What is a Knowledge Base? And how does RippleDown use it?
A Knowledge Base is a place to capture and store human knowledge in a computerised format. It contains a set of data – often in the form of rules – that describe the knowledge in a logical and consistent manner.
RippleDown offers a simplified system that extracts and interprets expert human knowledge from within a Knowledge Base. The RippleDown system can be taught by domain experts, not IT, using “rules”.
The system replicates an expert’s decision making process at scale, to create automated decisions, recommendations, reports and summaries. This delivers a more accurate, secure and efficient service to clinicians that ultimately enables better patient outcomes and increased customer satisfaction.
Do you provide a Knowledge Base with rules?
RippleDown is supplied without any Knowledge Base bases (i.e. validated rules). Domain experts configure their own customised Knowledge Bases to suit their organisation and their own requirements.
Following training, a user is equipped with all the skills required to tailor a Knowledge Base that will generate their own bespoke reports.
What is the maximum capacity for a Knowledge Base?
RippleDown Knowledge Bases are infinite, so there is no limit to their capacity.
Our technology is designed to be scalable, and in production we have many Knowledge Bases with tens of thousands of rules.
How long does it take to build a rule?
Using RippleDown, users can build rules in 1 minute, compared to the industry average of 2-5 rules per day.
Who builds the rules?
Any domain expert – whether that be a Pathologist, Scientist – can build rules and manage a RippleDown Knowledge Base.
The RippleDown system is designed to be managed by subject matter experts – not IT – to replicate their decision making process, at scale.
How are rules created and used?
Rules are created using the RippleDown Knowledge Builder application. This provides a user-friendly interface for adding rules, which can be built in a minute.
With RippleDown, rules can be built incrementally, and new rules does not break existing ones. RippleDown also has a safe rule building process with conflict checking.
Can RippleDown extract data from multiple sources?
Yes. RippleDown can query data from multiple sources simultaneously, then amalgamate results before providing a final interpretation.
Can I embed RippleDown with my existing applications?
Yes. RippleDown is highly customisable and can integrate with your existing software and infrastructure.
For example, RippleDown can interface directly to an instrument using a Knowledge Base prepared at a central site.
What are the options for deploying RippleDown Clients?
There are two main options for deploying RippleDown clients: “In-Cloud” or “In-House”.
The In-Cloud solution involves setting up a virtual server in the cloud to host the RippleDown applications. Whilst the In-House solution requires the RippleDown applications to be hosted on a physical server managed by the client. An In-House deployment means all client data remains behind the client’s own firewall.
Is RippleDown deployable as a SaaS?
Yes. RippleDown supports deployment in the cloud using web services interfaces.
Does RippleDown support multi-client access in the cloud?
Yes. RippleDown supports a “multi-tenant” SaaS. Each client would have their own set of Knowledge Bases, permissions and user accounts.
Will RippleDown supersede experts?
No. Human experts are critical in the RippleDown process workflow. They provide all the expertise which is then leveraged by RippleDown to automate the unique, human decision making process that’s critical to a laboratory.
RippleDown simply enables the experts to work more effectively – delivering a more accurate and efficient service, to enable better patient care.
Does RippleDown guarantee backwards compatibility?
Yes. PKS obtains the databases from each customer site prior to an upgrade, and runs a comprehensive set of automated regression tests at the PKS support facility to ensure that any upgrades will be backwards compatible with all existing databases.
How scalable is the RippleDown inferencing?
RippleDown’s internal rule-tree organisation is highly efficient when it comes to inferencing, with only a minimum number of nodes required to be evaluated for each interpretation.
The entire inference engine executes in-memory, with no database calls required in the main interpretation process. The use of immutable objects allows for concurrent interpretations by multiple threads. The limiting factor in transaction rate is the speed with which data can be presented to RippleDown from a client’s online information system.
How can RippleDown be incorporated in a system and how does it communicate?
The RippleDown server is typically deployed as a service/daemon on a Windows, Linux or Unix virtual server. It communicates in real-time with the client’s Online Information System using a standard protocol such as Web Services, HL7, XML, or DICOM. Custom interfaces can also be developed.
RippleDown can also be used without a real-time interface, for example, by processing client data which has been imported from a spreadsheet.
For a smaller installation, the RippleDown server can be co-hosted with other applications. It is implemented entirely in its own version of Java so it will not interfere with other Java installations. The RippleDown server also has its own zero-maintenance embedded database (Infinity) used for storing rules, so no external database is required.
RippleDown client applications (for rule building, validation and language translation) are typically deployed via Citrix or from a terminal server. They communicate to the RippleDown server via the network LAN.
What are the platform requirements for the RippleDown Server?
The minimum requirements for the RippleDown Server are:
• 64 bit virtual server 2 CPUs
• 16 GB memory
• Windows 2008 server 64 bit, or Linux or Unix 100 GB free disk space
• SMTP server access for email alerts
How much effort will be needed to install software updates?
Each update is supplied as an automated installer program taking less than 1 minute to execute. This can be done remotely by PKS support staff.
What security model does RippleDown use?
RippleDown security is modelled on the “Java Authentication and Authorization Service (JAAS)”.
It implements a secure user account framework which provides:
• User name and password credentials
• Fine-grained permissions facility which controls access to all RippleDown client applications, facilities within applications, and data stores
• User group facility allowing permissions to be flexibly and transparently allocated to user
The RippleDown security framework does not include a “single sign-on” facility with Windows or Linux.
Which Laboratory Information System data interfaces do you support?
RippleDown supports key interface standards, including HL7, REST, DICOM, XML and SOAP. IT also supports a number of proprietary standards.
RippleDown is interface compatible with all developed applications from PLS to SML, Labosys, Trak (InterSystems) Novius (Siemens), Cerner Millenium, OmniLab (Integrated Software Systems), Lunar Densitometer (GE), SOAP, Ultra (GE), DICOM (Hologic) and other various custom interfaces.
Can reports be tailored to specific doctors?
Yes. The RippleDown solutions are highly customisable, meaning report formats can be tailored to individual clinicians.
Does RippleDown only assess recent data?
No. RippleDown has the ability to assess all data from multiple sources. This means a patient’s full medical history can be analysed and interpreted against recent test results.
What sort of recommendations can RippleDown make?
The recommendations RippleDown can apply are really only limited to those recommended by the experts themselves. RippleDown can automate reflex testing and recommended treatments or diagnosis based on rules built by the experts within the Knowledge Base.
What types of pathology test results are suitable for interpretation by RippleDown?
RippleDown is highly customisable and can be tailored to specific client requirements. Typically RippleDown interpretations are best suited to Pathology domains where the numeric or textual test results are available electronically.
Can I tailor what information is sent to Specialists?
A high degree of control over the content of the released interpretations can be defined by adding rules to the Knowledge Base to define the nature of the reports. In particular, rules can easily be added so that a different report (or no report) is sent to a Specialist, whereas the full report for that patient is sent to the primary care physicians.