The pMD Blog

Welcome to the
pMD Blog...

where we cover interesting and relevant news, insights, events, and more related to the health care industry and pMD. Most importantly, this blog is a fun, engaging way to learn about developments in an ever-changing field that is heavily influenced by technology.

POSTS BY TAG | Artificial Intelligence

Recipe for Success

As more health data is being acquired, the utility of that data in improving patient outcomes is also increasing. Artificial intelligence (AI) improves patient care by integrating health data across different platforms. In particular, some of the key areas for AI to improve health care include diagnostic imaging, patient risk analysis, enrollment in clinical trials, and patient communication

AI’s first success in health care was in the field of diagnostic imaging. Machine-learning algorithms can find patterns in images and identify specific anatomical markers. These algorithms can spot details the human eye can’t catch, perhaps greatly improving a radiologist’s workflow. When a radiologist can validate and add to an AI assessment rather than scanning hundreds of images, she has more time to actually spend with her patients. A recent example of AI improvement in radiology is LabCorp and Mt. Sinai in New York City partnering to create the Mount Sinai Digital and Artificial Intelligence-Enabled Pathology Center of Excellence. This partnership will provide doctors and patients throughout the United States access to a digital pathology solution that has been developed at Mt. Sinai. When complex cases come from a larger network, expert pathologists will have more time to analyze and diagnose such anomalies. 

Another area where AI can improve health outcomes is through finer-tuned risk analysis. Self-monitoring devices are giving patients and their doctors access to much more information about a patient’s lifestyle that can be crucial to plans of care. This data can be used to assess if patients are at risk of developing conditions such as diabetes, or even if they are at risk for non-compliance to medication or lifestyle changes. Recently, three Johns Hopkins hospitals set up machine-learning-based systems that constantly scan EHR, lab tests, notes, and radiology tests of every admitted patient. If a patient crosses a threshold for having a high likelihood of developing sepsis, the system sends an alert to just the most relevant providers.

AI can also greatly improve patient enrollment in clinical trials. The current process for clinical trial enrollment often places the onus on the patient to research and find eligible trials. With increased data on a patient’s demographics and diagnosis, integrated with inclusion criteria for ongoing trials, the whole process can be automated to get patients into life-saving treatment. This change can also increase clinical trial success rates if patients are only enrolling in trials that are more well-suited for their conditions. 

Finally, AI has the potential to completely overhaul the way in which patients communicate with their doctors. While current systems of communication have no way of ensuring if patients are understanding the information that doctors are conveying to them, metrics on patient responses can add some clarity to these conversations. There are several methods that are emerging for analysis of conversations between providers and patients. AI can look at things such as the language that providers are using (is it all jargon that the patient may not understand?), turn-taking (is the provider sending too much information all at one time and not allowing for questions?), and analysis of tone and speed of spoken communication to give providers insight into the patient’s understanding of his or her own medical condition. This analysis can also be helpful in the setting when a translator is used, ensuring that nothing is being lost in translation.

Additionally, chatbots are being piloted as an alternative where patients can get access to information on medication, doctors’ schedules or clarifications of doctors’ instructions. This can replace the unreliable information that patients find online that is not specific to his or her own conditions or perhaps just inaccurate. One research report by Juniper Research predicts that the success of chatbot interactions where no human interventions take place will go up to 75% in 2022 from 12% in 2017. This study also predicts that chatbots alone will save $8 billion annually by 2022. 

Also, chatbots can actually be used to deliver Cognitive Behavioral Therapy in much the same way that a psychologist would for patients. This can be particularly beneficial for patients in rural areas who cannot access adequate mental health care or others who do not have the financial means to pay for a mental health provider.  

Dr. Albert Wu writes that “the great irony today is that artificial intelligence may prove to be the force that helps bring that human touch back to the forefront of medicine.”





Artificial intelligence (AI) is a term we have been hearing a lot lately. Most of the time it concerns the variety of problems AI is solving along with future solutions to those problems. In recent years, health care has adopted AI technology in a number of areas including disease diagnosis, imaging analytics, virtual telehealth services, and most recently, EMR integration. The possibilities for the application of AI are endless, especially when it comes to optimizing common processes in the health care industry, such as medical billing and coding.

What exactly is AI?

Robots and flashbacks to Terminator might ring a bell for some, but AI is actually a very broad term with many branches of application. According to the National Institute of Standards and Technology, or NIST, “Artificial Intelligence (AI) refers to computer systems that think and act like humans, and think and act rationally.” Simply, AI is the capability to learn and problem solve by machines, in contrast to how humans display these behaviors naturally. For machines, the ability to learn and make decisions require lots and lots of data. An algorithm is needed to then test this data, until eventually the machine learns how to tell a bee from a flower, or better yet, identify enlarged lymph nodes in a CT scan of the chest while simultaneously scanning the patient’s medical record to determine if they have lung cancer. These current applications work to make physicians’ workflow more efficient. What if there were a way for AI to similarly improve the efficiency of medical billing?

What problems are the medical billing and coding space facing today?

It depends on who you ask, but the error rates in medical billing are reportedly pretty high. According to a study conducted by the American Medical Association in 2013, the error rate for medical billing was at 7.1 percent, while a study by Nerdwallet in 2014 put the error rate for Medicare claims at a whopping 49 percent. This can be due to many factors such as upcoding, duplicate billing, and even the digit transposing of ICD and CPT codes. Most of this is at the hands of human error, but certainly not without reason. With the transition change from ICD-9 to ICD-10 codes in 2015, the amount of codes increased by five times the original amount. Now having roughly 68,000 ICD and 8,000 CPT codes to work with, medical billers and coders have had to learn and process tons of new information. In this period of getting used to a new system, there was a 50% loss in productivity at first, but the industry has since adapted. The system will surely change again, and I can’t help but wonder if machine learning might be able to work a little bit of its magic in the medical billing and coding space.

Using AI to assist with medical billing and coding

By using machine learning techniques to process data, it may be possible to deploy an application that can process medical codes and billing codes, identifying errors and making appropriate corrections. AI could be integrated as an assistant and allow for medical billers and coders to work more efficiently and accurately by flagging mistakes and suggesting possible fixes. Having an automated process where computers take care of the more tedious and common clerical tasks, will allow for medical billers and coders to spend their time on the more difficult and comprehensive issues, such as insurance appeals and the claims process. The end result of introducing AI to medical billing and coding: less errors, more money, and more time!

What’s on the horizon for AI in health care?

It took nearly a decade to develop ICD-10 codes and another decade to implement. Innovations in health care can sometimes be referred to as “slow moving giants”, so it’s possible we won’t see AI completely taking over right away, but we can expect to see the adoption of machine learning more frequently. According to Healthcare IT News’ reporting on HIMSS18 “Allscripts, Athenahealth, Cerner, eClinicalWorks and Epic revealed big plans for adding AI into the workflow in forthcoming iterations of their electronic health records platforms.” As you can see, AI is no longer this untouchable, sci-fi enigma, but now is something that is actually being utilized in our day-to-day. There’s much more on the horizon for AI’s role in health care, so we'll just have to stay tuned!

 If you'd like to find out more about pMD's suite of products, which includes our MIPS registry, charge capture, secure messaging, clinical communication, care navigation, and clinically integrated network software and services, please contact pMD.