"Doctors, emergency medicine and computers"

Dr Susan Clamp,
Director Yorkshire Centre for Health Informatics

Background

Dr Clamp began her talk by outlining her work in University and the Teaching Hospital.  She then described health informatics as “The knowledge, skills and tools which enable information to be collected, managed, used and shared to support the delivery of healthcare and promote health.”  The number of deaths caused by medical errors comes between those from obstructive lung disease and pneumonia and some of the errors are avoidable.  The amount of knowledge is increasing rapidly (doubles every five years) and is greater than the amount that can be learnt.  Experienced clinicians still have lots of new information to assimilate.  Evidence based medicine integrates individual expertise with clinical evidence from systematic research.  Things which can help reduce errors or improve access to relevant information in a timely manner can help clinicians.

The diagnostic process is essentially the same regardless of the problem; the clinician will ascertain the history, examine the patient and/or arrange investigations, analyse the information then make a decision based upon what is possible and available.  A clinical decision support system (DSS) can help by integrating a medical knowledge base (condition specific) with patient data and using an inference engine to give patient specific advice.

This started in 1970s with medical AI research and was developed during the 1980s but computers weren’t common then!  The system has continued to advance despite obstacles and may be used more in the future as IT infrastructure improves and medical practice becomes more formalized. 

The team has worked with emergency departments.  This is a very intense environment and patients present with varied symptoms (from a fracture to a severe headache).  Patients are assessed and prioritized, condition diagnosed and treated and the care pathway for future treatment drawn up.  IT is often fragmented in these departments e.g. administration system, picture archiving system, several discrete databases and some PDAs carried by keen individuals.  There may be some DSS used in an unstructured way.  For example there will be a specific pc with the TOXBASE web site loaded in case there is an overdose and there may be pieces of paper stuck on the walls e.g. resuscitation pathway and paracetamol graph.  There will also be some sophisticated DSS e.g. ECG machine which uses a neural network to print out comments about the individual’s ECG.

Decision support in acute abdominal pain (AAP)

Acute abdominal pain is a difficult clinical area accounting for about 6% of attendances in emergency departments (ED).  80% of cases can be diagnosed on clinical signs and symptoms if the clinician is experienced but averages 45-65% (UK).  If the clinician makes the right diagnosis then the decision about treatment is usually quite clear.  The software can help the inexperienced doctors.

As you can measure outcomes in AAP you can build up a database of cases.  Data has been gathered for 20 years and there has been no change in presentation; they are still using the same data collection forms and symptom definitions.  They started with 600 cases from Leeds surgical wards then gathered more data from ED then from Europe.  The information is now held in several databases e.g. A&E, wards, females, children.  The database holds the presenting symptoms and final diagnoses of patients with AAP.  Examples of completed forms can be found in the presentation, along with examples of how the probabilities of diagnoses are displayed.

The tool can be used as an educational tool and in a clinical setting.  It helps clinicians with recording the history and symptoms and can prompt the user to ask pertinent questions to help arrive at a diagnosis.  It analyses the most probable cause of the patient’s problem.  The patient’s symptoms are entered and the relevant database selected.  Bayesian analysis is performed and the database searched then the breakdown of outcomes of patients with similar clinical pictures is displayed.  This analysis can be used alongside other investigations to help inform the decision making process.  The doctor makes the diagnosis.

It is very difficult to do random control studies in this setting so the team did before and after studies.  All studies showed an improvement when the tool was in use.  Performance fell back after the system was taken out.

Problems identified include:

• Double entry (paper then IT) is needed

• Time taken to enter the information into the system

• Lack of integration with electronic patient record systems

•  Interpretation of the results displayed

•   Attitudes of some clinical staff to the tool including concerns of over reliance

The system may be able to be used in clinical decision units, by ambulance staff dealing with out of hours work and by nurse practitioners.  (Clinical decision units aim to reduce inappropriate admissions, reduce inappropriate discharges, reduce length of stay of admissions by optimizing care pathways.)  Various clinical protocols for investigation and care of certain conditions have been developed for these settings.  These protocols are often form based.  There is a move away from passive decision support to active DS.

The full presentation, including many examples of the materials referred to, is available on this web site.  Also, further background information and an online demonstration of the AAP tool is available on their web site (www.aaphelp.leeds.ac.uk).

Following Dr Clamp’s interesting talk we ended the evening with a lively debate.