- A web calculator based on a machine learning algorithm developed at Massachusetts General Hospital accurately predicts the need for nonhome discharge after surgery for lumbar spinal stenosis
- Variables considered are age, sex, body mass index, ASA class, functional status, number of levels included in surgery, fusion (yes/no), diabetes (presence and type), preoperative hematocrit and preoperative serum creatinine
- Using the calculator could lower costs, make hospital departments run more efficiently and reduce the complications associated with hospital stays, such as hospital-acquired infections
Discharge after orthopedic surgery is often delayed while patients wait to be placed in a rehabilitation or nursing facility. Delayed discharges incur unnecessary costs, hamper efficient delivery of care and increase the risk of complications such as hospital-acquired infections and adverse drug events.
Subscribe to the latest updates from Orthopaedics Advances in Motion
In a proof-of-principle study, Joseph H. Schwab, MD, chief of Orthopedic Spine Surgery at Massachusetts General Hospital, and colleagues showed that a machine learning algorithm could predict discharge placement for patients who underwent spondylolisthesis surgery. Machine learning, a form of artificial intelligence, allows algorithms to learn and self-improve from experience without explicit programming by a data scientist.
The same research team has now created a machine learning algorithm that provides individualized predictions for discharge placement after lumbar spinal stenosis surgery, according to a report in the European Spine Journal.
The researchers identified 28,600 patients in the American College of Surgeons National Surgical Quality Improvement Program database who underwent decompression, fusion or fixation for lumbar spinal stenosis between 2009 and 2016. The median age of the patients was 67 and 53% were male. The non-home discharge rate was 18.2%.
Developing and Testing the Algorithm
In random forest regression analysis, the following variables were independently significant: age, sex, body mass index, American Society of Anesthesiologists class, functional status, number of levels included in surgery, fusion (yes or no), diabetes (presence and type), preoperative hematocrit and preoperative serum creatinine.
Four types of machine learning algorithms (boosted decision tree, support vector machine, Bayes point machine and neural network) were trained with these variables to predict which patients were not discharged home. The neural network proved to be best in terms of discrimination, calibration and overall performance.
The dataset was split 80:20 into a training set and a testing set. The neural network was used in the testing set to predict discharge placement. These predictions were compared with the actual outcomes of the testing set to assess the performance of the algorithm outside the training set.
The researchers developed the algorithm into a web-based calculator. The user simply inputs the necessary variables to determine the probability of nonhome discharge. The calculator can be accessed here.
Integrating the Calculator into Practice
Accurate preoperative prediction of which patients will need a rehabilitation or nursing facility will allow reservation of a place in advance and earlier insurance precertification, the authors note.
In their earlier paper, they recommended that a nurse practitioner or case manager use the calculator during preoperative visits with the patient, after the visit to the surgeon and anesthesiologist when all variables are known. This would allow for patient education, and discharge planning might be started at that time.
The researchers now feel confident that their methodology can be implemented for a variety of other diseases and elective treatments, which could avoid risks associated with delayed discharge and lower costs. They add, though, that external validation will be crucial, especially if the algorithm is to be used outside the U.S.
Learn about the Orthopaedic Spine Center
Refer a Patient to the Department of Orthopaedics