Advisor for pain management in hospice setting

Description: Demonstration of an expert system for advising on management of pain in cancer patients, 1997

Publet Introduction:

An experimental system for advising on management of pain was developed in 1997 by Thomas Olsson under the supervision of Professor John Fox, Imperial Cancer Research Fund (now Cancer Research UK);  Dr. Robert Dunlop (St Chistopher’s hospice) and Professor Irene Higginson (King’s College London). The application was not intended for clinical use and has never been tested or trialed in routine care but was developed only to explore the technical and knowledge representation issues in a workflow and decision support service for the hospice setting.

The full system encapsulates best practice in pain control at the time of writing as a basis for assisting clinically qualified but non-specialist carers (e.g. general practitioners, Macmillan nurses).  The final system was planned to provide support from any point at which the patient’s diagnosis was known, curative efforts had been discontinued, but effective symptom control had not been achieved.

This protocol was intended to capture the knowledge routinely required at the point where St. Christopher’s Hospice staff first saw a patient, and the subsequent 1 or 2 reassessments of the patient.  The focus was initial management of pain using the WHO ladder that was current at the time, ancillary management of any problems due to side-effects of medications (e.g. nausea/vomiting; constipation/diarrhoea) and adjuvant therapy for symptoms due to tumour secondaries.

Following an initial elicitation of subjective data from the patient a cyclical process was begun, in which the patients pain is reviewed and if necessary additional/alternative analgesics are prescribed.  Simultaneously an objective data collection was initiated, in which the pain syndrome and the cancer characteristics were specified.  The combination of this information is used to decide the likelihood that the patient is suffering from tumour secondaries.  If there is a high likelihood of secondaries, the patient is referred to a specialist, and the end of the demonstrator protocol is reached.

 Information
Guideline Objectives

This project was intended to implement a fragment of the knowledge required for a comprehensive decision support system for management of terminally ill cancer patients.  The full system encapsulates best practice in pain control, as defined by Dr. Dunlop who at the time was the medical director of St. Christopher’s Hospice.

Target SettingHospice care
Target Users
  • Author
  • Release Date 26/07/2013
Overview

The main protocol developed is directly based on the flow diagram illustrated in Figure 1.  The only difference is that the cyclic follow-up procedure has not been implemented.  To cater for this, two distinct decision instances have been implemented; an initial one which takes the subjective data into consideration, and a second one which takes all gathered information into consideration.

The toplevel PROforma plan is illustrated in Figure 2. 

Figure 2: Toplevel PROforma procedure (plan)

The subjective data collection plan collects information regarding; pain severity, past treatments, current treatments, and other symptoms.  Data items used are listed in Appendix A.  All data items are currently mandatory and each sub-enquiry is done individually, so they each type of data can later be either distributed or collected from files currently held regarding the patient.  The subjective data is used to make an initial assessment and based on this one of three treatments (Paracetamol, NSAID’s, or Morphine) is prescribed for the patient.  Following this decision, further consideration is given to possible side-effects and possible co-analgesics are prescribed as well.  Once prescriptions have been printed, a more thorough information gathering task is initiated.

First the primary cancer location is specified, together with treatment and status, and then a number of hidden decisions are taken, including the likelihood of cancer spread to particular sites.  This is based on a knowledge grid (provided in the guideline in Appendix B) and is (currently) implemented as a plan with a number of decisions which the software can take autonomously.  This is illustrated in Figure 3. 

Figure 3: Hidden decisions

Each hidden decision has three candidates; likely, possible, and unlikely.  Based on the knowledge grid and the information at hand, each of these decisions is automatically taken and the system will after completion hold the likelihood of each secondary.  This information is later used in accessing the need to a referral. (The implementation is quite clumsy due to limitations of the version of PROforma used and will be refined in later versions.)

Once the cancer characteristics has been recorded, the characteristics of the patient’s pain are examined to establish a “pain syndrome” based on a number of characteristics such as character, severity and location of pain. There are two possible pain syndromes in this demonstrator; deep somatic pain, and neuropathic pain.

The patient’s pain severity is reviewed and the information gathered is combined to re-assess the initial treatment.  This may result in another prescribing decision for analgesics and co-analgesics or, if the pain is controlled by the original treatment, the patient can be discharged.

Finally, the pain characteristics and the cancer characteristics are combined to try and conclude if the patient’s pain is due to a possible spread of the cancer.  If this is the case, the patient is referred to a specialist.

Provenance7.No provenance has been assigned (default value)
Management

This project was intended to implement a fragment of the knowledge required for a comprehensive decision support system for management of terminally ill cancer patients.  The full system encapsulates best practice in pain control, as defined by Dr. Dunlop who at the time was the medical director of St. Christopher’s Hospice.

Safety CaseNone developed
Sources

Palliative care specialist

References

None provided