The Drug Data to Knowledge Pipeline: Large-Scale Claims Data Classification for Pharmacologic Insight.

In biomedical informatics, assigning drug codes to categories is a common step in the analysis pipeline. Unfortunately, incomplete mappings are the norm rather than the exception with coverage values less than 85% not uncommon. Here, we perform this linking task on a nationwide insurance claims database with over 13 million members who were dispensed, according to National Drug Codes (NDCs), over 50,000 unique product forms of medication. The chosen approach employs Cerner Multum's VantageRx and the U.S. National Library of Medicine's RxMix. As a result, 94.0% of the NDCs were successfully mapped to categories used by common drug terminologies, e.g., Anatomical Therapeutic Chemical (ATC). Implemented as an SQL database and scripts, the approach is generic and can be setup for a new data set in a few hours. Thus, the method is a viable option for large-scale drug classification.

Feasibility of Prioritizing Drug-Drug-Event Associations Found in Electronic Health Records.

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.

Provider Patient-Sharing Networks and Multiple-Provider Prescribing of Benzodiazepines.

BACKGROUND: Prescription benzodiazepine overdose continues to cause significant morbidity and mortality in the US. Multiple-provider prescribing, due to either fragmented care or "doctor-shopping," contributes to the problem. OBJECTIVE: To elucidate the effect of provider professional relationships on multiple-provider prescribing of benzodiazepines, using social network analytics. DESIGN: A retrospective analysis of commercial healthcare claims spanning the years 2008 through 2011. Provider patient-sharing networks were modelled using social network analytics. Care team cohesion was measured using care density, defined as the ratio between the total number of patients shared by provider pairs within a patient's care team and the total number of provider pairs in the care team. Relationships within provider pairs were further quantified using a range of network metrics, including the number and proportion of patients or collaborators shared. MAIN MEASURES: The relationship between patient-sharing network metrics and the likelihood of multiple prescribing of benzodiazepines. PARTICIPANTS: Patients between the ages of 18 and 64 years who received two or more benzodiazepine prescriptions from multiple providers, with overlapping coverage of more than 14 days. RESULTS: A total of 5659 patients and 1448 provider pairs were included in our study. Among these, 1028 patients (18.2 %) received multiple prescriptions of benzodiazepines, involving 445 provider pairs (30.7 %). Patients whose providers rarely shared patients had a higher risk of being prescribed overlapping benzodiazepines; the median care density was 8.1 for patients who were prescribed overlapping benzodiazepines and 10.1 for those who were not (p < 0.0001). Provider pairs who shared a greater number of patients and collaborators were less likely to co-prescribe overlapping benzodiazepines. CONCLUSIONS: Our findings demonstrate the importance of care team cohesion in addressing multiple-provider prescribing of controlled substances. Furthermore, we illustrate the potential of the provider network as a surveillance tool to detect and prevent adverse events that could arise due to fragmentation of care.

Using Nation-Wide Health Insurance Claims Data to Augment Lyme Disease Surveillance.

OBJECTIVE: Lyme disease (LD) is the most commonly reported tick-borne illness in North America. To improve LD surveillance, we explored claims data as an adjunct data source for monitoring trends in Lyme disease incidence. METHODS: We retrospectively analyzed claims from a nationwide US health insurance plan, identifying patients with newly diagnosed LD in 13 high-prevalence states over two time periods, 2004-2006 and 2010-2012. RESULTS: The average LD case incidence as estimated by using claims data in 2010-2012 (75.67 per 100,000 person-years, n = 3474) was 1.50 times higher than 2004-2006 (50.25 per 100,000 person-years, n = 1965) (p < 0.001) and higher than incidence reported by the states to the Centers for Disease Control and Prevention. Among the 13 highest-prevalence states, there were 11 states with increased LD incidence over time. CONCLUSIONS: Surveillance systems should explore a fusion of data sources, including payer claims that appear to be highly sensitive with limitations, with electronic laboratory data that afford high specificity, but appear to miss cases.

Incidence and Patterns of Extended-Course Antibiotic Therapy in Patients Evaluated for Lyme Disease.

BACKGROUND: Most patients with Lyme disease (LD) can be treated effectively with 2-4 weeks of antibiotics. The Infectious Disease Society of America guidelines do not currently recommend extended treatment even in patients with persistent symptoms. METHODS: To estimate the incidence of extended use of antibiotics in patients evaluated for LD, we retrospectively analyzed claims from a nationwide US health insurance plan in 14 high-prevalence states over 2 periods: 2004-2006 and 2010-2012. RESULTS: As measured by payer claims, the incidence of extended antibiotic therapy among patients evaluated for LD was higher in 2010-2012 (14.72 per 100 000 person-years; n = 684) than in 2004-2006 (9.94 per 100 000 person-years; n = 394) (P < .001). Among these patients, 48.8% were treated with ≥2 antibiotics in 2010-2012 and 29.9% in 2004-2006 (P < .001). In each study period, a distinct small group of providers (roughly 3%-4%) made the diagnosis in >20% of the patients who were evaluated for LD and prescribed extended antibiotic treatment. CONCLUSIONS: Insurance claims data suggest that the use of extended courses of antibiotics and multiple antibiotics in the treatment of LD has increased in recent years.

Concordance and predictive value of two adverse drug event data sets.

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.

Pharmacointeraction network models predict unknown drug-drug interactions.

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.

Predicting adverse drug events using pharmacological network models.

Early and accurate identification of adverse drug events (ADEs) is critically important for public health. We have developed a novel approach for predicting ADEs, called predictive pharmacosafety networks (PPNs). PPNs integrate the network structure formed by known drug-ADE relationships with information on specific drugs and adverse events to predict likely unknown ADEs. Rather than waiting for sufficient post-market evidence to accumulate for a given ADE, this predictive approach relies on leveraging existing, contextual drug safety information, thereby having the potential to identify certain ADEs earlier. We constructed a network representation of drug-ADE associations for 809 drugs and 852 ADEs on the basis of a snapshot of a widely used drug safety database from 2005 and supplemented these data with additional pharmacological information. We trained a logistic regression model to predict unknown drug-ADE associations that were not listed in the 2005 snapshot. We evaluated the model's performance by comparing these predictions with the new drug-ADE associations that appeared in a 2010 snapshot of the same drug safety database. The proposed model achieved an AUROC (area under the receiver operating characteristic curve) statistic of 0.87, with a sensitivity of 0.42 given a specificity of 0.95. These findings suggest that predictive network methods can be useful for predicting unknown ADEs.

Linking electronic health record-extracted psychosocial data in real-time to risk of readmission for heart failure.

BACKGROUND: Knowledge of psychosocial characteristics that helps to identify patients at increased risk for readmission for heart failure (HF) may facilitate timely and targeted care. OBJECTIVE: We hypothesized that certain psychosocial characteristics extracted from the electronic health record (EHR) would be associated with an increased risk for hospital readmission within the next 30 days. METHODS: We identified 15 psychosocial predictors of readmission. Eleven of these were extracted from the EHR (six from structured data sources and five from unstructured clinical notes). We then analyzed their association with the likelihood of hospital readmission within the next 30 days among 729 patients admitted for HF. Finally, we developed a multivariable predictive model to recognize individuals at high risk for readmission. RESULTS: We found five characteristics-dementia, depression, adherence, declining/refusal of services, and missed clinical appointments-that were associated with an increased risk for hospital readmission: the first four features were captured from unstructured clinical notes, while the last item was captured from a structured data source. CONCLUSIONS: Unstructured clinical notes contain important knowledge on the relationship between psychosocial risk factors and an increased risk of readmission for HF that would otherwise have been missed if only structured data were considered. Gathering this EHR-based knowledge can be automated, thus enabling timely and targeted care.

Measuring the effect of commuting on the performance of the Bayesian Aerosol Release Detector.

BACKGROUND: Early detection of outdoor aerosol releases of anthrax is an important problem. The Bayesian Aerosol Release Detector (BARD) is a system for detecting releases of aerosolized anthrax and characterizing them in terms of location, time and quantity. Modelling a population's exposure to aerosolized anthrax poses a number of challenges. A major difficulty is to accurately estimate the exposure level--the number of inhaled anthrax spores--of each individual in the exposed region. Partly, this difficulty stems from the lack of fine-grained data about the population under surveillance. To cope with this challenge, nearly all anthrax biosurveillance systems, including BARD, ignore the mobility of the population and assume that exposure to anthrax would occur at one's home administrative unit--an assumption that limits the fidelity of the model. METHODS: We employed commuting data provided by the U.S. Census Bureau to parameterize a commuting model. Then, we developed methods for integrating commuting into BARD's simulation and detection algorithms and conducted two studies to measure the effect. The first study (simulation study) was designed to assess how BARD's detection and characterization performance are impacted by incorporation of commuting in BARD's outbreak-simulation algorithm. The second study (detection study) was designed to measure the effect of incorporating commuting in BARD's outbreak-detection algorithm. RESULTS: We found that failing to account for commuting in detection (when commuting is present in simulation) leads to a deterioration in BARD's detection and characterization performance that is both statistically and practically significant. We found that a simplified approach to accounting for commuting in detection--simplified to maintain tractability of inference--nearly fully restored both detection and characterization performance of BARD detector. CONCLUSION: We conclude that it is important to account for commuting (and mobility in general) in BARD's simulation algorithm. Further, the proposed method for incorporating commuting in BARD's detection algorithm can successfully perform the necessary correction in the detection algorithm, while preserving BARD's practicality. In our future work, we intend to further study the problem of the trade-off between running time and accuracy of the computation in BARD's version that includes commuting and ultimately find the best such trade-off.

Mining aggregates of over-the-counter products for syndromic surveillance.

There has recently been a surge of research efforts aimed at very early detection of disease outbreaks. An important strategy for improving the timeliness of outbreak detection is to identify signals that occur early in the epidemic process. We have developed a novel algorithm to identify aggregates of "similar" over-the-counter products that have strong association with a given disease. This paper discusses the proposed algorithm and reports the results of an evaluation experiment. The experimental results show that this algorithm holds promise for discovering product aggregates with outbreak detection performance that is superior to that of predefined categories. We also found that the products extracted by the proposed algorithm were more strongly correlated with the disease data than the standard predefined product categories, while also being more strongly correlated with each other than the products in any predefined category.