Digital Clinical Decision Support for Malnutrition Risk

Malnutrition remains a significant challenge across healthcare systems, affecting recovery, physical function, immunity, and overall quality of life. It can occur when people consume insufficient nutrients, absorb them poorly, or experience increased nutritional requirements during illness. Nutritional problems may also involve excessive or imbalanced intake, although this article focuses primarily on undernutrition and disease-related malnutrition.

Early identification is essential because nutritional deterioration can develop before its consequences become obvious. Patients may experience unintended weight loss, reduced appetite, declining muscle strength, or difficulty completing everyday activities. These signs can be overlooked when healthcare teams concentrate on the primary illness.

Digital clinical decision support systems offer a way to improve this process. They combine patient information with screening rules, clinical guidelines, and predictive models. The resulting alerts can help healthcare professionals identify patients who require further nutritional assessment.

These systems can be particularly valuable when patients move between hospitals, outpatient clinics, community services, and long-term care facilities. Each setting collects different information and faces different practical constraints. A well-designed system can connect these sources while supporting decisions appropriate to each environment.

However, digital screening cannot replace professional assessment. Its purpose is to make important information more visible, improve the timing of referrals, and support consistent nutritional care.

Understanding Digital Clinical Decision Support for Malnutrition

From Patient Records to Clinical Recommendations

Clinical decision support systems use patient information and clinical knowledge to assist healthcare professionals during decision-making. In malnutrition care, these systems may examine weight changes, appetite, food intake, body mass index, medical diagnoses, and relevant laboratory results.

A basic system can apply established screening criteria to calculate a nutritional risk score. More advanced systems may integrate information from electronic health records and use machine learning to estimate the likelihood of malnutrition. Some can also recommend follow-up assessments or notify dietitians when predefined thresholds are reached.

For example, a patient admitted with pneumonia may have a history of unintended weight loss and reduced food intake. If these details appear in separate sections of the medical record, busy clinicians may not immediately recognise their combined significance. A decision support system can bring the information together and flag the need for nutritional screening.

The alert should explain the concern and recommend an appropriate next step. It should not automatically establish a diagnosis based solely on incomplete records.

Screening Is Different From Diagnosis

Nutritional screening identifies people who may be at risk and need further assessment. Diagnosis requires a more detailed evaluation of nutritional status, clinical history, physical findings, and relevant criteria.

The Global Leadership Initiative on Malnutrition framework provides an internationally recognised approach to diagnosing malnutrition. Following initial screening, the framework uses phenotypic criteria, such as weight loss, low body mass index, or reduced muscle mass, alongside etiologic criteria, such as reduced intake or disease-related inflammation. The GLIM framework and its clinical application support a more consistent approach across populations and healthcare settings.

Digital systems should preserve this distinction. A positive screening result should trigger assessment rather than an automatic diagnosis. Likewise, a low predicted risk should not override concerning clinical findings. This separation helps maintain professional accountability and reduces the danger of treating algorithmic predictions as definitive medical conclusions.

Why Early Nutritional Risk Identification Matters

Malnutrition Can Remain Hidden During Routine Care

Patients do not always appear visibly underweight when nutritional problems begin. Some may have a body mass index within a typical range while experiencing recent weight loss, reduced muscle mass, or inadequate food intake. Fluid retention can also obscure changes in body weight.

Clinical attention may focus on infections, surgery, cancer treatment, respiratory disease, or other immediate concerns. Nutritional risk can therefore remain undocumented unless screening is integrated into routine workflows. Digital decision support can help by identifying combinations of risk indicators that deserve attention. It can also remind healthcare teams to repeat screening when a patient’s condition changes or when an extended hospital stay increases the need for reassessment.

The 2024 European Society for Clinical Nutrition and Metabolism guidance recommends nutritional risk screening for polymorbid medical inpatients, followed by further assessment and individualised nutritional support when risk is detected. The practical guideline emphasises timely identification, treatment planning, and continued monitoring.

Earlier Identification Can Improve Continuity

Nutritional care frequently involves several professionals and multiple care settings. A patient may receive initial treatment in a hospital, attend outpatient appointments, and later recover at home. If nutritional concerns are not communicated between services, follow-up may become inconsistent.

Digital systems can preserve screening results, assessment findings, intervention plans, and review dates. This information can help the receiving team understand previous concerns and determine whether the patient’s nutritional condition has improved.

Continuity is especially important for people with chronic illnesses, older adults, patients recovering from major surgery, and individuals who experience repeated hospital admissions. Their nutritional requirements may change over time, making repeated assessment more useful than a single screening result.

Identifying Nutritional Risk Across Different Care Settings

Hospitals and Acute Care Units

Hospitals provide opportunities for systematic screening because patients generally undergo admission procedures and repeated clinical assessments. Electronic health records can incorporate nutritional screening into admission workflows, making it easier to identify risk early.

Decision support systems may review recent weight measurements, appetite changes, food intake, diagnoses, and documented symptoms. They can also identify patients whose nutritional status requires reassessment after surgery, intensive care, or prolonged illness.

In 2025, a study published in Clinical Nutrition ESPEN evaluated machine learning models that used electronic medical records to identify malnutrition risk among hospital inpatients. The researchers compared several algorithms and examined how structured records and information extracted from clinical text could support screening. Their findings suggested that automated approaches could assist early identification, although performance differed according to the data source and model.

In acute care, the system should prioritise timely alerts without overwhelming clinicians. A high-risk result should support referral for a comprehensive assessment, while repeated alerts for an unchanged condition should be managed carefully.

Outpatient Clinics and Primary Care

Nutritional risk can develop gradually outside hospitals. People with diabetes, kidney disease, gastrointestinal disorders, cancer, or other chronic conditions may experience changes in appetite, food tolerance, or nutrient requirements over time.

Primary care systems can monitor weight trajectories, medication changes, recurring symptoms, and previous nutritional assessments. If a patient’s weight declines across several appointments, the system could flag the trend even when individual measurements appear unremarkable.

This longitudinal approach may help clinicians recognise emerging nutritional problems before they become severe. It can also support referrals to dietitians and follow-up after nutritional interventions.

However, outpatient records are often incomplete. Patients may visit different providers, obtain measurements at home, or receive care from facilities that do not share compatible records. Decision support systems must therefore communicate uncertainty when information is missing and avoid interpreting absent documentation as proof that no risk exists.

Community Healthcare and Home-Based Care

Community healthcare services often work with people who face barriers to regular medical appointments. Older adults living alone, individuals with limited mobility, and people experiencing food insecurity may have difficulty accessing consistent nutritional support.

Community health workers can use mobile applications to record weight changes, appetite, food access, swallowing difficulties, and other relevant concerns. A digital system can identify warning signs and recommend referral to a qualified healthcare professional.

Home-based monitoring may also support patients after hospital discharge. With appropriate consent, patients or caregivers can record selected measurements and symptoms. Persistent deterioration can trigger a review rather than waiting until the next scheduled appointment.

These tools must remain practical for the people expected to use them. Complex forms, unreliable internet access, limited digital literacy, and language barriers can reduce participation. Systems designed for community settings should support simple interfaces, offline data entry where feasible, and accessible communication.

Long-Term Care Facilities

Residents of nursing homes and other long-term care facilities may face multiple nutritional risks. These can include reduced appetite, swallowing difficulties, cognitive impairment, medication-related symptoms, chronic illness, and dependence on others for meals.

Digital decision support can combine weight trends, meal consumption records, swallowing assessments, medication information, and changes in functional ability. It can help staff recognise persistent reductions in food intake or repeated weight loss. The system may also support regular reviews after changes in medication, acute illness, or deterioration in mobility. Alerts can direct staff toward assessment rather than relying on memory or informal handovers.

Nevertheless, long-term care decisions require individual judgement. A resident’s food intake may fluctuate for temporary reasons, while some interventions may conflict with swallowing safety, treatment goals, or personal preferences. Digital recommendations must support multidisciplinary discussion and respect the resident’s needs.

Paediatric and Maternal Care

Nutritional risk assessment requires particular care during pregnancy, infancy, and childhood because nutritional needs change with growth and development. Adult thresholds cannot simply be applied to children.

Paediatric systems may need age-appropriate growth measurements, growth trajectories, feeding histories, developmental information, and relevant medical conditions. In maternal care, decision support may incorporate gestational stage, maternal nutritional status, dietary intake, and other clinically relevant factors.

Digital tools can help identify concerning changes and prompt appropriate follow-up. However, growth measurements require correct interpretation, and results may be affected by measurement errors or incomplete histories.

Systems should use validated criteria for the intended population. Their performance must also be evaluated separately across relevant age groups rather than assumed to generalise from adult hospital data.

How Artificial Intelligence Strengthens Nutritional Decision Support

Combining Information From Multiple Sources

Traditional screening often depends on a limited set of questions and measurements. Machine learning can potentially combine a broader range of clinical information, including previous diagnoses, weight trajectories, food intake, medication records, and relevant laboratory results.

This approach may identify risk patterns that are difficult to recognise when each variable is considered separately. For instance, moderate weight loss combined with reduced intake and a serious underlying illness may warrant more attention than any single factor suggests. Natural language processing can also extract information from clinical notes. A clinician may document poor appetite or difficulty swallowing in free text rather than in a structured field. Automated text analysis could help surface these details for review.

However, laboratory results must be interpreted carefully. Measures such as albumin can be influenced by inflammation, illness severity, hydration, and other factors. They should not be treated as standalone proof of malnutrition. AI models should therefore use clinically justified features, be evaluated against appropriate assessment standards, and provide understandable reasons for their alerts whenever possible.

Predicting Deterioration Rather Than Only Detecting Existing Risk

Many screening systems identify patients who already show signs of nutritional deterioration. Predictive models could potentially extend this approach by estimating which patients are likely to develop nutritional problems.

For example, a patient with a chronic condition may currently maintain a stable weight but have repeated admissions, declining appetite, and increasing treatment burden. A predictive system could identify the combination as a reason for closer monitoring.

Such predictions may help healthcare teams intervene before substantial deterioration occurs. Nevertheless, future risk estimates are only useful when they are sufficiently accurate and linked to practical clinical actions. Models developed in hospital populations may not perform reliably in primary care or community settings. They require external validation in the populations and workflows where they will be used.

Supporting Individualised Care Plans

Once nutritional risk is identified, the next step is to determine the appropriate response. Digital systems can help clinicians organise assessment findings, document goals, and schedule follow-up.

A decision support system might prompt a dietitian to review energy and protein requirements, investigate swallowing concerns, or evaluate barriers to adequate food intake. It may also remind the team to reassess progress and update the care plan.

The system should not prescribe identical interventions to everyone with the same risk score. Nutritional needs depend on the underlying condition, clinical status, treatment goals, food preferences, and ability to eat safely.

Individualised recommendations require clinical assessment and coordination between relevant professionals. The technology should make this process easier rather than replace the expertise needed to make treatment decisions.

Designing Effective Digital Clinical Workflows

Connecting Alerts With Clear Actions

An alert has limited value if no one knows who should respond. Effective systems assign responsibility for reviewing positive screens, completing assessments, and arranging interventions.

For example, a high-risk result could generate a dietitian referral, while a moderate-risk result might prompt further screening by a trained healthcare professional. The exact workflow should reflect local staffing, validated protocols, and clinical priorities.

Systems should also record whether an alert was reviewed, what action followed, and whether the patient received the recommended care. This information helps healthcare organisations identify delays and evaluate whether the technology improves actual practice.

Avoiding Alert Fatigue

Clinicians already receive notifications about medication safety, abnormal test results, and numerous other clinical issues. Excessive or poorly targeted alerts can become easy to ignore.

Digital malnutrition systems should prioritise alerts according to clinical urgency and provide enough context to explain their relevance. They should avoid repeatedly presenting the same message when no meaningful change has occurred.

Organisations should regularly review alert frequency, response rates, missed cases, and unnecessary referrals. These measures can help refine the system without sacrificing sensitivity to important nutritional risks.

Limitations, Safety, and Ethical Considerations

Digital decision support can fail when data are incomplete, inaccurate, or poorly standardised. Weight measurements may be missing, food intake may be documented inconsistently, and clinical notes may contain ambiguous descriptions. These limitations can lead to missed cases or unnecessary alerts.

Bias is another concern. A model trained primarily on one population may perform differently among people with different ages, disabilities, cultural dietary practices, or socioeconomic circumstances. Healthcare organisations should assess performance across relevant groups and investigate unequal error rates.

Privacy protection is essential when systems combine clinical histories, nutritional information, and home-monitoring data. Access should be restricted appropriately, and information should be collected and retained according to applicable privacy requirements.

Finally, accountability must remain clear. Healthcare professionals should be able to review recommendations, challenge inappropriate alerts, and document their decisions. Patients should not be denied assessment because an algorithm assigns them a low risk score.

Evaluating the Impact of Digital Malnutrition Support

The success of a digital system should not be judged by prediction accuracy alone. Healthcare organisations must determine whether it improves the process and outcomes of care.

Evaluation can examine how quickly patients receive screening, how often high-risk cases receive comprehensive assessment, and whether appropriate nutritional interventions begin sooner. Researchers can also assess follow-up completion, changes in nutritional status, patient experience, and the workload created by alerts.

Comparisons before and after implementation may provide useful initial evidence, but they cannot always establish that the system caused an improvement. Differences in staffing, clinical protocols, and patient populations may also influence outcomes.

Stronger evaluations may include prospective studies, controlled implementation designs, and independent validation across multiple sites. These approaches help determine whether the benefits persist beyond the original development environment.

Conclusion

Digital clinical decision support can help healthcare professionals identify malnutrition risk more consistently across hospitals, outpatient clinics, community services, and long-term care facilities. By combining patient records, validated screening criteria, and carefully evaluated predictive models, these systems can make important nutritional warning signs easier to recognise.

Their greatest value lies in connecting risk detection with timely assessment, appropriate referrals, individualised care, and continued monitoring. Effective implementation also requires clear responsibilities, accessible interfaces, reliable data, and safeguards against unnecessary alerts.

Artificial intelligence can strengthen this process by identifying complex patterns and supporting earlier risk prediction. However, it cannot replace comprehensive nutritional assessment or professional judgement.

A successful digital approach treats malnutrition screening as part of continuous care rather than an isolated administrative task. When technology supports collaboration across healthcare settings, it can help make nutritional risks more visible and improve the consistency of care.