Building TreatGx Logic Trees for Personalized Prescribing Options

29 Nov Building TreatGx Logic Trees for Personalized Prescribing Options

TreatGx is an evidence-based medication decision support system that provides personalized prescribing options for the management of common conditions.

TreatGx logic trees include:
1. status of the condition being treated
2. previous non-pharmacological and pharmacological treatment(s)
3. current non-pharmacological and pharmacological treatment(s)
4. other diagnoses
5. pharmacogenetics
6. renal and hepatic function
7. other biophysical variables e.g. weight, INR

The TreatGx medication decision support system uses prescribing algorithms to offer personalized medication options. Each algorithm is condition-specific and informed by the highest levels of evidence. It takes more than 400 hours to develop an algorithm for each condition; including identifying the evidence, summarizing the evidence, consulting with experts, expanding the algorithm and releasing the condition to subscribers.

Identifying the evidence
• We create sensitive searches to find all published guidelines describing the pharmacological treatment for the specific condition/disease.
• We evaluate every guideline recommendation which usually involves going back to the original research.
• We identify evidence from guidelines from around the world for example, BC guidelines, the National Institute for Health and Care Excellence (NICE), American Medical Association (AMA), and the World Health Organisation (WHO).
• We also look at provincial guidelines if they exist.
• We review secondary sources such as the British National Formulary and RxFiles.
• We look at evidence from systematic reviews those found in the Cochrane Database.
• Where we identify gaps in guideline recommendations, we search for additional evidence from published trials, drug product monographs and drug-drug interaction databases.

Summarizing the evidence
• We use an evidence-based hierarchy, with all evidence being subjected to critical appraisal.
• All the data required to run the algorithm for the management of the condition is extracted and tabulated.
• The information is condition-specific and includes all the variables needed to identify the medication options and the drug doses. For example, to identify the optimal drug options in depression, prolonged QT interval is required, and for hypertension racial origin is required. All these variables, including the pharmacogenetics, determine which drugs can be safely and effectively used for that individual with the specific condition.
• Using the evidence, we build a preliminary logic tree/algorithm which incorporates relevant variables to help determine the best treatment options.

Consultation
• Preliminary logic trees/algorithms, along with all the evidence, are presented to an interdisciplinary team of expert pharmacists, geneticists, physicians, and epidemiologists.
• If necessary, we ask specialists to attend these meetings to review the evidence and give their expert opinion. The algorithms contain few surprises for clinicians reviewing the disease except in the complexity of the algorithms and the large number of variables that need to be considered for each condition.
• All processes are iterative, many searches are done, and evidence is presented to the team multiple times.
• A list of drugs is sent to an external pharmacist who uses more than one source, to document and categorize drug contraindications.
• All supplementary evidence is checked by the GenXys team and presented to the interdisciplinary team.
• All processes are documented to ensure that each decision node has clear supporting evidence.

Expanding Logic Trees
• Logic trees/algorithms are expanded and drugs that have the best levels of evidence for their use are added.
• Drug dosing and adjustments for pharmacogenetics as well as renal and/or hepatic insufficiency are made.
• Clinical considerations including drug availability in the jurisdiction that TreatGx is being deployed, drug formulation and dose availability are accounted for.
• The relative costs of drugs for each condition are determined and a price comparison is included.
• Expanded logic trees are presented to the interdisciplinary team.

Release
• All algorithms are extensively trialled prior to uploading to the TreatGx live site.
• No algorithm is ever considered finished.
• Minor updates are made quarterly as new evidence or guidelines become available and major updates and reviews are made annually for each algorithm.
• If there is new high level evidence that indicates a change should be made rapidly, the team will update the decision support system as soon as possible, usually within one week.