Article originally published on Healthcare IT Today
By Andy Oram
An earlier article, “Why Hospital Discharges Take So Long—And What We Can Do To Shorten Them,” explored the complexities of discharging and how hospitals are re-evaluating their workflows to get patients out faster. This article looks at some technical solutions.
Ali Parsa, CEO and founder of Babylon, an AI and digital health platform, says discharge planning doesn’t require “massive amounts of technology” or “rocket science”; just good planning from the beginning of the stay. Indeed, one research study (suggested to me by Dr. Geoffrey Rutledge, Chief Medical Officer and co-founder of HealthTap) achieved a significant time saving through low-tech measures such as checklists.
Other respondents, however, reported that their solutions were making a difference in discharges.
Weaknesses of Current Technology
With modern databases, it’s ridiculous for many care managers to be checking the availability and services of nursing facilities manually. They should be able to draw up a list of suitable facilities quickly and transfer data from the patient’s electronic record to a corresponding record at the chosen facility. And why can’t patients and caregivers have instant access to pictures, CMS ratings, and other information about the facility they’re considering?
Lu Zhang, founder and managing partner of Fusion Fund, a venture capital firm focused on investments in advancing healthcare tech, points out that the data with which many doctors work on a daily basis is still in Excel spreadsheets. Coding, she says, can take hours. AI can automate much of coding and billing. However, automated solutions don’t need AI to help clinicians enter data into the EHR.
Many hospitals also use pagers or telephones instead of asynchronous messaging systems.
Systems That Are Helping
Cynthia Davis, clinical transformation executive at Healthlink Advisors, says staff might have to fill out a form multiple times. Rutledge says that all the patient medications might have to be re-entered manually by the physician.
Donna Pritchard and Joy Avery of CipherHealth describe “centralized command centers” with large screens displaying all relevant information from different sources to help with admissions and discharges. They say that these command centers are used in many hospitals in the U.S. and U.K.
Gerry Miller, Founder and CEO of Cloudticity, described research at a multi-billion dollar hospital chain that analyzed millions of back histories and created a machine learning model to determine which patients are most likely to readmit. Case workers can be assigned to the most at-risk patients at the beginning of a stay instead of waiting until discharge is approaching. The chain achieved a 70% reduction in readmissions.
Tina Burbine, VP of Care Innovation and Enterprise Analytics at Healthlink Advisors, says, “We heal better in the home” and that analytics can help identify who is eligible for virtual care in the home.
Dr. Pallabi Sanyal-Dey, director of client services for inpatient beds at LeanTaaS, director suggests that hospital staff tend to rely on past experience. Data and AI can refine decisions through data gathering, predictive analytics, and automation. Systems developed through LeanTaaS can answer questions such as “What will the patient population look like a few hours from now?” and “Which departments will experience the greatest demand?”
She also says that the estimated date of discharge (EDD) is often chosen on the first or second day of a hospital stay, before all clinical data has been collected. Both Sanyal-Dey and Dr. Jason Cohen, Senior Director of Clinical Solutions at Qventus, say that if the EDD is not updated frequently based on new information, it loses value. These companies’ technologies, which look regularly at the level of care and how acute the patient’s conditions are, can help to update the EDD accurately.
Natural language processing (NLP) can extract data from the record that is relevant to discharge planning, such as medications the patient is taking. Cohen says that hospitals can achieve good accuracy with just three to six months of data, and should retrain their AI models each month. Qventus uses anonymized data from thousands of patients to determine what can reduce time to discharge. Data of discharge and disposition (such as skilled nursing facility, infusions, and home care) should be in place by the evening of the day after the procedure. Cohen reports that their processes eliminate 1.2 days on average from a hospital stay.
Automation can help plan each step toward discharge. For instance, the system can let the staff know that a patient with congestive heart failure hasn’t had an echo test for over a month, so they should schedule one before they are up against the discharge deadline. Or the payer requires three days to approve the admission to the skilled nursing facility.
Automating routine and mundane tasks also can free up clinicians to deliver more individualized care, spend time with those patients who need it most, conserve human resources, and reduce costs, adds Cindy Gaines, chief clinical transformation officer at Lumeon, a company that makes an automation platform to coordinate care.
EHRs now have fields to enter data related to social determinants of health, such as whether the patient lacks transportation or money for food and medication. Cohen expects that data to become more detailed and reliable. Such data can then feed into analytics to make discharge planning more accurate.
Dr. Darin Vercillo, CMO and co-founder of ABOUT, says that data and analytics can help make decisions. For instance, by scoring each patient along known indicators, hospital staff can know 90% of the time whether the patient will need to be discharged to a skilled nursing facility and can therefore plan on this well in advance of discharge.
Lisbeth Votruba, chief clinical officer at AvaSure, told me about their work with virtual nurses. No, these are not robots or AI creations; they are highly experienced nurses who work remotely to support bedside staff.
Like care managers, virtual nurses ensure that the hospital experience goes smoothly and that discharges take place as quickly as possible. They are particularly useful for documentation and patient education, but can also provide important emotional support to other nurses.
Votruba says that experienced nurses are in particularly short supply (as are all nurses). The field had a turn-over rate of 32% in 2021, and 52% of nurses are thinking about changing careers or retiring. So a virtual nurse can help with both speed and quality. Less experienced nurses appreciate having an experienced one to call up through two-way video when handling a difficult procedure.
Zhang and Miller point to the importance of accountability and explainability for AI. This kind of AI transparency is being demanded increasingly across industries, but is particularly important in health care where clinicians are making life-and-death decisions. If clinicians understand the strong and weak points of advanced IT such as AI, it can become more widespread in discharge planning.