On this page
Quick Reference
Overview and Recommendations
Background
- •Clinical Decision Support Systems (CDSS) are computer-based tools that integrate patient-specific clinical data with curated knowledge bases to deliver evidence-based recommendations, alerts, or risk estimates at the point of care, augmenting, not replacing, clinician judgment. Baseline medication error rates in hospitalized adults reach 6.83 per 1,000 patient-days, and adverse drug events occur in 8.9 per 100 admissions, framing the clinical opportunity for CDSS to improve safety and guideline adherence.
- •CDSS are broadly classified into knowledge-based (rule-driven, e.g., drug-lab interaction alerts generated by ATC drug class and lab thresholds) and non-knowledge-based (machine learning or AI-driven, e.g., gradient boosting models for sepsis prediction or videolaryngoscopy planning). A third emerging category uses large language models prompted with clinical inputs, though their outputs show prompt-sensitive threshold behavior and must be interpreted cautiously as quantitative risk estimates.
- •The clinical impact of CDSS is mediated by its interaction with human cognition and workflow. Unhelpful AI predictions increase clinician dwell time on original data by 19%-26%, and displaying AI uncertainty paradoxically lengthens cognitive processing time, meaning poorly calibrated systems can negate the efficiency gains CDSS is meant to provide.
- •The effectiveness of CDSS depends critically on the match between knowledge architecture and clinical task: rule-based alerts excel for predictable drug-lab interactions, while machine learning models capture complex non-linear risk. A Clinician Turing Test paradigm has been proposed as a Phase 1b validation step, testing whether clinicians can distinguish AI-generated from human-generated treatment recommendations, a measure not only of predictive performance but of clinical plausibility.
- •CDSS have evolved from generic rule-based systems to context-aware AI tools. Early systems suffered from poor workflow fit and excessive alerts, but tailored alerts (suppressing low-risk drug interactions) reduced administered high-risk combinations by 12% in ICU settings. The field now recognizes that CDSS consistently improve process measures (preventive services OR 1.42, clinical study ordering OR 1.72, prescribing OR 1.57) but have inconsistent effects on hard outcomes like mortality, which remains an active area of investigation.
Evaluation
- •Suspect that a CDSS may be useful when managing an undifferentiated patient with diagnostic uncertainty. Controlled testing shows that entering 3 to 6 key clinical findings into a diagnostic CDSS (e.g., Isabel) yields the correct diagnosis in 96% of complex cases, compared to 74% when the full history is pasted, a difference larger than most single diagnostic tests.
- •Examine the clinical context: CDSS are most helpful when the clinician already has a provisional differential and needs to broaden or confirm it, rather than when the presentation is genuinely obscure. In atypical presentations, CDSS may provide no additional insight and should not be relied upon.
- •Choose the right type of CDSS for the clinical task: use rule-based alerts for predictable drug-lab interactions (e.g., vitamin K antagonist plus elevated INR), machine learning models for non-linear risk prediction (e.g., necrotizing fasciitis prediction with AUC 0.809, pediatric sepsis), and LLM-based systems for differential generation, each has specific strengths but also limitations such as low positive predictive value in low-prevalence conditions.
- •Assess the quality of the underlying data before acting on a CDSS recommendation. Stale vital signs, incomplete medication lists, or inaccurate lab values generate misleading alerts, always verify that the CDSS is connected to a current, structured electronic health record.
- •Order a diagnostic CDSS proactively by entering a focused set of discriminative features extracted from the history and physical. For internal medicine cases, this process takes less than a minute with results in 2-3 seconds; the key is to enter discriminating positive and negative findings, not the full narrative.
- •Consider the cognitive phase of decision-making when selecting CDSS delivery mode. During early heuristic triage (System 1), passive, user-initiated tools (risk calculators, diagnostic checklists) are more effective than active interruptive alerts. During later analytical phases (System 2) when diagnostic uncertainty decreases and information quantity increases, active alerts (e.g., IV-to-oral switch prompts, medication reconciliation alerts) become appropriate.
- •Evaluate alert specificity: low-specificity alerts (e.g., pancreatic cancer screening models with positive predictive value below 1%) cause alert fatigue and erode clinician trust, leading to high override rates. The most effective CDSS reduce false positive rates by tailoring alerts to clinical context, for example, suppressing low-risk drug-drug interaction alerts in ICU settings.
- •Monitor the impact on workflow and cognitive load. CDSS should not increase burden; the TraumaFlow system for polytrauma improved documentation completeness without increasing NASA-RTLX workload scores. If a CDSS increases dwell time on original data by 19%-26% (as seen with unhelpful AI), it may hinder rather than help clinical decision-making.
- •Assess clinician acceptance and engagement before full deployment. In a large German trial, practices with lower CDSS usage intensity but higher change commitment showed larger intervention effects on hospitalization and mortality, indicating that implementation success depends more on attitudinal factors than on usage volume alone.
- •Use CDSS to support medication reconciliation at care transitions. Medication reconciliation using EHR-integrated CDSS reduces adverse drug events (OR 0.38) and improves follow-up adherence, in an HIV clinic, interactive alerts reduced 6-month suboptimal follow-up from 30.1 to 20.6 events per 100 patient-years.
- •For special populations, adjust CDSS expectations accordingly: in pregnancy, ensure CDSS includes teratogenicity alerts and pregnancy-adjusted vital sign thresholds (the PANDA system more than doubled quality of antenatal care); in the elderly, integrate STOPP/START criteria and fall-risk alerts; in pediatrics, weight-based dosing algorithms are essential; in immunocompromised patients, incorporate drug-drug interaction databases and adjusted infection marker thresholds.
- •Recognize that diagnostic CDSS achieves its highest sensitivity (96%) when clinicians enter 3 to 6 key findings rather than the full history; always cross-check the CDSS differential with local disease prevalence and patient-specific risk factors before ordering confirmatory tests.
- •When using CDSS for risk stratification (e.g., CHA2DS2-VASc for atrial fibrillation anticoagulation), ensure that all necessary data fields are complete and accurate, missing entries can misclassify patients and erode clinical trust.
- •For acute settings, deploy CDSS that provide real-time, time-critical recommendations. The Cerebri smartphone app for acute ischemic stroke reduced imaging-to-decision time from 22 to 6 minutes and improved overall guideline adherence from 73.9% to 96%.
Management
- •Implement CDSS as a process-improvement tool, not as a substitute for clinical judgment. Set realistic expectations: CDSS consistently improve prescribing, monitoring, and follow-up processes but have not shown consistent reductions in mortality or length of stay in meta-analyses.
- •Tailor drug-drug interaction alerts to high-risk combinations only. In a stepped-wedge trial across 9 Dutch ICUs, this approach reduced the number of administered high-risk drug combinations from 35.6 to 26.2 per 1000 drug administrations, a 12% reduction, without compromising safety.
- •Require clinicians to supply a reason when overriding CDSS advice. This practice increases the likelihood of the CDSS achieving its intended effect by an odds ratio of 11.23, as it forces cognitive engagement and accountability.
- •Integrate CDSS with complementary interventions such as point-of-care testing or pharmacist review. Bundled strategies achieve higher guideline adherence success rates (85%) than single-component CDSS alone.
- •For diagnostic CDSS, train clinicians to enter 3 to 6 key clinical findings rather than pasting full clinical notes, this simple technique raises diagnostic accuracy from 74% to 96%.
- •For acute time-critical conditions (stroke, polytrauma, sepsis), deploy CDSS that deliver real-time recommendations at the point of care. The Cerebri smartphone app reduced imaging-to-decision time by 16 minutes (from 22 to 6 minutes) and improved overall guideline adherence from 74% to 96%.
- •For chronic disease management, use CDSS to flag patients with suboptimal follow-up or therapy. In HIV care, interactive alerts increased the monthly CD4 count slope (0.0053 vs 0.0032 ×10⁹ cells/L) and reduced 6-month suboptimal follow-up events by 31.5%.
- •Implement deprescribing CDSS in older adults with polypharmacy. Use tools that integrate STOPP/START criteria, renal function-based dosing, and anticholinergic burden scoring to reduce potentially inappropriate medication initiation by up to 18%.
- •For antimicrobial stewardship, deploy CDSS as part of a digital intervention, 58% of studied programs use CDSS. CDSS consistently improves therapeutic appropriateness and reduces antibiotic consumption, but pair it with pharmacist review for maximal effect on clinical outcomes.
- •Avoid non-interruptive alerts for urgent safety issues that require immediate attention (e.g., severe drug allergy, critical lab value). Reserve non-interruptive modes for low-urgency preventive reminders (e.g., statin initiation in eligible patients) to minimize alert fatigue.
- •Monitor alert override rates as a quality metric. If override rates exceed 70%, reevaluate alert specificity and consider suppressing low-yield alerts or adjusting thresholds.
- •For pregnancy, ensure CDSS includes teratogenicity alerts with safer alternative suggestions, pregnancy-adjusted vital sign thresholds, and breastfeeding safety recommendations. The PANDA system in Burkina Faso increased excellent antenatal care quality more than twofold (RR 2.71).
- •For pediatric populations, implement weight-based dosing algorithms and age-adjusted normal ranges. Flag off-label prescribing and provide evidence-based dose adjustments when pediatric data exist.
- •Ensure data quality by implementing validation checks before CDSS inference. Stale or inaccurate EHR data produce misleading recommendations that undermine trust and may lead to clinical errors.
- •When CDSS is unavailable or fails (e.g., technical downtime), revert to standard institutional protocols and paper-based checklists. Have a contingency plan for acute settings where time pressure is highest.
- •Deploy CDSS with a multidisciplinary implementation team that includes clinicians, informaticians, and change management leaders. User-centered design, early stakeholder engagement, and workflow integration are the strongest predictors of long-term sustainability.
- •For risk stratification, use CDSS that embed validated clinical scores (e.g., CHA2DS2-VASc for atrial fibrillation, PI-RADS for prostate MRI) alongside machine learning models. Hybrid approaches combining traditional scores with ML offer the best balance of interpretability and predictive performance.
- •Consider cost-effectiveness: CDSS may reduce unnecessary medical visits (LabTest Checker reduced potential visits by 41.6%) and improve resource utilization, but economic evidence is still limited, incorporate local cost data before large-scale deployment.
Board Review — High Yield
- •3 to 6 key findings, Entering 3-6 discriminative clinical features into a diagnostic CDSS yields 96% accuracy vs 74% with full history paste.
- •Cognitive phase matching, Passive, user-initiated CDSS (risk calculators) are effective in early heuristic triage (System 1); active interruptive alerts suit later analytical phases (System 2).
- •Alert tailoring, Suppressing low-risk drug-drug interaction alerts reduces administered high-risk combinations by 12% in ICU settings.
- •Override justification, Requiring a reason for overriding CDSS advice increases success odds ratio to 11.23.
- •Process vs outcome, CDSS improves process measures (OR 1.42-1.72) but meta-analyses show no reduction in mortality (RR 0.97).
- •Deprescribing in elderly, CDSS reduces potentially inappropriate medication initiation by up to 18% in older adults.
- •Pregnancy CDSS, The PANDA system more than doubled excellent antenatal care quality (RR 2.71).
- •PPV limitation, AI pancreatic cancer screening has pooled AUC 0.785 but PPV <1%, limiting use to high-risk groups.
- •Data quality, Stale or incomplete EHR data undermine CDSS recommendations; always verify input currency.
- •Implementation science, Higher CDSS usage intensity does not guarantee better outcomes; attitudinal engagement and change commitment are stronger predictors.
Deep Dive — Evidence Details
Definition, Classification and Nomenclature
- ▸CDSS types: knowledge-based (rule-based) vs non-knowledge-based (ML) vs LLM-based
- ▸Bundled interventions (CDSS + POC testing) improve adherence to 85%
A Clinical Decision Support System (CDSS) integrates patient-specific data with curated knowledge to deliver evidence-based recommendations at the point of care [1]A1a. Synonyms include CDS, computerized decision support, alert systems, and order-set engines. Classification by knowledge source: knowledge-based (rule-based, e.g., drug-lab alerts [5]B2b) and non-knowledge-based (data-driven, e.g., gradient boosting for videolaryngoscopy [6]B2b or sleep apnea [4]C4). A third emerging category uses inference-only large language models (LLMs) for risk estimation [3]C4. Functionally, CDSS can be diagnostic, predictive, therapeutic, or pre-procedural. Most modern systems combine multiple functions; bundled interventions (CDSS + point-of-care testing) achieve higher guideline adherence (85%) than single-component strategies [1]A1a.
Pearl: CDSS effectiveness depends on the match between knowledge architecture and clinical task; rule-based alerts excel for predictable drug-lab interactions, while machine learning models capture complex non-linear risk, but both suffer from alert fatigue and threshold sensitivity that must be addressed during implementation [5]B2b[3]C4.
| CDSS Type | Knowledge Source | Key Examples | Representative Reference |
|---|---|---|---|
| Rule-based (knowledge-based) | Expert-crafted rules (ATC drug class + lab thresholds) | Alert generation for vitamin K antagonists, heparins | [5]B2b |
| Machine learning (non-knowledge-based) | Trained on structured clinical data (e.g., XGBoost, Random Forest) | Inpatient OSA prediction, videolaryngoscopy planning | [4]C4, [6]B2b |
| LLM-based (inference-only) | Pre-trained language model + prompt engineering | ICU mortality prediction | [3]C4 |
Pathophysiology and Mechanism
- ▸CDSS shifts clinician attention: 19-26% of fixations to AI region
- ▸Unhelpful AI increases cognitive load; uncertainty display adds burden
CDSS modulates human cognition and workflow. When AI-generated advice is available, pharmacists shift 19%-26% of visual fixations to the AI region [7]A1b. Unhelpful AI increases dwell time on reference images, signaling heightened cognitive processing [7]A1b. Displaying AI uncertainty paradoxically lengthens cognitive processing time [7]A1b. The technical architecture follows a closed-loop cycle: data acquisition → algorithmic inference → recommendation → clinician action → feedback. In oncology, a PRO alerting system triggered alerts on 7.82% of questionnaires; median response time was 1 hour, and red alerts were documented 56.83% of the time [10]C4. For stroke, a cerebrovascular AI-CDSS provides real-time imaging analysis and guideline-based recommendations, aiming for a 26% relative reduction in new vascular events at 3 months [9]D5. The Clinician Turing Test proposes a phase 1b safety validation where clinicians attempt to distinguish AI-generated from human-generated recommendations [8]D5.
Pearl: The most consequential mechanism of CDSS is the cognitive interaction with the clinician; unhelpful AI increases dwell time on original data by an estimated 19%-26% [7]A1b, and displaying uncertainty adds further cognitive load.
| Component | Mechanism | Evidence |
|---|---|---|
| Data acquisition | PRO questionnaires, imaging, lab values, vitals | Alert rate 7.82% in oncology [10]C4 |
| Algorithmic inference | Rule-based, ML, or AI models; federated aggregation | Adaptive aggregation maintains accuracy [13]D5 |
| Recommendation generation | AI-assigned imaging analysis, etiology, guideline-based suggestions | Cerebrovascular AI-CDSS: 26% relative risk reduction target [9]D5 |
| Clinician interaction | Visual fixations, dwell time, cognitive load | 19%-26% shift to AI region; unhelpful AI increases dwell time [7]A1b |
| Feedback loop | Alert documentation, response time, outcome | Median 1-hour response; red alerts 56.83% documented [10]C4 |
| Validation | Clinician Turing Test | Indistinguishability signals safety [8]D5 |
Epidemiology, Etiology and Risk Factors
- ▸Medication errors: 6.83/1000 patient-days; ADEs: 8.9/100 admissions
- ▸Low alert specificity (PPV <1%) is the main risk factor for CDSS failure
Baseline medication error rates in hospitalized adults are 6.83 per 1,000 patient-days, with adverse drug events (ADEs) in 8.9 per 100 admissions [14]A1b. Inappropriate abdominal imaging requests affect 6.19% of studies; ultrasound has the highest rate at 12.35% [15]B2b. Daily CDSS users rose from 42.1% to 88.7% over 12 months after AI enhancement [14]A1b. Pooled AUC for pancreatic cancer risk prediction is 0.785, but PPV remains <1% [19]B2a. Key risk factors for CDSS failure: low alert specificity (DDI alerts: OR 0.86 for ADE reduction) [20]B2a, low baseline clinician knowledge (pressure ulcer management: 4.3% correct without CDSS vs 49.3% with, aOR 29.1) [16]A1b, and low PPV in screening models eroding trust [19]B2a.
Pearl: The single most actionable risk factor for CDSS failure is low alert specificity; models with PPV < 1% generate excessive false alerts that erode clinician trust and adoption [19]B2a[20]B2a.
| Risk Factor | Associated Outcome | OR or Effect | Evidence Level | Source |
|---|---|---|---|---|
| Low alert specificity (high false positive rate) | No reduction in ADEs; alert fatigue | OR 0.86 (0.56-1.34) | High (meta-analysis) | [20]B2a |
| Low baseline clinician knowledge (pressure ulcers) | Incorrect treatment decision without CDSS | Baseline 4.3% correct | High (RCT) | [16]A1b |
| Low PPV in screening models (pancreatic cancer) | Over-alerting, low clinician trust | PPV < 1% | High (meta-analysis) | [19]B2a |
| Low initial clinician adoption | Incomplete CDSS benefit | Adoption 42.1% → 88.7% | Moderate (RCT) | [14]A1b |
Clinical Presentation
- ▸Enter 3-6 key findings into CDSS for 96% diagnostic accuracy (vs 74% with full history)
- ▸LabTest Checker: 100% sensitivity for emergency alerts, reduces unnecessary visits by 41.6%
Clinicians encounter CDSS in three scenarios: undifferentiated cases, atypical presentations, and routine test results needing risk stratification. For undifferentiated internal medicine, entering 3-6 key clinical findings into the Isabel web-based CDSS yields the correct diagnosis in 96% of cases, versus 74% when the full history is pasted [25]C4. Data entry takes <1 minute; results appear in 2-3 seconds [25]C4. In atypical presentations (e.g., pulmonary embolism), CDSS did not improve diagnostic accuracy compared to conventional methods [21]C4. For laboratory interpretation, the LabTest Checker achieved 74.3% pathology accuracy and 100% sensitivity for emergency alerts, potentially reducing unnecessary visits by 41.6% [26]B2b. AI-based CDSS for pediatric sepsis outperforms traditional scoring systems but false-positive rates vary by EHR infrastructure [23]B2a.
Pearl: When using a CDSS for an undifferentiated case, enter 3 to 6 key clinical findings rather than the full history; this raises the probability of seeing the correct diagnosis from 74% to 96%, a difference larger than most single diagnostic tests [25]C4.
| Domain | Tool | Key Metric | Source |
|---|---|---|---|
| Undifferentiated internal medicine | Isabel | 96% correct diagnosis with key findings [25]C4 | Graber 2008 [25]C4 |
| Laboratory diagnostics | LabTest Checker | 74.3% pathology accuracy, 100% emergency sensitivity [26]B2b | Szumilas 2024 [26]B2b |
| Sepsis (pediatric ICU) | AI-driven CDSS | Improved bundle adherence; false-positive rates vary by infrastructure [23]B2a | Fu 2026 [23]B2a |
| Pain management (head/neck RT) | Data-driven CDSS (concept) | Temporal prediction of pain trajectories (median AUCpain 16%) [27]B3b | Salama 2023 [27]B3b |
Diagnosis and Workup
- ▸Enter 3-6 key findings for 96% diagnostic accuracy; avoid full history paste
- ▸AI-assisted imaging CDSS: median sensitivity 93.6%, specificity 90.6%
The gold-standard approach for diagnostic CDSS is to enter 3-6 key clinical findings into a system like Isabel, yielding the correct diagnosis in 96% of cases vs 74% with full history paste [25]C4. AI-assisted CDSS across point-of-care imaging (ultrasound, chest X-ray, dermoscopy) achieved a median sensitivity of 93.6% (IQR 87-98%) and specificity of 90.6% (IQR 74.5-) across 20 studies (~78,000 patients) [34]B2a. For pediatric sepsis, ML models outperform traditional scoring systems but exact sensitivity/specificity not reported [23]B2a. The diagnostic algorithm: 1) Obtain history and physical; 2) Enter 3-6 key findings into CDSS; 3) Review differential; 4) Consider local prevalence and risk factors; 5) Order confirmatory tests; 6) Refine if uncertainty persists. CDSS that require a reason for overriding advice are more likely to succeed (OR 11.23) [29]A1a.
Pearl: The diagnostic accuracy of a CDSS depends critically on structured input of key findings; pasting full clinical notes reduces sensitivity from 96% to 74% [25]C4; always enter a focused set of discriminative features.
| Modality / Setting | CDSS Tool | Sensitivity | Specificity | Key Finding |
|---|---|---|---|---|
| General Medicine (complex cases) [25]C4 | Isabel web-based (structured entry) | 96% | Not reported | Manual entry of 3-6 key findings, results within 2-3 seconds |
| Point-of-care imaging (multimodal) [34]B2a | AI-assisted CDSS | 93.6% (IQR 87-98) | 90.6% (IQR 74.5-96.7) | Non-specialists achieved specialist-level performance after median 1 hour training |
Severity, Staging and Risk Stratification
- ▸APEL program: anticoagulation increased from 52.6% to 59.8% using CHA2DS2-VASc
- ▸Pancreatic cancer ML models: AUC 0.785 but PPV <1%
CDSS embed validated risk scores to automate stratification. In atrial fibrillation, the APEL program using CHA2DS2-VASc increased appropriate anticoagulation from 52.6% to 59.8% (p<0.001) [35]C4. For incidentally detected hepatic steatosis, the STIRRED protocol uses EHR-based risk scores to identify high-risk patients (n=616) for targeted follow-up, powered to detect a 5.6% absolute risk difference [41]D5. In prostate MRI, CDSS nomograms incorporating PI-RADS (82% of tools), PSA density (64%), age (64%), and PSA (41%) achieve AUC >0.80 [46]B2a. ML models for pancreatic cancer achieve pooled AUC 0.785 but PPV <1% [19]B2a. For prostate cancer, AI-CDSS models report AUROC ≈0.84 for recurrence and C-index up to 0.92 for progression-free survival [45]B2a. Challenges: low PPV in low-prevalence conditions, lack of external validation, and poor calibration reporting [43]B2a[45]B2a[46]B2a.
Pearl: CDSS risk stratification improves guideline-concordant care only when the underlying data are complete and the scores are validated in the target population; missing or inaccurate EHR entries can misclassify patients and erode clinical trust [35]C4[41]D5.
| Question | Position A | Position B | Strength | Implication |
|---|---|---|---|---|
| Which risk stratification approach is preferred? | Traditional validated scores (CHA2DS2-VASc, PI-RADS) are interpretable, widely validated, and guideline-endorsed [35]C4[46]B2a. | ML models (neural networks, gradient boosting) achieve higher AUC and capture temporal/dynamic risk [19]B2a[23]B2a[44]B2b. | Moderate; ML models are promising but lack prospective validation for most conditions. | Hybrid models combining traditional scores with ML may offer the best balance of interpretability and performance. |
Acute Management
- ▸Stroke CDSS: imaging-to-decision time reduced from 22 to 6 minutes; adherence 96% vs 74%
- ▸Medication reconciliation reduces ADEs (OR 0.38)
In acute settings, CDSS can deliver guideline-recommended actions at the point of care. Key evidence: Polytrauma (TraumaFlow): documentation completeness improved from 74% to 82% (P=0.002); 37% of prompts triggered new actions [53]C4. Medication reconciliation: reduces ADEs (OR 0.38, 95% CI 0.18-0.80) [48]A1a. UTI antibiotic prescribing: mixed results; 4 of 10 trials showed reduced prescribing rates, 2 showed increased appropriateness [49]B2a. Stroke (cerebri app): imaging-to-decision time reduced from 22 to 6 minutes; adherence increased from 74% to 96% [47]A1b. Necrotizing fasciitis: XGBoost model AUC 0.809 vs nomogram 0.724 [52]B3b. Acute alcoholic hallucinosis: multi-omics CDSS reduced ADRs (UKU day 6: 5.0 vs 12.0, P<0.01) [51]C4. What NOT to do: Do not assume CDSS replaces clinical judgment; verify data timeliness; deploy only after usability testing.
Pearl: Deploy CDSS in acute settings where time pressure is highest, stroke, polytrauma, sepsis, because even a few minutes saved (e.g., 16 minutes for stroke imaging-to-decision) translate directly into better guideline adherence and potentially improved outcomes [47]A1b[53]C4.
| Acute Condition | CDSS Type / Name | Key Outcome | Evidence Level |
|---|---|---|---|
| Acute ischemic stroke | Cerebri smartphone app | Imaging-to-decision time reduced by 16 min (6 vs 22 min); adherence increased from 74% to 96% [47]A1b | 1b |
| Polytrauma | TraumaFlow | Documentation completeness 82% vs 74%; 37% of prompts triggered new actions [53]C4 | 4 |
| Medication safety (hospital) | EHR-based medication reconciliation | ADE reduction OR 0.38, 95% CI 0.18-0.80 (moderate certainty) [48]A1a | 1a |
| UTI antibiotic prescribing | EHR-integrated CDSS | Mixed: reduced prescribing rates (4 studies), increased appropriateness (2 studies) [49]B2a | 2a |
| Sepsis/ARDS | AVA (AI ventilator assistant) | Preclinical; Turing test study ongoing [8]D5 | 5 |
| Necrotizing fasciitis | XGBoost ML model | AUC 0.809 vs nomogram 0.724 [52]B3b | 3b |
| Acute alcoholic hallucinosis | Multi-omics CDSS (pharmacogenetics, phenotyping, miRNA) | ADR reduction: UKU 5.0 vs 12.0 (P < 0.01) [51]C4 | 4 |
Long-term and Definitive Management
- ▸HIV: CD4 slope improved by 0.0021 ×10⁹/L/month; suboptimal follow-up reduced
- ▸Systematic review: process measures improved (OR 1.42-1.72) but clinical outcomes limited
CDSS sustain guideline-concordant therapy in chronic disease. In HIV, interactive alerts increased CD4 cell count slope (0.0053 vs 0.0032 ×10⁹/L/month, P=0.040) and reduced suboptimal follow-up (20.6 vs 30.1 per 100 patient-years) [28]A1b. In hypertension, CDSS increased treatment intensification from 11.6% to 47.3% (aOR 6.87) [31]B2b. For CVD, statin prescribing increased by 17.9% vs 5.5% nationally; CHD mortality fell 43% vs 25% [62]B2b. Atrial fibrillation: appropriate anticoagulation rose from 52.6% to 59.8% [35]C4. In diabetes, CHW-led tablet-based CDSS reduced HbA1c (adjusted mean difference -0.46%) [61]A1b. A systematic review of 148 RCTs found CDSS improved preventive services (OR 1.42), ordering (OR 1.72), and prescribing (OR 1.57) [54]A1a. Medication errors reduced (RR 0.24) [55]A1a. However, higher usage intensity does not always yield larger effects; attitudinal engagement matters [57]A1b.
Pearl: CDSS consistently improve process measures (prescribing, monitoring, follow-up) across chronic diseases, but evidence for hard clinical outcomes remains limited; successful long-term deployment hinges on clinician attitudinal engagement and practice context as much as on system design [28]A1b[54]A1a[57]A1b.
| Domain | Key study | Process outcome | Effect size | NNT / NNH |
|---|---|---|---|---|
| HIV care | [28]A1b (1b) | CD4 slope difference | 0.0021 ×10⁹/L/month (95% CI 0.0001-0.004) | Not calculable from reported data |
| CVD prevention | [62]B2b (2b) | CHD mortality reduction | 43% vs. 25% relative fall over 3 years | Not reported |
| AF anticoagulation | [35]C4 (4) | Anticoagulation rate increase (CHA₂DS₂-VASc ≥1) | 52.6% to 59.8% (P<0.001) | Not reported |
| Mental health | [36]B2b (2b) | Diagnosis rate increase per 1000 person-years | 58.8 (P=0.001) | Not calculable from reported data |
| Question | Position A | Position B | Strength | Implication |
|---|---|---|---|---|
| CDSS improves patient outcomes | Bright 2012 meta-analysis shows process improvements but limited clinical outcome data [54]A1a | Roumeliotis 2019 meta-analysis found no effect on mortality (RR 0.97) or length of stay [55]A1a | Moderate | CDSS should be viewed as a process-improvement tool; outcome benefits require further high-quality trials |
| Higher CDSS usage intensity improves outcomes | Common assumption that more use yields more benefit | Basten 2026 cluster analysis found opposite pattern: low-use cluster had largest effect [57]A1b | Strong | Implementation success depends on clinician engagement and context, not purely on usage volume |
History and Evolution of Treatment
- ▸Tailoring alerts reduced high-risk drug combinations by 12% in ICUs
- ▸AI-enhanced CDSS reduced medication errors by 49.2% and ADEs by 47.2%
Early rule-based CDSS suffered from alert fatigue and poor workflow fit [68]D5. Landmark trials: Samore et al. showed a 32% relative decrease in antibiotic prescribing for RTIs [67]A1b; Robbins et al. demonstrated improved CD4 counts in HIV [28]A1b. Tailoring alerts (e.g., Bakker et al. in ICUs) reduced high-risk drug combinations by 12% [66]A1b. AI-enhanced systems: Shakarbaev et al. reduced medication errors by 49.2% and ADEs by 47.2% [14]A1b. Rezende et al. shortened antibiotic courses using smartphone CDSS (6.0 vs 7.0 days) [71]A1b. Diagnostic applications: Isabel achieved 96% accuracy with structured entry [25]C4; LLM-assisted rheumatology diagnosis improved accuracy (aOR 7.0) [72]A1b. Stroke care: cerebri app reduced decision time from 22 to 6 minutes [47]A1b. Reporting standards: DECIDE-AI guideline (17 AI-specific items) [65]A1c. Implementation science: success depends on transparency, usability, and attitudinal factors more than usage volume [57]A1b[70]B2a.
Pearl: The evolution from generic rule-based alerts to context-aware AI systems has transformed CDSS from a source of alert fatigue to a measurable patient safety tool, but success still hinges on workflow integration, clinician trust, and complementary patient outreach strategies [66]A1b[70]B2a[75]B2b.
| Trial (Year) | Domain | Key Finding |
|---|---|---|
| Samore et al. (2005) [67]A1b | Antimicrobial prescribing | Prescribing rate decreased from 84.1 to 75.3 per 100 person-years; 32% relative reduction in never-indicated visits |
| Robbins et al. (2012) [28]A1b | HIV care | Greater CD4 increase (0.0053 vs 0.0032 ×10⁹ cells/L per month); reduced suboptimal follow-up (20.6 vs 30.1 events/100 pt-yr) |
| Bakker et al. (2024) [66]A1b | ICU drug-drug interactions | 12% decrease in high-risk drug combinations (26.2 vs 35.6 per 1000 administrations) |
| Shakarbaev et al. (2026) [14]A1b | Medication safety | 49.2% reduction in medication errors; 47.2% reduction in adverse drug events |
| Rezende et al. (2026) [71]A1b | Antibiotic stewardship | Median antibiotic duration 6.0 vs 7.0 days (P=0.015) |
| Bonura et al. (2025) [47]A1b | Acute stroke | Imaging-to-decision time reduced from 22 to 6 minutes; guideline adherence 96% vs 73.9% |
| Roemer et al. (2026) [72]A1b | Diagnostic accuracy (rheumatology) | Correct top diagnosis 77.5% vs 32.4% with LLM; OR 7.0 |
| Kostopoulou et al. (2015) [30]A1b | GP diagnostic accuracy | Absolute 6% improvement with early diagnostic suggestions (OR 1.31) |
Generalist Reasoning under Diagnostic Uncertainty, Point-of-Care Scores & Referral Thresholds
- ▸Passive CDSS for early heuristic phase; active alerts for analytical phase
- ▸CCHR reduced CT ordering by 21.1% absolute
CDSS must match the clinician's cognitive phase: early heuristic (System 1) benefits from passive tools; late analytical (System 2) benefits from active alerts [79]B2a. Evidence: CCHR tool reduced CT ordering for head trauma from 66.9% to 45.8% [76]A1b. CHICA system increased maternal depression screening (OR 2.06 for referral) [77]A1b. Uncertainty-aware AI improved pharmacist rejection of incorrect medications (96.1% vs 91.8% for black-box) but increased cognitive load [7]A1b[78]A1b. LLM-assisted diagnosis in gynecologic oncology showed 70% concordance with NCCN guidelines [80]B2a. Referral thresholds: STIRRED protocol uses EHR to identify incidental hepatic steatosis and prompt referral [41]D5.
Pearl: When designing or choosing a CDSS, match the delivery mode to the clinician's cognitive moment: passive tools support early heuristic triage, while active alerts are better reserved for later analytical decisions [79]B2a.
| Decision Phase | Cognitive Mode | Optimal CDSS Delivery | Example |
|---|---|---|---|
| Early (triage) | Heuristic (System 1) | Passive, user-initiated | Risk calculator lookup |
| Late (disposition) | Analytical (System 2) | Active, interruptive alert | IV-to-oral switch recommendation |
Multimorbidity, Polypharmacy & Deprescribing
- ▸PIM initiation reduced by up to 18%; PIM ordering in ED reduced by 40%
- ▸Tailored DDI alerts reduced high-risk combinations by 12%
CDSS reduces potentially inappropriate medications (PIMs) in older adults: systematic review of 16 RCTs showed up to 18% reduction in PIM initiation [85]A1a. In ED, CDSS reduced PIM ordering by 40% (OR 0.60) [86]A1a. Tailoring drug-drug interaction (DDI) alerts: cluster-randomized trial in ICUs reduced high-risk combinations by 12% (26.2 vs 35.6 per 1000 administrations) [66]A1b. Deprescribing: AMUSE trial tests CDSS-OPTIMED for patients with life expectancy ≤3 months [83]D5. Medication reconciliation reduces ADEs (OR 0.38) [48]A1a. Barriers: usability, poor interoperability, lack of provider engagement [84]B2a. Qualitative synthesis: clinicians accept CDSS when it provides relevant knowledge but reject simplistic systems that disrupt workflows [88]D5.
Pearl: Successful deprescribing CDSS requires user-centred design, workflow integration, and multidisciplinary collaboration; barriers include poor interoperability and lack of provider engagement [84]B2a[88]D5.
| Outcome | Setting | Effect estimate | Certainty | Source |
|---|---|---|---|---|
| High-risk drug combinations per 1000 administrations | ICU | 26.2 vs 35.6 (12% decrease) | High | [66]A1b |
| PIM initiation | Mixed (≥65 yrs) | Up to 18% reduction | Moderate | [85]A1a |
| PIM ordering | Emergency department | OR 0.60 (0.48-0.74) | Moderate | [86]A1a |
| PIM discontinuation | Mixed (≥65 yrs) | 55.4% | Moderate | [85]A1a |
| Adverse drug events | Mixed | OR 0.86 (0.56-1.34) | Low | [20]B2a |
| ADEs after medication reconciliation | Hospital | OR 0.38 (0.18-0.80) | Moderate | [48]A1a |
Complications
- ▸Delirium CDSS improved assessment from 31% to 73% but no reduction in incidence
- ▸Hypoglycemia and hyperkalemia reduced with CDSS-guided interventions
CDSS can preemptively reduce hospital-acquired complications. Respiratory monitoring: each unit increase in respiratory rate (HR 1.03) and decrease in saturation (HR 1.05) predict ICU admission [91]B2b. DVT/PE prophylaxis: CDSS improves compliance with prophylaxis protocols but no standalone DVT/PE reduction data [90]A1a. Delirium: 3D-DST mobile CDSS increased nurse assessment from 31% to 73% and recognition from 42% to 89%, but did not reduce incidence or length of stay [93]A1b. Hyperkalemia: AI-based dietary CDSS reduced events in CKD [92]B2a. Hypoglycemia: CDSS-guided insulin dosing decreased events [87]B2a. Postoperative nausea/vomiting: perioperative CDSS reduced nausea but not pain scores [90]A1a. Adverse transfusion reactions: AI prediction models exist but no active CDSS evaluated [69]B2a.
Pearl: CDSS reliably improve process adherence (e.g., delirium assessment, prophylaxis compliance) but have not yet shown consistent reductions in hard clinical outcomes; the greatest gains occur when CDSS are paired with multidisciplinary action rather than standalone alerts [93]A1b.
| Complication | Frequency (reported) | Prevention | Management |
|---|---|---|---|
| Hypoglycemia | Not reported in studies | CDSS-guided insulin dosing and antidiabetic drug prescribing [87]B2a | Hypoglycemia protocol with dextrose |
| Hyperkalemia | Not reported in studies | AI-based dietary recommendations [92]B2a | Potassium-binding resins, dietary restriction |
| Delirium | 8.5% intervention vs 11.4% control (not significant) [93]A1b | Routine CDSS assessment with 3D-DST [93]A1b | Multidisciplinary delirium care bundle |
| Postoperative Nausea/Vomiting | Not reported in studies | Perioperative CDSS alerts [90]A1a | Antiemetic protocol |
| Adverse Transfusion Reactions | Not reported in studies | AI prediction models [69]B2a | Hemovigilance protocols |
Prognosis and Natural History
- ▸Process improvements: OR 1.42-1.72 for preventive services, ordering, prescribing
- ▸Mortality: RR 0.97 (no effect); only 2/10 ICU RCTs showed benefit
CDSS consistently improves process outcomes: systematic review of 148 RCTs showed increased preventive services (OR 1.42), ordering (OR 1.72), and prescribing (OR 1.57) [54]A1a. Tailored DDI alerts reduced high-risk drug combinations by 12% [66]A1b. Smartphone CDSS for pressure ulcers improved correct decisions from 4.3% to 49.3% (aOR 29.1, NNT 2.2) [16]A1b. Patient-centered outcomes: electronic prescribing reduced ADEs (RR 0.52) but not mortality (RR 0.97) or length of stay [55]A1a. Medication reconciliation reduced ADEs (OR 0.38) [48]A1a. Only 2 of 10 ICU RCTs showed mortality reduction [95]A1a. Real-world predictive tools (e.g., Epic Sepsis Model) underperform with AUROC 0.65 and high cross-site heterogeneity (I² ≥93%) [94]A1a. Systems requiring reason for override are more likely to succeed (OR 11.23) [29]A1a.
Pearl: Clinicians should expect CDSS to consistently improve process measures (appropriate testing, prescribing, and diagnosis) but not rely on it to reduce mortality or length of stay, these outcomes demand rigorous local validation and attention to implementation fidelity.
| Outcome | Effect Size | NNT (if calculable) |
|---|---|---|
| Preventive services performance | OR 1.42 (1.27-1.58) [54]A1a | , |
| Appropriate ordering of studies | OR 1.72 (1.47-2.00) [54]A1a | , |
| Guideline-concordant prescribing | OR 1.57 (1.35-1.82) [54]A1a | , |
| Correct treatment decision (pressure ulcer) | aOR 29.1 (8.2-103) [16]A1b | 2.2 |
| Adverse drug events | RR 0.52 (0.40-0.68) [55]A1a | Not calculable |
| Mortality | RR 0.97 (0.79-1.19) [55]A1a | , |
| Length of stay | MD -0.18 days (-1.42 to 1.05) [55]A1a | , |
Special Populations and Pregnancy
- ▸Pregnancy: PANDA increased antenatal care quality (RR 2.71)
- ▸Elderly: CDSS improves medication appropriateness (RR 0.71) but not hard outcomes
Pediatric CDSS requires weight-based dosing, age-adjusted ranges, and growth-chart integration. Pregnancy: PANDA system in Burkina Faso increased antenatal care quality (RR 2.71) [22]A1b; Tommy's CDSS assesses preterm birth and placental dysfunction risk [100]C4. Essential features: teratogenicity alerts, pregnancy-adjusted vitals, dosing adjustments, breastfeeding safety. Elderly: CDSS improves medication appropriateness (RR 0.71) but no effect on hospital admissions, mortality, falls, or ADEs [97]A1a. OPERAM trial integrates STOPP/START criteria with shared decision-making [99]D5[101]D5. Immunocompromised: CDSS should incorporate drug-drug interaction databases, adjust infection marker thresholds, and flag attenuated vaccine contraindications. No randomized evidence for benefit in this group.
Pearl: When implementing CDSS in special populations, the greatest value comes not from replacing clinical judgment but from embedding population-specific safety rules, teratogenicity alerts in pregnancy, STOPP/START criteria in the elderly, and weight-based dosing in children, that are easily overlooked in routine care.
| Population | Key Evidence (Ref) | Essential CDSS Features |
|---|---|---|
| Pediatrics | Limited; no RCTs | Weight-based dosing, growth-chart integration, off-label alerts |
| Pregnancy | PANDA (RR 2.71 for ANC quality) [22]A1b; Tommy's Tool feasible [100]C4; scoping reviews [103]D5 | Teratogenicity alerts, pregnancy-adjusted reference ranges, breastfeeding safety |
| Elderly | 25 RCTs: improved appropriateness RR 0.71, no mortality/falls effect [97]A1a; OPERAM protocol [99]D5 | STOPP/START criteria, renal dosing, fall-risk alerts, anticholinergic burden |
| Immunocompromised | Exclusion in sepsis AI study [98]D5; no RCTs | Drug interaction databases, adjusted infection thresholds, vaccine alerts |
Prevention, Screening and Health Maintenance
- ▸Pancreatic cancer screening: AUC 0.785 but PPV <1%
- ▸Diabetes readmission prediction: non-adherence (aOR 4.37) and lack of follow-up (aOR 4.02) key
Primary prevention: INIRSHA-PC trial tests interruptive vs non-interruptive alerts for statin prescribing [105]D5. In LMICs, CHW-led tablet CDSS with metformin, atorvastatin, aspirin reduced HbA1c (adjusted mean difference -0.46%) [61]A1b. Secondary prevention: 30-day readmission prediction in diabetes (AUC 0.897) identified medication non-adherence (aOR 4.37) and lack of follow-up (aOR 4.02) as strongest predictors [37]B3b. GRU-D++ model predicts ICU readmission (AUROC 0.802) [108]B3b. Screening: AI models for pancreatic cancer (pooled AUC 0.785) but PPV <1% [19]B2a. Barriers differ: screening CDSS have fewer workflow issues (27% vs 64%) than symptomatic-assessment CDSS [104]D5. Patient education: emphasize adherence and follow-up as modifiable predictors.
Pearl: AI models using EHR data show promise for pancreatic cancer screening (pooled AUC 0.785) but low PPV (< 1%) limits use to high-risk populations; adherence and follow-up are the strongest modifiable predictors of readmission in diabetes [19]B2a[37]B3b.
References
- [1]
Gres E, Brigadoi G, Zamperetti E et al.. “Antibiotic stewardship and point-of-care testing for children in 25 low-income and lower-middle-income countries: a systematic review and meta-analysis.” EClinicalMedicine (2025). PMID: 41377907 ↗
L1SR_OBSCited in: Definition, Classification and Nomenclature - [2]
Rabinowitz EJ, Ouyang A, Guerriero R et al.. “Impaired Oxygen Delivery Risk Analytics in the 6 Hours Before Extracorporeal Membrane Oxygenation: Single-Center Retrospective Cohort Study in Infants, 2013-2017.” Pediatric critical care medicine : a journal of the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies (2026). PMID: 42042630 ↗
L3COHORTCited in: Definition, Classification and Nomenclature, Diagnosis and Workup - [3]
Kuo JR, Chen GY, Yap XV et al.. “Prompt-Sensitive Decision Behavior of Large Language Models in Intensive Care Unit Mortality Prediction for Spontaneous Intracerebral Hemorrhage: Comparative Benchmarking Study.” Journal of medical Internet research (2026). PMID: 42422967 ↗
L4OTHERCited in: Definition, Classification and Nomenclature, Diagnosis and Workup - [4]
Boltaboyeva A, Amangeldy B, Baigarayeva Z et al.. “Frailty-Driven Prediction of Inpatient Obstructive Sleep Apnea and Related Sleep Disorder Diagnoses Using Explainable AI.” Biomedicines (2026). PMID: 42351732 ↗
L4OTHERCited in: Definition, Classification and Nomenclature, Pathophysiology and Mechanism, Clinical Presentation, Complications - [5]
Dintilhac A, Lohan L, Laureau M et al.. “From bedside observations to clinical decision support system (CDSS) rules: using real-world adverse drug events (ADEs) data to identify high-risk iatrogenic situations.” International journal of medical informatics (2026). PMID: 42322885 ↗
L2OTHERCited in: Definition, Classification and Nomenclature - [6]
Fernández-Vaquero MA, Oyarzun-Silva R, Hernández-Hernández P et al.. “Airway coach project: development of a machine learning-based model using clinical and ultrasound parameters to support videolaryngoscopy strategy.” BMC anesthesiology (2026). PMID: 42316024 ↗
L2OTHERCited in: Definition, Classification and Nomenclature - [7]
Tsai CC, Kim JY, Chen Q et al.. “Effect of Artificial Intelligence Helpfulness and Uncertainty on Cognitive Interactions with Pharmacists: Randomized Controlled Trial.” Journal of medical Internet research (2025). PMID: 39888668 ↗
L1RCTCited in: Pathophysiology and Mechanism, Generalist Reasoning under Diagnostic Uncertainty, Point-of-Care Scores & Referral Thresholds - [8]
Angeli Gazola A, Bishop NS, Schmid BE et al.. “Evaluating AI-based comprehensive clinical decision support for sepsis and ARDS: protocol for a Clinician Turing Test.” BMJ open (2025). PMID: 41448698 ↗
L5TRIAL_NONRANDOMCited in: Pathophysiology and Mechanism, Epidemiology, Etiology and Risk Factors, Acute Management - [9]
Li Z, Zhang X, Ding L et al.. “Rationale and design of the GOLDEN BRIDGE II: a cluster-randomised multifaceted intervention trial of an artificial intelligence-based cerebrovascular disease clinical decision support system to improve stroke outcomes and care quality in China.” Stroke and vascular neurology (2024). PMID: 37699726 ↗
L5TRIAL_NONRANDOMCited in: Pathophysiology and Mechanism - [10]
Strachna O, Asan O, Stetson PD. “Managing Critical Patient-Reported Outcome Measures in Oncology Settings: System Development and Retrospective Study.” JMIR medical informatics (2022). PMID: 36326801 ↗
L4COHORTCited in: Pathophysiology and Mechanism - [11]
Zheng W, Wang Y, Lu J et al.. “From empirical hemostasis to precision reversal: clinical challenges, technological innovations, and individualized strategies in anticoagulant-associated bleeding.” Frontiers in cardiovascular medicine (2026). PMID: 42428488 ↗
L5REVIEW_NARRATIVECited in: Pathophysiology and Mechanism, Long-term and Definitive Management - [12]
Jeon H, Jang HR. “Emerging therapeutic strategies for acute kidney injury: a new dawn in renal medicine.” Kidney research and clinical practice (2025). PMID: 42298994 ↗
L5OTHERCited in: Pathophysiology and Mechanism - [13]
Sri AV, Pusapati B, Puli RS et al.. “Adaptive aggregation in federated learning for retinal vein occlusion detection: A dynamic approach based on data distribution.” Computerized medical imaging and graphics : the official journal of the Computerized Medical Imaging Society (2026). PMID: 42066515 ↗
L5OTHERCited in: Pathophysiology and Mechanism - [14]
Shakarbaev N. “Ai-enhanced clinical decision support reduces medication errors and adverse drug events in a multicenter teaching hospital network: A prospective randomized controlled trial.” International journal of medical informatics (2026). PMID: 42054932 ↗
L1RCTCited in: Epidemiology, Etiology and Risk Factors, History and Evolution of Treatment - [15]
Dijk SW, Mommersteeg M, Steltenpool S et al.. “Inappropriate imaging requests in abdominal imaging: observations from the MIDAS study, a multicenter randomized controlled trial in Germany.” Abdominal radiology (New York) (2025). PMID: 41417083 ↗
L2RCTCited in: Epidemiology, Etiology and Risk Factors, History and Evolution of Treatment - [16]
Ito T, Hirosawa T, Hayashi A et al.. “Utility of a Smartphone-Based Clinical Decision Support System for Pressure Ulcer Management by Physicians: Randomized Crossover Pilot Study.” JMIR formative research (2026). PMID: 41871340 ↗
L1RCTCited in: Epidemiology, Etiology and Risk Factors, Prognosis and Natural History - [17]
Correa EL, Strobel R, Canciglieri Junior O et al.. “Prioritizing Clinically Relevant Criteria for Longitudinal Obesity Management: A Systemic Framework to Support Decision-Making.” Obesity surgery (2026). PMID: 41989748 ↗
L5SR_OBSCited in: Epidemiology, Etiology and Risk Factors - [18]
Ferede Z, Patterson S, Perera J et al.. “Risk Factors for Long-Term Health-Related Quality-of-Life and Mental Health Outcomes in Traumatic Brain Injury: A Systematic Review and Meta-Analysis.” Journal of neurotrauma (2026). PMID: 41873521 ↗
L2SR_OBSCited in: Epidemiology, Etiology and Risk Factors, Clinical Presentation - [19]
Makiev GG, Samoylenko IV, Nazarova VV et al.. “The Efficacy of Electronic Health Record-Based Artificial Intelligence Models for Early Detection of Pancreatic Cancer: A Systematic Review and Meta-Analysis.” Cancers (2026). PMID: 41595234 ↗
L2SR_OBSCited in: Epidemiology, Etiology and Risk Factors, Severity, Staging and Risk Stratification, Prevention, Screening and Health Maintenance - [20]
Holbrook AM, Silva JM, Faruque JAY et al.. “Effect of electronic drug-drug interaction alerts on patient and clinician outcomes: a systematic review.” Journal of the American Medical Informatics Association : JAMIA (2025). PMID: 40853269 ↗
L2SR_OBSCited in: Epidemiology, Etiology and Risk Factors, Multimorbidity, Polypharmacy & Deprescribing - [21]
Kafke SD, Kuhlmey A, Schuster J et al.. “Can clinical decision support systems be an asset in medical education? An experimental approach.” BMC medical education (2023). PMID: 37568144 ↗
L4RCTCited in: Clinical Presentation - [22]
Coulibaly A, Sogo AE, Chikvaidze N et al.. “Assessing the effectiveness of the Pregnancy And Newborn Diagnostic Assessment system on the quality of antenatal care in Burkina Faso: A cluster-randomised controlled trial.” Digital health (2024). PMID: 39664758 ↗
L1RCTCited in: Clinical Presentation, Special Populations and Pregnancy - [23]
Fu S, Li F, Qian SY. “Artificial intelligence in pediatric intensive care units: current applications in sepsis management.” World journal of pediatrics : WJP (2026). PMID: 42265523 ↗
L2SR_OBSCited in: Clinical Presentation, Diagnosis and Workup, Severity, Staging and Risk Stratification, Long-term and Definitive Management - [24]
Clement J, Maldonado AQ. “Augmenting the Transplant Team With Artificial Intelligence: Toward Meaningful AI Use in Solid Organ Transplant.” Frontiers in immunology (2021). PMID: 34177958 ↗
L5SR_OBSCited in: Clinical Presentation - [25]
Graber ML, Mathew A. “Performance of a web-based clinical diagnosis support system for internists.” Journal of general internal medicine (2008). PMID: 18095042 ↗
L4OTHERCited in: Clinical Presentation, Diagnosis and Workup, History and Evolution of Treatment - [26]
Szumilas D, Ochmann A, Zięba K et al.. “Evaluation of AI-Driven LabTest Checker for Diagnostic Accuracy and Safety: Prospective Cohort Study.” JMIR medical informatics (2024). PMID: 39149851 ↗
L2COHORTCited in: Clinical Presentation - [27]
Salama V, Youssef S, Xu T et al.. “Temporal characterization of acute pain and toxicity kinetics during radiation therapy for head and neck cancer. A retrospective study.” Oral oncology reports (2023). PMID: 38638130 ↗
L3COHORTCited in: Clinical Presentation - [28]
Robbins GK, Lester W, Johnson KL et al.. “Efficacy of a clinical decision-support system in an HIV practice: a randomized trial.” Annals of internal medicine (2012). PMID: 23208165 ↗
L1RCTCited in: Diagnosis and Workup, Long-term and Definitive Management, History and Evolution of Treatment - [29]
Roshanov PS, Fernandes N, Wilczynski JM et al.. “Features of effective computerised clinical decision support systems: meta-regression of 162 randomised trials.” BMJ (Clinical research ed.) (2013). PMID: 23412440 ↗
L1SR_OBSCited in: Diagnosis and Workup, Prognosis and Natural History - [30]
Kostopoulou O, Rosen A, Round T et al.. “Early diagnostic suggestions improve accuracy of GPs: a randomised controlled trial using computer-simulated patients.” The British journal of general practice : the journal of the Royal College of General Practitioners (2015). PMID: 25548316 ↗
L1RCTCited in: Diagnosis and Workup, Long-term and Definitive Management, History and Evolution of Treatment - [31]
Song J, Liu Y, Shang Q et al.. “Clinical Decision Support System, Antihypertensive Treatment Intensification, and Blood Pressure Control: A Post Hoc Secondary Analysis of a Cluster Randomized Trial.” JAMA network open (2026). PMID: 42201734 ↗
L2RCTCited in: Diagnosis and Workup, Long-term and Definitive Management, History and Evolution of Treatment - [32]
Agarwal S, Chin WY, Vasudevan L et al.. “Digital tracking, provider decision support systems, and targeted client communication via mobile devices to improve primary health care.” The Cochrane database of systematic reviews (2025). PMID: 40193137 ↗
L1SR_OBSCited in: Diagnosis and Workup, Long-term and Definitive Management, Prognosis and Natural History, Special Populations and Pregnancy, Prevention, Screening and Health Maintenance - [33]
Boudra T, Idrissou A, Barakat O et al.. “Machine Learning in HIV Care and Antiretroviral Therapy: Systematic Review.” Journal of medical Internet research (2026). PMID: 42048495 ↗
L2SR_OBSCited in: Diagnosis and Workup, Prevention, Screening and Health Maintenance - [34]
Wadie P, Zakher B, Elgazzar K et al.. “AI in Point-of-Care Imaging for Clinical Decision Support: Systematic Review of Diagnostic Accuracy, Task-Shifting, and Explainability.” JMIR AI (2026). PMID: 42044298 ↗
L2SR_OBSCited in: Diagnosis and Workup, Prognosis and Natural History - [35]
Robson J, Dostal I, Mathur R et al.. “Improving anticoagulation in atrial fibrillation: observational study in three primary care trusts.” The British journal of general practice : the journal of the Royal College of General Practitioners (2014). PMID: 24771841 ↗
L4OTHERCited in: Diagnosis and Workup, Severity, Staging and Risk Stratification, Long-term and Definitive Management, Prognosis and Natural History - [36]
Laiteerapong N, Ham SA, Ari M et al.. “A Quasi-Experimental Evaluation of a Primary Care Behavioral Health Integration Program Based on the Chronic Care Model.” Journal of general internal medicine (2025). PMID: 40481385 ↗
L2OTHERCited in: Diagnosis and Workup, Long-term and Definitive Management - [37]
Hu X, Su X, Chen S et al.. “Predictors of 30-day readmission in young and middle-aged T2DM patients: a retrospective cohort study.” Endocrine (2026). PMID: 42068438 ↗
L3COHORTCited in: Diagnosis and Workup, Long-term and Definitive Management, Prevention, Screening and Health Maintenance - [38]
Yang M, Jin J, Liu YL et al.. “Advancements of artificial intelligence in Chinese herbal medicine recommendation: A comprehensive review of data-driven approaches and clinical applications form 2016 to 2025.” Medicine (2026). PMID: 42065178 ↗
L2SR_OBSCited in: Diagnosis and Workup, Long-term and Definitive Management - [39]
Usiskin IM, Danila MI, Cai T et al.. “Toward Artificial Intelligence-driven Clinical Decision Support Tools in Rheumatology.” Rheumatic diseases clinics of North America (2026). PMID: 42409432 ↗
L5REVIEW_NARRATIVECited in: Diagnosis and Workup - [40]
Chen Z, Tang Z, Ewing RM et al.. “Artificial intelligence in respiratory medicine: From diagnosis to treatment and future directions.” Chinese medical journal pulmonary and critical care medicine (2026). PMID: 42396189 ↗
L5REVIEW_NARRATIVECited in: Diagnosis and Workup - [41]
McCarthy DM, VanWagner LB, Rafferty MR et al.. “Improving diagnostic safety through STeatosis Identification, Risk stratification, and Referral pathway in the ED (STIRRED): Protocol for an effectiveness implementation trial.” Contemporary clinical trials (2025). PMID: 41407098 ↗
L5TRIAL_NONRANDOMCited in: Severity, Staging and Risk Stratification, Acute Management, Generalist Reasoning under Diagnostic Uncertainty, Point-of-Care Scores & Referral Thresholds - [42]
Zhang F, Yun W, Qin L. “Development and Application of a Nurse-Led Clinical Decision Support System for Safe Intravenous Medication Administration: A Nonrandomized Controlled Trial.” Applied clinical informatics (2026). PMID: 42091056 ↗
L2RCTCited in: Severity, Staging and Risk Stratification - [43]
Rucco A, Shumba AT, Montanaro T et al.. “Enhancing Chronic Heart Failure Monitoring, Prevention, and Management With IoT and AI: A Systematic Literature Review.” IEEE journal of biomedical and health informatics (2026). PMID: 41182927 ↗
L2SR_OBSCited in: Severity, Staging and Risk Stratification - [44]
Jang J, Kim WJ, Park SW et al.. “Development of explainable machine learning models to predict side effects in patients with rheumatoid arthritis taking methotrexate treatment: a nationwide multicentre cohort study.” BMJ open (2025). PMID: 41320203 ↗
L2COHORTCited in: Severity, Staging and Risk Stratification - [45]
Naderian S, Soleimanzadeh F, Nikniaz L et al.. “A Systematic Review of Artificial Intelligence-Based Clinical Decision Support Systems in Prostate Cancer Management.” Healthcare technology letters (2025). PMID: 41262209 ↗
L2SR_OBSCited in: Severity, Staging and Risk Stratification - [46]
Onwuharine EN, Clark AJ, McIntyre A et al.. “Clinical and MRI variables in decision support systems for prostate MRI: A systematic review of decision support tools, nomograms, and risk models.” Radiography (London, England : 1995) (2025). PMID: 41202669 ↗
L2SR_OBSCited in: Severity, Staging and Risk Stratification - [47]
Bonura A, Musotto G, Rossi SS et al.. “Cerebri: A Web-App to Reduce Door-to-Treatment Decision Time and Improve Guideline Adherence in Acute Ischemic Stroke.” Stroke (2025). PMID: 41208703 ↗
L1RCTCited in: Acute Management, History and Evolution of Treatment - [48]
Ciapponi A, Fernandez Nievas SE, Seijo M et al.. “Reducing medication errors for adults in hospital settings.” The Cochrane database of systematic reviews (2021). PMID: 34822165 ↗
L1SR_OBSCited in: Acute Management, Multimorbidity, Polypharmacy & Deprescribing, Prognosis and Natural History - [49]
Rangraz Jeddi F, Nabovati E, Sharif R et al.. “The effects of information technology interventions for optimizing antibiotic prescribing in urinary tract infections: a systematic review.” BMC infectious diseases (2025). PMID: 41462120 ↗
L2SR_OBSCited in: Acute Management - [50]
Beauchamp FO, Thériault J, Emeriaud G et al.. “Lung Recruitment Maneuvers During Invasive Mechanical Ventilation: Single-Center Retrospective PICU Cohort Study, 2016-2023.” Pediatric critical care medicine : a journal of the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies (2026). PMID: 42383786 ↗
L3COHORTCited in: Acute Management - [51]
Skryabin V, Masyakin A, Pozdniakov S et al.. “Personalizing Treatment for Acute Alcoholic Hallucinosis: Clinical Utility of Integrated Pharmacogenetic Testing, Metabolic Phenotyping, and microRNA Biomarkers.” Psychopharmacology bulletin (2026). PMID: 41532000 ↗
L4RCTCited in: Acute Management - [52]
Xu Z, Zhang R, Han Q et al.. “Explainable Machine Learning Models Using Routine Clinical Data for Early Prediction of Necrotizing Fasciitis: A Single-Center Retrospective Study.” ANZ journal of surgery (2025). PMID: 41277728 ↗
L3COHORTCited in: Acute Management - [53]
Bruckelt L, Neumann J, Keß A et al.. “Clinical evaluation of a computer-assisted decision support and documentation system for the primary care of polytrauma patients.” Frontiers in digital health (2026). PMID: 42434373 ↗
L4OTHERCited in: Acute Management, Long-term and Definitive Management - [54]
Bright TJ, Wong A, Dhurjati R et al.. “Effect of clinical decision-support systems: a systematic review.” Annals of internal medicine (2012). PMID: 22751758 ↗
L1SR_OBSCited in: Long-term and Definitive Management, Prognosis and Natural History - [55]
Roumeliotis N, Sniderman J, Adams-Webber T et al.. “Effect of Electronic Prescribing Strategies on Medication Error and Harm in Hospital: a Systematic Review and Meta-analysis.” Journal of general internal medicine (2019). PMID: 31396810 ↗
L1SR_OBSCited in: Long-term and Definitive Management, Prognosis and Natural History - [56]
Fan E, Laupacis A, Pronovost PJ et al.. “How to use an article about quality improvement.” JAMA (2010). PMID: 21098772 ↗
L5OTHERCited in: Long-term and Definitive Management - [57]
Basten J, Köberlein-Neu J, Ihle P et al.. “Patterns of CDSS adoption in primary care: a cluster analysis and predictive modelling study from a stepped wedge trial.” Implementation science : IS (2026). PMID: 42316212 ↗
L1RCTCited in: Long-term and Definitive Management, History and Evolution of Treatment, Prognosis and Natural History - [58]
Agarwal S, Glenton C, Tamrat T et al.. “Decision-support tools via mobile devices to improve quality of care in primary healthcare settings.” The Cochrane database of systematic reviews (2021). PMID: 34314020 ↗
L1SR_OBSCited in: Long-term and Definitive Management - [59]
Mandrina M, Gevorkyan T, Zvezda S et al.. “Gastric cancer survival prediction using artificial intelligence models based on electronic health records: a systematic review and meta-analysis.” Frontiers in digital health (2026). PMID: 42416807 ↗
L2SR_OBSCited in: Long-term and Definitive Management, Prognosis and Natural History - [60]
Cortés Sánchez CJ, Salazar González F, Gómez Portolés JM et al.. “The role of digital tools and artificial intelligence in supporting antimicrobial stewardship: a systematic review.” The Journal of antimicrobial chemotherapy (2026). PMID: 42313419 ↗
L2SR_OBSCited in: Long-term and Definitive Management - [61]
Gerber F, Gupta R, Sanchez-Samaniego G et al.. “Community health worker-led versus facility-based type 2 diabetes care in rural Lesotho: a cluster-randomized trial within the ComBaCaL cohort study.” BMC medicine (2026). PMID: 42174613 ↗
L1COHORTCited in: Long-term and Definitive Management, Prevention, Screening and Health Maintenance - [62]
Robson J, Hull S, Mathur R et al.. “Improving cardiovascular disease using managed networks in general practice: an observational study in inner London.” The British journal of general practice : the journal of the Royal College of General Practitioners (2014). PMID: 24771840 ↗
L2OTHERCited in: Long-term and Definitive Management, Prognosis and Natural History - [63]
Aksu BN, Gashi B, Schiefenhövel F et al.. “Effectiveness of Guideline-Based Clinical Decision Support Systems: Protocol for a Systematic Review.” JMIR research protocols (2026). PMID: 42275446 ↗
L5SR_OBSCited in: Long-term and Definitive Management - [64]
Dillen H, Dankaerts A, Snijders D et al.. “Physicians' preferences for the use of clinical decision support systems in the context of acutely ill children presenting to ambulatory care: a focus group study.” BMC medical informatics and decision making (2026). PMID: 42410436 ↗
L5OTHERCited in: Long-term and Definitive Management - [65]
Vasey B, Nagendran M, Campbell B et al.. “Reporting guideline for the early stage clinical evaluation of decision support systems driven by artificial intelligence: DECIDE-AI.” BMJ (Clinical research ed.) (2022). PMID: 35584845 ↗
L1GUIDELINECited in: History and Evolution of Treatment - [66]
Bakker T, Klopotowska JE, Dongelmans DA et al.. “The effect of computerised decision support alerts tailored to intensive care on the administration of high-risk drug combinations, and their monitoring: a cluster randomised stepped-wedge trial.” Lancet (London, England) (2024). PMID: 38262430 ↗
L1RCTCited in: History and Evolution of Treatment, Multimorbidity, Polypharmacy & Deprescribing, Prognosis and Natural History - [67]
Samore MH, Bateman K, Alder SC et al.. “Clinical decision support and appropriateness of antimicrobial prescribing: a randomized trial.” JAMA (2005). PMID: 16278358 ↗
L1RCTCited in: History and Evolution of Treatment - [68]
Rousseau N, McColl E, Newton J et al.. “Practice based, longitudinal, qualitative interview study of computerised evidence based guidelines in primary care.” BMJ (Clinical research ed.) (2003). PMID: 12574046 ↗
L5RCTCited in: History and Evolution of Treatment - [69]
ShojaeiBaghini M, Ghaemi MM, Ahmadipour A. “Artificial intelligence in the identification and prediction of adverse transfusion reactions(ATRs) and implications for clinical management: a systematic review of models and applications.” BMC medical informatics and decision making (2025). PMID: 41152861 ↗
L2SR_OBSCited in: History and Evolution of Treatment, Complications - [70]
Tun HM, Rahman HA, Naing L et al.. “Trust in Artificial Intelligence-Based Clinical Decision Support Systems Among Health Care Workers: Systematic Review.” Journal of medical Internet research (2025). PMID: 40772775 ↗
L2SR_OBSCited in: History and Evolution of Treatment - [71]
Rezende VMLR, Borges IN, Ravetti CG et al.. “Efficacy and safety of a digital clinical decision support system using a C-reactive protein-based algorithm and evidence-based stopping rules to guide antibiotic therapy duration: A single-center, open-label randomized controlled trial in a tertiary care hospital.” International journal of infectious diseases : IJID : official publication of the International Society for Infectious Diseases (2026). PMID: 41990873 ↗
L1RCTCited in: History and Evolution of Treatment, Prognosis and Natural History - [72]
Roemer A, Schlicker N, Kernder A et al.. “Large language models enhance diagnostic reasoning of medical students in rheumatology: a randomized controlled trial.” BMC medical education (2026). PMID: 41877128 ↗
L1RCTCited in: History and Evolution of Treatment, Prognosis and Natural History - [73]
Bolton WJ, Wilson R, Gilchrist M et al.. “The impact of artificial intelligence-driven decision support on uncertain antimicrobial prescribing: a randomised, multimethod study.” The Lancet. Digital health (2025). PMID: 41372053 ↗
L1RCTCited in: History and Evolution of Treatment, Generalist Reasoning under Diagnostic Uncertainty, Point-of-Care Scores & Referral Thresholds - [74]
Dijk SW, Wollny C, Kroencke T et al.. “Sex differences in inappropriate imaging requests: insights from the Medical Imaging Decision And Support (MIDAS) study.” European radiology (2025). PMID: 41196364 ↗
L4RCTCited in: History and Evolution of Treatment - [75]
Atlas SJ, Burdick TE, Wright A et al.. “Comparing clinical decision support systems for improving follow-up of abnormal cervical cancer screening test results.” Journal of biomedical informatics (2025). PMID: 40935221 ↗
L2RCTCited in: History and Evolution of Treatment - [76]
Gimbel RW, Pirrallo RG, Lowe SC et al.. “Effect of clinical decision rules, patient cost and malpractice information on clinician brain CT image ordering: a randomized controlled trial.” BMC medical informatics and decision making (2018). PMID: 29530029 ↗
L1RCTCited in: Generalist Reasoning under Diagnostic Uncertainty, Point-of-Care Scores & Referral Thresholds - [77]
Carroll AE, Biondich P, Anand V et al.. “A randomized controlled trial of screening for maternal depression with a clinical decision support system.” Journal of the American Medical Informatics Association : JAMIA (2012). PMID: 22744960 ↗
L1RCTCited in: Generalist Reasoning under Diagnostic Uncertainty, Point-of-Care Scores & Referral Thresholds - [78]
Lester C, Rowell B, Zheng Y et al.. “Effect of Uncertainty-Aware AI Models on Pharmacists' Reaction Time and Decision-Making in a Web-Based Mock Medication Verification Task: Randomized Controlled Trial.” JMIR medical informatics (2025). PMID: 40249341 ↗
L1RCTCited in: Generalist Reasoning under Diagnostic Uncertainty, Point-of-Care Scores & Referral Thresholds - [79]
Born C, Schwarz R, Böttcher TP et al.. “The role of information systems in emergency department decision-making-a literature review.” Journal of the American Medical Informatics Association : JAMIA (2024). PMID: 38781289 ↗
L2SR_OBSCited in: Generalist Reasoning under Diagnostic Uncertainty, Point-of-Care Scores & Referral Thresholds - [80]
Rosati A, Criscione M, Lilli L et al.. “Natural language processing as consultation service platform or clinical decision support system in gynecologic oncology: a systematic review.” International journal of gynecological cancer : official journal of the International Gynecological Cancer Society (2026). PMID: 41819636 ↗
L2SR_OBSCited in: Generalist Reasoning under Diagnostic Uncertainty, Point-of-Care Scores & Referral Thresholds - [81]
Alldred DP, Raynor DK, Hughes C et al.. “Interventions to optimise prescribing for older people in care homes.” The Cochrane database of systematic reviews (2013). PMID: 23450597 ↗
L1SR_OBSCited in: Multimorbidity, Polypharmacy & Deprescribing, Prognosis and Natural History - [82]
Alldred DP, Kennedy MC, Hughes C et al.. “Interventions to optimise prescribing for older people in care homes.” The Cochrane database of systematic reviews (2016). PMID: 26866421 ↗
L1SR_OBSCited in: Multimorbidity, Polypharmacy & Deprescribing, Prognosis and Natural History - [83]
van Hylckama Vlieg MAM, Pot IE, Visser HPJ et al.. “Appropriate medication use in Dutch terminal care: study protocol of a multicentre stepped-wedge cluster randomized controlled trial (the AMUSE study).” BMC palliative care (2024). PMID: 38172930 ↗
L5TRIAL_NONRANDOMCited in: Multimorbidity, Polypharmacy & Deprescribing - [84]
Vamadevan A, Vijayan V, Cole C et al.. “A NASSS framework-guided systematic review and exploratory modelling of digital health interventions for polypharmacy management in older adults.” BMC geriatrics (2025). PMID: 41272510 ↗
L2SR_OBSCited in: Multimorbidity, Polypharmacy & Deprescribing - [85]
Ng Y, Hsu JTY, Ng NNE et al.. “Evaluating the role of clinical decision support systems in medication safety for older people: a systematic review.” Age and ageing (2025). PMID: 40716041 ↗
L1SR_OBSCited in: Multimorbidity, Polypharmacy & Deprescribing - [86]
Skains RM, Hayes JM, Selman K et al.. “Emergency Department Programs to Support Medication Safety in Older Adults: A Systematic Review and Meta-Analysis.” JAMA network open (2025). PMID: 40067297 ↗
L1SR_OBSCited in: Multimorbidity, Polypharmacy & Deprescribing - [87]
Tlili NE, Robert L, Gerard E et al.. “A systematic review of the value of clinical decision support systems in the prescription of antidiabetic drugs.” International journal of medical informatics (2024). PMID: 39106772 ↗
L2SR_OBSCited in: Multimorbidity, Polypharmacy & Deprescribing, Complications - [88]
Chen W, O'Bryan CM, Gorham G et al.. “Barriers and enablers to implementing and using clinical decision support systems for chronic diseases: a qualitative systematic review and meta-aggregation.” Implementation science communications (2022). PMID: 35902894 ↗
L5SR_OBSCited in: Multimorbidity, Polypharmacy & Deprescribing - [89]
Bhatt M, Benterud E, Palechuk T et al.. “Advancing Community Care and Access to Follow-up After Acute Kidney Injury Hospitalization: Design of the AFTER AKI Randomized Controlled Trial.” Canadian journal of kidney health and disease (2024). PMID: 38495365 ↗
L5RCTCited in: Complications - [90]
Cai J, Li P, Li W et al.. “Outcomes of clinical decision support systems in real-world perioperative care: a systematic review and meta-analysis.” International journal of surgery (London, England) (2024). PMID: 39037722 ↗
L1SR_OBSCited in: Complications - [91]
Murzabekov M, Lidströmer N, Borg N et al.. “The association between vital signs at hospital admission and adverse outcomes in patients with COVID-19: a retrospective cohort study.” Frontiers in medicine (2025). PMID: 40678139 ↗
L2COHORTCited in: Complications - [92]
Palomares SM, Ferrara G, Sguanci M et al.. “The Impact of Artificial Intelligence Technologies on Nutritional Care in Patients With Chronic Kidney Disease: A Systematic Review.” Journal of renal nutrition : the official journal of the Council on Renal Nutrition of the National Kidney Foundation (2025). PMID: 40588047 ↗
L2SR_OBSCited in: Complications - [93]
Wang J, Wu Y, Huang Y et al.. “Comparative effectiveness of delirium recognition with and without a clinical decision assessment system on outcomes of hospitalized older adults: Cluster randomized controlled trial.” International journal of nursing studies (2024). PMID: 39700738 ↗
L1RCTCited in: Complications - [94]
Patel H, Crusco S, Hansen D et al.. “A Systematic Review and Meta-analysis of Externally Validated Epic Clinical Decision Support Tools.” Journal of general internal medicine (2026). PMID: 41917292 ↗
L1SR_OBSCited in: Prognosis and Natural History - [95]
Muñoz J, Fernández-Araujo NJ, Ruíz-Cacho R et al.. “Artificial intelligence and computerized decision support in adult intensive care: A systematic review of randomized controlled trials.” Journal of critical care (2026). PMID: 42048766 ↗
L1SR_MA_RCTCited in: Prognosis and Natural History - [96]
Zhang T, Shin N. “From interface to outcome: a 4I framework for AI-linked functionality in electronic health records.” BMC medical informatics and decision making (2026). PMID: 42174575 ↗
L5SR_OBSCited in: Prognosis and Natural History - [97]
Almutairi H, Stafford A, Etherton-Beer C et al.. “Optimisation of medications used in residential aged care facilities: a systematic review and meta-analysis of randomised controlled trials.” BMC geriatrics (2020). PMID: 32641005 ↗
L1SR_MA_RCTCited in: Special Populations and Pregnancy - [98]
Marko B, Palmowski L, Nowak H et al.. “Employing artificial intelligence for optimising antibiotic dosages in sepsis on intensive care unit: a study protocol for a prospective observational study (KI.SEP).” BMJ open (2024). PMID: 39672586 ↗
L5TRIAL_NONRANDOMCited in: Special Populations and Pregnancy - [99]
Crowley EK, Sallevelt BTGM, Huibers CJA et al.. “Intervention protocol: OPtimising thERapy to prevent avoidable hospital Admission in the Multi-morbid elderly (OPERAM): a structured medication review with support of a computerised decision support system.” BMC health services research (2020). PMID: 32183810 ↗
L5TRIAL_NONRANDOMCited in: Special Populations and Pregnancy - [100]
Carter J, Anumba D, Burden C et al.. “Tommy's Clinical Decision Support Tool: an intervention development and feasibility study to inform a future randomised controlled trial.” Pilot and feasibility studies (2026). PMID: 41742257 ↗
L4RCTCited in: Special Populations and Pregnancy - [101]
Drenth-van Maanen AC, Leendertse AJ, Jansen PAF et al.. “The Systematic Tool to Reduce Inappropriate Prescribing (STRIP): Combining implicit and explicit prescribing tools to improve appropriate prescribing.” Journal of evaluation in clinical practice (2017). PMID: 28776873 ↗
L5RCTCited in: Special Populations and Pregnancy - [102]
Kamaruzaman LH, Abdul Rahman H, Mohd Nor N'U et al.. “Integrating aqli (rationale) and naqli (revealed) knowledge in AI-driven clinical decision support systems to enhance healthcare delivery for guiding termination of pregnancy among Muslim patients: a systematic review (1).” BMC medical informatics and decision making (2026). PMID: 41742117 ↗
L5SR_OBSCited in: Special Populations and Pregnancy - [103]
Badisy IE, Assarag B, Belrhiti Z. “Interpretable clinical decision support systems in high-risk pregnancy: a scoping review of models, methods, and implementation.” BMC pregnancy and childbirth (2026). PMID: 41495684 ↗
L5SR_OBSCited in: Special Populations and Pregnancy, Prevention, Screening and Health Maintenance - [104]
Derksen C, Le Noury KA, Akbar AB et al.. “Why aren't they used? Systematic review of barriers to implementation of clinical decision support systems for early cancer detection in primary care.” The British journal of general practice : the journal of the Royal College of General Practitioners (2026). PMID: 41494775 ↗
L5SR_OBSCited in: Prevention, Screening and Health Maintenance - [105]
Wright AP, Choi L, Nairon KG et al.. “Interruptive versus Non-Interruptive Reminders for Statin tHerApy in Primary Care (INIRSHA-PC): protocol and statistical analysis plan for a randomised clinical trial.” BMJ open (2026). PMID: 41605591 ↗
L5TRIAL_NONRANDOMCited in: Prevention, Screening and Health Maintenance - [106]
Giordano L, Durand C, Ahmad R et al.. “Strategies and outcomes of CDSS implementation for antimicrobial stewardship in hospital settings: a systematic review.” Antimicrobial resistance and infection control (2026). PMID: 41992388 ↗
L5SR_OBSCited in: Prevention, Screening and Health Maintenance - [107]
Jain G, Bodade A, Pati S. “Impact of clinical decision support systems (CDSS) on clinical outcomes and healthcare delivery in low- and middle-income countries: protocol for a systematic review and meta-analysis.” BMJ open (2025). PMID: 41360443 ↗
L5SR_OBSCited in: Prevention, Screening and Health Maintenance - [108]
Heo Y, Kim M, Han SS et al.. “AI-Driven Predictions of Readmission and Mortality for Improved Discharge Decisions in Critical Care: A Retrospective Study.” Diagnostics (Basel, Switzerland) (2026). PMID: 41897607 ↗
L3COHORTCited in: Prevention, Screening and Health Maintenance