Sensors attached to an intelligent mechanism that decides whether or not some condition has been met.
This agent alerts other agents in the system such that they may respond to this event.
This has many applications within emergency management.
A lot of this has to do with mechanical sensors and exact measurements (parachuting forest fire sensors, earthquake monitors, floating tsunami sensors, nuclear reactor agents)
However, there are many cases where such measurements are inexact and relative...
Nealon and Moreno (2003) suggest that agents are useful in health care.
Agents in this field should use AI techniques and be...
Proactive
Sociable
Reactive
Autonomous
Palliative Care is for terminally ill patients who require...
Constant Care
Easing of Pain
Immediate, yet Unexpected Emergency Responses
Much of this is done manually and through paper evaluations
The use of agents and alarms can improve the process
Agents and sensors in healthcare systems have different concerns than mechanical sensors
Lives are at stake
Legal liability (and its variability across governments)
Moral concerns about information propagation
Moreno, et al (2005) introduce PalliaSys
A multi-agent system
Digitizes patient evaluations
Sophisticated alarm mechanisms to alert doctors
Secure information retrival for patients and families
Develop prediction models through machine learning
Database Wrapper
Doctor Agent
Patient Agent
Data Analyzer
Two types of alarms:
General
Apply to all patients
Designated by PCU head
Doctor-Specific
Apply to a subset of patients
Designated by assigned doctor
There is an interface in the Doctor Agent to create Alarms
They are based upon (0-10 scale) patient evaluations (from the Patient Agent) on certain criteria:
|
|
|
|
|
|
|
|
|
|
Doctors can create simple rules to trigger alarms per patient:
Extreme Weakness: (Weakness > 7) and (Pain > 8)
This may describe an abnormal condition for a particular patient.
When a patient evaluation is filed, it may trigger this alert.
The doctor will be notified via the user interface on the Doctor Agent.
Doctors are free to write more complex rules.
Dangerous Weakness: (Hunger < 3) and Extreme Weakness
In this case, it specifies a more severe rule based upon another one.
Some patients may need certain rules, but other patients may require more severe alerts to be monitored.
Evolution Alarms are monitors for events that occur over time.
Alarm rules, since human evaluated, are relative.
Therefore, set alarms for changes over time.
Fast Weakness Increase: d(Weakness) > 2
Alarm fires when a evaluation has increased by 2 from the previous evaluation.
Perhaps, a condition changes slightly over a period of time.
However, a doctor may still want to be alerted when evaluations show this trend.
Extreme Pain Increase: d(Pain, 28 days) > 4
Now, it is across all evaluations from the last 4 weeks.
To handle certain security and moral implications, the alarm is strictly defined.
Doctor Agent has a GUI to define alarms.
Alarms can be stored with the Patient Agent and secured locally.
All evaluations are only stored in the centralized database and encrypted.
Raised alarms are passed and stored in the Doctor Agent.
All reporting and checking is done automatically.
Patients can view their information from doctors securely (SSL, RSA)
Proactive healthcare approach
Information better exposed to patients
Intelligent agents can suggest doctor visits
More complete treatments
Processes are automated and faster
J. Nealon and A. Moreno. Agent-based Health care systems. Pages 1-18. Whitestein series on Agent Technology. Birkhauser, 2003.
A. Moreno and D. Riaño and A. Valls. Agent-based alarm management in a Palliative Care Unit. Pages 60-66. III Workshop on Agents Applied in Health Care en IJCAI2005 Edinburgh Actas del workshop, 2005.
/
#