Features is also available as a PDF here.
Diagnosis may be the most complex intellectual task in medicine. Diagnosis requires inverting knowledge learned by disease to reasoning from findings to likely disease candidates in a probabilistic way. Even more, diagnosis is necessarily iterative – the diseases under consideration determine the next most useful question to ask, which in turn affects the probability of the diseases under consideration in the differential diagnosis.
SimulConsult’s Phenome Analyzer is a Diagnostic Decision Support System (DDSS). The company’s mission is to make each clinician their best diagnostician,while also saving them time.
To do this, SimulConsult has curated medicine into a computational database. The database currently covers ~7,400 diseases, including all Mendelian genetics, pediatric neurology, pediatric rheumatology, everything in their differential diagnosis, and increasingly all of pediatrics.
The fundamental tasks performed by the DDSS are as follows:
- Computes differential diagnosis. Compares the findings (symptoms, signs and test results) you enter about your patient with every disease in the database and computes a measure of fit resulting a probability weighted list of possible diagnoses.
- Suggests useful findings. Computes which findings among its ~10,000 would most change the differential diagnosis and displays them in rank order, and then recalculates the differential diagnosis and useful findings iteratively each time a finding is entered.
- Provides tips to resources. Enables you to learn more and get right to the relevant resource, saving you time.
- Reveals its logic so you decide if you believe it. Research has shown that DDSS are only adopted if the physician can assess the logic, because no clinician wants to be in the position of answering the question “why did you diagnose x” with “the computer said so”.
Note: To reveal the logic means the DDSS does not use machine learning-based artificial intelligence, but rather manual curation by expert clinicians. Machine learning is lousy at unlearning while our understanding of medicine continues to change and many subtle facts about rare diseases have never been written down and are only known to specialists, so there are no training sets from which the machine could learn.
Time course of findings. Critical to the performance of the DDSS is the fact that the database has the time course of each finding in each disease, and you specify the time course in your patient when you enter findings, when known. The DDSS then uses time-based Bayesian logic to do a pattern match.
Pertinent negatives. Diagnoses of exclusion and pertinent negatives are a critical part of genetic diagnosis; the DDSS allows you to specify that a finding is absent.
The main screen. After you enter the patient age and sex (and family history where you have it), and add an initial finding, you will go to the main screen, “Diagnose”. It has four tabs:
- Differential. This shows the evolving differential diagnosis
- Add findings. This displays clinical findings not yet commented on, in order of usefulness
- Add tests. This displays test results not yet commented on, in order of usefulness. (you can select to see only one kind of test, and some test results are “bundled” such as imaging results
- Phenotype. These are the findings you have entered about the patient, both pertinent positives and negatives.
|Basic demographic||You must enter the age and chromosomal sex of the patient to get started. It gets entered on the initial screen|
|Search||Begin typing a finding and a list of findings matching that term will appear below and you can select the appropriate one.|
|Family history||On the initial screen, this feature allows you to specify if others in the family are affected, if known.|
|Core finding list||Found on the search page, these lists of findings are available in some specialties, like pediatric neurology but notably not genetics. The lists consist of ~40 findings describe the common findings in most conditions (most useful outside genetics where there is no good list of 40).|
|Onset / Absence||Next to each finding is a pull-down menu with a “?”. This is where you specify onset or absence, and “add the finding to the patient’s phenotype”. Be as specific as the information you have and to maximize efficiency, start with the most unusual finding.|
|Primary, secondary or tertiary||On the initial screen, this feature allows you to ignore the incidence of the disease, and is useful for when a patient has already had many conditions ruled out or where the disease may be newly characterized and the incidence not well known.|
|Require finding||Occasionally, a finding is so striking or so intense you want to be sure everything you consider has that finding. The asterisk before the onset allows you to specify required. Leaving a finding not required means the DDSS will consider the possibility that any particular finding is incidental to the diagnosis.|
|Tips||Click the button of the name of the finding and then click “Tips” up on the black bar to go to related resources about the finding. Hover over the name of the finding to see synonyms.|
|Profile||Click the button of the name of the finding and then click “Profile” up on the black bar to see all diseases with this finding|
|Assess||Click the button of the name of the finding and then click “Assess” up on the black bar to see all diseases in the differential diagnosis with this finding|
Diseases in the differential diagnosis
|Rule in or out||Click the button of the name of the disease and then click “Rule in or Rule out” at the top of the Add Findings or Add Tests tabs|
|Tips||Click the button of the name of the disease and then click “Tips” up on the black bar to go to related resources about the disease. Hover over the name of the disease to see synonyms.|
|Profile||Click the button of the name of the disease and then click “Profile” up on the black bar to see all the findings in this disease, and click younger or older to see how they emerge over time|
|Assess||Click the button of the name of the disease and then click “Assess” up on the black bar to see the fit between the patient and what the database knows about the disease – this is the reveal of the logic|
Tools on the Share Button (right-most on black bar)
|Save and reopen patient||You can save or reopen patient, either using a text string or using the summary report.|
|SOAP note||Automatically outputs a SOAP note format with the data you entered|
|Summary report||Assembles the pertinent positives and negatives, differential diagnosis and most useful tests into a graphical display. Document is also machine readable.|
|Summary report (findings only)||Assembles the pertinent positives and negatives, with Human Phenotype Ontology (HPO) codes, without showing differential diagnosis, so it can be saved to the electronic health record. Document is also machine readable.|
|Prognosis table||Useful for Patient Education and communication between specialists and primary care. Characterizes the natural progression of the selected disease.|