Anderson, Stephanie S.

Abstract Objective In 2024, the US News and World Report (USNWR)'s Best Hospitals began reporting complications following elective outpatient otolaryngology procedures in its ears, nose, and throat (ENT) ranking. However, this methodology has not been clinically adjudicated in a peer‐reviewed domain. We aimed to conduct manual health record adjudication to assess whether the USNWR methodology accurately attributes complications following elective outpatient otolaryngology procedures. Study Design Case‐cohort. Setting A case‐cohort of n = 80 elective, outpatient otolaryngology procedure/complication pairings from January 2019 to October 2023 across four academic medical centers using the 2024 USNWR otolaryngology procedure subgroups and Solventum Ambulatory Potentially Preventable Complications (AM‐PPC) software. Methods We used the AM‐PPC‐defined complication types as our standard of comparison for our clinical adjudication. Using the electronic health record, we adjudicated: (1) Was there agreement (yes/no) on complication type between our clinical adjudication and the Solventum AM‐PPC software? (2) Was the complication clinically related to, or a sequela of, the index otolaryngology procedure (yes/no)? Results The clinical adjudication agreed on complication type in 59/80 (74%) of patients, and 63/80 (79%) were judged to be clinically related to the index otolaryngology procedure. In total, 50/80 (63%) complications were adjudicated to be in agreement on both the complication type and clinical relatedness to the index procedure, while 9/80 (11%) complications did not agree with the complication type nor were clinically related to the index procedure. Conclusion Outpatient complications reported in the USNWR ENT rating lack the specificity to exclude patients whose complications are sequelae of chronic underlying conditions or related to concurrent, non‐otolaryngology procedures.

ABSTRACT Background Evaluating fairness in artificial intelligence (AI) is essential before clinical implementation to prevent unintended harm across diverse patient populations. This study assesses performance differences of an image-based surgical site infection (SSI) detection model across age, gender, geography, race, and skin color. Methods Model development/validation study using retrospective, multicenter data across surgical subspecialties from adults who submitted an image of their incision via the electronic health record within 30 days of their procedure between 2019-2024. Area under receiver operating curve (AUROC) was compared by age (<55 vs. ≥55 years old), gender (male vs. female), geography (Rochester, Minnesota vs. Jacksonville, Florida vs. Scottsdale, Arizona vs. regional Minnesota/Wisconsin areas), race (White/Caucasian vs. non-White/Caucasian), and skin color (Fitpatrick I-III vs. IV-VI and Monk A-E vs. F-J). Results Training data contained 13,702 images from 4,274 patients between 2019-2022 (50.1% ≥55 years old; 63.2% female; 91.6% White/Caucasian; 58.5% Rochester, 17.3% Florida, 10.5% Arizona; 96.3% light skin tone), of which 1,348 images (9.8%) had an SSI. No differences in SSI detection existed for age, gender, race, or skin color (AUROC 0.83-0.87, p>0.09), however geographical differences were significant: lowest AUROC 0.83 (95% CI 0.77-0.88) vs. highest AUROC 0.87 (95% CI 0.84-0.89, p=0.01). Temporal validation using 4,242 images from 1,353 patients between 2022-2023 showed intact fairness across subgroups (ΔAUROC ≤0.06) except geography (ΔAUROC ≤0.16). Conclusions The model behaved fairly across age, gender, race, and skin color, however geographical differences persisted. These findings have practical implications underscoring the importance of evaluating bias in clinical AI before deployment.

Objective: To develop an artificial intelligence–based pipeline to assess and triage patient-submitted postoperative wound images. Background: The rise of outpatient surgeries, remote monitoring, and patient-submitted wound images via online portals has contributed to a growing administrative workload on clinicians. Early identification of surgical site infection (SSI) is essential for reducing postoperative morbidity. Methods: Patients ≥18 years old who underwent surgery at 9 affiliated Mayo Clinic hospitals (2019-2022) and were captured by the National Surgical Quality Improvement Program (NSQIP) were included. Eligibility required submission of one image via the patient portal within 30 days after surgery. Images were independently screened in duplicate to determine the presence of an incision. SSI outcomes were obtained from NSQIP. The developed model consisted of 2 stages: incision detection and SSI detection in images with incisions. Four pretrained architectures were evaluated using 10-fold cross-validation, with upsampling and data augmentation to mitigate class imbalance. An end-to-end pipeline, image quality assessment and sensitivity analysis stratified by race were also performed. Results: Among 6060 patients, the median age was 54 years (interquartile range: 40–65), 61.4% (n=3805) were female, and 92.5% (n=5731) identified as White. SSIs were confirmed in 6.2% (n=386) images. Vision Transformer outperformed all others, achieving an incision detection accuracy of 0.94 (area under the curve=0.98) and an SSI detection accuracy of 0.73 (area under the curve=0.81). In addition, it demonstrated strong performance in assessing image quality. Sensitivity analysis revealed comparable performance across racial subgroups. Conclusion: This artificial intelligence pipeline demonstrates promising performance in automating wound image assessment and SSI detection, reducing clinical workload and improving postoperative care.

AbstractNational performance metrics ultimately enhance patient decision‐making and promote meaningful improvements in health care delivery, which makes having valid and reliable measures essential. This study examined US News and World Report metrics from 2019 to 2012 and used electronic health record data, combined with detailed chart review across 3 in‐system hospitals, to assess the provision of care compared to the attribution of patients assigned to the ear, nose, and throat (ENT) mortality group. Of the initial 47 ENT‐attributed deaths, 23 of those were verified, dimensioning the mortality rate from 1.7% to just 0.8%. These results underscore the necessity of rethinking measures and mortality attribution methodologies to be more accurate. Current methods use Medicare Severity Diagnosis Related Group billing coding to map the attribution. We suggest transitioning away from specialty ranking approaches and towards a procedure and condition “rating” approach to ensure that these ranking types capture data about the provision of care within a given encounter.

AbstractObjectiveTo evaluate the impact of a virtual registered nurse (ViRN) model on safety and care outcomes. ViRN is a telemedicine intervention that enables an experienced virtual nurse to assist the in‐person care team in providing care to patients.Data Sources and Study SettingElectronic health records data were utilized from the Mayo Clinic during the intervention (December 2020–November 2021) and historical periods (December 2018–November 2019). ViRN was implemented on general medical units at the Mayo Clinic Rochester. We used general medical units at the Mayo Clinic Arizona as the comparison group.Study DesignThis study used a difference‐in‐differences design to evaluate the impact of ViRN compared to usual care on transfer to the intensive care unit (ICU), inpatient mortality, and length of stay (LOS). We used logistic regression for transfer to the ICU and inpatient mortality and negative binomial regression for LOS. We controlled for demographics, patient interaction with the health system, clinical characteristics, and admission characteristics. We clustered standard errors to account for patients who have multiple admissions during the study period.Principal FindingsThere were no significant differences for transfer to the ICU (average marginal effect (AME) −0.08 percentage point [95% confidence interval (CI): −1.34, 1.18]), inpatient mortality (AME 0.43 percentage point [95% CI: −0.33, 1.18]), or LOS (AME −0.20 days [95% CI: −0.57, 0.17]). The findings were mostly consistent across the sensitivity analyses.ConclusionsOur results suggest that ViRN led to similar outcomes as usual care in general medical units. These findings support the potential to develop more advanced models of ViRN at the Mayo Clinic and the dissemination of the ViRN model to other systems. In the context of staffing shortages and other disruptions to the delivery of nursing care, it is critical to understand whether new models like ViRN provide nurse staffing alternatives without negatively affecting outcomes.

ObjectivesTo measure baseline human papillomavirus (HPV) vaccination rates among tertiary and community‐based Otolaryngology – Head and Neck Surgery (Oto‐HNS) clinic patients and to determine risk factors for under‐vaccination.MethodsRetrospective chart review of patients aged 9 to 26 years presenting to an Oto‐HNS clinic from 2017 to 2019. Patients were considered complete for HPV vaccination if they received two doses of HPV vaccine with the first dose received before age 15 years or three doses of HPV vaccine otherwise.Results8,532 unique patients met the criteria. At the index visit, 3,110 (36.5%) had completed the HPV series, 5,422 (63.5%) were due for one or more doses, with 4,981 (58.4%) eligible for vaccination at the time of their appointment. Of those dues, most (3,148/5,422 or 58%) were past due by age (≥13 years old). Of the 3,148 patients past due, 745 (23.7%) were partially vaccinated and 2,403 (76.3%) were vaccine naïve. Male sex and younger age were both independently associated with incomplete vaccination (p < 0.0001).ConclusionThis study demonstrates that the implementation of on‐site HPV vaccination has the potential to increase the opportunities for vaccination among vaccine‐eligible patients, especially among young males. Otolaryngologists have the potential to provide meaningful preventive services in the fight against HPV‐mediated disease.Level of Evidence3 Laryngoscope, 133:116–123, 2023

ABSTRACTBackgroundThe authors sought to estimate incremental economic impact of atrial fibrillation (AF) and the timing of its onset in myocardial infarction (MI) patients.HypothesisConcurrent AF and its timing are associated with higher costs in MI patients.MethodsThis retrospective cohort study included incident MI patients from Olmsted County, Minnesota, treated between November 1, 2002, and December 31, 2010. We compared inflation‐adjusted standardized costs accumulated between incident MI and end of follow‐up among 3 groups by AF status and timing: no AF, new‐onset AF (within 30 days after index MI), and prior AF. Multivariate adjustment of median costs accounted for right‐censoring in costs.ResultsThe final study cohort had 1389 patients, with 989 in no AF, 163 in new‐onset AF, and 237 in prior AF categories. Median follow‐up times were 3.98, 3.23, and 2.55 years, respectively. Mean age at index was 67 years, with significantly younger patients in the no AF group (64 years vs 76 and 77 years, respectively; P < 0.001). New‐onset and prior AF patients had more comorbid conditions (hypertension, heart failure, and chronic obstructive pulmonary disease). After accounting for differences in baseline characteristics, we found adjusted median (95% confidence interval) costs of $73 000 ($69 000‐$76 000) for no AF; $85 000 ($81 000‐$89 000) for new‐onset AF; and $97 000 ($94 000‐$100 000) for prior AF. Inpatient costs composed the largest share of total median costs (no AF, 82%; new‐onset AF, 84%; prior AF, 83%).ConclusionsAtrial fibrillation frequently coexists with MI and imposes incremental costs, mainly attributable to inpatient care. Timing of AF matters, as prior AF was found to be associated with higher costs than new‐onset AF.

Abstract Estrogen receptor (ER) is expressed in normal and malignant breast epithelium, and expression levels have been found to increase with age in normal breast epithelium but not in atypical hyperplasia (AH) and carcinoma in situ. Here we assess ER expression in AH and its association with later breast cancer. ER expression was assessed immunohistochemically in archival sections from 246 women with AH who had open benign breast biopsy from 1967 to 1991. The ACIS III (Dako) was utilized to calculate ER expression in all atypical foci. Using multivariate linear regression, we examined associations of ER expression with age at biopsy, indication for biopsy, type of atypia, number of atypical foci, involution status, and family history. Breast cancer risk across levels of ER expression was also assessed compared with the Iowa SEER control population. Among 246 women, 87 (35%) had atypical ductal hyperplasia (ADH), 141 (57%) had atypical lobular hyperplasia (ALH), and 18 (7%) had both. Forty-nine (20%) developed breast cancer (median follow-up of 14.4 years). Multivariate analysis indicated that type of atypia and age at diagnosis were significantly associated with ER percent staining and intensity (P &lt; 0.05). ER expression was increased in women with ADH and/or those over age 55. ER expression did not significantly impact breast cancer risk in patients diagnosed with atypia. We found increasing ER expression in AH with increasing age. ER expression in AH does not further discriminate breast cancer risk in women with atypia. Cancer Prev Res; 4(3); 435–44. ©2011 AACR.

Purpose Accurate breast cancer risk assessment is vital to personalize screening and risk reduction strategies. Women with atypical hyperplasia have a four-fold higher risk of breast cancer. We evaluated the performance of the Tyrer-Cuzick model, which was designed to predict 10-year risk of breast cancer development, in a well-defined cohort of women with atypia. Patients and Methods The Mayo Benign Breast Disease cohort includes 9,376 women who had a benign breast biopsy between 1967 and 1991. Among those, 331 women with atypia were identified by our study pathologists. Risk factor data for the Tyrer-Cuzick model were collated for each woman and used to predict individual risk of developing invasive breast cancer within 10 years. Results Over a median follow-up of 14.6 years, 64 (19%) of the 331 women developed invasive breast cancer. In the first 10 years after biopsy, 31 women developed invasive breast cancer whereas the Tyrer-Cuzick model predicted 58.9. The observed-to-predicted ratio was 0.53 (95% CI, 0.37 to 0.75). The concordance statistic was 0.540, revealing that the Tyrer-Cuzick model did not accurately distinguish, on an individual level, between women who developed invasive breast cancer and those who did not. Conclusion The Tyrer-Cuzick model significantly overestimated risk of breast cancer for women with atypia, and individual risk estimates showed poor concordance between predicted risk and invasive breast cancer development. Thus, we cannot recommend the use of the Tyrer-Cuzick model to predict 10-year breast cancer risk in women with atypical hyperplasia.