This retrospective study was approved by the Institutional Review Board (IRB 202300138) at Stony Brook University Hospital. The requirement for written informed consent was waived by the IRB.

Stony Brook Hospital is a tertiary care academic institution. Its ambulatory surgery center has 10 operating rooms and performs approximately 8–9000 procedures annually. Surgical specialties include orthopedics, plastic, general surgery, ophthalmology, ear nose and throat (ENT), urology, gynecology, colorectal, breast, and dental. Unplanned hospital admission was defined as patients who had an unplanned hospital admission within 24 h of surgery at our ASC. Admissions included any patient that was directly transferred from the ASC to the hospital as well as patients who were discharged home from the ASC but then were admitted within 24 h of their anesthetic care.

A query was submitted through our hospital’s Enterprise Reporting Department to obtain data for all adult patients ≥ 18 years. who underwent surgery at this ASC between January 1, 2016, to December 31, 2022. Data were extracted from the anesthesia EMR (Cerner, North Kansas City, Missouri). Missing data were treated as missing, i.e., not imputed. The query included patient information such as demographics including age, gender, Body mass index, duration of surgery (< 1 h, 1–3 h, and > 3 h), American Society of Anesthesiology (ASA 1–4) postoperative nausea and vomiting (PONV) score (0–5), type of anesthesia (general anesthesia, monitored anesthesia care, local, regional), surgical specialty (orthopedics, ophthalmology, urology, gynecology, plastic, breast, ENT, Dental, general and colorectal). We further condensed surgical specialty into General, (Colorectal, Breast, Plastic, Vascular, Surgical Oncology, Trauma), Ortho (Ortho, Pain), Gyn/Urology (Gyn, OB service, Urology), Ophthalmology, and others (Dental, OMFS, ENT, Neuro, Endo, Special procedures). Surgical duration was defined as surgical start time (time of incision) to surgical stop time (dressing applied). To avoid confounding, surgical duration was considered as a surrogate for surgical complexity with less than an hour as low, 1–3 h being moderate, and above 3 h as high. We collected comorbidities (defined as a history of the disorder during the patient’s lifetime) including diabetes mellitus, hypertension, obstructive sleep apnea, hyperlipidemia, coronary artery disease, peripheral vascular disease, atrial fibrillation, supraventricular tachycardia, anemia, asthma, valvular heart disease, congestive heart failure, chronic obstructive pulmonary disease, chronic kidney disease, thyroid disease, deep venous thrombosis, and cancer. Cases with age < 18 years, ASA > 4, and epidural, IV regional, and spinal anesthesia were excluded. Data were divided into those with (cases) and without (controls) admission to the hospital. Cases were studied by the author S. Shah to extract information for reasons of admission.

Chi-square tests with exact p values based on Monte Carlo simulation were utilized to examine the marginal association between categorical variables and unplanned readmission status. Wilcoxon rank sum tests were utilized to examine the marginal difference in continuous variables, such as age or surgical duration, between patients who did and did not have an unplanned admission following surgery. Due to the small number of admissions (49 cases), a maximum of five variables could be used in the multivariable logistic regression. Surgical duration time in mins, surgical specialty, anesthesia type, PVD, and DVT were significantly related to admission status based on univariate analyses with p value < 0.05. These variables were further adjusted in a multivariable logistic regression model. Firth’s correction was applied to mitigate the bias caused by rare events and low cell counts for unplanned admissions. Because there was an extremely small number of unplanned readmissions, Firth correction was used to adjust for bias allowing for more reliable estimates in the presence of the low event rate, while also providing more accurate confidence intervals and p values (D Firth 1993).

The odds ratio (OR) > 1 indicated a higher risk of unplanned admission, while an OR < 1 indicated a lower risk of unplanned admission. Statistical analyses were performed by a Biostatistician co-author (S. Stanley) using SAS 9.4.