“There are more than 9,000 billing codes for individual procedures and units of care. But there is not a single billing code for patient adherence or improvement, or for helping patients stay well.”

Clayton Christensen, Innovator’s Prescription

Here are additional problems in the primary care clinic that can be ameliorated with smart use of digital health and artificial intelligence (AI):

During the Clinic Visit

Problem: A clinician may not have all the information to make the correct diagnosis, especially outside of his/her subspecialty. 

Too often the clinician lacks all the data and information to consider the entire list of differential diagnoses, especially outside laboratory or diagnostic tests that have already been performed for the condition. There needs to be a system that “connects all the healthcare data dots”.

Solution: Knowledge graphs with machine learning and a graph database can connect the dots to expedite the workup that the clinicians have to gather for a diagnosis.

In order to expedite the workup, a knowledge graph with a machine learning tool can connect these dots for the clinicians as they work up the patients for certain diagnoses. For instance, if the patient had a syncopal episode, a family history for long QT syndrome and hypertrophic cardiomyopathy could already be gathered in case the clinicians do not remember to seek this information.

After the Clinic Visit

The primary care clinician will often order additional testing after the visit, but there can be logistical delays or obstacles. In addition, the test results are sometimes difficult to track and hard to review with the patients.

Problem: Procedures cannot be performed in the office due to authorization not being available at the visit. 

Too often a procedure such as continuous EKG monitoring cannot be completed due to lack of prior authorization. The inability to complete the procedure during the visit results in delays and affects quality of care.

Solution: Using robotic process automation (RPA), an automated application for prior authorization for procedures can be completed even before the visit. 

Based on the history and the practitioner’s practice, a patient’s approval for a procedure can be obtained prior to the visit in order to expedite the workup. For instance, a patient who is referred for frequent palpitations can have a continuous EKG monitor approved prior to the visit. RPA can be ideal for these predicable situations. 

Problem: Patients do not get the appropriate intervention that is recommended by guidelines or by standard of practice. 

There are many well-intentioned therapeutic offerings such as specialized follow-up clinics or recommended vaccines that fail to be executed because clinicians simply cannot remember to institute the measure, or the patients fail to show up.

Solution: Using RPA, an automated list of qualified interventional measures can be customized for each patient. 

Use of the aforementioned automated tools can be effective in putting together a list of appropriate follow-up clinics or therapies that patients qualify for. This results in better quality of care as these are the recommended guidelines and practices for patients who need to have these therapies, but human factors often do not result in perfect compliance.

Problem: Most results of laboratory tests or diagnostic studies are not forwarded to patients and families with explanations. 

Many tests that are borderline normal, such as a continuous rhythm monitor with few premature atrial contractions without SVT, are either not forwarded or not explained to patients and families. The abnormal tests are also not adequately triaged. These results create a significant burden for clinicians.  

Solution: Natural language processing (NLP) tools can triage the test results so that patients and families can be appropriately notified with information and guidance. 

Test results can be forwarded and coupled with an adequate explanation with NLP and machine learning so that patients and families can be reassured in a timely fashion. In addition, abnormal results can be tagged and forwarded to clinicians with a brief clinical history so that the patients and families can be informed.

In conclusion, AI as a resource is vastly under-leveraged in the primary care clinic, before, during, and after a visit. Future solutions to common problems in the clinic setting can be found in the smart use of AI tools such as machine and deep learning, NLP and RPA. All of these AI tools can be used singly or in combination to attain the Quintuple Aim for all patients, families, and clinicians.

Artificial intelligence in primary care settings, as well as other clinics, will be a topic of discussion at the in-person AIMed Global Summit 2023 scheduled for June 5-7 2023 in San Diego with the remainder of the week filled with exciting AI in medicine events like the Stanford AIMI Symposium on June 8th. Book your place now! 

Hope to see you there!


We at Ai-Med believe in changing healthcare one connection at a time. If you are interested in discussing the contents of this article or connecting, please drop me a line – [email protected]