Clinical decision support systems (CDSS) are increasingly recognized as vital tools in modern healthcare, designed to assist healthcare providers in making informed decisions and improving patient outcomes. Among their many benefits, a key advantage is their ability to significantly enhance adherence to clinical guidelines, ultimately leading to improved patient care by providers. This article explores how Decision Support Tools Increase Guidelines Care By Providers, examining their various applications and impacts across different areas of healthcare.
Patient safety remains a paramount concern in healthcare settings. Strategies aimed at reducing medical errors frequently leverage CDSS, as highlighted in studies and summarized in resources like Table 1 (referencing original article’s Table 1 for context). Medication errors, particularly those involving drug-drug interactions (DDIs), are a common and preventable issue. Research indicates that a substantial percentage of hospitalized patients may experience potentially harmful drug combinations. Modern Computerized Provider Order Entry (CPOE) systems are now equipped with sophisticated drug safety software. These systems incorporate safeguards to check for appropriate dosing, duplication of therapies, and potential DDIs. The alerts generated by these systems represent a widespread form of decision support in clinical practice.
Alt text: Table summarizing the benefits, potential harms, and evidence-based mitigation strategies associated with clinical decision support systems (CDSS) in healthcare.
However, the effectiveness of DDI alerts can be variable. Studies have observed inconsistencies in how these alerts are presented to providers, ranging from passive notifications to disruptive interventions. Prioritization of alerts and the underlying algorithms used to detect DDIs also differ across systems. This variability can lead to a significant number of irrelevant alerts, contributing to alert fatigue among clinicians and hindering the intended benefits of CDSS. Currently, there is a lack of standardization regarding the optimal implementation and prioritization of alerts for healthcare providers. Despite these challenges, efforts are underway to improve standardization. The US Office of the National Coordinator for Health Information Technology, for instance, has compiled a list of ‘high-priority’ DDIs for CDS. This list has gained traction and is being adopted and implemented in CDSS across various countries, including the UK, Belgium, and Korea, demonstrating a global move towards more consistent and effective decision support for medication safety.
Beyond DDIs, other systems contribute to patient safety, such as electronic drug dispensing systems (EDDS) and bar-code point-of-care (BPOC) medication administration systems. Often, these technologies are integrated to create a ‘closed-loop’ medication management process. In this system, each stage, from prescribing to administration, is computerized and interconnected. At the point of administration, medications are automatically identified using RFID or barcodes and verified against patient information and prescriptions. This presents another crucial area where CDSS can enhance safety by preventing medication administration errors at the bedside, complementing upstream safety measures. While adoption rates for these closed-loop systems are still growing, partly due to technological requirements and implementation costs, research indicates their effectiveness in reducing medication errors. Studies show that integrating BPOC with CPOE and CDSS further minimizes prescribing errors related to drug allergies, excessive dosing, and incomplete orders. It is crucial to acknowledge that even with advanced CDSS, errors can still occur if providers bypass or deliberately circumvent the technology, highlighting the importance of user training and workflow integration.
CDSS also plays a vital role in improving patient safety beyond medication management. Reminder systems within CDSS are used for various medical events. For example, a CDSS designed for blood glucose monitoring in intensive care units (ICUs) has proven successful in decreasing hypoglycemia events. This system automatically prompts nurses to perform glucose measurements based on predefined protocols, taking into account patient demographics and glucose level trends. This proactive approach ensures timely monitoring and intervention, directly enhancing patient safety.
Overall, CDSS aimed at enhancing patient safety, particularly through CPOE and related systems, have generally been successful in reducing prescribing and dosing errors, mitigating contraindications through automated warnings, and improving drug-event monitoring. Patient safety can be considered an inherent objective of nearly all types of CDSS, regardless of their primary intended function, underscoring the pervasive impact of decision support tools on improving healthcare quality.
Enhancing Clinical Management through Guideline Adherence
Studies consistently demonstrate that CDSS significantly improve adherence to clinical guidelines. This is a critical finding because traditional clinical guidelines and care pathways, despite their evidence-based nature, have historically faced challenges in practical implementation due to low clinician adherence rates. The assumption that healthcare practitioners will readily adopt and implement new guidelines simply by reading and internalizing them has not proven accurate in practice. However, the structured rules and recommendations embedded within clinical guidelines can be directly translated and encoded into CDSS.
These CDSS designed to promote guideline adherence can take various forms tailored to specific clinical needs. They can include standardized order sets for particular conditions, alerts triggered by patient-specific data indicating the need for a specific protocol, and reminders for necessary tests or follow-up actions. Furthermore, CDSS facilitate the management of patients enrolled in research or treatment protocols, streamlining tasks such as tracking progress, placing orders, scheduling follow-up appointments, and ensuring consistent preventative care measures are implemented.
CDSS can proactively alert clinicians to reach out to patients who are not adhering to their management plans or are due for scheduled follow-up appointments. They can also assist in identifying patients who meet specific criteria for research studies, streamlining the recruitment process. A notable example is a CDSS implemented at the Cleveland Clinic, which provides point-of-care alerts to physicians when a patient’s electronic health record (EHR) data matches the inclusion criteria for an ongoing clinical trial. This alert prompts the physician to complete a form to assess eligibility and obtain consent for contact, subsequently forwarding the patient’s chart to the study coordinator and generating a patient information sheet about the clinical trial. This proactive approach significantly enhances clinical trial recruitment and ensures patients are offered relevant research opportunities.
Cost Containment and Efficiency
The implementation of CDSS can lead to significant cost savings within healthcare systems. This cost-effectiveness is achieved through various mechanisms, including optimizing clinical interventions, reducing inpatient length-of-stay, suggesting more cost-effective medication alternatives integrated within CPOE systems, and minimizing unnecessary test duplication. For instance, a CPOE rule implemented in a pediatric cardiovascular ICU successfully limited the ordering of complete blood counts, chemistry panels, and coagulation panels to a 24-hour interval. This seemingly simple intervention resulted in a substantial reduction in laboratory resource utilization, projecting annual cost savings without negatively impacting patient length of stay or mortality rates.
CDSS can also provide real-time notifications to users about more affordable drug alternatives or highlight conditions that are covered by insurance, guiding providers towards cost-conscious prescribing practices. In Germany, where transitioning inpatients to drugs within hospital formularies is common, a CDSS was developed to address errors in drug substitutions. After identifying a significant error rate in manual substitutions, Heidelberg hospital implemented a drug-switch algorithm integrated into their CPOE system. This CDSS automated a large percentage of drug substitution consultations accurately and safely, reducing errors, minimizing provider workload, and ultimately lowering costs for healthcare providers and patients.
Streamlining Administrative Functions
Beyond direct patient care, CDSS offer valuable support for various administrative tasks within healthcare organizations. These include assisting with clinical and diagnostic coding, facilitating the ordering of procedures and tests, and optimizing patient triage processes. Sophisticated algorithms within CDSS can suggest refined lists of diagnostic codes, aiding physicians in selecting the most accurate and appropriate codes for billing and reporting purposes. To address inaccuracies in emergency department (ED) admission coding using ICD-9 (International Statistical Classification of Diseases codes), a CDSS was developed incorporating an anatomographical interface. This visual, interactive tool linked anatomical representations to ICD codes, enabling ED physicians to more quickly and accurately identify diagnostic admission codes.
CDSS also contribute to improving the quality and completeness of clinical documentation. For example, an obstetric CDSS with an enhanced prompting system significantly improved the documentation of indications for labor induction and estimated fetal weight compared to control hospitals. Accurate documentation is crucial as it directly informs clinical protocols and patient safety measures. Consider the example of vaccination protocols following splenectomy (spleen removal). A CDSS was implemented to ensure patients received appropriate vaccinations to mitigate the increased risk of infections associated with splenectomy. However, it was discovered that a significant proportion of patients with splenectomy in their EHR did not have it documented on their problem list, which was the trigger for the CDSS alert. To address this, a supplemental CDSS was developed to enhance problem list documentation of splenectomy, thereby improving the effectiveness of the original vaccination CDSS and ensuring patients received necessary preventative care.
Advancing Diagnostics Support
CDSS designed specifically for clinical diagnosis are known as diagnostic decision support systems (DDSS). Historically, DDSS functioned as computerized ‘consultants,’ processing patient data and user inputs to generate lists of potential or probable diagnoses. However, the adoption and impact of DDSS have lagged behind other types of CDSS due to factors such as physician skepticism, perceived inaccuracies, and challenges in system integration requiring manual data entry. These limitations are gradually being addressed through improved EHR integration and the adoption of standardized medical vocabularies like Snomed Clinical Terms, paving the way for more effective and user-friendly DDSS.
A successful example of a DDSS is a system developed for diagnosing peripheral neuropathy using fuzzy logic. This system, utilizing 24 input fields encompassing symptoms and diagnostic test results, achieved a high degree of accuracy in identifying different types of neuropathies and normal cases when compared to expert clinicians. This type of DDSS is particularly valuable in regions with limited access to specialist expertise. Furthermore, systems like DXplain serve as electronic references, providing probable diagnoses based on clinical manifestations. In a randomized controlled trial, family medicine residents using DXplain demonstrated significantly improved diagnostic accuracy compared to those without access to the system.
Given the recognized prevalence of diagnostic errors, especially in primary care settings, there is considerable optimism about the potential of CDSS and IT solutions to enhance diagnostic accuracy and efficiency. The emergence of non-knowledge-based techniques, such as machine learning, in diagnostic system development holds promise for even more accurate diagnostic capabilities. The Babylon AI-powered Triage and Diagnostic System in the UK represents the potential of AI in diagnostics, while also highlighting the ongoing development and refinement needed for these systems to achieve widespread clinical readiness.
Imaging and Precision Radiology
Knowledge-based imaging CDSS primarily focus on optimizing image ordering practices. These systems assist radiologists in selecting the most appropriate imaging tests, providing reminders of best practice guidelines, and alerting providers to contraindications, such as those related to contrast agents. An interventional CDS for image ordering at Virginia Mason Medical Center demonstrated substantial reductions in the utilization rates of lumbar MRI for low back pain, head MRI for headache, and sinus CT for sinusitis. This CDS required providers to answer a series of appropriateness questions prior to ordering images. Crucially, if an image was deemed inappropriate, the system suggested alternative imaging options. RadWise®, a commercial imaging CDS, guides clinicians to the most relevant imaging order by analyzing patient symptoms and matching them with a comprehensive database of diagnoses, providing point-of-care recommendations for appropriate imaging utilization.
Non-knowledge-based CDS, particularly those leveraging AI, are gaining significant traction in enhanced imaging and precision radiology, often referred to as ‘radiomics’. As medical imaging data volumes grow exponentially, requiring extensive manual interpretation, providers need advanced technologies to assist in data extraction, visualization, and interpretation. AI technologies, particularly deep learning (DL), are proving capable of extracting insights from image data that surpass human capabilities. Companies like IBM Watson Health, DeepMind, and Google are at the forefront of developing AI-powered tools for tumor detection, medical image interpretation, diabetic retinopathy diagnosis, Alzheimer’s diagnosis using multimodal feature learning, and numerous other applications. IBM Watson’s ‘Eyes of Watson’ exemplifies this trend, combining image recognition of brain scans with text recognition of case descriptions to provide comprehensive decision support, acting as a ‘cognitive assistant’ for radiologists.
Several AI-driven imaging systems have demonstrated performance that is comparable to or even surpasses that of human experts in specific tasks. For example, Google’s deep convolutional neural network (CNN) for detecting diabetic retinopathy achieved sensitivity and specificity on par with US board-certified ophthalmologists when trained on a large dataset of retinal images. Similarly, a CNN developed by Stanford researchers for detecting arrhythmias on electrocardiograms outperformed the average cardiologist in accuracy across various rhythm classes. While some experts speculate that computers may perform the majority of diagnostic imaging interpretation in the future, currently, these AI-powered systems should be viewed as valuable additions and augmentations to the clinician’s toolkit, enhancing their diagnostic capabilities rather than replacing them entirely.
Laboratory and Pathology Diagnostics
CDSS are also valuable in laboratory testing and interpretation. Alerts for abnormal lab results are already common in EHR systems. CDSS can extend the utility of lab tests to reduce the need for riskier or more invasive diagnostic procedures. In Hepatitis B and C testing, where liver biopsies are the traditional gold standard, AI models are being developed to combine multiple non-invasive tests (serum markers, imaging, gene tests) to achieve significantly improved diagnostic accuracy. CDSS also serve as interpretation tools, particularly when test reference ranges are highly personalized based on factors like age, sex, or disease subtypes.
Pathology reports are critical decision points in many medical specialties. CDSS are being utilized for automated tumor grading in pathology. For example, systems have been developed for urinary bladder tumor grading and recurrence prediction with high accuracy. Similar systems are being applied to brain tumor classification and grading. Numerous other applications exist, including computerized ECG analysis, automated arterial blood gas interpretation, protein electrophoresis reports, and CDSS for blood cell counting, demonstrating the broad utility of CDSS in enhancing laboratory and pathology diagnostics.
Empowering Patients with Decision Support
The rise of Personal Health Records (PHRs) has extended CDSS functionality to patients, empowering them as active participants in their own care. This patient-facing decision support aligns with the growing emphasis on patient-focused care and shared decision-making between patients and providers. CDSS-supported PHRs address the barrier of ‘lack of information’ that can hinder patient engagement in their healthcare decisions. PHRs are often integrated with commercial EHR software or exist as standalone web-based or mobile applications. When linked to EHRs, PHRs facilitate a bidirectional flow of information, allowing patients to view their EHR data and contribute information that can be accessed by their providers.
Early PHRs, like the “Patient Gateway,” primarily served as dashboards for patients to view medications, lab results, and communicate with their physicians. More advanced systems now enable patients to modify their own health records, with changes potentially affecting the EHR data. Vanderbilt University’s MyHealthAtVanderbilt PHR exemplifies this integration, offering disease-specific educational materials and tools like a ‘Flu Tool’ to help patients with flu-like symptoms determine the appropriate level of care and seek treatment. Symptom tracking and management are common features of PHRs, and the scope of data collected is expanding to include allergies, insurance information, and medication details. Furthermore, PHRs and patient monitoring applications are increasingly designed to integrate data from wearable health devices, generating actionable insights for providers.
Diabetes care provides a compelling example of patient-facing CDSS. Systems are being developed and implemented that utilize wearable glucose monitors to transmit data to patient devices and integrate with EHRs. The Stanford School of Medicine’s pilot program, using Apple HealthKit and Epic’s MyChart, allows providers to remotely monitor glucose trends in patients with diabetes, facilitating timely interventions and improved communication. This integration of PHRs, wearable technology, and CDSS is being extended to other medical fields, including cardiology for heart failure management, hypertension, sleep apnea, and palliative care, demonstrating the transformative potential of patient-facing decision support.
As PHRs evolve with advanced CDSS capabilities, the focus is shifting towards designing these systems as interactive tools that promote shared decision-making and enhance patient knowledge and involvement in their care. PHRs that function solely as repositories of health information are increasingly seen as inadequate, particularly by patients seeking active engagement in their healthcare journey. The future of CDSS lies in empowering both providers and patients with the information and tools needed to make the best possible healthcare decisions, ultimately leading to improved outcomes and a more patient-centered healthcare system.
