The early prediction of deterioration could have an important role in supporting healthcare professionals, as an estimated 11% of deaths in hospital follow a failure to promptly recognize and treat ...deteriorating patients
. To achieve this goal requires predictions of patient risk that are continuously updated and accurate, and delivered at an individual level with sufficient context and enough time to act. Here we develop a deep learning approach for the continuous risk prediction of future deterioration in patients, building on recent work that models adverse events from electronic health records
and using acute kidney injury-a common and potentially life-threatening condition
-as an exemplar. Our model was developed on a large, longitudinal dataset of electronic health records that cover diverse clinical environments, comprising 703,782 adult patients across 172 inpatient and 1,062 outpatient sites. Our model predicts 55.8% of all inpatient episodes of acute kidney injury, and 90.2% of all acute kidney injuries that required subsequent administration of dialysis, with a lead time of up to 48 h and a ratio of 2 false alerts for every true alert. In addition to predicting future acute kidney injury, our model provides confidence assessments and a list of the clinical features that are most salient to each prediction, alongside predicted future trajectories for clinically relevant blood tests
. Although the recognition and prompt treatment of acute kidney injury is known to be challenging, our approach may offer opportunities for identifying patients at risk within a time window that enables early treatment.
Ehlers-Danlos Syndrome (EDS) are a heterogeneous group of genetic connective tissue disorders, and typically manifests as weak joints that subluxate/dislocate, stretchy and/or fragile skin, ...organ/systems dysfunction, and significant widespread pain. Historically, this syndrome has been poorly understood and often overlooked. As a result, people living with EDS had difficulty obtaining an accurate diagnosis and appropriate treatment, leading to untold personal suffering as well as ineffective health care utilization. The GoodHope EDS clinic addresses systemic gaps in the diagnosis and treatment of EDS. This paper describes a leap forward--from lack of awareness, diagnosis, and treatment--to expert care that is tailored to meet the specific needs of patients with EDS. The GoodHope EDS clinic consists of experts from various medical specialties who work together to provide comprehensive care that addresses the multi-systemic nature of the syndrome. In addition, EDS-specific self-management programs have been developed that draw on exercise science, rehabilitation, and health psychology to improve physical and psychosocial wellbeing and overall quality of life. Embedded into the program are research initiatives to shed light on the clinical presentation, underlying mechanisms of pathophysiology, and syndrome management. We also lead regular educational activities for community health care providers to increase awareness and competence in the interprofessional management of EDS beyond our doors and throughout the province and country.
Objective Hospital-acquired acute kidney injury (HA-AKI) is a potentially preventable cause of morbidity and mortality. Identifying high-risk patients prior to the onset of kidney injury is a key ...step towards AKI prevention.
Materials and Methods A national retrospective cohort of 1,620,898 patient hospitalizations from 116 Veterans Affairs hospitals was assembled from electronic health record (EHR) data collected from 2003 to 2012. HA-AKI was defined at stage 1+, stage 2+, and dialysis. EHR-based predictors were identified through logistic regression, least absolute shrinkage and selection operator (lasso) regression, and random forests, and pair-wise comparisons between each were made. Calibration and discrimination metrics were calculated using 50 bootstrap iterations. In the final models, we report odds ratios, 95% confidence intervals, and importance rankings for predictor variables to evaluate their significance.
Results The area under the receiver operating characteristic curve (AUC) for the different model outcomes ranged from 0.746 to 0.758 in stage 1+, 0.714 to 0.720 in stage 2+, and 0.823 to 0.825 in dialysis. Logistic regression had the best AUC in stage 1+ and dialysis. Random forests had the best AUC in stage 2+ but the least favorable calibration plots. Multiple risk factors were significant in our models, including some nonsteroidal anti-inflammatory drugs, blood pressure medications, antibiotics, and intravenous fluids given during the first 48 h of admission.
Conclusions This study demonstrated that, although all the models tested had good discrimination, performance characteristics varied between methods, and the random forests models did not calibrate as well as the lasso or logistic regression models. In addition, novel modifiable risk factors were explored and found to be significant.
Acute kidney injury (AKI) occurs frequently after cardiac catheterization and percutaneous coronary intervention. Although a clinical risk model exists for percutaneous coronary intervention, no ...models exist for both procedures, nor do existing models account for risk factors prior to the index admission. We aimed to develop such a model for use in prospective automated surveillance programs in the Veterans Health Administration.
We collected data on all patients undergoing cardiac catheterization or percutaneous coronary intervention in the Veterans Health Administration from January 01, 2009 to September 30, 2013, excluding patients with chronic dialysis, end-stage renal disease, renal transplant, and missing pre- and postprocedural creatinine measurement. We used 4 AKI definitions in model development and included risk factors from up to 1 year prior to the procedure and at presentation. We developed our prediction models for postprocedural AKI using the least absolute shrinkage and selection operator (LASSO) and internally validated using bootstrapping. We developed models using 115 633 angiogram procedures and externally validated using 27 905 procedures from a New England cohort. Models had cross-validated C-statistics of 0.74 (95% CI: 0.74-0.75) for AKI, 0.83 (95% CI: 0.82-0.84) for AKIN2, 0.74 (95% CI: 0.74-0.75) for contrast-induced nephropathy, and 0.89 (95% CI: 0.87-0.90) for dialysis.
We developed a robust, externally validated clinical prediction model for AKI following cardiac catheterization or percutaneous coronary intervention to automatically identify high-risk patients before and immediately after a procedure in the Veterans Health Administration. Work is ongoing to incorporate these models into routine clinical practice.
Background Anemia is a risk factor for adverse cardiovascular disease outcomes. However, there is limited information concerning the association of hemoglobin concentration and new onset of ...clinically recognized coronary artery disease (CAD). Methods An historical cohort study was conducted with patients from Veterans Affairs medical centers. Baseline hemoglobin determinations were evaluated with respect to CAD using data from records of 25 622 subjects with no known heart disease. Coronary artery disease was identified from a new diagnosis based on the International Classification of Diseases, Ninth Edition , coding or a new prescription for nitroglycerin. Models were adjusted for age, sex, body mass index, smoking, systolic blood pressure, diastolic blood pressure, fasting glucose, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, triglycerides, creatinine clearance, and use of statin or β-blocker. Results Among the cohort, 5297 (20.7%) subjects developed CAD over 73 895 person-years of follow-up. Compared with control hemoglobin levels of 15.0 to 17.0 g/dL, the multivariable-adjusted risk of CAD increased with lower hemoglobin levels: an adjusted hazard ratio (HR) of 1.47 and 95% confidence interval (CI) of 1.18 to 1.84 for hemoglobin levels of 9.0 to 11.0 g/dL; an HR of 1.34 and 95% CI of 1.20 to 1.49 for 11.0 to 13.0 g/dL; and an HR of 1.07 and 95% CI of 1.01 to 1.13 for 13.0 to 15.0 g/dL. Hemoglobin levels ≥17.0 g/dL were also associated with increased risk for CAD (adjusted HR 1.22, 95% CI 1.08-1.37). Conclusions Hemoglobin levels ≥17 or <15 g/dL are independently associated with increase risk for new cardiac events.
Emerging evidence indicates that perinatal infection and inflammation can influence the developing immune system and may ultimately affect long-term health and disease outcomes in offspring by ...perturbing tissue and immune homeostasis. We posit that perinatal inflammation influences immune outcomes in offspring by perturbing (1) the development and function of fetal-derived immune cells that regulate tissue development and homeostasis, and (2) the establishment and function of developing hematopoietic stem cells (HSCs) that continually generate immune cells across the lifespan. To disentangle the complexities of these interlinked systems, we propose the cochlea as an ideal model tissue to investigate how perinatal infection affects immune, tissue, and stem cell development. The cochlea contains complex tissue architecture and a rich immune milieu that is established during early life. A wide range of congenital infections cause cochlea dysfunction and sensorineural hearing loss (SNHL), likely attributable to early life inflammation. Furthermore, we show that both immune cells and bone marrow hematopoietic progenitors can be simultaneously analyzed within neonatal cochlear samples. Future work investigating the pathogenesis of SNHL in the context of congenital infection will therefore provide critical information on how perinatal inflammation drives disease susceptibility in offspring.
OBJECTIVE:--Nondiabetic patients were studied to determine whether increasing blood glucose is associated with subsequent incidence of heart failure. RESEARCH DESIGN AND METHODS--Baseline morning ...blood glucose determinations were evaluated with respect to subsequent heart failure using records from 20,810 nondiabetic patients. The onset of heart failure >1 year after initial glucose determinations was evaluated for patients who had 2-12 years of care. Patients were excluded if they had ever had the diagnosis of diabetes, had a diagnosis of heart failure <1 year after initial blood glucose determinations, had a blood glucose determination >125 mg/dl, or used corticosteroids, loop diuretics, insulin, or oral hypoglycemics. RESULTS:--Of the 20,810 patients studied, 916 patients developed heart failure over a total analysis time of 71,890 years at risk. Higher baseline morning glucose levels were associated with increased heart failure from 3.5% (glucose <90 mg/dl) to 3.8% (90-99 mg/dl) to 4.8% (100-109 mg/dl) to 6% (110-125 mg/dl) over a mean 4- to 5-year evaluation period. The incidence rate increased from 7.5 cases per 1,000 person-years (glucose <90 mg/dl) to 8.4 (90-99 mg/dl, NS) to 11.1 (100-109 mg/dl, P < 0.001) to 13.7 (110-125 mg/dl, P < 0.0001), an 83% increase in heart failure if baseline glucose was >109 mg/dl compared with <90 mg/dl. A Cox proportionate hazards model including age, sex, BMI, creatinine, hypertension, lipids, smoking, medications, and coronary disease showed a progressive increase in hazard ratio from 1.25 (glucose 90-99 mg/dl, P < 0.05) to 1.46 (100-109 mg/dl, P < 0.001) to 1.55 (110-125 mg/dl, P < 0.001) compared with glucose <90 mg/dl. Kaplan-Meier analysis showed increased glucose- associated risk with time. CONCLUSIONS:--Patients with higher baseline blood glucose levels in the absence of diabetes and after adjustment for covariants have a significantly increased risk of heart failure.
Background Surveillance at hospital admission for multidrug-resistant (MDR) gram-negative bacteria (GNB) is not often performed, potentially leaving patients carrying these organisms unrecognized and ...not placed in transmission precautions until they develop infection. Veterans Affairs (VA) facilities screen all admissions for methicillin-resistant Staphylococcus aureus (MRSA) and place positive patients in contact precautions. We assessed how often patients with MDR GNB in clinical cultures obtained within 30 days following admission would have been in contact precautions because of a positive MRSA admission screen. Methods MRSA screening and MDR GNB culture results were extracted from a database of patients admitted to all VA acute care medical facilities from January 2009-December 2012. Results Of patients with MDR GNB-positive cultures within 30 days following admission, up to 44.3% (dependent on bacterial species) would have been in contact precautions because of a clinical positive admission MRSA nasal screen. Admissions with a positive MRSA screen had odds for MDR GNB in a culture 2.5 times greater than those with a negative screen (95% confidence interval CI, 2.4-2.6). Odds ratios were 2.4 (95% CI, 2.3-2.5) for MDR Enterobacteriaceae , 2.7 (95% CI, 2.5-2.9) for MDR Pseudomonas aeruginosa , and 4.3 (95% CI, 3.8-4.8) for MDR Acinetobacter spp. Conclusions Patients may be serendipitously placed in contact precautions for MDR GNB when isolated for a positive admission MRSA screen.
Patients with heart failure (HF) are at high risk of hospitalization or death. The objective of this study was to develop prediction models to identify patients with HF at highest risk for ...hospitalization or death. Using clinical and administrative databases, we identified 198,460 patients who received care from the Veterans Health Administration and had ≥1 primary or secondary diagnosis of HF that occurred within 1 year before June 1, 2009. We then tracked their outcomes of hospitalization and death during the subsequent 30 days and 1 year. Predictor variables chosen from 6 clinically relevant categories of sociodemographics, medical conditions, vital signs, use of health services, laboratory tests, and medications were used in multinomial regression models to predict outcomes of hospitalization and death. In patients who were in the ≥95th predicted risk percentile, observed event rates of hospitalization or death within 30 days and 1 year were 27% and 80% respectively, compared to population averages of 5% and 31%, respectively. The c-statistics for the 30-day outcomes were 0.82, 0.80, and 0.80 for hospitalization, death, and hospitalization or death, respectively, and 0.82, 0.76, and 0.77, respectively, for 1-year outcomes. In conclusion, prediction models using electronic health records can accurately identify patients who are at highest risk for hospitalization or death. This information can be used to assist care managers in selecting patients for interventions to decrease their risk of hospitalization or death.
Accurate information is needed to direct healthcare systems' efforts to control methicillin-resistant Staphylococcus aureus (MRSA). Assembling complete and correct microbiology data is vital to ...understanding and addressing the multiple drug-resistant organisms in our hospitals.
Herein, we describe a system that securely gathers microbiology data from the Department of Veterans Affairs (VA) network of databases. Using natural language processing methods, we applied an information extraction process to extract organisms and susceptibilities from the free-text data. We then validated the extraction against independently derived electronic data and expert annotation.
We estimate that the collected microbiology data are 98.5% complete and that methicillin-resistant Staphylococcus aureus was extracted accurately 99.7% of the time.
Applying natural language processing methods to microbiology records appears to be a promising way to extract accurate and useful nosocomial pathogen surveillance data. Both scientific inquiry and the data's reliability will be dependent on the surveillance system's capability to compare from multiple sources and circumvent systematic error. The dataset constructed and methods used for this investigation could contribute to a comprehensive infectious disease surveillance system or other pressing needs.
PDF
