Scalable and accurate deep learning with electronic health records.

Predictive modeling with electronic health record (EHR) data is anticipated to drive personalized medicine and improve healthcare quality. Constructing predictive statistical models typically requires extraction of curated predictor variables from normalized EHR data, a labor-intensive process that discards the vast majority of information in each patient's record. We propose a representation of patients' entire raw EHR records based on the Fast Healthcare Interoperability Resources (FHIR) format. We demonstrate that deep learning methods using this representation are capable of accurately predicting multiple medical events from multiple centers without site-specific data harmonization. We validated our approach using de-identified EHR data from two US academic medical centers with 216,221 adult patients hospitalized for at least 24 h. In the sequential format we propose, this volume of EHR data unrolled into a total of 46,864,534,945 data points, including clinical notes. Deep learning models achieved high accuracy for tasks such as predicting: in-hospital mortality (area under the receiver operator curve [AUROC] across sites 0.93-0.94), 30-day unplanned readmission (AUROC 0.75-0.76), prolonged length of stay (AUROC 0.85-0.86), and all of a patient's final discharge diagnoses (frequency-weighted AUROC 0.90). These models outperformed traditional, clinically-used predictive models in all cases. We believe that this approach can be used to create accurate and scalable predictions for a variety of clinical scenarios. In a case study of a particular prediction, we demonstrate that neural networks can be used to identify relevant information from the patient's chart.

Proton Pump Inhibitor Usage and the Risk of Myocardial Infarction in the General Population.

BACKGROUND AND AIMS: Proton pump inhibitors (PPIs) have been associated with adverse clinical outcomes amongst clopidogrel users after an acute coronary syndrome. Recent pre-clinical results suggest that this risk might extend to subjects without any prior history of cardiovascular disease. We explore this potential risk in the general population via data-mining approaches. METHODS: Using a novel approach for mining clinical data for pharmacovigilance, we queried over 16 million clinical documents on 2.9 million individuals to examine whether PPI usage was associated with cardiovascular risk in the general population. RESULTS: In multiple data sources, we found gastroesophageal reflux disease (GERD) patients exposed to PPIs to have a 1.16 fold increased association (95% CI 1.09-1.24) with myocardial infarction (MI). Survival analysis in a prospective cohort found a two-fold (HR = 2.00; 95% CI 1.07-3.78; P = 0.031) increase in association with cardiovascular mortality. We found that this association exists regardless of clopidogrel use. We also found that H2 blockers, an alternate treatment for GERD, were not associated with increased cardiovascular risk; had they been in place, such pharmacovigilance algorithms could have flagged this risk as early as the year 2000. CONCLUSIONS: Consistent with our pre-clinical findings that PPIs may adversely impact vascular function, our data-mining study supports the association of PPI exposure with risk for MI in the general population. These data provide an example of how a combination of experimental studies and data-mining approaches can be applied to prioritize drug safety signals for further investigation.

Progress towards Millennium Development Goals 4 and 5 on maternal and child mortality: an updated systematic analysis.

BACKGROUND: With 4 years until 2015, it is essential to monitor progress towards Millennium Development Goals (MDGs) 4 and 5. Although estimates of maternal and child mortality were published in 2010, an update of estimates is timely in view of additional data sources that have become available and new methods developed. Our aim was to update previous estimates of maternal and child mortality using better data and more robust methods to provide the best available evidence for tracking progress on MDGs 4 and 5. METHODS: We update the analyses of the progress towards MDGs 4 and 5 from 2010 with additional surveys, censuses, vital registration, and verbal autopsy data. For children, we estimate early neonatal (0-6 days), late neonatal (7-28 days), postneonatal (29-364 days), childhood (ages 1-4 years), and under-5 mortality. We use an improved model for estimating mortality by age under 5 years. For maternal mortality, our updated analysis includes greater than 1000 additional site-years of data. We tested a large set of alternative models for maternal mortality; we used an ensemble model based on the models with the best out-of-sample predictive validity to generate new estimates from 1990 to 2011. FINDINGS: Under-5 deaths have continued to decline, reaching 7·2 million in 2011 of which 2·2 million were early neonatal, 0·7 million late neonatal, 2·1 million postneonatal, and 2·2 million during childhood (ages 1-4 years). Comparing rates of decline from 1990 to 2000 with 2000 to 2011 shows that 106 countries have accelerated declines in the child mortality rate in the past decade. Maternal mortality has also continued to decline from 409,100 (uncertainty interval 382,900-437,900) in 1990 to 273,500 (256,300-291,700) deaths in 2011. We estimate that 56,100 maternal deaths in 2011 were HIV-related deaths during pregnancy. Based on recent trends in developing countries, 31 countries will achieve MDG 4, 13 countries MDG 5, and nine countries will achieve both. INTERPRETATION: Even though progress on reducing maternal and child mortality in most countries is accelerating, most developing countries will take many years past 2015 to achieve the targets of the MDGs 4 and 5. Similarly, although there continues to be progress on maternal mortality the pace is slow, without any overall evidence of acceleration. Immediate concerted action is needed for a large number of countries to achieve MDG 4 and MDG 5. FUNDING: Bill & Melinda Gates Foundation.

Neonatal, postneonatal, childhood, and under-5 mortality for 187 countries, 1970-2010: a systematic analysis of progress towards Millennium Development Goal 4.

BACKGROUND: Previous assessments have highlighted that less than a quarter of countries are on track to achieve Millennium Development Goal 4 (MDG 4), which calls for a two-thirds reduction in mortality in children younger than 5 years between 1990 and 2015. In view of policy initiatives and investments made since 2000, it is important to see if there is acceleration towards the MDG 4 target. We assessed levels and trends in child mortality for 187 countries from 1970 to 2010. METHODS: We compiled a database of 16 174 measurements of mortality in children younger than 5 years for 187 countries from 1970 to 2009, by use of data from all available sources, including vital registration systems, summary birth histories in censuses and surveys, and complete birth histories. We used Gaussian process regression to generate estimates of the probability of death between birth and age 5 years. This is the first study that uses Gaussian process regression to estimate child mortality, and this technique has better out-of-sample predictive validity than do previous methods and captures uncertainty caused by sampling and non-sampling error across data types. Neonatal, postneonatal, and childhood mortality was estimated from mortality in children younger than 5 years by use of the 1760 measurements from vital registration systems and complete birth histories that contained specific information about neonatal and postneonatal mortality. FINDINGS: Worldwide mortality in children younger than 5 years has dropped from 11.9 million deaths in 1990 to 7.7 million deaths in 2010, consisting of 3.1 million neonatal deaths, 2.3 million postneonatal deaths, and 2.3 million childhood deaths (deaths in children aged 1-4 years). 33.0% of deaths in children younger than 5 years occur in south Asia and 49.6% occur in sub-Saharan Africa, with less than 1% of deaths occurring in high-income countries. Across 21 regions of the world, rates of neonatal, postneonatal, and childhood mortality are declining. The global decline from 1990 to 2010 is 2.1% per year for neonatal mortality, 2.3% for postneonatal mortality, and 2.2% for childhood mortality. In 13 regions of the world, including all regions in sub-Saharan Africa, there is evidence of accelerating declines from 2000 to 2010 compared with 1990 to 2000. Within sub-Saharan Africa, rates of decline have increased by more than 1% in Angola, Botswana, Cameroon, Congo, Democratic Republic of the Congo, Kenya, Lesotho, Liberia, Rwanda, Senegal, Sierra Leone, Swaziland, and The Gambia. INTERPRETATION: Robust measurement of mortality in children younger than 5 years shows that accelerating declines are occurring in several low-income countries. These positive developments deserve attention and might need enhanced policy attention and resources. FUNDING: Bill & Melinda Gates Foundation.

Worldwide mortality in men and women aged 15-59 years from 1970 to 2010: a systematic analysis.

BACKGROUND: Adult deaths are a crucial priority for global health. Causes of adult death are important components of Millennium Development Goals 5 and 6. However, adult mortality has received little policy attention, resources, or monitoring efforts. This study aimed to estimate worldwide mortality in men and women aged 15-59 years. METHODS: We compiled a database of 3889 measurements of adult mortality for 187 countries from 1970 to 2010 using vital registration data and census and survey data for deaths in the household corrected for completeness, and sibling history data from surveys corrected for survival bias. We used Gaussian process regression to generate yearly estimates of the probability of death between the ages of 15 years and 60 years (45q15) for men and women for every country with uncertainty intervals that indicate sampling and non-sampling error. We showed that these analytical methods have good predictive validity for countries with missing data. FINDINGS: Adult mortality varied substantially across countries and over time. In 2010, the countries with the lowest risk of mortality for men and women are Iceland and Cyprus, respectively. In Iceland, male 45q15 is 65 (uncertainty interval 61-69) per 1000; in Cyprus, female 45q15 is 38 (36-41) per 1000. Highest risk of mortality in 2010 is seen in Swaziland for men (45q15 of 765 [692-845] per 1000) and Zambia for women (606 [518-708] per 1000). Between 1970 and 2010, substantial increases in adult mortality occurred in sub-Saharan Africa because of the HIV epidemic and in countries in or related to the former Soviet Union. Other regional trends were also seen, such as stagnation in the decline of adult mortality for large countries in southeast Asia and a striking decline in female mortality in south Asia. INTERPRETATION: The prevention of premature adult death is just as important for global health policy as the improvement of child survival. Routine monitoring of adult mortality should be given much greater emphasis. FUNDING: Bill & Melinda Gates Foundation.

The essential role of Broca's area in imitation.

The posterior sector of Broca's area (Brodmann area 44), a brain region critical for language, may have evolved from neurons active during observation and execution of manual movements. Imaging studies showing increased Broca's activity during execution, imagination, imitation and observation of hand movements support this hypothesis. Increased Broca's activity in motor task, however, may simply be due to inner speech. To test whether Broca's area is essential to imitation, we used repetitive transcranial magnetic stimulation (rTMS), which is known to transiently disrupt functions in stimulated areas. Subjects imitated finger key presses (imitation) or executed finger key presses in response to spatial cues (control task). While performing the tasks, subjects received rTMS over the left and right pars opercularis of the inferior frontal gyrus (where Brodmann area 44 is probabilistically located) and over the occipital cortex. There was significant impairment in imitation, but not in the control task, during rTMS over left and right pars opercularis compared to rTMS over the occipital cortex. This suggests that Broca's area is a premotor region essential to finger movement imitation.