The ASTI project

1
The ASTI project

• H. Falcoff - Société de Formation Thérapeutique
  du Généraliste (SFTG),
• A. Venot - LimBio Paris XIII - B. Seroussi et J.
  Bouaud - Paris VI

2
The quality problem

• Evidence about frequency of suboptimal or
  inappropriate drug prescriptions
• Two main reasons - explosion of therapeutic
  information - human errors
• Computerized decision support systems (CDSS) for
  drug prescribing is it a solution ?

3
CDSS 1st generation

• They improve security- guide to choose de right
  dose - control of drug interactions et drug
  contra indications- generate reminders
• They may save money
• Doctors like them
• They are based on structured drug databases
• Exchanges with the EMR limited (age,
  weight,contra indications, allergies), data have
  to be coded
• No help to choose the drug class

4
CDSS 2nd generation

• They aim to help the GP to decide what is the
  best drug for each patient, taking account of all
  the health problems and the drugs received before
• They are based on clinical practice guidelines
• Developpment in progress Prodigy (UK), ASTI
  (France)
• Several problems to solve

5
The ASTI project

• ASTI Aide à la Stratégie Thérapeutique
  Informatisée
• Started in 1999
• Public funding
• Consortium - medical informatics academics
  Pr Alain Venot - PARIS13, Dr Brigitte Seroussi
  et Mr Jacques Bouaud - PARIS VI, Dr Christine
  Riou - Rennes- software developers Mr Gérard
  Simon - Banque Claude Bernard, Mr Christian
  Simon - Silk Informatique, éO- GPs SFTG
  (éO users)

6
Principles of ASTI

• Two modes- Critic mode operates as a
  background process to control the physicians
  prescriptions it displays reminders when
  physician prescrpition differs from
  guidelines recommendations- Guided mode
  operates on demand, has to be triggered by
  the GP when he needs support to establish
  the drug prescription
• Knowledge base built from the 1999 Canadian
  guidelines for hypertension

7
Display of ASTI
8
Critic mode main functional components,
knowledge ressources and data flow
Patient categorisation rules
Abstracted patient data
EMR
Reminder !
Critic mode
Abstracted prescription
Prescription
Guideline based rules (if - then - else)
Drug database
Guideline
9
Guided mode main functional components,
knowledge ressources and data flow
Guideline
Experts
Patient categorisation rules
Collection of scenarios
Doctor
Abstracted patient data
EMR
Guided mode
Abstracted prescription
Prescription
Suggestion
Drug database
10
The two modes are complementary

• Preliminary evaluation of ASTI (10 GPs confronted
  with clinical cases)
• Reminder-based interaction seems to be
  appropriate for simple cases
• GPs want to use on-demand system as clinical
  situations become more complex

11
Some problems...

• Quality of guidelines (complete, accurate, not
  ambiguous)
• Updating guidelines
• Quality of the EMR (well structured, coded data,
  updated)
• Therapeutic history taking account of each drug
  prescribed in the past, the reason of its
  prescription and its effectiveness
• How many guidelines can a GP cope with ?

12
ASTI 2 - ASTI 3

• ASTI 2 - based on French guidelines-
  hypertension type 2 diabetes- assessment of
  the impact of the system on the quality of the
  prescriptions (RCT)
• ASTI 3 assessment of the genericity of the
  model different guidelines and different EMR
  softwares