Technical details
Biosafe is a software program written for Windows 95, 98, 2000, NT and XP.  Biosafe is also available as a Widows CE program to run on pocket PC’s.  BioSafe needs only minimal computing power and RAM; running at a respectable speed on 400Mhz or less machines with a minimum of 128Mb of RAM.

Biosafe consists of an expert system engine “MEDEX” written in C++ that calculates probabilities from a separate database of conditions and symptoms.  This description relates to the BioSafe database module that contains information on all known agents of biological and chemical warfare and all agents known to have been used or have been threatened to be used by terrorists.

Medex uses a combination of calculations using Bayesian probability methods and rule based pattern matching to diagnose what a patient has been exposed to from his or her symptoms.  Using the frequency with which a symptom is found in a particular condition and the frequency that it is shown in all the conditions in the database it produces a “differential diagnosis” list.  This list shows all the likely conditions and the probability that they were responsible for the symptoms displayed by the patient.

Program Installation
The program is provided on a CD in either a cassette, DVD case or retail “shelf” box.  For pocket Pc’s it can be provided on SD, Compact Flash or Sony memory stick.  Alternatively the program can be downloaded from the web site.  Installation is automatic with an icon placed on the desktop or Start bas as required.

Running the program
Initially the user selects a database, in this case the Biosafe database.  The program is designed to be used in different ways depending on the level of the user’s medical expertise and whether the user has some idea of what agent has been used.

If the user has little expertise it can allow the program to select the symptoms to look for.  The program selects questions in an order that will most rapidly identify the agent used from its list of more than 80 suspicious agents.  The user can answer yes/no or don’t know to any question.  If the user needs more information a click on the “reference” button brings up lay descriptions, or photographs if the symptom has visible signs.  A diagram or video may be presented if the user needs to perform a particular procedure to identify the symptom.

A more experienced user, a well trained paramedic or clinician for example, may wish to use a faster method of data entry.  They can select to identify the symptoms themselves and tick them off directly on a list of symptoms.  While this is occurring they can also see the differential diagnosis list with all the possible agents listed in order of likelihood as symptoms are entered.

As the questions are asked numbers appear against the list of known agents.  The most likely candidate rises to the top of the list.  If the patient shows any unique symptoms found only in one condition a red star appears next to that condition.  Conversely if the patient does not exhibit a symptom shown by all people exposed to a particular agent a black star is placed against that agent.  At any time the user can ask for more information about a symptom or agent simply by clicking on the name in the list.

As the symptoms start to point towards one agent the user may decide to temporarily switch to examining that particular agent.  Clicking on “narrow” search changes the questions to those related to whatever condition is at the top of the diagnosis list, this may change, as symptoms are “ticked or crossed” by a mouse click. 

As the symptoms are checked off, the diagnosis may become more certain, as indicated by its score rising to the top of the list of possible agents, and possibly by a red star as well if a unique symptom has been identified.  However, it may be that as more symptoms are examined the diagnosis falls, at any point the user can go back to examine another condition or return to the questioning

If an agent is identified by its high score and red star, the user may wish to “lock on” to that agent and proceed to the next stage.  Clicking on the button marked “pursue” brings up a sub list of all the symptoms linked to the agent in question.  Those already selected are marked and the user can then examine whether the rest of the symptoms are present.  Numbers alongside each symptom indicate what percentage of people exposed to the agent show the particular symptom.  At all times the user can call up information and illustrations of the symptom simply by right clicking on the symptom required.

If the user has some reason to suspect a particular agent has been used (a warning phone call, use of the agent elsewhere etc.) the user may wish to omit the first stages and go straight to “pursuing” that “known” agent.  BioSafe provides the user with a description and illustrations of the agent’s effects along with a list of symptoms and the frequency they are shown by individuals exposed to the agent.  This method is best for eliminating or confirming exposure to a known agent.

When the user wishes to terminate the enquiry (because of shortage of time, or identification of an agent by its high score) the user can “accept the diagnosis”.  At this point the user is presented with a summary of all the symptoms checked positively or negatively or “unknown”, these are shown in a list of the symptoms related to the condition or not normally shown by the condition.  A list of all the symptoms in the database is also shown alongside.  At this point the user can survey the position and edit any of the answers by a simple mouse click or by dragging a new symptom across from the major list.  If this does not substantially change the diagnosis list the questioning is complete.

Although this verbal description sounds lengthy, the whole process takes less time in practice than it does to examine the patient for the symptoms.  It is also a simple process with menus and choices kept to a minimum.  A user can see how to use the program without any manual or training.  No typing is required, everything operating by mouse click.

At the last stage the user can save the patients record and print a report.  The saved record can be e-mailed to the A&E ward, taken on disk with the patient or stored in a pocket PC.  The record is useful for follow up, when the patient is next examined, the record can be loaded and the user can then see if the previously entered symptoms have reduced or new symptoms have appeared as the progress of the condition continues.  This may confirm or reduce the likelihood of the selected agent being responsible. 

The print record lists all the symptoms shown by the patient along with the percentage of patients that normally exhibit that symptom when exposed to the suspected agent.  It also describes the agent, sources of reference and any other symptoms shown by the patient along with the date, time and any other data the user wishes to enter in the notes about the patient.  This record can be printed off to form a paper record of the “computerised consultation” to add to the patient’s records.

Data Verification
All the data included in this program has been extracted and verified from as many sources as possible.  Our research team has combined 60 years of experience of analysis of technical and research papers in the biomedical sciences.  The preferred data is a study of multiple patients in a reputable, peer-reviewed journal.  Unfortunately (or fortunately) for many of the rare agents we have very little information of their effect on humans.  Sometimes this is because the results are not published (the agents may have only been used by a certain country, which may even deny the agent’s use) or because they have yet to be used against humans and all the data available comes from animal studies or from theoretical estimates of their effect from studies of comparable agents.

To cope with such uncertainty all data is coded with a “security” tag that identifies how reliable the data is based on the concept of a “notional” exposed population.  The more secure the data, the higher the score that will be obtained when those symptoms are identified.  The program “remembers” the actual patients it has seen and the symptoms exhibited.  If the actual patients show different symptoms to the “classical” picture, the program can (if the user wishes) use this information to alter the database.  Depending on how secure the original data was and the number of patients seen the database will change to reflect the actual symptoms shown on the field.  In this way the program can adapt to any particular different scenario (if for example the agent has been modified or the people exposed are unusual – old for example).