August 04, 2022: – For all their benefits, EHRs aren’t so reasonable at detecting modifications in patient status that divide those being monitored and those who end up admitted. The nursing staff is often left to comb and click through papers in an endless game of finding and filing documents.
But at Doylestown Health, AI and algorithmic technology are producing this more efficiently.
The suburban Philadelphia healthcare network, positioned around an autonomous 270-bed hospital, uses predictive analytics technology from XSOLIS to enhance medical utilization management. In the first six months of use, administrators say they have improved observation swiftness by 20% and observation to inpatient conversion rates by 37%. And after three years, the initial return on investment of 4.6x has now enhanced to 7.3x.
Mary Beth Mitchell, MSN, RN, CPHQ, CCM, SSBB, senior executive director of care transformation strategies at Doylestown Health, headed this transformation, as well as hospice/palliative care and clinical documentation improvement while steering the hospital’s case management department. Talking to Health Leaders, Mitchell says hospitals would like to be able to admit all presenting patients, but payers demand observation status as a less-costly alternative depending on how sick the patient is. That generally does not last more than 48 hours.
“We are needed contractually to review and assure that we have the patient in the correct status so that when we bill the insurer, we are billing appropriately,” she expresses.
To be sure they are in the proper status, Mitchell says utilization review (UR) nurses must review every patient admitted and placed in a bed, whether on observation status or inpatient status. These nurses create patient summaries that are sent to the payer, who then can decide or disagree with the hospital’s position assigned to the patient.
Before embracing the XSOLIS technology platform, those nurses would, daily, begin at one end of the 270-patient roster, either by a payer or by floor, and work their path through to the other end of the floor, one chart at a time, to look for differences in patient status that rise to the verge of changing position from observation to inpatient or vice versa, Mitchell sounds.
“I could peek at a chart in the morning, and the patient looks suitable for observation,” she says. “But during the period of the day, lots of things happens to patients. But [UR nurses] are not going to glance at that chart again till the next day, because this is a manual operation.”
Mitchell states that some hospitals start with specific diagnoses but still assume what they will find in those charts.
The technology platform “allocates a severity for us, and through their AI platform [we] are able to utilize that severity to foretell that the patient should be inpatient or observation status,” she says.
The technology regularly combs through each chart, looking for events entered by clinicians and informing UR nurses when those events grow to the level of indicating a change in status, Mitchell adds. “It can be termed as an assistant, re-reviewing your charts regularly,” she says.