Electronic Health Records

Use & Optimization News

Google Study Uses Entire Patient EHR for Predictive Analytics

Researchers at Google showed using all data contained within a patient EHR can improve accuracy for predictive analytics.

Comprehensive Patient EHR Data Improves Predictive Analytics

Source: Thinkstock

By Kate Monica

- A recent study by researchers from Google, University of California San Francisco (UCSF), Stanford University, and University of Chicago Medicine (UCM) found representations of a comprehensive patient EHR using Fast Healthcare Interoperability Resources (FHIR) can be used for more accurate predictive analytics.

Researchers used de-identified EHR data from UCSF and UCM gathered from 2009-2016 during inpatient and outpatient encounters. Datasets included patient demographics, provider orders, diagnoses, procedures, medications, lab values, vital signs, and flowsheet data. In total, the study included data about 216,221 hospitalizations involving 114,003 patients.

Researchers then used a single data structure to predict health outcomes instead of requiring custom datasets for each new prediction.

“This approach represents the entire EHR in temporal order: data are organized by patient and by time,” noted researchers in the report. “To represent events in a patient’s timeline, we adopted the FHIR standard.”

Ultimately, researchers determined deep learning could produce valid predictions across a variety of clinical problems and health outcomes. Researchers were able to predict outcomes ranging from mortality rates to readmissions, as well as length of stay and diagnoses. Additionally, the single predictive model could be used at different medical facilities.

“A deep learning approach that incorporated the entire electronic health record, including free-text notes, produced predictions for a wide range of clinical problems and outcomes that outperformed state-of-the-art traditional predictive models,” wrote researchers.

The study showed the potential for advancements in the scalability of predictive analytics models in clinical care, researchers said.

“First, our study’s approach uses a single data-representation of the entire EHR as a sequence of events, allowing this system to be used for any prediction that would be clinically or operationally useful with minimal additional data preparation,” they stated.

Additionally, using a patients entire EHR for every prediction allows clinicians to incorporate more information for improved prediction accuracy.

“For predictions made at discharge, our deep learning models considered more than 46 billion pieces of EHR data and achieved more accurate predictions, earlier in the hospital stay, than did traditional models,” said researchers.

While the predictive model was largely successful in improving the accuracy with which clinicians are able to predict patient health outcomes and diagnoses, researchers said predicting a patient’s full suite of discharge diagnoses proved difficult.

“First, a patient may have between 1 and 228 diagnoses, and the number is not known at the time of prediction,” wrote researchers. “Second, each diagnosis may be selected from approximately 14,025 ICD-9 diagnoses codes, which makes the total number of possible combinations exponentially large.”

Despite potential problems associated with predicting patient discharge diagnoses, researchers stated the study serves as a proof-of-concept for gaining a diagnosis from routine EHR data. The ability to predict health outcomes or diagnoses from routine EHR data could help to improve clinical decision support or clinical trial recruitment.

“Accurate predictive models can be built directly from EHR data for a variety of important clinical problems with explanations highlighting evidence in the patient’s chart,” researchers concluded.

Findings from this study could be used to encourage clinicians to leverage comprehensive patient EHRs for improved predictive analytics and health outcomes.

Another recent study published in the American Journal of Managed Care (AJMC) showed the value of utilizing diagnostic EHR data to lower patient mortality rates and improve health outcomes. Researchers found sharing diagnostic EHR data within health systems and hospitals is associated with lower rates of patient mortality and reduced hospital readmission rates for patients with heart failure and pneumonia.

The exchange of radiology images, lab results, and other kinds of diagnostic data allows for more accurate, informed patient care. 



Sign up to continue reading and gain Free Access to all our resources.

Sign up for our free newsletter and join 60,000 of your peers to stay up to date with tips and advice on:

EHR Optimization
EHR Interoperability

White Papers, Webcasts, Featured Articles and Exclusive Interviews

Our privacy policy

no, thanks

Continue to site...