Identifying signs and symptoms of urinary tract infection from emergency department clinical notes using large language models.

BACKGROUND: Natural language processing (NLP) tools including recently developed large language models (LLMs) have myriad potential applications in medical care and research, including the efficient labeling and classification of unstructured text such as electronic health record (EHR) notes. This opens the door to large-scale projects that rely on variables that are not typically recorded in a structured form, such as patient signs and symptoms. OBJECTIVES: This study is designed to acquaint the emergency medicine research community with the foundational elements of NLP, highlighting essential terminology, annotation methodologies, and the intricacies involved in training and evaluating NLP models. Symptom characterization is critical to urinary tract infection (UTI) diagnosis, but identification of symptoms from the EHR has historically been challenging, limiting large-scale research, public health surveillance, and EHR-based clinical decision support. We therefore developed and compared two NLP models to identify UTI symptoms from unstructured emergency department (ED) notes. METHODS: The study population consisted of patients aged ≥ 18 who presented to an ED in a northeastern U.S. health system between June 2013 and August 2021 and had a urinalysis performed. We annotated a random subset of 1250 ED clinician notes from these visits for a list of 17 UTI symptoms. We then developed two task-specific LLMs to perform the task of named entity recognition: a convolutional neural network-based model (SpaCy) and a transformer-based model designed to process longer documents (Clinical Longformer). Models were trained on 1000 notes and tested on a holdout set of 250 notes. We compared model performance (precision, recall, F1 measure) at identifying the presence or absence of UTI symptoms at the note level. RESULTS: A total of 8135 entities were identified in 1250 notes; 83.6% of notes included at least one entity. Overall F1 measure for note-level symptom identification weighted by entity frequency was 0.84 for the SpaCy model and 0.88 for the Longformer model. F1 measure for identifying presence or absence of any UTI symptom in a clinical note was 0.96 (232/250 correctly classified) for the SpaCy model and 0.98 (240/250 correctly classified) for the Longformer model. CONCLUSIONS: The study demonstrated the utility of LLMs and transformer-based models in particular for extracting UTI symptoms from unstructured ED clinical notes; models were highly accurate for detecting the presence or absence of any UTI symptom on the note level, with variable performance for individual symptoms.

Identifying incarceration status in the electronic health record using large language models in emergency department settings.

Background: Incarceration is a significant social determinant of health, contributing to high morbidity, mortality, and racialized health inequities. However, incarceration status is largely invisible to health services research due to inadequate clinical electronic health record (EHR) capture. This study aims to develop, train, and validate natural language processing (NLP) techniques to more effectively identify incarceration status in the EHR. Methods: The study population consisted of adult patients (≥ 18 y.o.) who presented to the emergency department between June 2013 and August 2021. The EHR database was filtered for notes for specific incarceration-related terms, and then a random selection of 1,000 notes was annotated for incarceration and further stratified into specific statuses of prior history, recent, and current incarceration. For NLP model development, 80% of the notes were used to train the Longformer-based and RoBERTa algorithms. The remaining 20% of the notes underwent analysis with GPT-4. Results: There were 849 unique patients across 989 visits in the 1000 annotated notes. Manual annotation revealed that 559 of 1000 notes (55.9%) contained evidence of incarceration history. ICD-10 code (sensitivity: 4.8%, specificity: 99.1%, F1-score: 0.09) demonstrated inferior performance to RoBERTa NLP (sensitivity: 78.6%, specificity: 73.3%, F1-score: 0.79), Longformer NLP (sensitivity: 94.6%, specificity: 87.5%, F1-score: 0.93), and GPT-4 (sensitivity: 100%, specificity: 61.1%, F1-score: 0.86). Conclusions: Our advanced NLP models demonstrate a high degree of accuracy in identifying incarceration status from clinical notes. Further research is needed to explore their scaled implementation in population health initiatives and assess their potential to mitigate health disparities through tailored system interventions.

Big data in Earth science: Emerging practice and promise.

Improvements in the number and resolution of Earth- and satellite-based sensors coupled with finer-resolution models have resulted in an explosion in the volume of Earth science data. This data-rich environment is changing the practice of Earth science, extending it beyond discovery and applied science to new realms. This Review highlights recent big data applications in three subdisciplines-hydrology, oceanography, and atmospheric science. We illustrate how big data relate to contemporary challenges in science: replicability and reproducibility and the transition from raw data to information products. Big data provide unprecedented opportunities to enhance our understanding of Earth's complex patterns and interactions. The emergence of digital twins enables us to learn from the past, understand the current state, and improve the accuracy of future predictions.

Automated HEART score determination via ChatGPT: Honing a framework for iterative prompt development.

Objectives: This study presents a design framework to enhance the accuracy by which large language models (LLMs), like ChatGPT can extract insights from clinical notes. We highlight this framework via prompt refinement for the automated determination of HEART (History, ECG, Age, Risk factors, Troponin risk algorithm) scores in chest pain evaluation. Methods: We developed a pipeline for LLM prompt testing, employing stochastic repeat testing and quantifying response errors relative to physician assessment. We evaluated the pipeline for automated HEART score determination across a limited set of 24 synthetic clinical notes representing four simulated patients. To assess whether iterative prompt design could improve the LLMs' ability to extract complex clinical concepts and apply rule-based logic to translate them to HEART subscores, we monitored diagnostic performance during prompt iteration. Results: Validation included three iterative rounds of prompt improvement for three HEART subscores with 25 repeat trials totaling 1200 queries each for GPT-3.5 and GPT-4. For both LLM models, from initial to final prompt design, there was a decrease in the rate of responses with erroneous, non-numerical subscore answers. Accuracy of numerical responses for HEART subscores (discrete 0-2 point scale) improved for GPT-4 from the initial to final prompt iteration, decreasing from a mean error of 0.16-0.10 (95% confidence interval: 0.07-0.14) points. Conclusion: We established a framework for iterative prompt design in the clinical space. Although the results indicate potential for integrating LLMs in structured clinical note analysis, translation to real, large-scale clinical data with appropriate data privacy safeguards is needed.

Correction: How Does ChatGPT Perform on the United States Medical Licensing Examination (USMLE)? The Implications of Large Language Models for Medical Education and Knowledge Assessment.

[This corrects the article DOI: 10.2196/45312.].

The effects of double-filtration plasmapheresis on coagulation profiles and the risk of bleeding.

BACKGROUND/PURPOSE: Double-filtration plasmapheresis (DFPP) can be used to remove circulating pathogenic molecules. By reclaiming filtered albumin, DFPP reduces the need for albumin and plasma replacement. Large proteins, such as fibrinogen, are removed. Our institution adopts a DFPP treatment protocol consisting of active surveillance of coagulation profiles and prophylactic supplementation of blood products containing fibrinogen. This study aims to investigate the effects of consecutive DFPP treatments on serial coagulation profiles and the risk of bleeding under this protocol. METHODS: Serial laboratory data and bleeding events at a single tertiary medical center were prospectively collected. Prophylactic transfusion of cryoprecipitate or fresh frozen plasma (FFP) was instituted if significant coagulopathy or a clinically evident bleeding event was observed. RESULTS: After the first treatment session, plasma fibrinogen levels decreased from 332 ± 106 mg/dL to 96 ± 44 mg/dL in the 37 study patients. In the following sessions, plasma fibrinogen levels were maintained at around 100 mg/dL under prophylactic transfusion. No major bleeding events were recorded, but five (14%) patients experienced minor bleeding. CONCLUSION: DFPP treatment might be performed safely along with active monitoring of coagulation profiles and prophylactic transfusion of cryoprecipitate or FFP.

Sodium-Glucose Cotransport Protein 2 Inhibitors in Patients With Type 2 Diabetes and Acute Kidney Disease.

Importance: Sodium-glucose cotransport protein 2 inhibitors (SGLT-2is) have demonstrated associations with positive kidney-related and cardiovascular outcomes in patients with type 2 diabetes. However, the association of SGLT-2is with outcomes among patients with type 2 diabetes and acute kidney disease (AKD) remains unclear. Objective: To examine the long-term associations of SGLT-2is with mortality, major adverse kidney events (MAKEs), and major adverse cardiovascular events (MACEs) in patients with type 2 diabetes and AKD. Design, Setting, and Participants: This cohort study used global health care data (the TriNetX database) spanning from September 30, 2002, to September 30, 2022. Propensity score matching was used to select a cohort of patients, and follow-up was conducted with a maximum duration of 5 years (completed on September 30, 2022) or until the occurrence of an outcome or death. Intervention: The use of SGLT-2is. Main Outcomes and Measures: The primary outcomes measured were mortality, MAKEs, and MACEs. Adjusted hazard ratios (AHR) with 95% CIs were calculated to compare the risks between SGLT-2i users and nonusers, representing the mean treatment effect among the treated patients. Results: A total of 230 366 patients with AKD (mean [SD] age, 67.1 [16.4] years; 51.8% men and 48.2% women) were enrolled in the study, which had a median follow-up duration of 2.3 (IQR, 1.2-3.5) years. Among these, 5319 individuals (2.3%) were identified as SGLT-2i users. Among nonusers, the incidence of mortality was 18.7%, the incidence of MAKEs was 21.0%, and the incidence of MACEs was 25.8%. After propensity score matching, the absolute differences between SGLT-2i users and nonusers for incidence of mortality, MAKEs, and MACEs were 9.7%, 11.5%, and 12.3%, respectively. Based on the treated population, SGLT-2i use was associated with a significantly lower risk of mortality (AHR, 0.69 [95% CI, 0.62-0.77]), MAKEs (AHR, 0.62 [95% CI, 0.56-0.69]), and MACEs (AHR, 0.75 [95% CI, 0.65-0.88]) compared with nonuse. External validation using a multicenter cohort data set of 1233 patients with AKD patients who were SGLT-2i users confirmed the observed beneficial outcomes. Notably, the risk reduction associated with SGLT-2is remained significant even among patients without hypertension, those with advanced chronic kidney disease, and those not receiving other hypoglycemic agents. Conclusions and Relevance: In this cohort study of patients with type 2 diabetes and AKD, administration of SGLT-2is was associated with a significant reduction in all-cause mortality, MAKEs, and MACEs when compared with nonuse, underscoring the importance of SGLT-2is in care after acute kidney injury. These findings emphasize the potential benefits of SGLT-2is in managing AKD and mitigating the risks of major cardiovascular and kidney diseases.

Low levels of mouse sperm chromatin fragmentation delay embryo development.

We previously demonstrated that MnCl2 induces double-stranded DNA breaks in sperm in a process that we term as sperm chromatin fragmentation. Here, we tested if the levels of double-stranded DNA breaks were corelated to the concentration of MnCl2, and we compared this to another agent that causes single-stranded DNA breaks, H2O2. We found that both methods have the advantage of inducing DNA breaks in a concentration-dependent manner. Mouse sperm were treated with varying concentrations of either H2O2 or MnCl2, and the DNA damage was assessed by pulse-field gel electrophoresis, and the alkaline and neutral comet assays. Oocytes were injected with either treated sperm and the resulting embryos analyzed with an embryoscope to detect subtle changes in embryonic development. We confirmed that H2O2 treatment induced primarily single-stranded DNA breaks and MnCl2 induced primarily double-stranded DNA breaks, indicating different mechanisms of damage. These sperm were injected into oocytes, and the development of the resulting embryos followed with an embryoscope equipped with time lapse recording. We found that aberrations in early embryonic development by day 2 with even the lowest levels of DNA damage and that the levels of embryonic aberrations correlated to the concentration of either H2O2 or MnCl2. Low levels of H2O2 caused significantly more aberrations in embryonic development than low levels of MnCl2 even though the levels of DNA damage as measured by comet assays were similar. These data demonstrate that even low levels of sperm DNA damage cause delays and arrests in embryonic development.

QDTrack: Quasi-Dense Similarity Learning for Appearance-Only Multiple Object Tracking.

Similarity learning has been recognized as a crucial step for object tracking. However, existing multiple object tracking methods only use sparse ground truth matching as the training objective, while ignoring the majority of the informative regions in images. In this paper, we present Quasi-Dense Similarity Learning, which densely samples hundreds of object regions on a pair of images for contrastive learning. We combine this similarity learning with multiple existing object detectors to build Quasi-Dense Tracking (QDTrack), which does not require displacement regression or motion priors. We find that the resulting distinctive feature space admits a simple nearest neighbor search at inference time for object association. In addition, we show that our similarity learning scheme is not limited to video data, but can learn effective instance similarity even from static input, enabling a competitive tracking performance without training on videos or using tracking supervision. We conduct extensive experiments on a wide variety of popular MOT benchmarks. We find that, despite its simplicity, QDTrack rivals the performance of state-of-the-art tracking methods on all benchmarks and sets a new state-of-the-art on the large-scale BDD100K MOT benchmark, while introducing negligible computational overhead to the detector.

Comparison of euploid blastocyst expansion with subgroups of single chromosome, multiple chromosome, and segmental aneuploids using an AI platform from donor egg embryos.

PURPOSE: This retrospective observational study compares how different classes of blastocyst genotypes from egg donor cycles differentially blastulate and expand using a standard assay. METHODS: Quantitative measurements of expansion utilized a customized neural network that segments all sequential time-lapse images during the first 10 h of expansion. RESULTS: Analyses were performed using two developmental time perspectives using time-lapse imaging. The first was the time to blastocyst formation (tB), which broadly reflects variations in developmental rate. Euploidy peaked at 100-115 h from fertilization. In contrast, aneuploidy peaks flanked this interval bi-modally. These distributions limit ploidy discrimination based upon traditional standard grading features when assessed in real time. In contrast, from the second perspective of progressive blastocyst expansion that is normalized to each individual blastocyst's tB time, euploidy was significantly increased at expansion values > 20,000µ2 across all tB intervals studied. A Cartesian coordinate plot graphically summarizes information useful to rank order blastocysts within cohorts for transfer. Defined aneuploidy subgroups, distinguished by the number and complexity of chromosomes involved, also showed distributive differences from both euploids and from each other. A small subset of clinically significant trisomies did not show discriminating features separating them from other euploids. CONCLUSION: A standard assay of blastocyst expansion normalized to each individual blastocyst's time of blastocyst formation more usefully discriminates euploidy from aneuploidy than real-time expansion comparisons using absolute developmental time from fertilization.

How Does ChatGPT Perform on the United States Medical Licensing Examination (USMLE)? The Implications of Large Language Models for Medical Education and Knowledge Assessment.

BACKGROUND: Chat Generative Pre-trained Transformer (ChatGPT) is a 175-billion-parameter natural language processing model that can generate conversation-style responses to user input. OBJECTIVE: This study aimed to evaluate the performance of ChatGPT on questions within the scope of the United States Medical Licensing Examination (USMLE) Step 1 and Step 2 exams, as well as to analyze responses for user interpretability. METHODS: We used 2 sets of multiple-choice questions to evaluate ChatGPT's performance, each with questions pertaining to Step 1 and Step 2. The first set was derived from AMBOSS, a commonly used question bank for medical students, which also provides statistics on question difficulty and the performance on an exam relative to the user base. The second set was the National Board of Medical Examiners (NBME) free 120 questions. ChatGPT's performance was compared to 2 other large language models, GPT-3 and InstructGPT. The text output of each ChatGPT response was evaluated across 3 qualitative metrics: logical justification of the answer selected, presence of information internal to the question, and presence of information external to the question. RESULTS: Of the 4 data sets, AMBOSS-Step1, AMBOSS-Step2, NBME-Free-Step1, and NBME-Free-Step2, ChatGPT achieved accuracies of 44% (44/100), 42% (42/100), 64.4% (56/87), and 57.8% (59/102), respectively. ChatGPT outperformed InstructGPT by 8.15% on average across all data sets, and GPT-3 performed similarly to random chance. The model demonstrated a significant decrease in performance as question difficulty increased (P=.01) within the AMBOSS-Step1 data set. We found that logical justification for ChatGPT's answer selection was present in 100% of outputs of the NBME data sets. Internal information to the question was present in 96.8% (183/189) of all questions. The presence of information external to the question was 44.5% and 27% lower for incorrect answers relative to correct answers on the NBME-Free-Step1 (P<.001) and NBME-Free-Step2 (P=.001) data sets, respectively. CONCLUSIONS: ChatGPT marks a significant improvement in natural language processing models on the tasks of medical question answering. By performing at a greater than 60% threshold on the NBME-Free-Step-1 data set, we show that the model achieves the equivalent of a passing score for a third-year medical student. Additionally, we highlight ChatGPT's capacity to provide logic and informational context across the majority of answers. These facts taken together make a compelling case for the potential applications of ChatGPT as an interactive medical education tool to support learning.

CCNet: Criss-Cross Attention for Semantic Segmentation.

Contextual information is vital in visual understanding problems, such as semantic segmentation and object detection. We propose a criss-cross network (CCNet) for obtaining full-image contextual information in a very effective and efficient way. Concretely, for each pixel, a novel criss-cross attention module harvests the contextual information of all the pixels on its criss-cross path. By taking a further recurrent operation, each pixel can finally capture the full-image dependencies. Besides, a category consistent loss is proposed to enforce the criss-cross attention module to produce more discriminative features. Overall, CCNet is with the following merits: 1) GPU memory friendly. Compared with the non-local block, the proposed recurrent criss-cross attention module requires 11× less GPU memory usage. 2) High computational efficiency. The recurrent criss-cross attention significantly reduces FLOPs by about 85 percent of the non-local block. 3) The state-of-the-art performance. We conduct extensive experiments on semantic segmentation benchmarks including Cityscapes, ADE20K, human parsing benchmark LIP, instance segmentation benchmark COCO, video segmentation benchmark CamVid. In particular, our CCNet achieves the mIoU scores of 81.9, 45.76 and 55.47 percent on the Cityscapes test set, the ADE20K validation set and the LIP validation set respectively, which are the new state-of-the-art results. The source codes are available at https://github.com/speedinghzl/CCNethttps://github.com/speedinghzl/CCNet.

Angiopoietin-2 inhibition attenuates kidney fibrosis by hindering chemokine C-C motif ligand 2 expression and apoptosis of endothelial cells.

Plasma levels of angiopoietin-2 are increased in patients with chronic kidney disease (CKD). Moreover, mouse models of progressive kidney disease also demonstrate increased angiopoietin-2 in both plasmas and kidneys. The role of dysregulated angiopoietins in the progression of kidney disease has not been thoroughly investigated. Here, we found in a cohort of 319 patients with CKD that plasma angiopoietin-2 and angiopoietin-2/angiopoietin-1 ratios were positively associated with the development of kidney failure. In mice with progressive kidney disease induced by either ureteral obstruction or ischemia-reperfusion injury, overexpression of human angiopoietin-1 in the kidney tubules not only reduced macrophage infiltration in the initial stage post-injury but also attenuated endothelial cell apoptosis, microvascular rarefaction, and fibrosis in the advanced disease stage. Notably, angiopoietin-1 attenuated chemokine C-C motif ligand 2 (CCL2) expression in the endothelial cells of the fibrosing kidneys, and these protective effects led to attenuation of functional impairment. Mechanistically, angiopoietin-1 reduced CCL2-activated macrophage migration and protected endothelial cells against cell apoptosis induced by angiopoietin-2 and Wnt ligands. Based on this, we applied L1-10, an angiopoietin-2 inhibitor, to the mouse models of progressive kidney disease and found inhibitory effects on macrophage infiltration, microvascular rarefaction, and fibrosis. Thus, we defined the detrimental impact of increased angiopoietin-2 on kidney survival of patients with CKD which appears highlighted by angiopoietin-2 induced endothelial CCL2-activated macrophage infiltration and endothelial cell apoptosis in their kidneys undergoing fibrosis.

Distinct Subtyping of Successful Weaning from Acute Kidney Injury Requiring Renal Replacement Therapy by Consensus Clustering in Critically Ill Patients.

Background: Clinical decisions regarding the appropriate timing of weaning off renal replacement therapy (RRT) in critically ill patients are complex and multifactorial. The aim of the current study was to identify which critical patients with acute kidney injury (AKI) may be more likely to be successfully weaned off RRT using consensus cluster analysis. Methods: In this study, critically ill patients who received RRT at three multicenter referral hospitals at several timepoints from August 2016 to July 2018 were enrolled. An unsupervised consensus clustering algorithm was used to identify distinct phenotypes. The outcomes of interest were the ability to wean off RTT and 90-day mortality. Results: A total of 124 patients with AKI requiring RRT (AKI-RRT) were enrolled. The 90-day mortality rate was 30.7% (38/124), and 49.2% (61/124) of the patients were successfully weaned off RRT for over 90 days. The consensus clustering algorithm identified three clusters from a total of 45 features. The three clusters had distinct features and could be separated according to the combination of urinary neutrophil gelatinase-associated lipocalin to creatinine ratio (uNGAL/Cr), Sequential Organ Failure Assessment (SOFA) score, and estimated glomerular filtration rate at the time of weaning off RRT. uNGAL/Cr (hazard ratio [HR] 2.43, 95% confidence interval [CI]: 1.36-4.33) and clustering phenotype (cluster 1 vs. 3, HR 2.7, 95% CI: 1.11-6.57; cluster 2 vs. 3, HR 44.5, 95% CI: 11.92-166.39) could predict 90-day mortality or re-dialysis. Conclusions: Almost half of the critical patients with AKI-RRT could wean off dialysis for over 90 days. Urinary NGAL/Cr and distinct clustering phenotypes could predict 90-day mortality or re-dialysis.

Urinary Biomarkers Can Predict Weaning From Acute Dialysis Therapy in Critically Ill Patients.

CONTEXT.­: Critically ill patients with acute kidney injury (AKI) requiring renal replacement therapy (RRT) have a poor prognosis. Several urinary AKI biomarkers have been proposed to predict renal recovery, but with limited discriminatory ability. OBJECTIVE.­: To validate the predictive performances of novel biomarkers to identify which critical patients with AKI may successfully wean from RRT. DESIGN.­: We prospectively recorded and analyzed clinical variables at several time points: (1) before starting RRT, (2) at the time of weaning off RRT, and (3) 24 hours after stopping RRT. A total of 140 critically ill patients who received RRT at a multicenter referral hospital from August 2016 to January 2019 were enrolled. The outcomes of interest were the ability to wean from RRT and 90-day mortality. RESULTS.­: The 90-day mortality rate was 13.6% (19 of 140), and 47.9% (67 of 140) of the patients were successfully weaned from RRT. Cluster analysis showed that the following biomarkers were correlated with estimated glomerular filtration rate at the time of weaning off RRT: urinary neutrophil gelatinase-associated lipocalin, kidney injury molecule 1, hemojuvelin, C-C motif chemokine ligand 14, interleukin 18, and liver-type fatty acid-binding protein (L-FABP). Among these, urinary L-FABP/creatinine (uL-FABP/Cr) at the time of weaning off RRT showed the best predictive performance for mortality (area under the receiver operating characteristic curve = 0.79). Taking mortality as a competing risk, Cox proportional hazards analysis indicated that a low uL-FABP/Cr (log) level was an independent prognostic factor for weaning from RRT (subdistribution hazard ratio, 0.35; P = .01). CONCLUSIONS.­: uL-FABP/Cr at the time of weaning off RRT could predict weaning from RRT and 90-day mortality.

Delayed Onset of Subclavian Artery Pseudoaneurysm With Brachial Plexus Compression Following Gunshot Wound Injury.

Early diagnosis of brachial plexus injuries is crucial to prevent long-term morbidity and improve outcomes. We present a unique case of delayed onset of brachial plexus compression two months following a traumatic gunshot injury causing multiple injuries including a T1 vertebral body comminuted fracture and pneumothorax. The patient experienced significant pain and progressive neurological examination changes during follow-up visits, and thus duplex ultrasound and computed tomography (CT) angiography were performed, which demonstrated a left subclavian artery pseudoaneurysm. This was managed operatively by evacuation and interposition bypass. Injuries to the cervical and upper thoracic spine are complex, and when patients present with new-onset neurological findings, axillary swelling, or significant uncontrolled postoperative pain, secondary complications should be suspected. Patients at a high risk of vascular reinjury should be routinely monitored at follow-up to prevent the development of progressive neurological damage to the brachial plexus.

Spectrum of cancer patients receiving renal biopsy.

BACKGROUND: The frontier of onco-nephrology, particularly renal complications of cancer and treatment, remains unexplored. We revisit the fundamental tool of diagnosing kidney disease, renal biopsy, in cancer patients with renal manifestation. METHODS: Patients who received renal biopsy from July 2015 to July 2019 were analyzed. Primary outcomes included end-stage renal disease (ESRD), mortality, and catastrophic outcome defined as either ESRD or mortality. A Cox proportional hazards model and Kaplan-Meier technique were used to assess the association with outcome measurements and survival analyses. Immunosuppression after renal biopsy and response to the treatment were evaluated. RESULTS: Among the 77 patients, the median age was 66 years (interquartile range [IQR] 59-73 years) and 46 (59.7%) were male. At the time of renal biopsy, 57 patients (74%) had various degrees of renal insufficiency. Tubulointerstitial damage score, quantified by renal pathology, were associated with higher hazards of ESRD (hazard ratio [HR], 1.77; 95% confidence interval [95% CI], 1.20 to 2.61; P = 0.004) and catastrophic outcome (HR, 1.30; 95% CI, 0.99 to 1.70; P = 0.058). The response rate to immunosuppression was lower in those diagnosed with tubulointerstitial nephritis (1 of 4 patients, 25%) than those with glomerulopathy (10 of 20 patients, 50%). CONCLUSION: Renal biopsy may improve diagnostic accuracy and assist in treatment guidance of cancer patients with renal manifestation. Renal biopsy should be encouraged with clinical indication. Collaboration between oncologists and nephrologists is of paramount importance to provide more comprehensive care for caner patients.