Decision support tools increasingly integrate clinical knowledge such as medication indications and contraindications with electronic health record (EHR) data to support clinical care and patient safety. The availability of this encoded information and patient data provides an opportunity to develop measures of clinical decision complexity that may be of value for quality improvement and research efforts. We investigated the feasibility of using encoded clinical knowledge and EHR data to develop a measure of comorbidity interrelatedness (the degree to which patients’ co-occurring conditions interact to generate clinical complexity). Using a common clinical scenario—decisions about blood pressure medications in patients with hypertension—we quantified comorbidity interrelatedness by calculating the number of indications and contraindications to blood pressure medications that are generated by patients’ comorbidities (e.g., diabetes, gout, depression). We examined properties of comorbidity interrelatedness using data from a decision support system for hypertension in the Veterans Affairs Health Care System.
Tele-education systems are increasingly being utilized in medical education worldwide. Due to limited human resources in healthcare in low and middle-income countries, developing online systems that are accessible to medical trainees in underserved areas potentially represents a highly efficient and effective method of improving the quantity and quality of the health care workforce. We developed, implemented, and evaluated an interactive web-based tele-education system (based on internationally accepted, image-based guidelines) for the diagnosis of retinopathy of prematurity among ophthalmologists-in-training in Brazil, Mexico, and the Philippines. We demonstrate that participation in this tele-education program improved diagnostic accuracy and reliability, and was preferred to standard pedagogical methods. This system may be employed not only in training, but also in international certification programs, and the process may be generalizable to other image-based specialties, such as dermatology and radiology.
Dengue is a viral disease of expanding global incidence without cures. Here we present a drug repositioning system (DenguePredict) leveraging upon a unique drug treatment database and vast amounts of disease- and drug-related data. We first constructed a large-scale genetic disease network with enriched dengue genetics data curated from biomedical literature. We applied a network-based ranking algorithm to find dengue-related diseases from the disease network. We then developed a novel algorithm to prioritize FDA-approved drugs from dengue-related diseases to treat dengue. When tested in a de-novo validation setting, DenguePredict found the only two drugs tested in clinical trials for treating dengue and ranked them highly: chloroquine ranked at top 0.96% and ivermectin at top 22.75%. We showed that drugs targeting immune systems and arachidonic acid metabolism-related apoptotic pathways might represent innovative drugs to treat dengue. In summary, DenguePredict, by combining comprehensive disease- and drug-related data and novel algorithms, may greatly facilitate drug discovery for dengue.
Using longitudinal data in electronic health records (EHRs) for post-marketing adverse drug event (ADE) detection allows for monitoring patients throughout their medical history. Machine learning methods have been shown to be efficient and effective in screening health records and detecting ADEs. How best to exploit historical data, as encoded by clinical events in EHRs is, however, not very well understood. In this study, three strategies for handling temporality of clinical events are proposed and evaluated using an EHR database from Stockholm, Sweden. The random forest learning algorithm is applied to predict fourteen ADEs using clinical events collected from different lengths of patient history. The results show that, in general, including longer patient history leads to improved predictive performance, and that assigning weights to events according to time distance from the ADE yields the biggest improvement.
Family history is considered a core element of clinical care. In this study we assessed the quality of family history data captured in an established commercial electronic health record (EHR) at a large academic medical center. Because the EHR had no centralized location to store family history information, it was collected as part of clinical notes in structured or free-text format. We analyzed differences between 10,000 free-text and 9,121 structured family history observations. Each observation was classified according to disease presence/absence and family member affected (e.g., father, mother, etc.). The structured notes did not collect a complete family history as defined by standards endorsed by the U.S. Agency for Healthcare Research and Quality; the free-text notes contained more information than the structured notes, but still not enough to be considered “complete.” Several barriers remain for collecting complete, useful family history data in electronic health records.
The Generalizability Index for Study Traits (GIST) has been proposed recently for assessing the population representativeness of a set of related clinical trials using eligibility features (e.g., age or BMI), one each time. However, GIST has not yet been evaluated. To bridge this knowledge gap, this paper reports a simulation-based validation study for GIST. Using the National Health and Nutrition Examination Survey (NHANES) data, we demonstrated the effectiveness of GIST at quantifying the population representativeness of a set of related trials that differ in disease domains, study phases, sponsor types, and study designs, respectively. We also showed that among seven example medical conditions, the GIST of age increases from Phase I trials to Phase III trials in the seven disease domains and is the lowest in asthma trials. We concluded that GIST correlates with simulation-based generalizability results and is a valid metric for quantifying population representativeness of related clinical trials.
Patient-reported functional status is widely recognized as an important patient-centered outcome that adds value to medical care, research, and quality improvement. Functional status outcomes are, however, not routinely or uniformly collected in the medical record, except in certain small patient populations (e.g. geriatrics, nursing home residents). To utilize patient reported functional status for clinical research and practice, we manually collected 2,763 terms from clinical records and social media sites and modeled them on the widely used Short Form-36 Health Survey. We then examined the coverage of the Unified Medical Language System (UMLS) for these functional status terms through automated mapping. Most terms (85.9%) did not have exact matches in the UMLS. The partial matches were prevalent, however, they typically did not capture the terms’ exact semantics. Our study suggests that there is a need to extend existing standard terminologies to incorporate functional status terms used by patients and clinicians.
Over the last few decades, machine learning and data mining have been increasingly used for clinical prediction in ICUs. However, there is still a huge gap in making full use of the time-series data generated from ICUs. Aiming at filling this gap, we propose a novel approach entitled Time Slicing Cox regression (TS-Cox), which extends the classical Cox regression into a classification method on multi-dimensional time-series. Unlike traditional classifiers such as logistic regression and support vector machines, our model not only incorporates the discriminative features derived from the time-series, but also naturally exploits the temporal orders of these features based on a Cox-like function. Empirical evaluation on MIMIC-II database demonstrates the efficacy of the TS-Cox model. Our TS-Cox model outperforms all other baseline models by a good margin in terms of AUC_PR, sensitivity and PPV, which indicates that TS-Cox may be a promising tool for mortality prediction in ICUs.
Diabetes management is a complex, dynamic process that is largely incumbent on patient choices and behavior. We explore how health-management needs—and the needs for technological support—change over time for individuals with diabetes. Through interviews and a focus group, we found that after initial diagnosis, individuals face acute information needs and chiefly turn to mobile applications and Internet resources to help understand the diabetes-specific factors that affect their health. Over time their focus shifts from highly regimented routines to more flexible ones that enable them to maintain a quality of life. Our results suggest that long-term engagement with health technology does not necessarily require continuous, sustained use: routine disease management could lead to a decrease in use, until a new event occurs. Our findings point to a need for tools that help patients with diabetes to effectively manage their health as their bodies, treatment and circumstances change over time.
Nutrition care and metabolic control contribute to clinical patient outcomes. Biomedical informatics applications represent a way to potentially improve quality and efficiency of nutrition management. We performed a systematic literature review to identify clinical decision support and computerized provider order entry systems used to manage nutrition care. Online research databases were searched using a specific set of keywords. Additionally, bibliographies were referenced for supplemental citations. Four independent reviewers selected sixteen studies out of 364 for review. These papers described adult and neonatal nutrition support applications, blood glucose management applications, and other nutrition applications. Overall, results indicated that computerized interventions could contribute to improved patient outcomes and provider performance. Specifically, computer systems in the clinical setting improved nutrient delivery, rates of malnutrition, weight loss, blood glucose values, clinician efficiency, and error rates. In conclusion, further investigation of informatics applications on nutritional and performance outcomes utilizing rigorous study designs is recommended.
Postmarketing drug surveillance is critical to assessing adverse events associated with medications, because prelaunch clinical trials frequently miss negative drug effects. The Informatics for Integrating Biology and the Bedside platform (i2b2) has been used effectively for this. However, previous work suffers from incomplete medical data present in electronic health record (EHR) systems. Here, we develop a system to integrate non-traditional data sources with EHR data: pharmacy dispensing information and patient-reported data. We implement and validate a toolset to gather medication data from a Pharmacy Benefit Manager network, import it into an i2b2 EHR repository using a standard data format, merge it with the EHR data, and present it to for annotation with results returned to i2b2. This toolkit is enabling studies on medication list data quality, adherence, and adverse event detection.
Accurate temporal identification and normalization is imperative for many biomedical and clinical tasks such as generating timelines and identifying phenotypes. A major natural language processing challenge is developing and evaluating a generalizable temporal modeling approach that performs well across corpora and institutions. Our long-term goal is to create such a model. We initiate our work on reaching this goal by focusing on temporal expression (TIMEX3) identification. We present a systematic approach to 1) generalize existing solutions for automated TIMEX3 span detection, and 2) assess similarities and differences by various instantiations of TIMEX3 models applied on separate clinical corpora. When evaluated on the 2012 i2b2 and the 2015 Clinical TempEval challenge corpora, our conclusion is that our approach is successful – we achieve competitive results for automated classification, and we identify similarities and differences in TIMEX3 modeling that will be informative in the development of a simplified, general temporal model.
Orthographic and grammatical errors are a common feature of informal texts written by lay people. Health-related questions asked by consumers are a case in point. Automatic interpretation of consumer health questions is hampered by such errors. In this paper, we propose a method that combines techniques based on edit distance and frequency counts with a contextual similarity-based method for detecting and correcting orthographic errors, including misspellings, word breaks, and punctuation errors. We evaluate our method on a set of spell-corrected questions extracted from the NLM collection of consumer health questions. Our method achieves a F1 score of 0.61, compared to an informed baseline of 0.29, achieved using ESpell, a spelling correction system developed for biomedical queries. Our results show that orthographic similarity is most relevant in spelling error correction in consumer health questions and that frequency and contextual information are complementary to orthographic features.
With improvements in early detection and treatment, the number of cancer survivors has been on the rise. Studies suggest that cancer survivors do not often receive proper follow-up care despite existing guidelines. Patient engagement is key to healthy survivorship, and mHealth provides a viable platform to empower survivors with just- in-time personalized support. However, our understanding of existing mHealth solutions in cancer survivorship is limited. In this paper, we use Patient Engagement Framework to investigate existing apps to bridge this knowledge gap. App features are mapped to the framework components to determine the level of engagement facilitated. Ability to record treatment summaries has been found in five out of seven apps examined. While collaborative care and social engagement are found minimally, the majority of features (95%) are limited to information and way finding, e-tools, and interactive forms. Limitations of the existing apps and possible improvements to the framework are discussed.
Asthma is the most common chronic childhood disease and has seen increasing prevalence worldwide. While there is existing evidence of familial and other risk factors for pediatric asthma, there is a need for further studies to explore and understand interactions among these risk factors. The goal of this study was to develop an approach for mining, visualizing, and evaluating association rules representing pairwise interactions among potential familial risk factors based on information documented as part of a patient’s family history in the electronic health record. As a case study, 10,260 structured family history entries for a cohort of 1,531 pediatric asthma patients were extracted and analyzed to generate family history associations at different levels of granularity. The preliminary results highlight the potential of this approach for validating known knowledge and suggesting opportunities for further investigation that may contribute to improving prediction of asthma risk in children.
Pregnancy is a time when expectant mothers may have numerous questions about their unborn children, especially when congenital anomalies are diagnosed prenatally. We sought to characterize information needs of pregnant women seen in the Vanderbilt Children’s Hospital Fetal Center. Participants recorded questions from diagnosis through delivery. Questions were categorized by two researchers using a hierarchical taxonomy describing consumer health information needs. Consensus category assignments were made, and inter-rater reliability was measured with Cohen’s Kappa. Sixteen participants reported 398 questions in 39 subcategories, of which the most common topics were prognosis (53 questions; 13.3%) and indications for intervention (31 questions; 7.8%). Inter-rater reliability of assignments showed moderate (κ=0.57) to substantial (κ=0.75) agreement for subcategories and primary categories, respectively. Pregnant women with prenatal diagnoses have diverse unmet information needs; a taxonomy of consumer health information needs may improve the ability to meet such needs through content and system design.
We examined a large body of research study documents (protocols) to identify mentions of drug concepts and established base concepts and roles needed to characterize the semantics of these instances. We found these concepts in three general situations: background knowledge about the drug, study procedures involving the drug, and other roles of the drug in the study. We identified 18 more specific contexts (e.g., adverse event information, administration and dosing of the drug, and interactions between the study drug and other drugs). The ontology was validated against a test set of protocol documents from NIH and ClinicalTrial.gov. The goal is to support the automated extraction of drug information from protocol documents to support functions such as study retrieval, determination of subject eligibility, generation of order sets, and creation of logic for decision support alerts and reminders. Further work is needed to formally extend existing ontologies of clinical research.
Primary care practices have been limited in their ability to leverage electronic health records (EHRs) and health information exchange (HIE) to improve care coordination, but will soon be incentivized to do so under proposed Stage 3 meaningful use criteria. We use mixed methods to understand how primary care practices manage, share and reconcile electronic patient information across care settings, and identify innovations in EHR design to support enhanced care coordination. Opportunities identified by practices focused on availability and usability of features that facilitate (1) generation of customized summary of care records, (2) team-based care approaches, and (3) management of the increased volume of electronic information generated and exchanged during care transitions. More broadly, vendors and policymakers need to continue to work together to improve interoperability as the key to effective care coordination. If these EHR innovations were widespread, the value of meeting the proposed Stage 3 care coordination criteria would be substantially enhanced.
Given growing interest in mobile phones for health management (mHealth), we surveyed consumer perceptions of mHealth in security, privacy, and healthcare quality using national random-digit-dial telephone surveys in 2013 and 2014. In 2013, 48% thought that using a mobile phone to communicate data with a physician’s electronic health record (EHR) would improve the quality of health care. By 2014, the proportion rose to 57% (p < .001). There were no similar changes in privacy concerns yet nearly two-thirds expressed privacy concerns. In 2013 alone, respondents were more likely to express privacy concerns about medical data on mobile phones than they were to endorse similar concerns with EHRs or health information exchange (HIE). Consumers increasingly believe that mHealth improves healthcare quality, but security and privacy concerns need to be addressed for quality improvement to be fully realized.
FDA-approved prescribing information (also known as product labeling or labels) contain critical safety information for health care professionals. Drug labels have often been criticized, however, for being overly complex, difficult to read, and rife with overwarning, leading to high cognitive load. In this project, we aimed to improve the usability of drug labels by increasing the ‘signal-to-noise ratio’ and providing meaningful information to care providers based on patient-specific comorbidities and concomitant medications. In the current paper, we describe the design process and resulting web application, known as myDrugLabel. Using the Structured Product Label documents as a base, we describe the process of label personalization, readability improvements, and integration of diverse evidence sources, including the medical literature from PubMed, pharmacovigilance reports from FDA adverse event reporting system (FAERS), and social media signals directly into the label.
User frustration research has been one way of looking into clinicians’ experience with health information technology use and interaction. In order to understand how clinician frustration with Health Information Technology (HIT) use occurs, there is the need to explore Human-Computer Interaction (HCI) literature that addresses both frustration and HIT use. In the past three decades, HCI frustration research has increased and expanded. Researchers have done a lot of work to understand emotions, end-user frustration and affect. This paper uses a historical literature review approach to review the origins of emotion and frustration research and explore the research question; Does HCI research on frustration provide insights on clinicians’ frustration with HIT interfaces? From the literature review HCI research on emotion and frustration provides additional insights that can indeed help explain user frustration in HIT. Different approaches and HCI perspectives also help frame HIT user frustration research as well as inform HIT system design. The paper concludes with a suggested directions on how future design and research may take.
Frustration; Emotional design; Affective Computing; Health Information Technology (HIT); Human-Computer Interaction (HCI); Interface design; Clinicians; User Experience (UX); H5.2. Information interfaces and presentation: User Interfaces
Detection of early symptoms in cervical cancer is crucial for early treatment and survival. To find symptoms of cervical cancer in clinical text, Named Entity Recognition is needed. In this paper the Clinical Entity Finder, a machine-learning tool trained on annotated clinical text from a Swedish internal medicine emergency unit, is evaluated on cervical cancer records. The Clinical Entity Finder identifies entities of the types body part, finding and disorder and is extended with negation detection using the rule-based tool NegEx, to distinguish between negated and non-negated entities. To measure the performance of the tools on this new domain, two physicians annotated a set of clinical notes from the health records of cervical cancer patients. The inter-annotator agreement for finding, disorder and body part obtained an average F-score of 0.677 and the Clinical Entity Finder extended with NegEx had an average F-score of 0.667.
The annual influenza vaccine is one of the most common public health interventions and is universally recommended for all individuals older than six months. Vaccine composition depends on viruses circulating over the past flu season and are estimated to be the most prevalent and representative strains in the current season. Here, we use clinical data outfitted with viral genetics to characterize confirmed influenza cases from the past two flu seasons and genetically compare them to the strains that they were vaccinated against that year. We show that case similarities to vaccine strains differ by geographic region and that the vaccines appear to have different levels of effectiveness by region. This study demonstrates the value of merging viral genetics with clinical data. Further research is needed to formally evaluate whether this improves biosurveillance efforts and enhances efficacy of influenza vaccines.
Despite their widespread use, audit and feedback (A&F) interventions show variable effectiveness on improving professional performance. Based on known facilitators of successful A&F interventions, we developed a web-based A&F intervention with indicator-based performance feedback, benchmark information, action planning and outreach visits. The goal of the intervention was to engage with multidisciplinary teams to overcome barriers to guideline concordance and to improve overall team performance in the field of cardiac rehabilitation (CR). To assess its effectiveness we conducted a cluster-randomized trial in 18 CR clinics (14,847 patients) already working with computerized decision support (CDS). Our preliminary results showed no increase in concordance with guideline recommendations regarding prescription of CR therapies. Future analyses will investigate whether our intervention did improve team performance on other quality indicators.