Published on

Authors

# Smarter EMRs: Raising the Bar for Hypertensive Patient Care

The field of medicine and patient health was not immune to the benefits and innovations that came with the digital age. A key innovation in medicine has been the development and improvement of Electronic Medical Records (EMRs for short) that have massively improved the productivity and overall performance of healthcare providers and their organizations. It is time now, as we are slowly witnessing, for the Electronic Medical Record to undergo another transformation with the age of Artificial Intelligence. This time they are getting smarter!

To set some ground work, let's define what an Electronic Medical Record is and does. Not to be confused with an "Electronic Health Record", an EMR is simply the digital version of what you would find in a patient chart/file. A medical record contains a patients diagnoses, allergies, treatment histories, hospitalization records, immunizations, and other medical history. These medical records are usually specific to individual health facilities (hospitals and clinics), and are rarely accessible outside of that facility.

Tangent: The fragmentation of patient medical records leads to a "broken" and "incomplete" patient history, as a patient can have multiple medical records with different health facilities and no information is shared between the two. A later post will show how Elsa is built from the ground up to tackle this challenge by shifting the dynamic and making patients the owners and custodians of their own data.

## Understanding Hypertension - High Blood Pressure

Hypertension, defined as elevated blood pressure, is a serious condition where the pressure inside your arteries is high. Blood pressure is measured with two numbers:

1. Systolic Pressure: This is the pressure in your arteries when your heart "beats" or "contracts" - this is the active part of your heart beat.
2. Diastolic Pressure: This is the pressure in your arteries when your heart rests between contractions and beats - this is the inactive/relaxed part of your heart beat.
\begin{align} \frac{Systolic Pressure}{Diastolic Pressure} \end{align}

The two numbers can be self measured at home and are often correlated. When measured, they give healthcare providers a glimpse into your overall cardiovascular health.

Hypertension can be diagnosed if your systolic pressure is above 140 mmHg and/ or your diastolic pressure is greater than 90 mmHg when measured on two different days. It is important to know that many things can affect your blood pressure at the time of measurement, including caffeine consumption, climbing a set of stairs, or simply being nervous about the results of the readings.

## Understanding Artificial Intelligence and (its often-confused-with child) Machine Learning

Artificial Intelligence is changing how we interact with the world and is completely shifting how work and services are being delivered. From the simple nice-to-haves like face unlock/ facial recognition on your iPhone Ref, to the life and death applications of surgery Ref, it is clear to us at Elsa Health that Artificial Intelligence is here to stay. Simply put, Artificial Intelligence is non-natural/ man-made intelligence using knowledge from multiple disciplines like mathematics, computer science, information theory, and more.

Machine Learning is one way to build artificially intelligent systems where the AI developer shows the computer/ "machine" some examples and guides it to "learn" from those examples in the hopes that after it has learnt, it can be used autonomously or to augment human performance. Generally speaking, the more examples (ie: data), the better the "machine" can "learn", meaning it will be able to perform better in the real world as it would have seen more examples of a varying nature.

It is worth noting that in the grand scheme of things, and with respect to the potential of these technologies, both AI and ML are very young and new innovations that push the envelope daily. This also means there is some disagreement in the community and ecosystem over many things, including the very definitions of the terms.

## Leveraging the Data Troves of EMRs

It should come as no surprise that Electronic Medical Record systems can have massive amounts of data that, more often than not, is only used for historical record keeping for the hospital and for reporting purposes.

The usual challenges, more common in LMICs (Low and Middle Income Countries), of undigitized data are almost completely solved by EMR systems (provided adequate technical literacy and following best practices).

Assuming the best case scenario where proper care has been used when recording patient visits and histories, the data stored in hospitals and other health facilities can be extremely beneficial to patient care and in decision support for care takers and doctors.

Machine learning systems are showing promise in the both the early identification of Hypertension and in supporting the management of patients with Hypertension. This promise can be realized when applied to medical record systems, allowing them to be more supportive to the care providers and patients.

## ML/ AI in Identification and Prediction of Hypertension

Early identification, or better yet, ahead-of-time onset prediction of Hypertension can be crucial in management and intervention planning for both patients with Hypertension and those at high risk of developing Hypertension over the coming months.

Researchers have validated the use of Machine & Statistical learning from EMR databases using popular AI techniques like Gradient boosting Ref, Artificial Neural Networks Ref, Logistic Regressions Ref, and more. All these techniques have one thing in common, they use existing data to develop a mathematical/statistical representation of how the patient information relates to their hypertensive status. The learning of the representation from data is called "training" the model.

Once the model is trained, and assuming its performance is acceptable Ref, the algorithms can then be used to mine and learn from EMR records and either:

1. Flag patients who might have hypertension, but do not know their status, or
2. Flag patients at risk of developing hypertension over the next few months/years Ref

Two studies that research the use of algorithms (Gradient boosting using XGBoost Ref) show great performance:

## ML/ AI in the Management of Hypertension

The combination of powerful Machine Learning algorithms and Big Data Ref means management of Hypertension can be better planned and even more personalized.

Learning algorithms can analyse thousands of other patients' management plans to predict the most effective treatment plans for a specific patient. Furthermore, these algorithms can monitor patient progress and perform continuous risk assessment while providing recommendations for any alterations to the treatment plans.

Meta learners Ref have been successful in achieving exactly this as shown in this study by Liu C et. al: Clinical Value of Predicting Individual Treatment Effects for Intensive Blood Pressure Therapy

## How we, Elsa Health, "smarten-up" EMR systems.

The digital health ecosystem is full of "systems" and "platforms" that help health facilities manage their patients data. This ecosystem is also extremely fragmented, with many developers and maintainers of the digital tools developing only their systems and not prioritizing interoperability of health data. This results in a messy, siloed, inconsistent, and incoherent medical data landscape where the patients ultimately pay the price.

Because of this, Elsa is prioritizing the development of easily composable and applied health algorithms that any EMR service can use to easily add decision support and automatic information extraction to their databases.

### Reach out to us

To learn more about our work, or if you are interested in working together, please reach out to us through our website, or follow us on social media!