Susceptibility Loci for Type 2 Diabetes in the Ethnically Endogamous Indian Sindhi Population: A Pooled Blood Genome-Wide Association Study.

Type 2 diabetes (T2D) is a complex metabolic derangement that has a strong genetic basis. There is substantial population-specificity in the association of genetic variants with T2D. The Indian urban Sindhi population is at a high risk of T2D. The genetic basis of T2D in this population is unknown. We interrogated 28 pooled whole blood genomes of 1402 participants from the Diabetes In Sindhi Families In Nagpur (DISFIN) study using Illumina's Global Screening Array. From a total of 608,550 biallelic variants, 140 were significantly associated with T2D after adjusting for comorbidities, batch effects, pooling error, kinship status and pooling variation in a random effects multivariable logistic regression framework. Of the 102 well-characterized genes that these variants mapped onto, 70 genes have been previously reported to be associated with T2D to varying degrees with known functional relevance. Excluding open reading frames, intergenic non-coding elements and pseudogenes, our study identified 22 novel candidate genes in the Sindhi population studied. Our study thus points to the potential, interesting candidate genes associated with T2D in an ethnically endogamous population. These candidate genes need to be fully investigated in future studies.

Polysomnographic identification of anxiety and depression using deep learning.

Anxiety and depression are common psychiatric conditions associated with significant morbidity and healthcare costs. Sleep is an evolutionarily conserved health state. Anxiety and depression have a bidirectional relationship with sleep. This study reports on the use of analysis of polysomnographic data using deep learning methods to detect the presence of anxiety and depression. Polysomnography data on 940 patients performed at an academic sleep center during the 3-year period from 01/01/2016 to 12/31/2018 were identified for analysis. The data were divided into 3 subgroups: 205 patients with Anxiety/Depression, 349 patients with no Anxiety/Depression, and 386 patients with likely Anxiety/Depression. The first two subgroups were used for training and testing of the deep learning algorithm, and the third subgroup was used for external validation of the resulting model. Hypnograms were constructed via automatic sleep staging, with the 12-channel PSG data being transformed into three-channel RGB (red, green, blue channels) images for analysis. Composite patient images were generated and utilized for training the Xception model, which provided a validation set accuracy of 0.9782 on the ninth training epoch. In the independent test set, the model achieved a high accuracy (0.9688), precision (0.9533), recall (0.9630), and F1-score (0.9581). Classification performance of most other mainstream deep learning models was comparable. These findings suggest that machine learning techniques have the potential to accurately detect the presence of anxiety and depression from analysis of sleep study data. Further studies are needed to explore the utility of these techniques in the field of psychiatry.

The bleeding risk treatment paradox at the physician and hospital level: Implications for reducing bleeding in patients undergoing percutaneous coronary intervention.

BACKGROUND: Bleeding is a common and costly complication of percutaneous coronary intervention (PCI). Bleeding avoidance strategies (BAS) are used paradoxically less in patients at high-risk of bleeding: "bleeding risk-treatment paradox" (RTP). We determined whether hospitals and physicians, who do not align BAS to PCI patients' bleeding risk (ie, exhibit a RTP) have higher bleeding rates. METHODS: We examined 28,005 PCIs from the National Cardiovascular Data Registry CathPCI Registry for 7 hospitals comprising BJC HealthCare. BAS included transradial intervention, bivalirudin, and vascular closure devices. Patients' predicted bleeding risk was based on National Cardiovascular Data Registry CathPCI bleeding model and categorized as low (<2.0%), moderate (2.0%-6.4%), or high (≥6.5%) risk tertiles. BAS use was considered risk-concordant if: at least 1 BAS was used for moderate risk; 2 BAS were used for high risk and bivalirudin or vascular closure devices were not used for low risk. Absence of risk-concordant BAS use was defined as RTP. We analyzed inter-hospital and inter-physician variation in RTP, and the association of RTP with post-PCI bleeding. RESULTS: Amongst 28,005 patients undergoing PCI by 103 physicians at 7 hospitals, RTP was observed in 12,035 (43%) patients. RTP was independently associated with a higher likelihood of bleeding even after adjusting for predicted bleeding risk, mortality risk and potential sources of variation (OR 1.66, 95% CI 1.44-1.92, P < .001). A higher prevalence of RTP strongly and independently correlated with worse bleeding rates, both at the physician-level (Wilk's Lambda 0.9502, F-value 17.21, P < .0001) and the hospital-level (Wilk's Lambda 0.9899, F-value 35.68, P < .0001). All the results were similar in a subset of PCIs conducted since 2015 - a period more reflective of the contemporary practice. CONCLUSIONS: Bleeding RTP is a strong, independent predictor of bleeding. It exists at the level of physicians and hospitals: those with a higher rate of RTP had worse bleeding rates. These findings not only underscore the importance of recognizing bleeding risk upfront and using BAS in a risk-aligned manner, but also inform and motivate national efforts to reduce PCI-related bleeding.

Protocol for CLASSIC PBB: comparison of lower airway sampling strategies in children with protracted bacterial bronchitis.

BACKGROUND: Protracted bacterial bronchitis (PBB) is an endobronchial infection and a the most common cause of chronic wet cough in young children. It is treated with antibiotics, which can only be targeted if the causative organism is known. As most affected children do not expectorate sputum, lower airway samples can only be obtained by bronchoalveolar lavage (BAL) samples taken during flexible bronchoscopy (FB-BAL). This is invasive and is therefore reserved for children with severe or relapsing cases. Most children with PBB are treated empirically with broad spectrum antibiotics. CLASSIC PBB will compare the pathogen yield from two less invasive strategies with that from FB-BAL to see if they are comparable. METHODS: 131 children with PBB from four UK centres referred FB-BAL will be recruited. When attending for FB-BAL, they will have a cough swab and an induced sputum sample obtained. The primary outcome will be the discordance of the pathogen yield from the cough swab and the induced sputum when compared with FB-BAL. Secondary outcomes will be the sensitivity of each sampling strategy, the success rate of the induced sputum in producing a usable sample and the tolerability of each of the three sampling strategies. DISCUSSION: If either or both of the two less invasive airway sampling strategies are shown to be a useful alternative to FB-BAL, this will lead to more children with PBB having lower airway samples enabling targeted antibiotic prescribing. It would also reduce the need for FB, which is known to be burdensome for children and their families. TRIAL REGISTRATION NUMBER: ISRCTN79883982.

Genetic association of anthropometric traits with type 2 diabetes in ethnically endogamous Sindhi families.

BACKGROUND: Ethnically endogamous populations can shed light on the genetics of type 2 diabetes. Such studies are lacking in India. We conducted this study to determine the genetic and environmental contributions of anthropometric traits to type 2 diabetes risk in the Sindhi families in central India. METHODS: We conducted a family study in Indian Sindhi families with at least one case of type 2 diabetes. Variance components methods were used to quantify the genetic association of 18 anthropometric traits with eight type 2 diabetes related traits. Univariate and bivariate polygenic models were used to determine the heritability, genetic and environmental correlation of anthropometric traits with type 2 diabetes related traits. RESULTS: We included 1,152 individuals from 112 phenotyped families. The ascertainment-bias corrected prevalence of type 2 diabetes was 35%. Waist circumference, hip circumference and the biceps, triceps, subscapular and medial calf skinfold thicknesses were polygenically and significantly associated with type 2 diabetes. The range of heritability of the anthropometric traits and type 2 diabetes related traits was 0.27-0.73 and 0.00-0.39, respectively. Heritability of type 2 diabetes as a discrete trait was 0.35. Heritability curves demonstrated a substantial local influence of type 2 diabetes related traits. Bivariate trait analyses showed that biceps and abdominal skinfold thickness and all waist-containing indexes were strongly genetically correlated with type 2 diabetes. CONCLUSIONS: In this first study of Sindhi families, we found evidence for genetic and environmental concordance of anthropometric traits with type 2 diabetes. Future studies need to probe into the genetics of type 2 diabetes in this population.

Costs associated with transradial access and same-day discharge after percutaneous coronary intervention: a systematic review and meta-analysis.

Transradial access for PCI (TRI) along with same day discharge (SDD) is associated with varying estimates of cost savings depending on the population studied, the clinical scenario and application to low-risk vs high-risk patients. A summary estimate of the true cost savings of TRI and SDD are unknown. We searched the PubMed, EMBASE®, CINAHL® and Google Scholar® databases for published studies on hospitalization costs of TRI and SDD. Primary outcome of interest in all included studies was the cost saving with TRI (or SDD), inflation-corrected US$ 2018 values using the medical consumer price index. For meta-analytic synthesis, we used Hedges' summary estimate (g) in a random-effects framework of the DerSimonian and Laird model, with inverse variance weights. Heterogeneity was quantified using the I2 statistic. The cost savings of TRI from four US studies of 349,757 patients reported a consistent and significant cost saving associated with TRI after accounting for currency inflation, of US$ 992 (95% CI US$ 850-1,134). The cost savings of SDD from six US studies of 1,281,228 patients, after inflation-correcting to the year 2018, were US$ 3,567.58 (95% CI US$ 2,303-4,832). In conclusion, this meta-analysis demonstrates that TRI and SDD are associated with mean cost reductions of by approximately US$ 1,000/patient and US$ 3,600/patient, respectively, albeit with wide heterogeneity in the cost estimates. When combined with the safety of TRI and SDD, this meta-analysis underscores the value of combining TRI and SDD pathways and calls for a wide-ranging practice change in the direction of TRI and SDD.

Transradial Access for High-Risk Percutaneous Coronary Intervention: Implications of the Risk-Treatment Paradox.

[Figure: see text].

Deep learning model to predict the need for mechanical ventilation using chest X-ray images in hospitalised patients with COVID-19.

OBJECTIVES: There exists a wide gap in the availability of mechanical ventilator devices and their acute need in the context of the COVID-19 pandemic. An initial triaging method that accurately identifies the need for mechanical ventilation in hospitalised patients with COVID-19 is needed. We aimed to investigate if a potentially deteriorating clinical course in hospitalised patients with COVID-19 can be detected using all X-ray images taken during hospitalisation. METHODS: We exploited the well-established DenseNet121 deep learning architecture for this purpose on 663 X-ray images acquired from 528 hospitalised patients with COVID-19. Two Pulmonary and Critical Care experts blindly and independently evaluated the same X-ray images for the purpose of validation. RESULTS: We found that our deep learning model predicted the need for mechanical ventilation with a high accuracy, sensitivity and specificity (90.06%, 86.34% and 84.38%, respectively). This prediction was done approximately 3 days ahead of the actual intubation event. Our model also outperformed two Pulmonary and Critical Care experts who evaluated the same X-ray images and provided an incremental accuracy of 7.24%-13.25%. CONCLUSIONS: Our deep learning model accurately predicted the need for mechanical ventilation early during hospitalisation of patients with COVID-19. Until effective preventive or treatment measures become widely available for patients with COVID-19, prognostic stratification as provided by our model is likely to be highly valuable.

A Machine-Learning Framework to Identify Distinct Phenotypes of Aortic Stenosis Severity.

OBJECTIVES: The authors explored the development and validation of machine-learning models for augmenting the echocardiographic grading of aortic stenosis (AS) severity. BACKGROUND: In AS, symptoms and adverse events develop secondarily to valvular obstruction and left ventricular decompensation. The current echocardiographic grading of AS severity focuses on the valve and is limited by diagnostic uncertainty. METHODS: Using echocardiography (ECHO) measurements (ECHO cohort, n = 1,052), we performed patient similarity analysis to derive high-severity and low-severity phenogroups of AS. We subsequently developed a supervised machine-learning classifier and validated its performance with independent markers of disease severity obtained using computed tomography (CT) (CT cohort, n = 752) and cardiovascular magnetic resonance (CMR) imaging (CMR cohort, n = 160). The classifier's prognostic value was further validated using clinical outcomes (aortic valve replacement [AVR] and death) observed in the ECHO and CMR cohorts. RESULTS: In 1,964 patients from the 3 multi-institutional cohorts, 1,346 (68%) subjects had either nonsevere or discordant AS severity. Machine learning identified 1,117 (57%) patients as having high-severity and 847 (43%) as having low-severity AS. High-severity patients in CT and CMR cohorts had higher valve calcium scores and left ventricular mass and fibrosis, respectively than the low-severity group. In the ECHO cohort, progression to AVR and progression to death in patients who did not receive AVR was faster in the high-severity group. Compared with the conventional classification of disease severity, machine-learning-based severity classification improved discrimination (integrated discrimination improvement: 0.07; 95% confidence interval: 0.02 to 0.12) and reclassification (net reclassification improvement: 0.17; 95% confidence interval: 0.11 to 0.23) for the outcome of AVR at 5 years. For both ECHO and CMR cohorts, we observed prognostic value of the machine-learning classifications for subgroups with asymptomatic, nonsevere or discordant AS. CONCLUSIONS: Machine learning can integrate ECHO measurements to augment the classification of disease severity in most patients with AS, with major potential to optimize the timing of AVR.

Development and Validation of a Deep Learning Model to Quantify Interstitial Fibrosis and Tubular Atrophy From Kidney Ultrasonography Images.

Importance: Interstitial fibrosis and tubular atrophy (IFTA) is a strong indicator of decline in kidney function and is measured using histopathological assessment of kidney biopsy core. At present, a noninvasive test to assess IFTA is not available. Objective: To develop and validate a deep learning (DL) algorithm to quantify IFTA from kidney ultrasonography images. Design, Setting, and Participants: This was a single-center diagnostic study of consecutive patients who underwent native kidney biopsy at John H. Stroger Jr. Hospital of Cook County, Chicago, Illinois, between January 1, 2014, and December 31, 2018. A DL algorithm was trained, validated, and tested to classify IFTA from kidney ultrasonography images. Of 6135 Crimmins-filtered ultrasonography images, 5523 were used for training (5122 images) and validation (401 images), and 612 were used to test the accuracy of the DL system. Kidney segmentation was performed using the UNet architecture, and classification was performed using a convolution neural network-based feature extractor and extreme gradient boosting. IFTA scored by a nephropathologist on trichrome stained kidney biopsy slide was used as the reference standard. IFTA was divided into 4 grades (grade 1, 0%-24%; grade 2, 25%-49%; grade 3, 50%-74%; and grade 4, 75%-100%). Data analysis was performed from December 2019 to May 2020. Main Outcomes and Measures: Prediction of IFTA grade was measured using the metrics precision, recall, accuracy, and F1 score. Results: This study included 352 patients (mean [SD] age 47.43 [14.37] years), of whom 193 (54.82%) were women. There were 159 patients with IFTA grade 1 (2701 ultrasonography images), 74 patients with IFTA grade 2 (1239 ultrasonography images), 41 patients with IFTA grade 3 (701 ultrasonography images), and 78 patients with IFTA grade 4 (1494 ultrasonography images). Kidney ultrasonography images were segmented with 91% accuracy. In the independent test set, the point estimates for performance matrices showed precision of 0.8927 (95% CI, 0.8682-0.9172), recall of 0.8037 (95% CI, 0.7722-0.8352), accuracy of 0.8675 (95% CI, 0.8406-0.8944), and an F1 score of 0.8389 (95% CI, 0.8098-0.8680) at the image level. Corresponding estimates at the patient level were precision of 0.9003 (95% CI, 0.8644-0.9362), recall of 0.8421 (95% CI, 0.7984-0.8858), accuracy of 0.8955 (95% CI, 0.8589-0.9321), and an F1 score of 0.8639 (95% CI, 0.8228-0.9049). Accuracy at the patient level was highest for IFTA grade 1 and IFTA grade 4. The accuracy (approximately 90%) remained high irrespective of the timing of ultrasonography studies and the biopsy diagnosis. The predictive performance of the DL system did not show significant improvement when combined with baseline clinical characteristics. Conclusions and Relevance: These findings suggest that a DL algorithm can accurately and independently predict IFTA from kidney ultrasonography images.

Independent association of meteorological characteristics with initial spread of Covid-19 in India.

Whether weather plays a part in the transmissibility of the novel Coronavirus Disease-19 (COVID-19) is still not established. We tested the hypothesis that meteorological factors (air temperature, relative humidity, air pressure, wind speed and rainfall) are independently associated with transmissibility of COVID-19 quantified using the basic reproduction rate (R0). We used publicly available datasets on daily COVID-19 case counts (total n = 108,308), three-hourly meteorological data and community mobility data over a three-month period. Estimated R0 varied between 1.15 and 1.28. Mean daily air temperature (inversely), wind speed (positively) and countrywide lockdown (inversely) were significantly associated with time dependent R0, but the contribution of countrywide lockdown to variability in R0 was over three times stronger as compared to that of temperature and wind speed combined. Thus, abating temperatures and easing lockdown may concur with increased transmissibility of COVID-19 in India.

Artificial neural network-based prediction of prolonged length of stay and need for post-acute care in acute coronary syndrome patients undergoing percutaneous coronary intervention.

BACKGROUND: Prolonged length of stay (LOS) and post-acute care after percutaneous coronary intervention (PCI) is common and costly. Risk models for predicting prolonged LOS and post-acute care have limited accuracy. Our goal was to develop and validate models using artificial neural networks (ANN) to predict prolonged LOS > 7days and need for post-acute care after PCI. METHODS: We defined prolonged LOS as ≥7 days and post-acute care as patients discharged to: extended care, transitional care unit, rehabilitation, other acute care hospital, nursing home or hospice care. Data from 22 675 patients who presented with ACS and underwent PCI was shuffled and split into a derivation set (75% of dataset) and a validation dataset (25% of dataset). Calibration plots were used to examine the overall predictive performance of the MLP by plotting observed and expected risk deciles and fitting a lowess smoother to the data. Classification accuracy was assessed by a receiver-operating characteristic (ROC) and area under the ROC curve (AUC). RESULTS: Our MLP-based model predicted prolonged LOS with an accuracy of 90.87% and 88.36% in training and test sets, respectively. The post-acute care model had an accuracy of 90.22% and 86.31% in training and test sets, respectively. This accuracy was achieved with quick convergence. Predicted probabilities from the MLP models showed good (prolonged LOS) to excellent calibration (post-acute care). CONCLUSIONS: Our ANN-based models accurately predicted LOS and need for post-acute care. Larger studies for replicability and longitudinal studies for evidence of impact are needed to establish these models in current PCI practice.

Early identification of preterm neonates at birth with a Tablet App for the Simplified Gestational Age Score (T-SGAS) when ultrasound gestational age dating is unavailable: A validation study.

BACKGROUND: In low resource settings recall of the date of the mother's last menstrual period may be unreliable and due to limited availability of prenatal ultrasound, gestational age of newborns may not be assessed reliably. Preterm babies are at high risk of morbidity and mortality so an alternative strategy is to identify them soon after birth is needed for early referral and management. OBJECTIVE: The objective of this study was to assess the accuracy in assessing prematurity of newborn, over and above birthweight, using a pictorial Simplified Gestational Age Score adapted for use as a Tablet App. METHODS: Two trained nurse midwives, blinded to each other's assessment and the actual gestational age of the baby used the app to assess gestational age at birth in 3 hospitals based on the following 4 parameters-newborn's posture, skin texture, breast and genital development. Inter-observer variation was evaluated and the optimal scoring cut-off to detect preterm birth was determined. Sensitivity and specificity of gestational age score using the tablet was estimated using combinations of last menstrual period and ultrasound as reference standards to assess preterm birth. The predictive accuracy of the score using the area under a receiver operating characteristic curve was also determined. To account for potential reference standard bias, we also evaluated the score using latent class models. RESULTS: A total of 8,591 live singleton births whose gestational age by last menstrual period and ultrasound was within 1 weeks of each other were enrolled. There was strong agreement between assessors (concordance correlation coefficient 0.77 (95% CI 0.76-0.78) and Fleiss' kappa was 0.76 (95% CI 0.76-0.78). The optimal cut-off for the score to predict preterm was 13. Irrespective of the reference standard, the specificity of the score was 90% and sensitivity varied from 40-50% and the predictive accuracy between 74%-79% for the reference standards. The likelihood ratio of a positive score varied between 3.75-4.88 while the same for a negative likelihood ratio consistently varied between 0.57-0.72. Latent class models showed similar results indicating no reference standard bias. CONCLUSION: Gestational age scores had strong inter-observer agreement, robust prediction of preterm births simplicity of use by nurse midwives and can be a useful tool in resource-limited scenarios. TRIAL REGISTRATION: The Tablet App for the Simplified Gestational Age Score (T-SGAS) study was registered at ClinicalTrials.gov NCT02408783.

Incremental Prognostic Value of Anemia in Acute Coronary Syndrome from A Rural Hospital in India.

Background: Anemia is highly prevalent in low- and middle-income countries, where prevalence of acute coronary syndrome (ACS) is also rising. Evidence indicates that baseline anemia status can prognosticate ACS. However, the Global Registry of Acute Coronary Events (GRACE) score that is popularly used all over the world does not include information on anemia. Objectives: Our objective was to investigate if anemia at admission, along with the GRACE score, improves the prediction of adverse outcomes within 6 months in rural Indian patients of ACS. Methods: We enrolled 200 ACS patients at the Acharya Vinoba Bhave Rural Hospital-a rural, tertiary care hospital in central India. Patients were followed for 6 months for death and major adverse cardiac event (MACE). Improvement in the prediction of adverse events by including anemia in addition to the GRACE score was quantified using area under the receiver operating characteristic curve (AUC), integrated discrimination improvement (IDI) and the net reclassification index (NRI). Results: There were 31 deaths due to MACE and an additional 28 non-fatal MACE events during follow-up. Baseline hemoglobin was strongly and independently associated with both outcomes even after adjusting for a multivariable propensity score. For the outcome of death and death/MACE there was a moderate improvement in the AUC of 1% and 6%, respectively. However, for these outcomes the IDI for baseline hemoglobin was 6% (p = 0.03) and 12% (p << 0.0001), respectively, while the NRI was 0.50 (p = 0.01) and 0.78 (p << 0.0001), respectively. Conclusions: Inclusion of baseline anemia in addition to the GRACE score improves prognostication of ACS patients.

Improving Care Pathways for Acute Coronary Syndrome: Patients Undergoing Percutaneous Coronary Intervention.

Acute coronary syndrome (ACS) admissions are common and costly. The association between comprehensive ACS care pathways, outcomes, and costs are lacking. From 434,172 low-risk, uncomplicated ACS patients eligible for early discharge (STEMI 35%, UA/NSTEMI 65%) from the Premier database, we identified ACS care pathways, by stratifying low-risk, uncomplicated STEMI and UA/NSTEMI patients by access site for PCI (trans-radial intervention [TRI] vs transfemoral intervention [TFI]) and by length of stay (LOS). Associations with costs and outcomes (death, bleeding, acute kidney injury, and myocardial infarction at 1-year) were tested using hierarchical, mixed-effects regression, and projections of cost savings with change in care pathways were obtained using modeling. In low-risk uncomplicated STEMI patients, compared with TFI and LOS ≥3 days, a strategy of TRI with LOS <3 days and TFI with LOS <3 days were associated with cost savings of $6,206 and $4,802, respectively. Corresponding cost savings for UA/NSTEMI patients were $7,475 and $6,169, respectively. These care-pathways did not show an excess risk of adverse outcomes. We estimated that >$300 million could be saved if prevalence of the TRI with LOS <3 days and TFI with LOS <3 days strategies are modestly increased to 20% and 70%, respectively. In conclusion, we demonstrate the potential opportunity of cost savings by repositioning ACS care pathways in low-risk and uncomplicated ACS patients, toward transradial access and a shorter LOS without an increased risk of adverse outcomes.

The Evolving Landscape of Impella Use in the United States Among Patients Undergoing Percutaneous Coronary Intervention With Mechanical Circulatory Support.

BACKGROUND: Impella was approved for mechanical circulatory support (MCS) in 2008, but large-scale, real-world data on its use are lacking. Our objective was to describe trends and variations in Impella use, clinical outcomes, and costs across US hospitals in patients undergoing percutaneous coronary intervention (PCI) treated with MCS (Impella or intra-aortic balloon pump). METHODS: From the Premier Healthcare Database, we analyzed 48 306 patients undergoing PCI with MCS at 432 hospitals between January 2004 and December 2016. Association analyses were performed at 3 levels: time period, hospital, and patient. Hierarchical models with propensity adjustment were used for association analyses. We examined trends and variations in the proportion of Impella use, and associated clinical outcomes (in-hospital mortality, bleeding requiring transfusion, acute kidney injury, stroke, length of stay, and hospital costs). RESULTS: Among patients undergoing PCI treated with MCS, 4782 (9.9%) received Impella; its use increased over time, reaching 31.9% of MCS in 2016. There was wide variation in Impella use across hospitals (>5-fold variation). Specifically, among patients receiving Impella, there was a wide variation in outcomes of bleeding (>2.5-fold variation), and death, acute kidney injury, and stroke (all ≈1.5-fold variation). Adverse outcomes and costs were higher in the Impella era (years 2008-2016) versus the pre-Impella era (years 2004-2007). Hospitals with higher Impella use had higher rates of adverse outcomes and costs. After adjustment for the propensity score, and accounting for clustering of patients by hospitals, Impella use was associated with death: odds ratio, 1.24 (95% CI, 1.13-1.36); bleeding: odds ratio, 1.10 (95% CI, 1.00-1.21); and stroke: odds ratio, 1.34 (95% CI, 1.18-1.53), although a similar, nonsignificant result was observed for acute kidney injury: odds ratio, 1.08 (95% CI, 1.00-1.17). CONCLUSIONS: Impella use is rapidly increasing among patients undergoing PCI treated with MCS, with marked variability in its use and associated outcomes. Although unmeasured confounding cannot be ruled out, when analyzed by time periods, or at the hospital level or the patient level, Impella use was associated with higher rates of adverse events and costs. More data are needed to define the appropriate role of MCS in patients undergoing PCI.