HOMI16-CDSS

Clinical Decision Support System Definitions:Shortliffe Definition - Any computer program that helps health professionals make clinical decisionsWyatt and Speigelhalter: Active knowledge systems which use 2 or more items of patient data to generate case specific adviceExamples:1) DeDombal's System for abdominal pain2) ECG interpretation 

Categorizing CDSS:By Knowledge Representation (i.e. Decision trees, truth tables, bayesian statistics, rules, semantic frames, etc)By decision type - (ie. diagnosis, therapy, ordering etc)By medical domain - (intermal medicine, surgery, obstetrics, laboratory etc)For Clinicians , we can characterize them as to whether advice is solicited or not:1) Solicited Advice - systems that provide advice only they are instructed to2) Un-solicited Advice - Systems provide advice independently of a request forma clinician 3) Autonomous Systems - monitor incoming patient data and take action without human intervention

Solicited Advice - QMR is an example and descendant of the Internist Research system - Usually, these are standalone systems. 600 diseases, 4500 clinical findings. Past Medical history, Symptoms of current illness, findings on physical examinations, simple inexpensive lab tests, moderately expensive or invasive laboratory tests, very expensive or labor intensive lab tests. THere are 2 relationships between a diseases and clinical findings . - Evoking strength and frequencyEvoking strength - how strong is the finding suggests the presence of the disease --> correlates with subjective positive predictive value or conditional probabilty of a disease  given a finding Frequency - how often the findings is seen with the disease, ie. a subjective measure for the sensitivity

Unsolicited Advice  - Systems that provide unsolicited advice do not require the initiative of the clinician to generate advice. Rely on Software systems and relevant patient data and generate advice independently of the request from the clinicianThese systems are integrated with EHRs . THe CARE system by Clem McDonald . Uses a care syntax to write rules to remind the clinician of diagnosis or problems that might be overlooked. Typically these are rules use simple boolean logic. There is a risk of generating excessive advice and false positive  advice or incorrect alerts may generate antagonistic responses. Want systems to have high predictive value. Challenge is that increasing the sensitivity of advice will likely decrease the predictive value, and amount of false positive advice will increase. EXAMPLE: Arden Syntax - sharing of medical knowledge --> medical logic modules- each one encodes a discrete segment of medical knowledge

Autonomous systems- Advice is directly applied to the system - only prototypes of autonomous systems are developed --> anesthesia, ventilator control, drug delivery.

Clinical DSS