RESUMEN
Background: Perioperative fluid management is critical in neurosurgery as over perfusion can lead to brain edema whereas under perfusion may lead to brain hypoperfusion or ischemia. We aimed to determine the effectiveness of intraoperative goal-directed fluid therapy (GDFT) in patients undergoing intracranial surgeries. Methods: We searched MEDLINE, Cochrane, and PubMed databases and forward-backward citations for studies published between database inception and February 22, 2024. Randomized controlled trials where intraoperative GDFT was performed in neurosurgery and compared to the conventional regime were included in the study. GDFT was compared with the conventional regime as per primary outcomes - total intraoperative fluid requirement, serum lactate, hemodynamics, brain relaxation, urine output, serum biochemistry, and secondary outcomes - intensive care unit and hospital length of stay. The quality of evidence was assessed with the Cochrane risk of bias tool. This study is registered on PROSPERO (CRD42024518816). Results: Of 75 records identified, eight were eligible, the majority of which had a low to moderate risk of overall bias. In four studies, more fluid was given in the control group. No difference in postoperative lactate values was noted in 50% of studies. In the remaining 50%, lactate was more in the control group. Three out of four studies did not find any significant difference in the incidence of intraoperative hypotension, and four out of six studies did not find a significant difference in vasopressor requirement. The majority of studies did not show significant differences in urine output, brain relaxation, and length of stay between both groups. None found any difference in acid base status or electrolyte levels. Conclusion: GDFT, when compared to the conventional regime in neurosurgery, showed that the total volume of fluids administered was lesser in the GDFT group with no increase in serum lactate. There was no difference in the hemodynamics, urine output, brain relaxation, urine output, length of stay, and biochemical parameters.
RESUMEN
OBJECTIVE: Life expectancy models are useful tools to support clinical decision-making. Prior models have not been used widely in clinical practice for patients with renal masses. We sought to develop and validate a model to predict life expectancy following the detection of a localized renal mass suspicious for renal cell carcinoma. MATERIALS AND METHODS: Using retrospective data from 2 large centers, we identified patients diagnosed with clinically localized renal parenchymal masses from 1998 to 2018. After 2:1 random sampling into a derivation and validation cohort stratified by site, we used age, sex, log-transformed tumor size, simplified cardiovascular index and planned treatment to fit a Cox regression model to predict all-cause mortality from the time of diagnosis. The model's discrimination was evaluated using a C-statistic, and calibration was evaluated visually at 1, 5, and 10 years. RESULTS: We identified 2,667 patients (1,386 at Corewell Health and 1,281 at Johns Hopkins) with renal masses. Of these, 420 (16%) died with a median follow-up of 5.2 years (interquartile range 2.2-8.3). Statistically significant predictors in the multivariable Cox regression model were age (hazard ratio [HR] 1.04; 95% confidence interval [CI] 1.03-1.05); male sex (HR 1.40; 95% CI 1.08-1.81); log-transformed tumor size (HR 1.71; 95% CI 1.30-2.24); cardiovascular index (HR 1.48; 95% CI 1.32-1.67), and planned treatment (HR: 0.10, 95% CI: 0.06-0.18 for kidney-sparing intervention and HR: 0.20, 95% CI: 0.11-0.35 for radical nephrectomy vs. no intervention). The model achieved a C-statistic of 0.74 in the derivation cohort and 0.73 in the validation cohort. The model was well-calibrated at 1, 5, and 10 years of follow-up. CONCLUSIONS: For patients with localized renal masses, accurate determination of life expectancy is essential for decision-making regarding intervention vs. active surveillance as a primary treatment modality. We have made available a simple tool for this purpose.
Asunto(s)
Neoplasias Renales , Modelos de Riesgos Proporcionales , Humanos , Neoplasias Renales/mortalidad , Neoplasias Renales/cirugía , Neoplasias Renales/patología , Masculino , Femenino , Estudios Retrospectivos , Anciano , Persona de Mediana Edad , Causas de Muerte , Carcinoma de Células Renales/mortalidad , Carcinoma de Células Renales/patología , Carcinoma de Células Renales/cirugíaRESUMEN
OBJECTIVE: Understanding the relationship between comorbidities and life expectancy is important in cancer patients who carry risks of cancer and noncancer-related mortality. Comorbidity indices (CI) are tools to provide an objective measure of competing risks of death. We sought to determine which CI might be best incorporated into clinical practice for patients with suspected renal cancer. MATERIALS AND METHODS: 1572 patients diagnosed with renal masses (stage I-IV) between 1998 and 2016 were analyzed for this study. Patient data were gathered from a community-based health center. Comorbidities were evaluated individually, and with 1 of 4 CI: Charlson (CCI), updated CCI (uCCI), age-adjusted CCI (aCCI), and simplified cardiovascular index (CVI). Cox-proportional hazard analysis of all-cause mortality was performed using the four CI, adjusting for the 4 CI, adjusting for age, gender, race, tumor size, and tumor stage. RESULTS: Univariable analyses revealed the four CI were significant predictors of mortality (P < 0.05), as were age, gender, tumor size, and stage. Comorbid conditions at diagnosis included hypertension (47.8%), diabetes mellitus (47.2%), coronary artery disease (41.1%), chronic kidney disease (31.8%), peripheral vascular disease (8.0%), congestive heart failure (5.7%), chronic obstructive pulmonary disease (5.7%), and cerebrovascular disease (2.0%). When analyzing the 4 CI in multivariable survival analyses accounting for factors available at diagnosis, and analyses incorporating pathologic and recurrence data, only CVI score and uCCI remained statistically significant (P < 0.05). Limitations of this work are the retrospective nature of data collection and data from a single institution, limiting the generalizability. CONCLUSION: Increasing comorbidity, age, tumor size, and cM stage are predictors of ACM for suspected renal cancer patients. CVI appears to provide comparable information to various iterations of CCI (uCCI, aCCI) while being the simplest to use. Utilization of CVI may assist clinicians and patients when considering between interventional and noninterventional approaches for suspected renal cancer.
Asunto(s)
Carcinoma de Células Renales , Diabetes Mellitus , Neoplasias Renales , Humanos , Estudios Retrospectivos , ComorbilidadRESUMEN
BACKGROUND: Substantial effort has been directed toward demonstrating uses of predictive models in health care. However, implementation of these models into clinical practice may influence patient outcomes, which in turn are captured in electronic health record data. As a result, deployed models may affect the predictive ability of current and future models. OBJECTIVE: To estimate changes in predictive model performance with use through 3 common scenarios: model retraining, sequentially implementing 1 model after another, and intervening in response to a model when 2 are simultaneously implemented. DESIGN: Simulation of model implementation and use in critical care settings at various levels of intervention effectiveness and clinician adherence. Models were either trained or retrained after simulated implementation. SETTING: Admissions to the intensive care unit (ICU) at Mount Sinai Health System (New York, New York) and Beth Israel Deaconess Medical Center (Boston, Massachusetts). PATIENTS: 130 000 critical care admissions across both health systems. INTERVENTION: Across 3 scenarios, interventions were simulated at varying levels of clinician adherence and effectiveness. MEASUREMENTS: Statistical measures of performance, including threshold-independent (area under the curve) and threshold-dependent measures. RESULTS: At fixed 90% sensitivity, in scenario 1 a mortality prediction model lost 9% to 39% specificity after retraining once and in scenario 2 a mortality prediction model lost 8% to 15% specificity when created after the implementation of an acute kidney injury (AKI) prediction model; in scenario 3, models for AKI and mortality prediction implemented simultaneously, each led to reduced effective accuracy of the other by 1% to 28%. LIMITATIONS: In real-world practice, the effectiveness of and adherence to model-based recommendations are rarely known in advance. Only binary classifiers for tabular ICU admissions data were simulated. CONCLUSION: In simulated ICU settings, a universally effective model-updating approach for maintaining model performance does not seem to exist. Model use may have to be recorded to maintain viability of predictive modeling. PRIMARY FUNDING SOURCE: National Center for Advancing Translational Sciences.
Asunto(s)
Lesión Renal Aguda , Inteligencia Artificial , Humanos , Unidades de Cuidados Intensivos , Cuidados Críticos , Atención a la SaludRESUMEN
BACKGROUND: The aim of this study was to evaluate the use of machine learning to predict persistent opioid use after hand surgery. METHODS: We trained two algorithms to predict persistent opioid use, first using a general surgery dataset and then using a hand surgery dataset, resulting in four trained models. Next, we tested each model's performance using hand surgery data. Participants included adult surgery patients enrolled in a cohort study at an academic center from 2015-2018. The first algorithm (Michigan Genomics Initiative model) was designed to accommodate patient-reported data and patients with or without prior opioid use. The second algorithm (claims model) was designed for insurance claims data from patients who were opioid-naïve only. The main outcome was model discrimination, measured by area under the receiver operating curve (AUC). RESULTS: Of 889 hand surgery patients, 49% were opioid-naïve and 21% developed persistent opioid use. Most patients underwent soft tissue procedures (55%) or fracture repair (20%). The MGI model had AUCs of 0.84 when trained only on hand surgery data, and 0.85 when trained on the full cohort of surgery patients. The claims model had AUCs of 0.69 when trained only on hand surgery data, and 0.52 when trained on the opioid-naïve cohort of surgery patients. CONCLUSION: Opioid use is common after hand surgery. Machine learning has the potential to facilitate identification of patients who are at risk for prolonged opioid use, which can promote early interventions to prevent addiction.
RESUMEN
Introduction: Fine needle aspiration cytology (FNAC) is considered the first line investigation of choice for evaluating head and neck swellings as it is a quick, safe, and rapid diagnostic procedure. Material and Methods: This is a retrospective study that included 242 cases of head and neck lesions in the Department of Pathology, Maharaja Agrasen Medical College, Agroha. FNAC was performed by aspiration and non aspiration techniques, and cytological diagnosis was given and correlated with clinical findings and investigations. Results: The most common age group affected was 21-30 years. Male to female ratio was 1:1.49. Out of 242 cases, maximum lesions were found in lymph nodes (128 cases), followed by thyroid gland in 81 cases, salivary gland in 23 cases, and miscellaneous group in 10 cases. Maximum number of cases reported were inflammatory (55.37%), followed by benign (29.33%) and malignant (11.15%) cases. Most swellings occurring in the head and neck region are inflammatory in nature. Conclusion: Our study concluded that FNAC is a simple, safe, and minimal invasive technique that differentiates between neoplastic and non neoplastic lesions and avoids unnecessary surgeries.
RESUMEN
BACKGROUND: Decentralized, digital health studies can provide real-world evidence of the lasting effects of COVID-19 on physical, socioeconomic, psychological, and social determinant factors of health in India. Existing research cohorts, however, are small and were not designed for longitudinal collection of comprehensive data from India's diverse population. Data4Life is a nationwide, digitally enabled, health research initiative to examine the post-acute sequelae of COVID-19 across individuals, communities, and regions. Data4Life seeks to build an ethnically and geographically diverse population of at least 100,000 participants in India. METHODS: Here we discuss the feasibility of developing a completely decentralized COVID-19 cohort in India through qualitative analysis of data collection procedures, participant characteristics, participant perspectives on recruitment and reported study motivation. RESULTS: As of June 13th, 2022, more than 6,000 participants from 17 Indian states completed baseline surveys. Friend and family referral were identified as the most common recruitment method (64.8%) across all demographic groups. Helping family and friends was the primary reason reported for joining the study (61.5%). CONCLUSIONS: Preliminary findings support the use of digital technology for rapid enrollment and data collection to develop large health research cohorts in India. This demonstrates the potential for expansion of digitally enabled health research in India. These findings also outline the value of person-to-person recruitment strategies when conducting digital health research in modern-day India. Qualitative analysis reveals opportunities to increase diversity and retention in real time. It also informs strategies for improving participant experiences in the current Data4Life initiative and future studies.
Due to the vast geographical size and ethnic diversity of the population, India represents a huge challenge for conducting research studies. The Data4Life study was set up to understand if digital tools can be an effective way to study long-term effects of COVID-19 across India. We studied different ways of collecting the relevant information from participants, the background of each participant, reasons, and motivation of each participant for joining the study. The results showed that friend and family referrals were the most common recruitment reason. Helping family and friends was reported as the main motivation for joining the study. Overall, the findings support the use of digital tools as an effective recruitment method for research studies in India.
RESUMEN
Acute kidney injury (AKI), which is a common complication of acute illnesses, affects the health of individuals in community, acute care and post-acute care settings. Although the recognition, prevention and management of AKI has advanced over the past decades, its incidence and related morbidity, mortality and health care burden remain overwhelming. The rapid growth of digital technologies has provided a new platform to improve patient care, and reports show demonstrable benefits in care processes and, in some instances, in patient outcomes. However, despite great progress, the potential benefits of using digital technology to manage AKI has not yet been fully explored or implemented in clinical practice. Digital health studies in AKI have shown variable evidence of benefits, and the digital divide means that access to digital technologies is not equitable. Upstream research and development costs, limited stakeholder participation and acceptance, and poor scalability of digital health solutions have hindered their widespread implementation and use. Here, we provide recommendations from the Acute Disease Quality Initiative consensus meeting, which involved experts in adult and paediatric nephrology, critical care, pharmacy and data science, at which the use of digital health for risk prediction, prevention, identification and management of AKI and its consequences was discussed.
Asunto(s)
Lesión Renal Aguda , Nefrología , Adulto , Niño , Humanos , Enfermedad Aguda , Consenso , Lesión Renal Aguda/diagnóstico , Lesión Renal Aguda/terapia , Lesión Renal Aguda/etiología , Cuidados CríticosRESUMEN
Astroblastoma is a rare tumor, which is mostly found in pediatric population. Due to scarcity of literature, the data about treatment is lacking. We are reporting case of brainstem astroblastoma in an adult female. A 45-year-old lady presented with complaint of headache, vertigo, vomiting, and nasal regurgitation for 3 months. On examination, she had weak gag, left hemiparesis. Magnetic resonance imaging brain reported medulla oblongata mass, dorsally exophytic. She underwent suboccipital craniotomy and decompression of mass. Histopathology confirmed diagnosis of astroblastoma. She underwent radiotherapy and recovered well. Brainstem astroblastoma is an extremely rare entity. The surgical resection is possible due to well-defined plane. For best outcome, maximum resection and radiation are indicated.
RESUMEN
Importance: Neighborhood-level social risk factors may contribute to health disparities in microbial keratitis (MK) disease presentation. Understanding neighborhood-level factors may identify areas for revised health policies to address inequities that impact eye health. Objective: To investigate if social risk factors were associated with presenting best-corrected visual acuity (BCVA) for patients with MK. Design, Setting, and Participants: This was a cross-sectional study of patients with a diagnosis of MK. Patients presenting to the University of Michigan with a diagnosis of MK between August 1, 2012, and February 28, 2021, were included in the study. Patient data were obtained from the University of Michigan electronic health record. Main Outcomes and Measures: Individual-level characteristics (age, self-reported sex, self-reported race and ethnicity), presenting log of the minimum angle of resolution (logMAR) BCVA, and neighborhood-level factors, including measures on deprivation, inequity, housing burden, and transportation at the census block group, were obtained. Univariate associations of presenting BCVA (< 20/40 vs ≥20/40) with individual-level characteristics were assessed with 2-sample t, Wilcoxon, and χ2 tests. Logistic regression was used to test associations of neighborhood-level characteristics with the probability of presenting BCVA worse than 20/40 after adjustment for patient demographics. Results: A total of 2990 patients with MK were identified and included in the study. Patients had a mean (SD) age of 48.6 (21.3) years, and 1723 were female (57.6%). Patients self-identified with the following race and ethnicity categories: 132 Asian (4.5%), 228 Black (7.8%), 99 Hispanic (3.5%), 2763 non-Hispanic (96.5%), 2463 White (84.4%), and 95 other (3.3%; included any race not previously listed). Presenting BCVA had a median (IQR) value of 0.40 (0.10-1.48) logMAR units (Snellen equivalent, 20/50 [20/25-20/600]), and 1508 of 2798 patients (53.9%) presented with BCVA worse than 20/40. Patients presenting with logMAR BCVA less than 20/40 were older than those who presented with 20/40 or higher (mean difference, 14.7 years; 95% CI, 13.3-16.1; P < .001). Furthermore, a larger percentage of male vs female sex patients presented with logMAR BCVA less than 20/40 (difference, 5.2%; 95% CI, 1.5-8.9; P = .04), as well as Black race (difference, 25.7%; 95% CI, 15.0%-36.5%;P < .001) and White race (difference, 22.6%; 95% CI, 13.9%-31.3%; P < .001) vs Asian race, and non-Hispanic vs Hispanic ethnicity (difference, 14.6%; 95% CI, 4.5%-24.8%; P = .04). After adjusting for age, self-reported sex, and self-reported race and ethnicity, worse Area Deprivation Index (odds ratio [OR], 1.30 per 10-unit increase; 95% CI, 1.25-1.35; P < .001), increased segregation (OR, 1.44 per 0.1-unit increase in Theil H index; 95% CI, 1.30-1.61; P < .001), higher percentage of households with no car (OR, 1.25 per 1 percentage point increase; 95% CI, 1.12-1.40; P = .001), and lower average number of cars per household (OR, 1.56 per 1 less car; 95% CI, 1.21-2.02; P = .003) were associated with increased odds of presenting BCVA worse than 20/40. Conclusion and Relevance: Findings of this cross-sectional study suggest that in a sample of patients with MK, patient characteristics and where they live were associated with disease severity at presentation. These findings may inform future research on social risk factors and patients with MK.
Asunto(s)
Equidad en Salud , Queratitis , Oftalmología , Humanos , Masculino , Femenino , Persona de Mediana Edad , Estudios Transversales , Factores de Riesgo , Agudeza VisualRESUMEN
OBJECTIVE: To develop and validate a model to predict whether patients undergoing ureteroscopy (URS) will receive a stent. METHODS: Using registry data obtained from the Michigan Urological Surgery Improvement Collaborative Reducing Operative Complications from Kidney Stones initiative, we identified patients undergoing URS from 2016 to 2020. We used patients' age, sex, body mass index, size and location of the largest stone, current stent in place, history of any kidney stone procedure, procedure type, and acuity to fit a multivariable logistic regression model to a derivation cohort consisting of a random two-thirds of episodes. Model discrimination and calibration were evaluated in the validation cohort. A sensitivity analysis examined urologist variation using generalized mixed-effect models. RESULTS: We identified 15,048 URS procedures, of which 11,471 (76%) had ureteral stents placed. Older age, male sex, larger stone size, the largest stone being in the ureteropelvic junction, no prior stone surgery, no stent in place, a planned procedure type of laser lithotripsy, and urgent procedure were associated with a higher risk of stent placement. The model achieved an area under the receiver operating characteristic curve of 0.69 (95% CI 0.67, 0.71). Incorporating urologist-level variation improved the area under the receiver operating characteristic curve to 0.83 (95% CI 0.82, 0.84). CONCLUSION: Using a large clinical registry, we developed a multivariable regression model to predict ureteral stent placement following URS. Though well-calibrated, the model had modest discrimination due to heterogeneity in practice patterns in stent placement across urologists.
Asunto(s)
Cálculos Renales , Litotripsia por Láser , Litotricia , Uréter , Cálculos Ureterales , Humanos , Masculino , Ureteroscopía/métodos , Cálculos Ureterales/terapia , Cálculos Renales/cirugía , Uréter/cirugía , Stents , Resultado del Tratamiento , Litotricia/métodosRESUMEN
This cohort study uses data from electronic health records to assess variability in a sepsis prediction model across 9 hospitals.
Asunto(s)
Modelos Estadísticos , Sepsis , Humanos , Pronóstico , Sepsis/diagnóstico , Hospitales , Atención al PacienteRESUMEN
OBJECTIVES: Implementing a predictive analytic model in a new clinical environment is fraught with challenges. Dataset shifts such as differences in clinical practice, new data acquisition devices, or changes in the electronic health record (EHR) implementation mean that the input data seen by a model can differ significantly from the data it was trained on. Validating models at multiple institutions is therefore critical. Here, using retrospective data, we demonstrate how Predicting Intensive Care Transfers and other UnfoReseen Events (PICTURE), a deterioration index developed at a single academic medical center, generalizes to a second institution with significantly different patient population. DESIGN: PICTURE is a deterioration index designed for the general ward, which uses structured EHR data such as laboratory values and vital signs. SETTING: The general wards of two large hospitals, one an academic medical center and the other a community hospital. SUBJECTS: The model has previously been trained and validated on a cohort of 165,018 general ward encounters from a large academic medical center. Here, we apply this model to 11,083 encounters from a separate community hospital. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: The hospitals were found to have significant differences in missingness rates (> 5% difference in 9/52 features), deterioration rate (4.5% vs 2.5%), and racial makeup (20% non-White vs 49% non-White). Despite these differences, PICTURE's performance was consistent (area under the receiver operating characteristic curve [AUROC], 0.870; 95% CI, 0.861-0.878), area under the precision-recall curve (AUPRC, 0.298; 95% CI, 0.275-0.320) at the first hospital; AUROC 0.875 (0.851-0.902), AUPRC 0.339 (0.281-0.398) at the second. AUPRC was standardized to a 2.5% event rate. PICTURE also outperformed both the Epic Deterioration Index and the National Early Warning Score at both institutions. CONCLUSIONS: Important differences were observed between the two institutions, including data availability and demographic makeup. PICTURE was able to identify general ward patients at risk of deterioration at both hospitals with consistent performance (AUROC and AUPRC) and compared favorably to existing metrics.
Asunto(s)
Cuidados Críticos , Habitaciones de Pacientes , Humanos , Estudios Retrospectivos , Curva ROC , Hospitales ComunitariosRESUMEN
OBJECTIVE: The objective of this study is to provide a method to calculate model performance measures in the presence of resource constraints, with a focus on net benefit (NB). MATERIALS AND METHODS: To quantify a model's clinical utility, the Equator Network's TRIPOD guidelines recommend the calculation of the NB, which reflects whether the benefits conferred by intervening on true positives outweigh the harms conferred by intervening on false positives. We refer to the NB achievable in the presence of resource constraints as the realized net benefit (RNB), and provide formulae for calculating the RNB. RESULTS: Using 4 case studies, we demonstrate the degree to which an absolute constraint (eg, only 3 available intensive care unit [ICU] beds) diminishes the RNB of a hypothetical ICU admission model. We show how the introduction of a relative constraint (eg, surgical beds that can be converted to ICU beds for very high-risk patients) allows us to recoup some of the RNB but with a higher penalty for false positives. DISCUSSION: RNB can be calculated in silico before the model's output is used to guide care. Accounting for the constraint changes the optimal strategy for ICU bed allocation. CONCLUSIONS: This study provides a method to account for resource constraints when planning model-based interventions, either to avoid implementations where constraints are expected to play a larger role or to design more creative solutions (eg, converted ICU beds) to overcome absolute constraints when possible.
Asunto(s)
Hospitalización , Unidades de Cuidados Intensivos , Humanos , Aprendizaje AutomáticoRESUMEN
PURPOSE: The purpose of this study was to predict visual acuity (VA) 90 days after presentation for patients with microbial keratitis (MK) from data at the initial clinical ophthalmic encounter. METHODS: Patients with MK were identified in the electronic health record between August 2012 and February 2021. Random forest (RF) models were used to predict 90-day VA < 20/40 [visual impairment (VI)]. Predictors evaluated included age, sex, initial VA, and information documented in notes at presentation. Model diagnostics are reported with 95% confidence intervals (CIs) for area under the curve (AUC), misclassification rate, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). RESULTS: One thousand seven hundred ninety-one patients were identified. The presenting logMAR VA was on average 0.86 (Snellen equivalent and standard deviation = 20/144 ± 12.6 lines) in the affected or worse eye, and 43.6% with VI. VI at 90-day follow-up was present in the affected eye or worse eye for 26.9% of patients. The RF model for predicting 90-day VI had an AUC of 95% (CI: 93%-97%) and a misclassification rate of 9% (7%-12%). The percent sensitivity, specificity, PPV, and NPV were 86% (80%-91%), 92% (89%-95%), 81% (74%-86%), and 95% (92%-97%), respectively. Older age, worse presenting VA, and more mentions of "penetrating keratoplasty" and "bandage contact lens" were associated with increased probability of 90-day VI, whereas more mentions of "quiet" were associated with decreased probability of 90-day VI. CONCLUSIONS: RF modeling yielded good sensitivity and specificity to predict VI at 90 days which could guide clinicians about the risk of poor vision outcomes for patients with MK.
Asunto(s)
Queratitis , Baja Visión , Humanos , Agudeza VisualRESUMEN
PURPOSE: There is a need to understand physicians' diagnostic uncertainty in the initial management of microbial keratitis (MK). This study aimed to understand corneal specialists' diagnostic uncertainty by establishing risk thresholds for treatment of MK that could be used to inform a decision curve analysis for prediction modeling. METHODS: A cross-sectional survey of corneal specialists with at least 2 years clinical experience was conducted. Clinicians provided the percentage risk at which they would always or never treat MK types (bacterial, fungal, herpetic, and amoebic) based on initial ulcer sizes and locations (<2 mm2 central, <2 mm2 peripheral, and >8 mm2 central). RESULTS: Seventy-two of 99 ophthalmologists participated who were 50% female with an average of 14.7 (SD = 10.1) years of experience, 60% in academic practices, and 38% outside the United States. Clinicians reported they would "never" and "always" treat a <2 mm2 central MK infection if the median risk was 0% and 20% for bacterial (interquartile range, IQR = 0-5 and 5-50), 4.5% and 27.5% for herpetic (IQR = 0-10 and 10-50), 5% and 50% for fungal (IQR = 0-10 and 20-75), and 5% and 50.5% for amoebic (IQR = 0-20 and 32-80), respectively. Mixed-effects models showed lower thresholds to treat larger and central infections (P < 0.001, respectively), and thresholds to always treat differed between MK types for the United States (P < 0.001) but not international clinicians. CONCLUSIONS: Risk thresholds to treat differed by practice locations and MK types, location, and size. Researchers can use these thresholds to understand when a clinician is uncertain and to create decision support tools to guide clinicians' treatment decisions.