Person:

Reis, Ben

Drug-drug interactions (DDIs) can lead to serious and potentially lethal adverse events. In recent years, several drugs have been withdrawn from the market due to interaction-related adverse events (AEs). Current methods for detecting DDIs rely on the accumulation of sufficient clinical evidence in the post-market stage – a lengthy process that often takes years, during which time numerous patients may suffer from the adverse effects of the DDI. Detection methods are further hindered by the extremely large combinatoric space of possible drug-drug-AE combinations. There is therefore a practical need for predictive tools that can identify potential DDIs years in advance, enabling drug safety professionals to better prioritize their limited investigative resources and take appropriate regulatory action. To meet this need, we describe Predictive Pharmacointeraction Networks (PPINs) – a novel approach that predicts unknown DDIs by exploiting the network structure of all known DDIs, together with other intrinsic and taxonomic properties of drugs and AEs. We constructed an 856-drug DDI network from a 2009 snapshot of a widely-used drug safety database, and used it to develop PPIN models for predicting future DDIs. We compared the DDIs predicted based solely on these 2009 data, with newly reported DDIs that appeared in a 2012 snapshot of the same database. Using a standard multivariate approach to combine predictors, the PPIN model achieved an AUROC (area under the receiver operating characteristic curve) of 0.81 with a sensitivity of 48% given a specificity of 90%. An analysis of DDIs by severity level revealed that the model was most effective for predicting “contraindicated” DDIs (AUROC = 0.92) and less effective for “minor” DDIs (AUROC = 0.63). These results indicate that network based methods can be useful for predicting unknown drug-drug interactions.

Background: Emergency department (ED) based syndromic surveillance systems identify abnormally high visit rates that may be an early signal of a bioterrorist attack. For example, an anthrax outbreak might first be detectable as an unusual increase in the number of patients reporting to the ED with respiratory symptoms. Reliably identifying these abnormal visit patterns requires a good understanding of the normal patterns of healthcare usage. Unfortunately, systematic methods for determining the expected number of (ED) visits on a particular day have not yet been well established. We present here a generalized methodology for developing models of expected ED visit rates. Methods: Using time-series methods, we developed robust models of ED utilization for the purpose of defining expected visit rates. The models were based on nearly a decade of historical data at a major metropolitan academic, tertiary care pediatric emergency department. The historical data were fit using trimmed-mean seasonal models, and additional models were fit with autoregressive integrated moving average (ARIMA) residuals to account for recent trends in the data. The detection capabilities of the model were tested with simulated outbreaks. Results: Models were built both for overall visits and for respiratory-related visits, classified according to the chief complaint recorded at the beginning of each visit. The mean absolute percentage error of the ARIMA models was 9.37% for overall visits and 27.54% for respiratory visits. A simple detection system based on the ARIMA model of overall visits was able to detect 7-day-long simulated outbreaks of 30 visits per day with 100% sensitivity and 97% specificity. Sensitivity decreased with outbreak size, dropping to 94% for outbreaks of 20 visits per day, and 57% for 10 visits per day, all while maintaining a 97% benchmark specificity. Conclusions: Time series methods applied to historical ED utilization data are an important tool for syndromic surveillance. Accurate forecasting of emergency department total utilization as well as the rates of particular syndromes is possible. The multiple models in the system account for both long-term and recent trends, and an integrated alarms strategy combining these two perspectives may provide a more complete picture to public health authorities. The systematic methodology described here can be generalized to other healthcare settings to develop automated surveillance systems capable of detecting anomalies in disease patterns and healthcare utilization.

Background: Internet search patterns have emerged as a novel data source for monitoring infectious disease trends. We propose that these data can also be used more broadly to study the impact of health policies across different regions in a more efficient and timely manner. Methods: As a test use case, we studied the relationships between abortion-related search volume, local abortion rates, and local abortion policies available for study. Results: Our initial integrative analysis found that, both in the US and internationally, the volume of Internet searches for abortion is inversely proportional to local abortion rates and directly proportional to local restrictions on abortion. Conclusion: These findings are consistent with published evidence that local restrictions on abortion lead individuals to seek abortion services outside of their area. Further validation of these methods has the potential to produce a timely, complementary data source for studying the effects of health policies.

Background: Accurate prediction of adverse drug events (ADEs) is an important means of controlling and reducing drug-related morbidity and mortality. Since no single “gold standard” ADE data set exists, a range of different drug safety data sets are currently used for developing ADE prediction models. There is a critical need to assess the degree of concordance between these various ADE data sets and to validate ADE prediction models against multiple reference standards. Methods: We systematically evaluated the concordance of two widely used ADE data sets – Lexi-comp from 2010 and SIDER from 2012. The strength of the association between ADE (drug) counts in Lexi-comp and SIDER was assessed using Spearman rank correlation, while the differences between the two data sets were characterized in terms of drug categories, ADE categories and ADE frequencies. We also performed a comparative validation of the Predictive Pharmacosafety Networks (PPN) model using both ADE data sets. The predictive power of PPN using each of the two validation sets was assessed using the area under Receiver Operating Characteristic curve (AUROC). Results: The correlations between the counts of ADEs and drugs in the two data sets were 0.84 (95% CI: 0.82-0.86) and 0.92 (95% CI: 0.91-0.93), respectively. Relative to an earlier snapshot of Lexi-comp from 2005, Lexi-comp 2010 and SIDER 2012 introduced a mean of 1,973 and 4,810 new drug-ADE associations per year, respectively. The difference between these two data sets was most pronounced for Nervous System and Anti-infective drugs, Gastrointestinal and Nervous System ADEs, and postmarketing ADEs. A minor difference of 1.1% was found in the AUROC of PPN when SIDER 2012 was used for validation instead of Lexi-comp 2010. Conclusions: In conclusion, the ADE and drug counts in Lexi-comp and SIDER data sets were highly correlated and the choice of validation set did not greatly affect the overall prediction performance of PPN. Our results also suggest that it is important to be aware of the differences that exist among ADE data sets, especially in modeling applications focused on specific drug and ADE categories.

Background: Physician notes routinely recorded during patient care represent a vast and underutilized resource for human disease studies on a population scale. Their use in research is primarily limited by the need to separate confidential patient information from clinical annotations, a process that is resource-intensive when performed manually. This study seeks to create an automated method for de-identifying physician notes that does not require large amounts of private information: in addition to training a model to recognize Protected Health Information (PHI) within private physician notes, we reverse the problem and train a model to recognize non-PHI words and phrases that appear in public medical texts. Methods: Public and private medical text sources were analyzed to distinguish common medical words and phrases from Protected Health Information. Patient identifiers are generally nouns and numbers that appear infrequently in medical literature. To quantify this relationship, term frequencies and part of speech tags were compared between journal publications and physician notes. Standard medical concepts and phrases were then examined across ten medical dictionaries. Lists and rules were included from the US census database and previously published studies. In total, 28 features were used to train decision tree classifiers. Results: The model successfully recalled 98% of PHI tokens from 220 discharge summaries. Cost sensitive classification was used to weight recall over precision (98% F10 score, 76% F1 score). More than half of the false negatives were the word “of” appearing in a hospital name. All patient names, phone numbers, and home addresses were at least partially redacted. Medical concepts such as “elevated white blood cell count” were informative for de-identification. The results exceed the previously approved criteria established by four Institutional Review Boards. Conclusions: The results indicate that distributional differences between private and public medical text can be used to accurately classify PHI. The data and algorithms reported here are made freely available for evaluation and improvement.

Objective: To determine whether longitudinal data in patients’ historical records, commonly available in electronic health record systems, can be used to predict a patient’s future risk of receiving a diagnosis of domestic abuse. Design: Bayesian models, known as intelligent histories, used to predict a patient’s risk of receiving a future diagnosis of abuse, based on the patient’s diagnostic history. Retrospective evaluation of the model’s predictions using an independent testing set. Setting: A state-wide claims database covering six years of inpatient admissions to hospital, admissions for observation, and encounters in emergency departments. Population: All patients aged over 18 who had at least four years between their earliest and latest visits recorded in the database (561 216 patients). Main outcome measures: Timeliness of detection, sensitivity, specificity, positive predictive values, and area under the ROC curve. Results: 1.04% (5829) of the patients met the narrow case definition for abuse, while 3.44% (19 303) met the broader case definition for abuse. The model achieved sensitive, specific (area under the ROC curve of 0.88), and early (10-30 months in advance, on average) prediction of patients’ future risk of receiving a diagnosis of abuse. Analysis of model parameters showed important differences between sexes in the risks associated with certain diagnoses. Conclusions: Commonly available longitudinal diagnostic data can be useful for predicting a patient’s future risk of receiving a diagnosis of abuse. This modelling approach could serve as the basis for an early warning system to help doctors identify high risk patients for further screening.

Background and Objective Several studies have demonstrated the ability to detect adverse events potentially related to multiple drug exposure via data mining. However, the number of putative associations produced by such computational approaches is typically large, making experimental validation difficult. We theorized that those potential associations for which there is evidence from multiple complementary sources are more likely to be true, and explored this idea using a published database of drug–drug-adverse event associations derived from electronic health records (EHRs). Methods: We prioritized drug–drug-event associations derived from EHRs using four sources of information: (1) public databases, (2) sources of spontaneous reports, (3) literature, and (4) non-EHR drug–drug interaction (DDI) prediction methods. After pre-filtering the associations by removing those found in public databases, we devised a ranking for associations based on the support from the remaining sources, and evaluated the results of this rank-based prioritization. Results: We collected information for 5983 putative EHR-derived drug–drug-event associations involving 345 drugs and ten adverse events from four data sources and four prediction methods. Only seven drug–drug-event associations (<0.5 %) had support from the majority of evidence sources, and about one third (1777) had support from at least one of the evidence sources. Conclusions: Our proof-of-concept method for scoring putative drug–drug-event associations from EHRs offers a systematic and reproducible way of prioritizing associations for further study. Our findings also quantify the agreement (or lack thereof) among complementary sources of evidence for drug–drug-event associations and highlight the challenges of developing a robust approach for prioritizing signals of these associations. Electronic supplementary material The online version of this article (doi:10.1007/s40264-015-0352-2) contains supplementary material, which is available to authorized users.

To evaluate the effectiveness of the BNT162b2 mRNA vaccine among pregnant women, we conducted an observational cohort study of pregnant women 16 years or older, with no history of SARS-CoV-2, who were vaccinated between December 20, 2020 and June 3, 2021. 10,861 vaccinated pregnant women were matched to 10,861 unvaccinated control women using demographic and clinical characteristics. Study outcomes included documented infection with SARS-CoV-2, symptomatic COVID-19, COVID-19-related hospitalization, severe illness and death. Estimated vaccine effectiveness from 7 through 28 days after the second dose was 97% (95% CI 91%-100%) for any documented infection, 96% (86-100%) for infections with documented symptoms, and 85% (32%-100%) for COVID-19-related hospitalization. Only one event of severe illness was observed in the unvaccinated group, and no deaths were observed in either group. In summary, the BNT162b2 mRNA vaccine was estimated to have high vaccine effectiveness among pregnant women, similar to the effectiveness estimated in the general population.