Clinical Decision Support

Systems (CDSS)
Ajay Vamadevan
PGDM – HCM | Term 3
Outline
• Introduction
• Classification
• Components
• Applications
• Steps in Developing CDSS
• Challenges
• Examples from India
Clinical Decision Support System (CDSS)

• For improving healthcare delivery by enhancing medical decisions with

targeted clinical knowledge, patient information, & other health info

• Primarily used at the point-of-care, for the clinician to combine/support

their knowledge with the information provided by the CDSS

• Current Trend: Aims to leverage data and observations otherwise

unobtainable or uninterpretable by humans
CDSS
• Historically Paper-based tools
• Charts
• Calculators
• Clinical Protocols
CDSS
• Can manifest as: • Assist in:
• Computerized guidelines • Clinical decision-making: Diagnostics,

• Computerized alerts & reminders Disease management, Prescription (Rx)

• Patient data reports • Alarm systems, Reduce medical errors

• Documentation templates • Support clinical coding

• Order sets • Authorize procedures/referral

• Clinical workflow tools • Monitor medication/test orders

CDSS: Uptake
• CDSS have been endorsed by the US Government’s Health and Medicare
acts, financially incentivizing CDS implementation into EHRs

• In 2013, ~41% of US hospitals with an EHR, also had a CDSS

• In 2017, ~40% of US hospitals had advanced CDSS capability

• Multiple solution providers:

• Philips Healthcare, Wolters Kluwer Health, Hearst Health, Elsevier B.V, IBM,
Meditech, Cerner Corporation, Epic Systems Corporation, Allscripts, Mckesson
India: NCD care under Comprehensive
Primary Healthcare (CPHC)
• Revised Operational Guidelines National Programme For Prevention &

Control Of Cancer, Diabetes, Cardiovascular Diseases & Stroke

• Recommends CDSS-enabled CPHC-NCD care as a good practice

• States can opt for the CDSS-enabled CPHC NCD system for use at

government health facilities

CDSS: Classification
• Knowledge-based & Non-knowledge based

• Knowledge-based CDSS
• Rules (IF-THEN statements) are created, with the system retrieving
data to evaluate the rule, and producing an action or output

• Rules can be made using literature-based, practice-based, or patient-

directed evidence
Examples:
TB Management Protocol
Examples:
Immunization
Schedule
 Colon cancer and Breast Cancer are the #1 malignancy
for men and women respectively in Saudi Arabia

Embedded a
CDSS - Health
Maintenance
Tab (HMT) -
within the EMR
to incorporate
best practices,
evidence based
guidelines and
regulations

https://www.himss.org/sites/hde/files/media/file/2020/01/17/2019davies_kingfaisal_cancerscreeningcasestudy.pdf
 Non-knowledge based CDSS:
Need data sources: Longitudinal medical records
Personal
Medical
Lab results Genetic Patient Database
History
Profile
Best Evidences

Drug-Drug
Interaction
Drug – Gene
Interaction

Decision leverages artificial intelligence (AI), machine learning (ML), for

statistical pattern recognition
Non-knowledge based CDSS
• Met with challenges including problems understanding the logic that AI
uses to produce recommendations, and problems with data availability

• Yet to reach widespread adoption

 CDSS: Components
• Base: Rules that are programmed
into the system (knowledge-based)
OR the algorithm used to model the
decision (non-knowledge based), as
well as the data available
• Inference engine: Takes the
programmed or AI-determined rules,
and data structures, and applies
them to the patient’s clinical data to
generate an output or action, which
is presented to the end-user
• Communication mechanism: The
website, application, or EHR frontend
interface, with which the end-user
interacts with the system
How the CDSS services are provided?

• Embedded with EHR

• Stand-alone

• Integration through APIs

Applications of CDSS

A. Patient Safety

B. Diagnosis Decision Support (DDS)

C. Clinical Management

D. Patient Facing Decision Support

E. Cost Containment

F. Clinical Research
 A. Patient safety (1)
• Strategies
Drug 1 to reduceDrug
medication
2 errors Outcome

Furosemide Amikacin Potentiate the risk of oto- and nephrotoxicity

• Errors involving drug-drug interactions (DDI) are common and
Omeprazole Clopidogrel Decreased effectiveness of Clopidogrel
preventable
Aspirin Clopidogrel Increased platelet inhibition effect

Nifedipine Magnesium sulfate Hypotension and neuromuscular blockade

Furosemide Digoxin Predispose patients to digitalis induced arrhythmias

Amikacin Megnesium sulfate Severe and/or prolonged respiratory depression

Hydrocortisone Moxifloxacin Potentiate the risk of tendinitis and tendon rupture

A. Patient safety (1)
• CPOE systems in EHRs are now designed with drug safety algorithms
that safeguard for dosing, duplication of therapies, and DDI checking

• The US Office of the National Coordinator for Health Information

Technology has developed a list of ‘high-priority’ list of DDIs for CDSS,
which has reached various levels of endorsement and deployment in
CDSS’ even in other countries: UK, Belgium, and Korea
A. Patient safety (2)
• Prevention of medication administration errors occurring at the ‘bedside’
• Electronic drug dispensing systems (EDDS), and bar-code point-of-care
(BPOC) medication administration systems
• Often implemented together to create a ‘closed loop’, where each step of the
process (prescribing, transcribing, dispensing, administering) is computerized and
occurs within a connected system
• At administration, the medication is automatically identified through radio-
frequency identification (RFID) or barcodes and cross-checked with patient
information and prescriptions
• These systems can be integrated with CPOE and CDSS simultaneously,
with reduced prescribing error rates for drug allergy detection, excessive
dosing, and incomplete or unclear ordering
Electronic drug
dispensing
systems (EDDS)

Bar-code point-
of-care (BPOC)
medication
administration
systems
A. Patient safety (3)
• CDSS also improve patient safety through reminder systems for other
medical events

• E.g., To prevent hypoglycemia events in ICUs: CDSS prompts for blood

glucose measurement according to glucose monitoring protocol,
considering patient demographics & previous glucose levels/trends
B. Diagnostic Decision Support System (DDSS)
• A computerized ‘consultation’ or filtering step, generating a list of possible

or probable diagnoses

• Diagnosis using fuzzy logic involving symptoms

• May also suggest diagnostic tests

• High accuracy compared to experts

• Great utility in low-resource setting

B. Diagnostic Decision Support System (DDSS)
• DXplain is an electronic reference-based DDSS that provides a
probable diagnosis based on clinical manifestations
• DXplain is owned by Massachusetts General Hospital (MGH) and
access after executing an institutional license with MGH

• New approaches with non-knowledge-based techniques – AI/ML –

may pave the way for more accurate diagnosis
• The Babylon AI-powered Triage and Diagnostic System in the U.K
 The Fall of Babylon Is a Warning for AI
Unicorns | WIRED UK
• An AI-powered symptom checker, which would even automate diagnoses
• Since 2016, the hype around Babylon just kept growing
• Picked up contracts with the NHS and British health insurance providers
• Chinese tech giant Tencent signed a deal to offer services through WeChat
• Saudi Arabia’s sovereign wealth fund invested $550 million
• Went public on the New York Stock Exchange in 2021, Babylon was valued
at $4.2 billion
B. Diagnostics support: Imaging
• Knowledge-based imaging CDSS:
• CDSS can aid radiologists in selecting the most appropriate Imaging, or alerting
contraindications to contrast, etc.

• Non-knowledge Based CDSS:

• A commercialized example is RadWise®, which guides clinicians to the most
relevant imaging order by analyzing patient symptoms and matching them with a
large database of diagnoses

• Substantially decrease the utilization rate of expensive imaging studies

and reduce cost
 B. Diagnostics support: Imaging (2)

• Non-knowledge Based CDSS helps in enhanced imaging and precision radiology

• Aid in extracting, visualizing, interpreting images and providing insights beyond

what humans can

• IBM Watson Health, Google DeepMind are developing tech for tumour
detection, medical imaging interpretation, diabetic retinopathy
diagnosis, Alzheimer’s diagnosis etc.
EHR Integrated Clinical Decision
Support (iCDSS) for Skin Cancer
Detection
• Oracle Cloud Infrastructure (OCI) Artificial
Intelligence (AI) Services

• Analyzing patient data for precise skin cancer

detection

• Seamlessly integrated with an electronic health

record (EHR)

• Integrated solutions with OCI

B. Diagnostics support: Laboratory & Pathology
• Alerts/reminders for abnormal lab results are ubiquitous in EHR systems
• CDSS can extend the utility of lab-based tests for avoiding riskier invasive
diagnostics
• In Hepatitis B and C testing, liver biopsies are considered the gold standard for
diagnosis, while non-invasive lab tests are not accurate enough to be accepted
• AI models are being developed that combine multiple tests (serum markers, imaging,
and gene tests) to produce much greater accuracy
• CDSS can be used for automated tumour grading from pathology reports
• Urinary bladder tumour grading and estimating recurrence, with up to 93% accuracy
• CDSS for computerized ECG analysis, automated arterial blood gas
interpretation, protein electrophoresis reports, blood cell counting, etc.
C. Clinical management
• The assumption that practitioners will read, internalize, and implement new
guidelines has NOT held true
• Traditional clinical guidelines and care pathways are difficult to implement in
practice with low clinician adherence
• CDSS can increase adherence to clinical guidelines
• The rules implicitly encoded in guidelines can be encoded into CDSS
• Alerts to a specific protocol
• Generating prescriptions
• Reminders for testing
• Tracking and placing orders in CPOE
• Follow-up for referrals | ensuring preventative care
• Alert to reach out to patients who don’t adhere management plans | due for follow-up
 D. Patient-facing decision support
• With ‘PHR’, patients become the end-user or ‘manager’ of the data
• When connected to EHRs, PHRs can provide a two-way relationship and aid in the self-
management of patients
• Vanderbilt University’s MyHealthAtVanderbilt has a Flu Tool for patients with flu-like symptoms to
decide the level of care they need and then help them seek treatment

• PHRs and connected wearables to create actionable insights

• Wearable glucose monitors can transmits data to an Apple device (HealthKit).
• Apple has made HealthKit interoperable with the Epic EHR and Epic PHR, “MyChart”.
• Allows providers to monitor glucose trends in their patient’s in-between visits, and contact them for
follow up or urgent recommendations

• Self-monitoring in heart failure, hypertension, sleep apnea, palliative/elder care

E. Cost containment
• CDSS can decrease inpatient length-of-stay, CPOE-integrated systems
suggesting cheaper medication alternatives, or reducing test duplication
• Heidelberg hospital in Germany developed a drug-switch algorithm and integrated it
into their existing CPOE system. The CDSS could switch 91.6% of 202 medication
consultations automatically, with no errors, increasing safety, reducing workload and
reducing cost for providers

• CPOE-rule can limit the scheduling of blood count/chemistry/coagulation panels to a

24-h interval reducing resource utilization without increasing the length of stay or
mortality
F. Clinical Research

• Helps to identify patients eligible for research based on specific criteria

• Point-of-care alert to physicians when a patient’s record matches clinical

trial criteria

• The alert prompts the user to complete a form that establishes eligibility
and consent-to-contact, forwards the patient’s chart to the study
coordinator, and prints a clinical trial patient information sheet
Steps in Developing CDSS

• Technical validation: Translation of clinical rules to software algorithms

• Therapeutic Validation: Check whether the alerts produced by the CDSS

are clinically relevant, useful and actionable

• Pre-implementation Prospective Validation: Feasibility in clinical setting

• Post-implementation Prospective Validation: Effectiveness

• Adaption in Practice
https://www.ncbi.nlm.nih.gov/books/NBK543516/
CDSS: Challenges?
CDSS: Challenges
• Fragmented workflows: CDSS can disrupt clinician workflow,
especially in the case of stand-alone systems
• Alert fatigue and inappropriate alerts: If physicians are presented with
excessive/unimportant alerts, they can suffer from alert fatigue
• Disruptive alerts should be limited to more life-threatening or
consequential contraindications, such as serious allergies.
• Medication alerts can also be specialty-specific
• Alert against using broad-spectrum antibiotics may be inappropriate in ICU
• Alert against duplicate medications may be inappropriate in inflammatory
bowel disease clinics, where the same class of drug can be applied through
different administration routes for increased effect
CDSS: Challenges (2)
• Lack of transportability and interoperability
• Many CDSS exist as cumbersome stand-alone systems or exist in a system that
cannot communicate effectively with other systems.
• Financial challenges
• Outset costs to set up and integrate new systems can be substantial.
• Ongoing costs can continue to be an issue indefinitely as new staff need to be
trained to use the system, and system updates are required to keep pace with
current knowledge.
CDSS: Example from India
 mPower Heart mHealth System
Hypertension Diabetes Dyslipidaemia COPD

Screening of Cancers
Depression Alcohol & Tobacco Asthma
(Oral/Cervical/Breast)

Comorbidities –
Acidity, Arthritis, MCH – PIH, GDM
Blindness, backache
 mPower Heart mHealth System
Features/Functionalities
Prompts: optimal drug/
Screening for HTN & DM Generate personalized Electronic Health
dosage; Contrindications;
10 year risk for CVD clinical management plan Record/Big data analytics
Lifestyle advices

Profile visualizer for Push ever evolving clinical Stable platform designed
trending and quick Case data sharing/Quality
management guidelines to cover other diseases in
decision making Assurance
remotely future

Android and Offline version ensures

Data Interoperability as Secured central server
desktop/windows connectivity in remote
per national standards with data back-up
versions settings as well
 Implementation of CDSS workflow in Govt. Health Facilities

Nurse / Care Coordinator MO

Patient Enters Initial Assessment and examination Gets printed CDSS
walks-in or details into Nurse Portal recommended action
gets
referred by
ANM

Accept
CDSS Integrated Patient Pathway Rewrite
Medical Officer
MO
Patient Accepts
walks out recommendation or
with rewrite prescription
prescription Nurse / Care Coordinator
Updates Patient record in Nurse App after MO
40
decision
mPower Heart Project: Results from
Himachal Pradesh
21 5 6

Time Period Government Outpatient

(months) Hospitals Clinics

21,000 6,800 50%

Patients Patients First Time

Screened treated Detections

Results sustained after 18 months

15 7 50
mmHg mmHg mg/dl

Drop in Systolic BP Drop in Diastolic BP Drop in Glucose Levels

Vamadevan et al. JAHA 2016
mPower Health: State of Tripura & Mizoram

• Funded by the National Health Mission and State-wide implementation

• 56 hospitals: 2 PHCs, 24 CHCs, 15 sub-district hospitals, 14 district hospitals & 1 State Hospital
• Technical Coordination Cum Support Team:
• State Health Officials and Project team
• Software updates
• Operational updates
• Server Access for state government
• Patients enrolled/benefited: ~195,000
• HTN-67313, DM-21849, HTN&DM-12127, COPD-2152
• https://www.clinally.com/
• https://www.clinally.com/
• https://www.youtube.com/watch?v=kirrgYcl6Ro
• https://www.nature.com/articles/s41746-020-0221-y
• https://bmcmedresmethodol.biomedcentral.com/articles/10.1186/s1
2874-018-0587-6
• https://www.himss.org/sites/hde/files/media/file/2020/01/17/2019d
avies_kingfaisal_cancerscreeningcasestudy.pdf
• https://www.ncbi.nlm.nih.gov/books/NBK543516/
Clinical Practice Research
Datalink (CPRD), UK
Ajay Vamadevan
PGDM – HCM | Term 3
Clinical Practice Research Datalink (CPRD), UK
• CPRD maintains the repository of anonymised patient data from a
network of GP practices across the UK

• Primary care data are linked to a range of other health-related data to

provide a longitudinal, representative UK population health dataset

• A real-world research service supporting retrospective and

prospective public health and clinical studies
• Jointly sponsored by the Medicines and Healthcare products Regulatory Agency, UK and the National Institute for Health Research
(NIHR), as part of the Department of Health and Social Care, UK.
Clinical Practice Research Datalink (CPRD) - 2
• The data encompass 60 million patients, including 18 million currently
registered patients.
• For more than 30 years, research using CPRD data and services has
informed clinical guidance and best practice, resulting in over 3000
peer-reviewed publications investigating:
• Drug safety,
• Use of medicines,
• Effectiveness of health policy,
• Health care delivery and
• Disease risk factors
Primary care and linked data flow
• NHS Digital, a statutory body, receive
identifiable patient data for linkage
• EHR providers submit de-identified data to
CPRD and patient identifiers to NHS Digital
• Data custodians for external datasets also
send patient identifiers to NHS Digital
• NHS Digital uses identifiers to link the
datasets
• CPRD provides researchers with anonymised
linked data and metadata detailing the
linkage process
• Linking CPRD data to other sources improves
the scope & quality of research studies
Linked datasets currently available include:
Data from NHS Digital: Small area level data:
• Hospital Episode Statistics (HES) Admitted • Patient postcode linked deprivation measures
Patient Care (HES APC) data • Practice postcode linked deprivation measures
• HES Outpatient (HES OP) data • Index of Multiple Deprivation (IMD)
• HES Accident and Emergency (HES A&E) data • Townsend Deprivation Index
• HES Diagnostic Imaging Dataset (HES DID) • Carstairs Index
• Death registration data from the Office for
National Statistics (ONS) • Rural-Urban Classification
• Mental Health Dataset (MHDS) COVID-19 data
Cancer data from Public Health England (PHE): • PHE Second Generation Surveillance System
(SGSS) COVID-19 positive virology test data
• Cancer Registration • COVID-19 Hospitalisation in England
• Systemic Anti-Cancer Therapy (SACT) Dataset Surveillance System (CHESS)
• National Radiotherapy Dataset (RTDS)
How linked, de-identified dataset inform
evidence-based clinical practice
Use cases
• Preventing whooping cough in new-born babies

• Influenza and diabetes

Safety of pertussis vaccination in pregnant
women in UK -1
• Whooping cough can cause serious and fatal complications in newborn
babies and young children.

• Babies are routinely vaccinated early, from two months of age, but can
still be at risk if a mother catches whooping cough whilst pregnant

• After an outbreak of whooping cough in 2012, a national programme was

introduced to give pregnant mothers a vaccine to protect their baby

https://www.bmj.com/content/349/bmj.g4219.long
 Safety of pertussis vaccination in pregnant
women in UK -2
• Objective: To examine the safety of pertussis vaccination in pregnancy

• Design: Observational RETROSPECTIVE cohort study, using the UK CPRD,

involving 20,074 pregnant women with a median age of 30 who received the
pertussis vaccine and a matched historical unvaccinated control group.

• Main outcome measure: Adverse events identified from clinical diagnoses

during pregnancy. The primary event of interest was stillbirth (intrauterine
death after 24 weeks gestation).

https://www.bmj.com/content/349/bmj.g4219.long
 Safety of pertussis vaccination in pregnant
women in UK -3
• Results: No evidence of an increased risk of stillbirth in the 14 days
immediately after vaccination (incidence rate ratio 0.69; CI: 0.23 - 1.62)
or later in pregnancy (0.85, CI: 0.44 - 1.61) compared with historical
national rates

• Conclusion: In women given pertussis vaccination in the 3rd trimester,

there is no evidence of an increased risk of any adverse events related to
pregnancy.

https://www.bmj.com/content/349/bmj.g4219.long
 Influenza and diabetes -1

• Seasonal influenza accounts for a significant proportion of excess winter

mortality

• Imperial College London used CPRD data to investigate the effectiveness of

the influenza vaccine at reducing cardiovascular and respiratory hospital
admissions and mortality in patients with type-2 diabetes
 Influenza and diabetes - 2
• Methods: A RETROSPECTIVE cohort study using CPRD data, over a 7-year period
between 2003-04 & 2009-10, enrolling 124,503 adults with type 2 diabetes

• Results: Vaccination was associated with significantly lower admission rates in

Influenza season for:
• Stroke (IRR 0.70, 95% CI: 0.53-0.91)
• Heart failure (IRR 0.78, 95% CI: 0.65-0.92)
• Pneumonia (IRR 0.85, 95% CI: 0.74-0.99)
• All-cause death (IRR 0.76, 95% CI: 0.65-0.83)
• Acute MI (IRR 0.81, 95% CI: 0.62-1.04)
IRR: Incidence Rate Ratio; CI: Confidence Interval
• Interpretation: In type 2 diabetes, influenza vaccination was associated with
reductions in hospital admission rates for specific cardiovascular events
Effectiveness of the influenza vaccine in preventing admission to hospital and death in people with type 2 diabetes - PubMed (nih.gov)