Real-Time Prediction Algorithm Reduces Overnight Vital Checks, Promotes Sleep in Patients

Sleeping man, OSA
Sleeping man, OSA
Researchers sought to determine whether a real-time clinical decision support tool can help physicians identify clinically stable patients and safely discontinue overnight vital sign checks.

A real-time prediction algorithm can help physicians identify clinically stable patients that can be safely removed from overnight vital sign checks to improve sleep, according to a study published in JAMA Internal Medicine.

Insomnia is common among hospitalized patients, often because of nighttime vital sign checks. Prior research has shown that these iatrogenic interruptions could be reduced in low-risk medical inpatients. Recognizing the effect of sleep promotion clinical decision support (CDS) required the use of a randomized control group.

The objective of this randomized clinical trial was to determine whether an algorithm in a real-time CDS tool could help physicians appropriately select low-risk patients and successfully promote their sleep.

Researchers analyzed vital signs of inpatient encounters for general medicine services at the University of California, San Francisco, Medical Center to develop the prediction algorithm. A logistic regression model trained on 70% of the data set was able to correctly predict most vital signs (84% normal nighttime vital signs; 70% abnormal vital signs). The model triggered a sleep promotion vitals (SPV) CDS alert in electronic health records (EHR) to indicate to the primary team members during the day that the patient was 90% likely to have normal vital signs during the upcoming night. The doctor could then choose to order SPV, delay notification for 1 hour, or delay until the next day.

A total of 1930 inpatient encounters from March 11 to November 24, 2019 were randomized to physician notification (n=966; 41% women aged 53±15 years) and without notification (n=964).

Physicians who received notification could order SPV. Encounters with this order nearly doubled (770 vs 430; P <.001) before a 31% reduction in encounter-level blood pressure checks per night (0.97 vs 1.41; P <.001).

Delirium (Nursing Delirium Screening Scale (Nu-DESC) score of at least 2) was similar between the intervention and usual care groups (108 vs 123).

Individuals in the intervention arm had more sleep opportunity compared with the control individuals (4.95±1.45 vs 4.57±1.30; P <.001).

The 5% of patients (53 intervention group individuals; 49 control individuals) who completed a discharge survey did not report significantly different noise in or around their room at night. Safety outcomes were similar among the groups.

Study limitations include the reliance of the predictive algorithm on data derived from the researchers’ institution, rate of carrying out SPV orders, and encounter-level randomization.

“While this randomized clinical trial found no difference between groups in the primary outcome, delirium incidence, the secondary findings indicate that a real-time prediction algorithm embedded within a clinical decision support tool in the electronic health record can help physicians identify clinically stable patients who can forgo routine vital sign checks, safely giving them greater opportunity to sleep,” the researchers stated.

They also noted that the algorithm may be useful for assessing need for continuous cardiac monitoring, higher level of care, and routine daily blood tests.


Najafi N, Robinson A, Pletcher MJ, et al. Effectiveness of an analytics-based intervention for reducing sleep interruption in hospitalized patients: a randomized clinical trial. JAMA Int Med. Published online December 28, 2021. doi:10.1001/jamainternmed.2021.7387