation, software versions or data integration may also affect performance and require further evaluation.

Local validation should demonstrate that the particular system performs acceptably in its intended setting. It should not be treated as proof that similar AI technologies are universally safe or effective.

Vietnamese Professional and Legal Considerations

AI-assisted AKI recognition must be implemented within Viet Nam's applicable framework for medical examination and treatment, professional clinical practice, artificial intelligence, medical devices and personal data protection.

These frameworks address different responsibilities. Compliance with one does not automatically establish compliance with another or demonstrate that a particular clinical application is safe.

Professional Clinical Requirements

The Law on Medical Examination and Treatment No. 15/2023/QH15, effective January 1, 2024, as subsequently amended, provides the broader legal foundation for professional healthcare practice, patient rights, medical records and healthcare organizational responsibilities.

For kidney-related clinical guidance, Ministry of Health Decision No. 3931/QĐ-BYT, issued on September 21, 2015, introduced the professional document Guidelines for Diagnosis and Treatment of Certain Kidney and Urological Diseases.

The document includes sections addressing acute kidney injury and acute renal failure.

Ministry of Health Decision No. 2388/QĐ-BYT, issued on August 12, 2024, subsequently introduced updated guidance on chronic kidney disease and certain kidney conditions.

Article 3 of Decision No. 2388 expressly repealed four sections of the 2015 guidance concerning diabetic kidney disease, IgA nephropathy, lupus nephritis and chronic kidney disease.

It did not expressly repeal the 2015 sections addressing acute kidney injury and acute renal failure.

Nevertheless, the remaining 2015 material should not automatically be treated as reflecting every development in contemporary AKI care.

Healthcare organizations should consider its continuing applicability alongside current evidence, professional requirements and locally approved clinical protocols.

International guidance from KDIGO and NICE can inform clinical practice and quality improvement, but it does not automatically become binding Vietnamese law.

Artificial Intelligence Regulation

Viet Nam's Law on Artificial Intelligence No. 134/2025/QH15 took effect on March 1, 2026.

Decree No. 142/2026/NĐ-CP, effective May 1, 2026, provides implementing provisions, including those concerning AI risk classification and responsibilities associated with AI systems.

Decision No. 33/2026/QĐ-TTg, effective August 15, 2026, establishes the List of High-Risk Artificial Intelligence Systems.

An application used to predict AKI or support its recognition should not automatically be assigned a particular regulatory classification merely because it operates in healthcare.

The assessment depends on the product's intended purpose, actual functions, degree of automation, deployment arrangements and applicable legal criteria.

A system used to retrieve laboratory information for professional review may raise different concerns from software designed to make clinical predictions or recommend treatment decisions.

The absence of an application's name from a particular high-risk list should not be interpreted as proof that every product performing that function is legally low-risk or exempt from applicable requirements.

Clinical risk also requires separate assessment.

A system that misses severe AKI, generates misleading risk classifications or contributes to inappropriate medication changes may pose substantial patient-safety risks regardless of its statutory classification.

Healthcare organizations should therefore examine both the applicable regulatory obligations and the practical clinical consequences of the intended use.

Medical Device Requirements

Some healthcare AI applications may fall within Viet Nam's medical device regulatory framework, depending on their intended medical purpose and functions.

Decree No. 98/2021/NĐ-CP on medical device management, as amended, provides part of the applicable framework.

Ministry of Health Circular No. 24/2026/TT-BYT, effective July 1, 2026, addresses risk determination and management measures for products classified as medical devices and amends relevant implementing provisions.

The use of AI does not automatically determine whether software is a medical device.

Conversely, describing a product as an administrative or workflow application does not necessarily exclude medical device requirements if its intended functions have a medical purpose.

Product classification and applicable obligations should be assessed for the particular technology and deployment context.

Authorization or certification obtained in another jurisdiction may provide relevant evidence but does not independently establish compliance with Vietnamese requirements.

Protecting Patient Data and Preserving Accountability

AI-assisted kidney care may involve identifiable laboratory results, clinical diagnoses, medication histories, physiological measurements and other health information.

These data require appropriate protection.

The Law on Personal Data Protection No. 91/2025/QH15 and Decree No. 356/2025/NĐ-CP, both effective January 1, 2026, provide relevant requirements concerning personal data processing and protection.

Healthcare organizations should establish the applicable lawful basis and purposes for processing, including relevant consent requirements or legally applicable exceptions.

Access should be restricted to authorized personnel, with appropriate information-security controls, retention arrangements and responsibilities for external technology providers.

Where a supplier processes identifiable patient information, the healthcare organization should examine data hosting, access permissions, system maintenance, secondary use and any applicable cross-border processing requirements.

Information collected for clinical care should not automatically be reused to train AI models or support unrelated commercial development without assessing the relevant legal basis and safeguards.

These considerations should be addressed before identifiable clinical information is transferred to an external AI service.

Professional accountability also requires reliable records of clinically important decisions.

This does not mean that every technical calculation needs to be reproduced in the patient's medical record.

Rather, the organization should preserve sufficient information to understand when an AI-generated signal materially contributed to clinical review or decision-making, consistent with its approved procedures.

AI does not independently transfer professional responsibility for AKI assessment, medication decisions, treatment or follow-up from authorized healthcare professionals to a technology provider.

Technology providers and healthcare organizations may also have their own applicable legal and contractual obligations.

The role of AI is to support clinical care within an accountable organizational process.

A Practical Starting Point for Vietnamese Hospitals

A hospital considering AI-assisted AKI recognition should begin by examining its existing kidney safety process.

A focused review may include patients who developed AKI during hospitalization and those who arrived with possible community-acquired AKI.

The organization can examine how cases were identified, whether appropriate baseline information was available, whether likely causes were assessed and whether medication-related risks were reviewed.

The review should also consider what happened after discharge.

Was the AKI episode clearly documented? Did the discharge summary explain relevant changes in kidney function and medication treatment? Was follow-up assessment arranged when appropriate?

These questions may identify weaknesses that do not require AI.

One hospital may already have reliable laboratory information but lack a consistent process for assigning responsibility after an abnormal result.

Another may identify AKI appropriately during admission but fail to communicate recovery needs to outpatient services.

A third may experience delays in medication review because laboratory and prescribing systems are not adequately integrated.

Each problem requires a different response.

If AI is expected to add value, the hospital should select a limited and clinically meaningful use case.

One option is evaluating a prediction model that identifies adult inpatients at increased risk of AKI before creatinine-defined injury develops.

Another is testing whether a criteria-based electronic system reliably recognizes patients who already meet accepted AKI definitions and supports timely professional review.

These functions should be evaluated separately.

An initial silent evaluation may allow the system to generate predictions or alerts without influencing live clinical decisions.

This provides an opportunity to assess data quality, missed cases, unnecessary alerts and technical reliability while established care continues.

If the results are acceptable, a supervised clinical pilot may be considered.

The pilot should define who reviews alerts, how urgent findings are escalated, what assessment is expected and how decisions are documented.

It should also preserve the hospital's ability to recognize and manage AKI when the technology is unavailable.

Treating physicians, nephrologists, clinical pharmacists, nurses, laboratory medicine specialists, information technology professionals and quality management teams may all contribute to implementation.

The organization should monitor whether the system improves clinically appropriate actions and whether additional workload or unintended harms emerge.

Results should be reviewed through established clinical quality and patient-safety governance processes, with authority to modify, restrict, suspend or discontinue the intervention when necessary.

Expansion to other departments or healthcare facilities should depend on demonstrated clinical reliability and practical benefit rather than technical performance alone.

The most useful implementation is one that strengthens kidney care across the patient pathway, including recovery and follow-up after discharge.

From AKI Recognition to Safer Long-Term Kidney Care

Acute kidney injury illustrates both the potential and the limitations of AI in healthcare.

AI may help identify patients who are likely to develop kidney injury or recognize changes in clinical information that require timely attention. Electronic systems may also support medication review, kidney function monitoring and follow-up after hospitalization.

However, recognition alone does not establish the cause of AKI, determine the safest intervention or ensure recovery.

Clinical research published in 2026 demonstrates that even when a machine-learning model identifies a high-risk population, the intervention that follows may not produce measurable improvement in patient outcomes.

For healthcare organizations in Viet Nam, responsible implementation should therefore begin with reliable clinical criteria, accurate data, professional assessment and appropriate treatment of underlying causes.

Medication review should balance kidney-related risks against the benefits of necessary treatment. Follow-up after hospitalization should recognize that improving serum creatinine does not always mean kidney recovery is complete.

Hospitals should also evaluate whether AI provides meaningful benefit beyond established clinical processes and simpler electronic decision-support tools.

A well-governed system should help healthcare professionals recognize important changes, respond appropriately and maintain continuity of care after the acute episode.

Its success should not be defined by the number of AKI alerts generated or the predictive accuracy of a model alone.

The more important measure is whether patients receive timely assessment, potentially avoidable harm is reduced, necessary treatments are managed safely and recovery is followed through to an appropriate clinical conclusion.

AI can support the recognition of kidney injury. Protecting patients depends on the quality of care that follows.

Selected References and Regulatory Sources

Vietnamese Clinical Guidance

Ministry of Health of Viet Nam. Decision No. 3931/QĐ-BYT, dated September 21, 2015, issuing the professional guidance Diagnosis and Treatment of Certain Kidney and Urological Diseases (Hướng dẫn chẩn đoán và điều trị một số bệnh về thận – tiết niệu).

Ministry of Health of Viet Nam. Decision No. 2388/QĐ-BYT, dated August 12, 2024, issuing the professional guidance Diagnosis and Treatment of Chronic Kidney Disease and Certain Kidney Diseases (Hướng dẫn chẩn đoán và điều trị bệnh thận mạn và một số bệnh lý thận).

International Clinical Guidelines and Quality Standards

Kidney Disease: Improving Global Outcomes (KDIGO) Acute Kidney Injury Work Group. KDIGO Clinical Practice Guideline for Acute Kidney Injury. Kidney International Supplements. 2012;2(1):1–138. DOI: 10.1038/kisup.2012.1.

Kidney Disease: Improving Global Outcomes. KDIGO 2026 Clinical Practice Guideline for Acute Kidney Injury and Acute Kidney Disease. Public review draft released March 2026. Final guideline publication pending as of November 1, 2026.

National Institute for Health and Care Excellence. Acute Kidney Injury: Prevention, Detection and Management. NICE Guideline NG148. Published December 18, 2019; last updated October 16, 2024.

National Institute for Health and Care Excellence. Acute Kidney Injury. NICE Quality Standard QS76. Quality Statement 6: Clinical Review After Hospital Discharge. Updated March 23, 2023.

International Society of Nephrology. Acute Kidney Injury Toolkit. Clinical assessment, medication safety and post-AKI follow-up resources.

AI, Electronic Alerts and Clinical Implementation Evidence

Churpek MM, Fatima A, Anjorin O, et al. Early Nephrology Consultation and Acute Kidney Injury in Hospitalized Patients: A Randomized Clinical Trial. JAMA Network Open. 2026;9(7). DOI: 10.1001/jamanetworkopen.2026.22554.

Thongprayoon C, Pesce F, Cheungpasitporn W. AI-Enabled Acute Kidney Injury Prediction and the Challenge of Prevention. JAMA Network Open. 2026;9(7). DOI: 10.1001/jamanetworkopen.2026.22565.

Su CH, Hsu CN. Development and Validation of an Automated Acute Kidney Injury E-Alert System Integrated with Clinical Decision Support for Hospitalized Patients. Journal of Medical Systems. 2026;50:102. DOI: 10.1007/s10916-026-02428-8.

Li T, Wu B, Li L, et al. Automated Electronic Alert for the Care and Outcomes of Adults With Acute Kidney Injury: A Randomized Clinical Trial. JAMA Network Open. 2024;7(1). DOI: 10.1001/jamanetworkopen.2023.51710.

Post-AKI Recovery and Continuity of Care

Mok PLH, Sawhney S, Blakeman T, et al. Examining Variations in Adherence to Guideline-Recommended Post-Acute Kidney Injury Care in Primary Care in England: A Population-Based Cohort Study. BMC Medicine. 2026;24:406. DOI: 10.1186/s12916-026-05042-0.

Selby NM, De Seigneux S, Saritas T, et al. Top Ten Tips to Manage Patients After Acute Kidney Injury. Clinical Kidney Journal. 2026;19(4). DOI: 10.1093/ckj/sfag104.

Relevant Vietnamese Legal and Regulatory Framework

National Assembly of Viet Nam. Law on Medical Examination and Treatment No. 15/2023/QH15, effective January 1, 2024, as subsequently amended.

National Assembly of Viet Nam. Law on Artificial Intelligence No. 134/2025/QH15, effective March 1, 2026.

Government of Viet Nam. Decree No. 142/2026/NĐ-CP, detailing certain provisions and measures for implementing the Law on Artificial Intelligence, effective May 1, 2026.

Prime Minister of Viet Nam. Decision No. 33/2026/QĐ-TTg, promulgating the List of High-Risk Artificial Intelligence Systems, effective August 15, 2026.

Government of Viet Nam. Decree No. 98/2021/NĐ-CP on Medical Device Management, as amended.

Ministry of Health of Viet Nam. Circular No. 24/2026/TT-BYT on Risk Determination and Management Measures for Medical Devices and Amendments to Related Implementing Provisions, effective July 1, 2026.

National Assembly of Viet Nam. Law on Personal Data Protection No. 91/2025/QH15, effective January 1, 2026.

Government of Viet Nam. Decree No. 356/2025/NĐ-CP, detailing certain provisions and measures for implementing the Law on Personal Data Protection, effective January 1, 2026.

Editorial and Clinical Notice

This article is intended for professional education, responsible healthcare AI governance and healthcare quality-improvement discussion. The opening clinical scenario is illustrative and does not describe an identified patient or documented clinical incident.

The article summarizes selected Vietnamese professional guidance, international clinical recommendations, published research and relevant legal and regulatory considerations. It also describes proposed organizational approaches for AKI recognition, clinical assessment, medication safety, post-discharge follow-up and performance monitoring.

These approaches do not constitute a mandatory Vietnamese AI implementation standard, an independently validated AKI management protocol or a legal determination of any particular AI product's regulatory classification.

The article does not provide patient-specific medical advice, authorize automated medication changes or replace professional clinical judgment.

AKI assessment, diagnostic confirmation, investigation of underlying causes, medication decisions, treatment, specialist referral and follow-up must remain under the responsibility of appropriately qualified healthcare professionals, consistent with applicable Vietnamese requirements, current clinical guidance and locally approved procedures.

AI predictions and electronic alerts must not delay urgent assessment or necessary treatment. Clinical decisions should not rely solely on a risk score, an automated laboratory comparison or an unverified estimate of kidney function.

The research discussed does not establish that AI-assisted AKI prediction or recognition necessarily reduces severe kidney injury, kidney replacement therapy, mortality or long-term kidney complications in Vietnamese healthcare settings. Local validation, meaningful professional oversight, supervised implementation and ongoing evaluation remain necessary before claims of clinical effectiveness can be made.

The KDIGO 2026 AKI/AKD guideline remained in the publication preparation stage at the date of this review. Proposals contained in the consultation draft should be distinguished from final published clinical guidance.

Digital Medicine Vietnam (DMVN) | Evidence and regulatory review: November 1, 2026.