EHRs: healthcare’s crystal ball

Authors:
Xiongcai Cai, Oscar Perez-Concha, Enrico Coiera, Fernando Martin-Sanchez,
Richard Day, David Roffe, Blanca Gallego

What’s the research –
This study tested a model for estimating the daily probability that a hospitalized patient will remain in the hospital, be discharged, readmitted, or die, regardless of their medical condition. The model uses the results of each new pathology test recorded in the EHR system to estimate how much longer a patient will stay in hospital, whether they are likely to readmit or die for each day up to 7 days in advance.

Why it’s important –
We hear it time and time again that hospital inpatient services cannot keep up with demand. At the same time hospitals are under pressure to be more efficient, reduce lengths of stay and increase patient throughput. Of course, discharging patients too early, when it is not clinically safe to do so, increases the chances of patients returning to hospital or worse, dying. With the volume of data that electronic medical record systems can aggregate quickly and in real time, clinicians and managers can very powerfully and with greater accuracy determine which patients are ready to go home and those who are not. Standardised care plans are a great tool to ensure consistency of care. But what if you could blend care plans with actual data from the patient’s own record, while the patient is still in hospital to make safer, more personalised decisions about discharge. For the benefit of patient care, a model that can predict with a high degree of accuracy future patient outcome is welcome.

Key Learnings –
This paper demonstrates use case for EHRs in predictive analytics. Provided there is good quality data held within EHRs, it is possible to forecast future daily probabilities of patient outcome in relation to length of stay, timing for discharge, or dying. So with the help of EHRs, deciding when a patient is ready to go home after their hospital stay needs to also consider the chances that they will return within a few days. Using this model, it is possible to predict with 80-93% accuracy, the risk of death occurring within the following week.

Who should read this paper –
Hospital administrators who need to make decisions about how busy and how full to run their hospitals. Discharge planners and treating clinicians who ultimately decide when a patient is ready to go home tend to use standardised care plans to make the call. Now they can use metrics from the patient’s own EMR to determine readiness for discharge.

To read the entire journal article, click here.