Automating Medical Risk Calculators

As all health providers know, there can never be enough tools in the toolbox to help deliver the best quality care. However, providers will also tell you that not all tools are created equal and while some are great, others make their workflow more cumbersome. From published guidelines to algorithms to risk assessments, the modern provider is not lacking in information, but lacking in hours of the day in which to choose and implement the best tools and solutions.

Your provider is at the crossroad of a multitude of information and data streams, and they are expected to wear different hats that change as fast as new situations arise. From administrative to clinical roles, their tasks can encompass an entire spectrum of needs, requiring more efficient and effective tools in order to best fulfill their duties. More often than not, when care is not optimally delivered it is not due to a lack of diligence and effort, but instead is a reflection of the bombardment of responsibilities providers face.

Big Data vs. Usable Data

At Lexigram, we are singularly focused in how data can work for the medical provider in their everyday tasks. While working with providers, we began to see how every piece of medical data should not only be accurate, but timely and readily available as well. We saw how often providers utilized tools such as risk scoring algorithms and medical calculators, and were surprised to discover how often the patient data required to utilize these tools is buried deep in the heart of the electronic medical record, often requiring a significant amount of time and dedication to uncover. We thought, “That’s unacceptable.”

We committed to looking for a medical use case that would not only provide a useful tool, but also serve as an indispensable part of a provider’s practice. To that end, we have applied our technology, and implemented a proof of concept version of the Opioid Risk Tool For Narcotic Abuse.

The Opioid Risk Tool For Narcotic Abuse

The use and abuse of opiates has increased dramatically in the last 15 years, causing a commiserate increase in deaths due to opiate use. During this 15 year period, the number of prescription opioids sold in the U.S. has quadrupled, to the degree that 91 people per day die due to opioid overdoses 1.

If you are a provider on the frontlines of this epidemic, one of the key ways to prevent opioid addiction is knowing the likelihood that a given patient is at risk for abuse. This allows the provider to seek alternative treatments or follow a patient more closely than they might have otherwise. One invaluable tool was created by Dr. Lynn Webster, and is aptly called the Opioid Risk Tool 2. The tool requires the provider to assess the patient’s risk based on five questions, the answers of which, can be found within the medical records. Indeed, the depth at which this data is embedded within the records is perhaps the single most difficult part of using the tool.

Screenshot of Lexigram's version of the Opioid Risk Tool

Lexigram’s version of the Opioid Risk Tool combs through unstructured patient data – from notes, reports, medical codes etc. – looking for the information that are known risk factors. As the provider validates the data Lexigram surfaces, the patient’s risk score is updated. Providers can also supply information that they know about the patient that wasn’t in the unstructured data or, if seeing the patient, they can ask follow up questions as needed. We have created a real-time process that allows providers to review and validate the clinical data, instead of spending that time searching for a needle in the medical haystack.

Next Steps

Lexigram’s version of the Opioid Risk Tool is a proof of concept, but it does show the power of utilizing clinical NLP and other machine intelligence approaches to aid in the automation of routine health care tasks. This is only the first step toward our goal of transforming Big Data into usable data.

We see opportunities for Lexigram’s technology to power other commonly used provider scoring tools, such as the Framingham Coronary Heart Disease Risk Score, Cardiac Risk Index for Pre-Operative Risk, and CHADS2 scoring. There are examples in nearly every specialty where the accumulation of data has allowed similar score calculations to be performed. From calculating the risk of cardiovascular events to the likelihood that a patient will develop diabetes, most of the data points are already documented somewhere within the unstructured medical data.

Lexigram is eager to work with health organizations who would like to improve their quality of care and make their clinical staff more efficient and effective, either with the Opioid Risk Tool For Narcotic Abuse calculator or another tool that might be relevant to your setting.


2: Webster LR, Webster R. Predicting aberrant behaviors in Opioid-treated patients: preliminary validation of the Opioid risk tool. Pain Med. 2005;6(6):432

The Flow of Data In Healthcare (Part 2)

In Part 1, I discussed how data is generated by patients and at the point of care and a few different ways that data can flow between the different processes that regulate the flow of clinical information. This post will outline three types of organization that exist outside of the patient/caregiver relationship but still have access to almost every piece of data that the healthcare system generates about you.

Ancillary Services

Healthcare providers rely on a variety of vendors to assist in delivering efficient care. These include outside labs, pharmacies, medical coders and transcribers, and technology providers. Each of these organizations touch a slice of the data.

Medical coders review the patient’s chart to translate the data generated by care providers into codes that are used to track patients through the diagnostic and billing process. Many health systems employ their own coders while some choose to outsource.

Technology providers get a wide range of access to data. Some work only with codes while others process the entirety of the medical record or handle all data transmissions, including emails, faxes, and telephone calls.

Labs and pharmacies both get orders that are relatively structured and tracked with a common set of codes. Labs get actual living samples from patients, a unique type of data in the system. This group of service providers also communicate directly with insurance companies or other payer organizations.


Health insurance companies and government payers (CMS and state-level payers) have the broadest view of a patient’s interactions with the system but, just like when you zoom out on a map, their view isn’t very detailed.

The most frequent kind of patient data that a payer will interact with are billing and diagnostic codes. Diagnostic codes are used to track the diagnoses a provider makes when a patient comes to them with a problem. When a doctor performs a procedure, such as an office visit or a mole removal, these are tracked with a procedure code. The payer uses the diagnostic and procedure codes to determine how the provider will get reimbursed.

Payers can also request additional data from the provider in support of a claim. They will also routinely audit patients using data collected from providers. Payers need to collect data from all of the providers who a patient sees, adding complexity to their data transmission process.

Outside Agencies

Along with the rapid adoption of EHRs due to the HITECH Act, health information exchanges began to spring up to address the need for information sharing between provider organizations and, in some cases, insurance companies. There have been varying levels of adoption when it comes to data sharing

Government agencies can also request access to health records. This includes law enforcement organizations in the course of investigating a crime or locating a suspect or as a part of a terrorist investigation. Government agencies responsible for handling disability claims, worker’s compensation, or Medicare/Medicaid will also have access to medical data.

The Flow of Data in Healthcare (Part 1)

Data moves through different parts of the healthcare system in obscure and confusing ways that range from modern to arcane. It’s a struggle to know exactly what data gets produced, where it gets sent, and who has access to it. Continue reading “The Flow of Data in Healthcare (Part 1)”