Search

Objective: The increasing translation of artificial intelligence (AI)/machine learning (ML) models into clinical practice brings an increased risk of direct harm from modeling bias; however, bias remains incompletely measured in many medical AI applications. This article aims to provide a framework for objective evaluation of medical AI from multiple aspects, focusing on binary classification models.Materials and Methods: Using data from over 56 thousand Mass General Brigham (MGB) patients with confirmed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), we evaluate unrecognized bias in four AI models developed during the early months of the pandemic in Boston, Massachusetts that predict risks of hospital admission, ICU admission, mechanical ventilation, and death after a SARS-CoV-2 infection purely based on their pre-infection longitudinal medical records. Models were evaluated both retrospectively and prospectively using model-level metrics of discrimination, accuracy, and reliability, and a novel individual-level metric for error.Results: We found inconsistent instances of model-level bias in the prediction models. From an individual-level aspect, however, we found most all models performing with slightly higher error rates for older patients.Discussion: While a model can be biased against certain protected groups (i.e., perform worse) in certain tasks, it can be at the same time biased towards another protected group (i.e., perform better). As such, current bias evaluation studies may lack a full depiction of the variable effects of a model on its subpopulations.Conclusion: Only a holistic evaluation, a diligent search for unrecognized bias, can provide enough information for an unbiased judgment of AI bias that can invigorate follow-up investigations on identifying the underlying roots of bias and ultimately make a change. [ABSTRACT FROM AUTHOR]

Nonunion is a clinically significant complication of fracture associated with worse outcomes, including increased pain, disability, and higher healthcare costs. The risk for nonunion is likely to be complex and multifactorial, and as such, the biology underlying such risk remains poorly understood. Genetic studies represent one approach to identify implicated biology for further investigation, but to date the lack of large cohorts for study has limited such efforts. We utilized the electronic health records of two large academic medical centers in Boston to identify individuals with fracture nonunion and control individuals with fracture but no evidence of nonunion. We conducted a genomewide association study among 1760 individuals of Northern European ancestry with upper or lower extremity fracture, including 131 with nonunion, to examine whether common variants were associated with nonunion in this cohort. In all, one locus in the Calcyon (CALY) gene exceeded a genomewide threshold for statistical significance (p = 1.95e–8), with eight additional loci associated with p < 5e–7. Previously reported candidate genes were not supported by this analysis. Electronic health records should facilitate identification of common genetic variations associated with adverse orthopedic outcomes. The loci we identified in this small cohort require replication and further study to characterize mechanism of action, but represent a starting point for the investigation of genetic liability for this costly outcome. © 2018 American Society for Bone and Mineral Research. [ABSTRACT FROM AUTHOR]

Background: Carbon fiber-reinforced (CFR) polymer implants have theoretical advantages over titanium plates.Questions/Purposes: The aim of the present study was to assess our early outcomes with CFR plates in lateral opening-wedge distal femoral osteotomy to correct valgus lower limb malalignment. We asked the following: (1) Did the CFR polymer implant change time to union when compared with the titanium implant? (2) Did the incidence of displacement of medial cortical fractures differ between the implants? (3) Did the incidence of complications differ between the two techniques, and did other factors, such as bone graft material used, affect healing?Methods: A retrospective review of 16 limbs treated with an opening-wedge distal femoral osteotomy for genu valgum using either titanium (n = 10) or CFR plates (n = 6) was performed. Patient and clinical covariates as well as the primary outcome of time to union and secondary outcome of fracture displacement were collected and analyzed.Results: Those treated with CFR plates had longer times to union than did those in the titanium-treated group (median, 121.5 vs 81.5 days, respectively). The incidence of fracture displacement was higher in the CFR plate-treated group (CFR, n = 5/6; titanium, n = 1/10). Although the CFR plate-treated group had a 33% nonunion incidence while the titanium group had no nonunions, the study lacked the power to show significance. Bone graft material used did not affect outcome. Complication rates were higher in the CFR plate-treated patients.Conclusion: The CFR plate was associated with a longer time to unite and higher fracture displacement rate than the titanium plate. As this is a retrospective case series, further research is required to confirm these results and clarify best practices in plating of distal femoral osteotomy for deformity correction. [ABSTRACT FROM AUTHOR]

Background: To shorten the time required to bring new treatments to clinics, recent efforts have focused on repurposing existing Food and Drug Administration (FDA)-approved drugs with established safety data for new indications. We hypothesized that adverse effect profiles might aid in prioritizing compounds for investigation in central nervous system (CNS) applications by providing an indication of their abilities to cross the blood-brain barrier. Methods: Data were drawn from an investigation of similarity of adverse effect profiles, utilizing pre- and post-marketing data. A panel of known CNS-active drugs was utilized to estimate aggregate similarity profiles for all other FDA drugs in the database. Permutations were used to test whether similarities for any given drug exceeded that expected under the null hypothesis. To estimate the performance of algorithms using such profiles, manually-curated lists of known CNS-active and -inactive medications were classified using logistic regression. Algorithms with and without this similarity data were compared for prediction of CNS penetrance. Results: Models incorporating adverse effect similarity data exhibited greater discrimination of brain-penetrant and non-penetrant drugs than models without this data. A visualization tool was developed to allow any medication to be evaluated for adverse effect similarity to the CNS panel or a custom panel. Conclusions: Consideration of adverse effect profiles allows in silico prioritization of compounds for follow-up investigation for CNS indications. In concert with chemical screening approaches, this may accelerate repurposing efforts for putative CNS-active medications.

Background: The use of restraints in hospitalized patients is associated with physical and psychological adversity for patients and staff. The minimization of restraint use is a key goal in the hospital setting. Reaching this goal requires an accurate assessment of existing patterns of use across clinical settings.This study reports the rate of physical restraints among patients hospitalized within a multi-entity healthcare network along with stratification by care context, diagnostic, and demographic factors, and examines the sensitivity and specificity of ICD-10 code Z78.1 “physical restraint status” for defining physical restraints relative to electronic health record (EHR) documentation.The EHR was used for a retrospective analysis of all adults hospitalized between 2017 and 2022.Hospitalized adults.Patient demographics, structured diagnostic information, care area, length of stay, and in-hospital mortality, Z78.1 coding for physical restraints, restraint documentation in orders and flowsheets.Among 742,607 hospitalizations, 6.3% (n=47,041) involved the use of physical restraint based on coding or EHR documentation. Treatment in the intensive care unit (ICU) included restraint in 39% of encounters whereas treatment outside the ICU included restraint use in 1.3% of encounters. Besides critical illness, demographic factors including increasing age (adjusted odds ratio (aOR)=1.21 [1.19–1.23]), male gender (aOR=1.56, [1.52–1.60]), unknown race (aOR=1.27 [1.19–1.35]), and preferred language other than English (aOR=1.24, [1.18–1.29]) were associated with higher odds of restraint utilization. As compared to EHR orders or documentation of restraint, the ICD-10 code for physical restraint had a sensitivity of 1.5% and a specificity of 99.99%.Among adults admitted to acute care hospitals, clinical, demographic, and operational factors were associated with increased odds of restraint, with care in the ICU associated with greatly increased odds of restraint. Research into restraint utilization using coded administrative claims data is likely limited by the sensitivity of physical restraint coding.Objective: The use of restraints in hospitalized patients is associated with physical and psychological adversity for patients and staff. The minimization of restraint use is a key goal in the hospital setting. Reaching this goal requires an accurate assessment of existing patterns of use across clinical settings.This study reports the rate of physical restraints among patients hospitalized within a multi-entity healthcare network along with stratification by care context, diagnostic, and demographic factors, and examines the sensitivity and specificity of ICD-10 code Z78.1 “physical restraint status” for defining physical restraints relative to electronic health record (EHR) documentation.The EHR was used for a retrospective analysis of all adults hospitalized between 2017 and 2022.Hospitalized adults.Patient demographics, structured diagnostic information, care area, length of stay, and in-hospital mortality, Z78.1 coding for physical restraints, restraint documentation in orders and flowsheets.Among 742,607 hospitalizations, 6.3% (n=47,041) involved the use of physical restraint based on coding or EHR documentation. Treatment in the intensive care unit (ICU) included restraint in 39% of encounters whereas treatment outside the ICU included restraint use in 1.3% of encounters. Besides critical illness, demographic factors including increasing age (adjusted odds ratio (aOR)=1.21 [1.19–1.23]), male gender (aOR=1.56, [1.52–1.60]), unknown race (aOR=1.27 [1.19–1.35]), and preferred language other than English (aOR=1.24, [1.18–1.29]) were associated with higher odds of restraint utilization. As compared to EHR orders or documentation of restraint, the ICD-10 code for physical restraint had a sensitivity of 1.5% and a specificity of 99.99%.Among adults admitted to acute care hospitals, clinical, demographic, and operational factors were associated with increased odds of restraint, with care in the ICU associated with greatly increased odds of restraint. Research into restraint utilization using coded administrative claims data is likely limited by the sensitivity of physical restraint coding.Design: The use of restraints in hospitalized patients is associated with physical and psychological adversity for patients and staff. The minimization of restraint use is a key goal in the hospital setting. Reaching this goal requires an accurate assessment of existing patterns of use across clinical settings.This study reports the rate of physical restraints among patients hospitalized within a multi-entity healthcare network along with stratification by care context, diagnostic, and demographic factors, and examines the sensitivity and specificity of ICD-10 code Z78.1 “physical restraint status” for defining physical restraints relative to electronic health record (EHR) documentation.The EHR was used for a retrospective analysis of all adults hospitalized between 2017 and 2022.Hospitalized adults.Patient demographics, structured diagnostic information, care area, length of stay, and in-hospital mortality, Z78.1 coding for physical restraints, restraint documentation in orders and flowsheets.Among 742,607 hospitalizations, 6.3% (n=47,041) involved the use of physical restraint based on coding or EHR documentation. Treatment in the intensive care unit (ICU) included restraint in 39% of encounters whereas treatment outside the ICU included restraint use in 1.3% of encounters. Besides critical illness, demographic factors including increasing age (adjusted odds ratio (aOR)=1.21 [1.19–1.23]), male gender (aOR=1.56, [1.52–1.60]), unknown race (aOR=1.27 [1.19–1.35]), and preferred language other than English (aOR=1.24, [1.18–1.29]) were associated with higher odds of restraint utilization. As compared to EHR orders or documentation of restraint, the ICD-10 code for physical restraint had a sensitivity of 1.5% and a specificity of 99.99%.Among adults admitted to acute care hospitals, clinical, demographic, and operational factors were associated with increased odds of restraint, with care in the ICU associated with greatly increased odds of restraint. Research into restraint utilization using coded administrative claims data is likely limited by the sensitivity of physical restraint coding.Participants: The use of restraints in hospitalized patients is associated with physical and psychological adversity for patients and staff. The minimization of restraint use is a key goal in the hospital setting. Reaching this goal requires an accurate assessment of existing patterns of use across clinical settings.This study reports the rate of physical restraints among patients hospitalized within a multi-entity healthcare network along with stratification by care context, diagnostic, and demographic factors, and examines the sensitivity and specificity of ICD-10 code Z78.1 “physical restraint status” for defining physical restraints relative to electronic health record (EHR) documentation.The EHR was used for a retrospective analysis of all adults hospitalized between 2017 and 2022.Hospitalized adults.Patient demographics, structured diagnostic information, care area, length of stay, and in-hospital mortality, Z78.1 coding for physical restraints, restraint documentation in orders and flowsheets.Among 742,607 hospitalizations, 6.3% (n=47,041) involved the use of physical restraint based on coding or EHR documentation. Treatment in the intensive care unit (ICU) included restraint in 39% of encounters whereas treatment outside the ICU included restraint use in 1.3% of encounters. Besides critical illness, demographic factors including increasing age (adjusted odds ratio (aOR)=1.21 [1.19–1.23]), male gender (aOR=1.56, [1.52–1.60]), unknown race (aOR=1.27 [1.19–1.35]), and preferred language other than English (aOR=1.24, [1.18–1.29]) were associated with higher odds of restraint utilization. As compared to EHR orders or documentation of restraint, the ICD-10 code for physical restraint had a sensitivity of 1.5% and a specificity of 99.99%.Among adults admitted to acute care hospitals, clinical, demographic, and operational factors were associated with increased odds of restraint, with care in the ICU associated with greatly increased odds of restraint. Research into restraint utilization using coded administrative claims data is likely limited by the sensitivity of physical restraint coding.Main Measures: The use of restraints in hospitalized patients is associated with physical and psychological adversity for patients and staff. The minimization of restraint use is a key goal in the hospital setting. Reaching this goal requires an accurate assessment of existing patterns of use across clinical settings.This study reports the rate of physical restraints among patients hospitalized within a multi-entity healthcare network along with stratification by care context, diagnostic, and demographic factors, and examines the sensitivity and specificity of ICD-10 code Z78.1 “physical restraint status” for defining physical restraints relative to electronic health record (EHR) documentation.The EHR was used for a retrospective analysis of all adults hospitalized between 2017 and 2022.Hospitalized adults.Patient demographics, structured diagnostic information, care area, length of stay, and in-hospital mortality, Z78.1 coding for physical restraints, restraint documentation in orders and flowsheets.Among 742,607 hospitalizations, 6.3% (n=47,041) involved the use of physical restraint based on coding or EHR documentation. Treatment in the intensive care unit (ICU) included restraint in 39% of encounters whereas treatment outside the ICU included restraint use in 1.3% of encounters. Besides critical illness, demographic factors including increasing age (adjusted odds ratio (aOR)=1.21 [1.19–1.23]), male gender (aOR=1.56, [1.52–1.60]), unknown race (aOR=1.27 [1.19–1.35]), and preferred language other than English (aOR=1.24, [1.18–1.29]) were associated with higher odds of restraint utilization. As compared to EHR orders or documentation of restraint, the ICD-10 code for physical restraint had a sensitivity of 1.5% and a specificity of 99.99%.Among adults admitted to acute care hospitals, clinical, demographic, and operational factors were associated with increased odds of restraint, with care in the ICU associated with greatly increased odds of restraint. Research into restraint utilization using coded administrative claims data is likely limited by the sensitivity of physical restraint coding.Key Results: The use of restraints in hospitalized patients is associated with physical and psychological adversity for patients and staff. The minimization of restraint use is a key goal in the hospital setting. Reaching this goal requires an accurate assessment of existing patterns of use across clinical settings.This study reports the rate of physical restraints among patients hospitalized within a multi-entity healthcare network along with stratification by care context, diagnostic, and demographic factors, and examines the sensitivity and specificity of ICD-10 code Z78.1 “physical restraint status” for defining physical restraints relative to electronic health record (EHR) documentation.The EHR was used for a retrospective analysis of all adults hospitalized between 2017 and 2022.Hospitalized adults.Patient demographics, structured diagnostic information, care area, length of stay, and in-hospital mortality, Z78.1 coding for physical restraints, restraint documentation in orders and flowsheets.Among 742,607 hospitalizations, 6.3% (n=47,041) involved the use of physical restraint based on coding or EHR documentation. Treatment in the intensive care unit (ICU) included restraint in 39% of encounters whereas treatment outside the ICU included restraint use in 1.3% of encounters. Besides critical illness, demographic factors including increasing age (adjusted odds ratio (aOR)=1.21 [1.19–1.23]), male gender (aOR=1.56, [1.52–1.60]), unknown race (aOR=1.27 [1.19–1.35]), and preferred language other than English (aOR=1.24, [1.18–1.29]) were associated with higher odds of restraint utilization. As compared to EHR orders or documentation of restraint, the ICD-10 code for physical restraint had a sensitivity of 1.5% and a specificity of 99.99%.Among adults admitted to acute care hospitals, clinical, demographic, and operational factors were associated with increased odds of restraint, with care in the ICU associated with greatly increased odds of restraint. Research into restraint utilization using coded administrative claims data is likely limited by the sensitivity of physical restraint coding.Conclusion: The use of restraints in hospitalized patients is associated with physical and psychological adversity for patients and staff. The minimization of restraint use is a key goal in the hospital setting. Reaching this goal requires an accurate assessment of existing patterns of use across clinical settings.This study reports the rate of physical restraints among patients hospitalized within a multi-entity healthcare network along with stratification by care context, diagnostic, and demographic factors, and examines the sensitivity and specificity of ICD-10 code Z78.1 “physical restraint status” for defining physical restraints relative to electronic health record (EHR) documentation.The EHR was used for a retrospective analysis of all adults hospitalized between 2017 and 2022.Hospitalized adults.Patient demographics, structured diagnostic information, care area, length of stay, and in-hospital mortality, Z78.1 coding for physical restraints, restraint documentation in orders and flowsheets.Among 742,607 hospitalizations, 6.3% (n=47,041) involved the use of physical restraint based on coding or EHR documentation. Treatment in the intensive care unit (ICU) included restraint in 39% of encounters whereas treatment outside the ICU included restraint use in 1.3% of encounters. Besides critical illness, demographic factors including increasing age (adjusted odds ratio (aOR)=1.21 [1.19–1.23]), male gender (aOR=1.56, [1.52–1.60]), unknown race (aOR=1.27 [1.19–1.35]), and preferred language other than English (aOR=1.24, [1.18–1.29]) were associated with higher odds of restraint utilization. As compared to EHR orders or documentation of restraint, the ICD-10 code for physical restraint had a sensitivity of 1.5% and a specificity of 99.99%.Among adults admitted to acute care hospitals, clinical, demographic, and operational factors were associated with increased odds of restraint, with care in the ICU associated with greatly increased odds of restraint. Research into restraint utilization using coded administrative claims data is likely limited by the sensitivity of physical restraint coding. [ABSTRACT FROM AUTHOR]