JMIR Medical Informatics

Clinical informatics, decision support for health professionals, electronic health records, and ehealth infrastructures.

Putting Meaning into Meaningful Use: A Roadmap to successful Integration of Evidence at the Point of Care

Pressures to contain healthcare costs, personalize patient care, use of big data, and enhance healthcare quality have highlighted the need for integration of evidence at the point of care.(1-7) But the application of EBM has great promise in the era of electronic health records (EHRs) and health technology. Most successful integration of evidence into EHRs has been complex decision tools that trigger at a critical point of the clinical visit and include patient specific recommendations. Incorporation of complex CDS tools into the EMR is equally complex and continues to challenge health service researchers and implementation scientists. Poor adoption and sustainably of EBM guidelines and CDS tools at the point of care has persisted and continue to document low rates of usage. The barriers cited by physicians include efficiency, perception of usefulness, information content, user interface, and over-triggering. Building on the traditional EHR implementation frameworks we review keys strategies for successful CDSs: 1) the quality of the evidence, 2) the potential to reduce unnecessary care, 3) ease of integrating evidence at the point of care, 4) the evidence’s consistency with clinician perceptions and preferences, 5) incorporating bundled sets or automated documentation and 6) shared decision making tools.

Journal Description

JMIR Medical Informatics (JMI, ISSN 2291-9694) focusses on clinical informatics, big data in health and health care, decision support for health professionals, electronic health records, ehealth infrastructures and implementation. It has a focus on applied, translational research, with a broad readership including clinicians, CIOs, engineers, industry and health informatics professionals.

Published by JMIR Publications, publisher of the Journal of Medical Internet Research (JMIR), the leading eHealth/mHealth journal (Impact Factor 2013: 4.7), JMIR Med Inform has a different scope (emphasizing more on applications for clinicians and health professionals rather than consumers/citizens, which is the focus of JMIR), publishes even faster, and also allows papers which are more technical or more formative than what would be published in the Journal of Medical Internet Research.

JMIR Medical Informatics journal features a rapid and thorough peer-review process, professional copyediting, professional production of PDF, XHTML, and XML proofs (ready for deposit in PubMed Central/PubMed). The site is optimized for mobile and iPad use.

JMIR Medical Informatics adheres to the same quality standards as JMIR and all articles published here are also cross-listed in the Table of Contents of JMIR, the worlds' leading medical journal in health sciences / health services research and health informatics (


Recent Articles:

  • An image from authors.

    Comprehensive Evaluation of Electronic Medical Record System Use and User Satisfaction at Five Low-Resource Setting Hospitals in Ethiopia


    Background: Electronic medical record (EMR) systems are increasingly being implemented in hospitals of developing countries to improve patient care and clinical service. However, only limited evaluation studies are available concerning the level of adoption and determinant factors of success in those settings. Objective: The objective of this study was to assess the usage pattern, user satisfaction level, and determinants of health professional’s satisfaction towards a comprehensive EMR system implemented in Ethiopia where parallel documentation using the EMR and the paper-based medical records is in practice. Methods: A quantitative, cross-sectional study design was used to assess the usage pattern, user satisfaction level, and determinant factors of an EMR system implemented in Ethiopia based on the DeLone and McLean model of information system success. Descriptive statistical methods were applied to analyze the data and a binary logistic regression model was used to identify determinant factors. Results: Health professionals (N=422) from five hospitals were approached and 406 responded to the survey (96.2% response rate). Out of the respondents, 76.1% (309/406) started to use the system immediately after implementation and user training, but only 31.7% (98/309) of the professionals reported using the EMR during the study (after 3 years of implementation). Of the 12 core EMR functions, 3 were never used by most respondents, and they were also unaware of 4 of the core EMR functions. It was found that 61.4% (190/309) of the health professionals reported over all dissatisfaction with the EMR (median=4, interquartile range (IQR)=1) on a 5-level Likert scale. Physicians were more dissatisfied (median=5, IQR=1) when compared to nurses (median=4, IQR=1) and the health management information system (HMIS) staff (median=2, IQR=1). Of all the participants, 64.4% (199/309) believed that the EMR had no positive impact on the quality of care. The participants indicated an agreement with the system and information quality (median=2, IQR=0.5) but strongly disagreed with the service quality (median=5, IQR=1). The logistic regression showed a strong correlation between system use and dissatisfaction (OR 7.99, 95% CI 5.62-9.10) and service quality and satisfaction (OR 8.23, 95% CI 3.23-17.01). Conclusions: Health professionals’ use of the EMR is low and they are generally dissatisfied with the service of the implemented system. The results of this study show that this dissatisfaction is caused mainly and strongly by the poor service quality, the current practice of double documentation (EMR and paper-based), and partial departmental use of the system in the hospitals. Thus, future interventions to improve the current use or future deployment projects should focus on improving the service quality such as power infrastructure, user support, trainings, and more computers in the wards. After service quality improvement, other departments (especially inter-dependent departments) should be motivated and supported to use the EMR to avoid the dependency deadlock.

  • A screenshot of ECG diagnosis using the telesurveillance system. The ECG waveform and the corresponding classification suggestions are revealed on the screen. The suggested heartbeat classification is marked with a blue dot. Health professionals can make decisions using this information in clinical practice.

    A Telesurveillance System With Automatic Electrocardiogram Interpretation Based on Support Vector Machine and Rule-Based Processing


    Background: Telehealth care is a global trend affecting clinical practice around the world. To mitigate the workload of health professionals and provide ubiquitous health care, a comprehensive surveillance system with value-added services based on information technologies must be established. Objective: We conducted this study to describe our proposed telesurveillance system designed for monitoring and classifying electrocardiogram (ECG) signals and to evaluate the performance of ECG classification. Methods: We established a telesurveillance system with an automatic ECG interpretation mechanism. The system included: (1) automatic ECG signal transmission via telecommunication, (2) ECG signal processing, including noise elimination, peak estimation, and feature extraction, (3) automatic ECG interpretation based on the support vector machine (SVM) classifier and rule-based processing, and (4) display of ECG signals and their analyzed results. We analyzed 213,420 ECG signals that were diagnosed by cardiologists as the gold standard to verify the classification performance. Results: In the clinical ECG database from the Telehealth Center of the National Taiwan University Hospital (NTUH), the experimental results showed that the ECG classifier yielded a specificity value of 96.66% for normal rhythm detection, a sensitivity value of 98.50% for disease recognition, and an accuracy value of 81.17% for noise detection. For the detection performance of specific diseases, the recognition model mainly generated sensitivity values of 92.70% for atrial fibrillation, 89.10% for pacemaker rhythm, 88.60% for atrial premature contraction, 72.98% for T-wave inversion, 62.21% for atrial flutter, and 62.57% for first-degree atrioventricular block. Conclusions: Through connected telehealth care devices, the telesurveillance system, and the automatic ECG interpretation system, this mechanism was intentionally designed for continuous decision-making support and is reliable enough to reduce the need for face-to-face diagnosis. With this value-added service, the system could widely assist physicians and other health professionals with decision making in clinical practice. The system will be very helpful for the patient who suffers from cardiac disease, but for whom it is inconvenient to go to the hospital very often.

  • Single-word text cloud created in TagCrowd from all free text comments.

    Web-Based Textual Analysis of Free-Text Patient Experience Comments From a Survey in Primary Care


    Background: Open-ended questions eliciting free-text comments have been widely adopted in surveys of patient experience. Analysis of free text comments can provide deeper or new insight, identify areas for action, and initiate further investigation. Also, they may be a promising way to progress from documentation of patient experience to achieving quality improvement. The usual methods of analyzing free-text comments are known to be time and resource intensive. To efficiently deal with a large amount of free-text, new methods of rapidly summarizing and characterizing the text are being explored. Objective: The aim of this study was to investigate the feasibility of using freely available Web-based text processing tools (text clouds, distinctive word extraction, key words in context) for extracting useful information from large amounts of free-text commentary about patient experience, as an alternative to more resource intensive analytic methods. Methods: We collected free-text responses to a broad, open-ended question on patients’ experience of primary care in a cross-sectional postal survey of patients recently consulting doctors in 25 English general practices. We encoded the responses to text files which were then uploaded to three Web-based textual processing tools. The tools we used were two text cloud creators: TagCrowd for unigrams, and Many Eyes for bigrams; and Voyant Tools, a Web-based reading tool that can extract distinctive words and perform Keyword in Context (KWIC) analysis. The association of patients’ experience scores with the occurrence of certain words was tested with logistic regression analysis. KWIC analysis was also performed to gain insight into the use of a significant word. Results: In total, 3426 free-text responses were received from 7721 patients (comment rate: 44.4%). The five most frequent words in the patients’ comments were “doctor”, “appointment”, “surgery”, “practice”, and “time”. The three most frequent two-word combinations were “reception staff”, “excellent service”, and “two weeks”. The regression analysis showed that the occurrence of the word “excellent” in the comments was significantly associated with a better patient experience (OR=1.96, 95%CI=1.63-2.34), while “rude” was significantly associated with a worse experience (OR=0.53, 95%CI=0.46-0.60). The KWIC results revealed that 49 of the 78 (63%) occurrences of the word “rude” in the comments were related to receptionists and 17(22%) were related to doctors. Conclusions: Web-based text processing tools can extract useful information from free-text comments and the output may serve as a springboard for further investigation. Text clouds, distinctive words extraction and KWIC analysis show promise in quick evaluation of unstructured patient feedback. The results are easily understandable, but may require further probing such as KWIC analysis to establish the context. Future research should explore whether more sophisticated methods of textual analysis (eg, sentiment analysis, natural language processing) could add additional levels of understanding.

  • A processing pipeline that transforms verbal clinical handover information into electronic structured records automatically. H Suominen, 2014, Creative Commons – Attribution Alone.

    Benchmarking Clinical Speech Recognition and Information Extraction: New Data, Methods, and Evaluations


    Background: Over a tenth of preventable adverse events in health care are caused by failures in information flow. These failures are tangible in clinical handover; regardless of good verbal handover, from two-thirds to all of this information is lost after 3-5 shifts if notes are taken by hand, or not at all. Speech recognition and information extraction provide a way to fill out a handover form for clinical proofing and sign-off. Objective: The objective of the study was to provide a recorded spoken handover, annotated verbatim transcriptions, and evaluations to support research in spoken and written natural language processing for filling out a clinical handover form. This dataset is based on synthetic patient profiles, thereby avoiding ethical and legal restrictions, while maintaining efficacy for research in speech-to-text conversion and information extraction, based on realistic clinical scenarios. We also introduce a Web app to demonstrate the system design and workflow. Methods: We experiment with Dragon Medical 11.0 for speech recognition and CRF++ for information extraction. To compute features for information extraction, we also apply CoreNLP, MetaMap, and Ontoserver. Our evaluation uses cross-validation techniques to measure processing correctness. Results: The data provided were a simulation of nursing handover, as recorded using a mobile device, built from simulated patient records and handover scripts, spoken by an Australian registered nurse. Speech recognition recognized 5276 of 7277 words in our 100 test documents correctly. We considered 50 mutually exclusive categories in information extraction and achieved the F1 (ie, the harmonic mean of Precision and Recall) of 0.86 in the category for irrelevant text and the macro-averaged F1 of 0.70 over the remaining 35 nonempty categories of the form in our 101 test documents. Conclusions: The significance of this study hinges on opening our data, together with the related performance benchmarks and some processing software, to the research and development community for studying clinical documentation and language-processing. The data are used in the CLEFeHealth 2015 evaluation laboratory for a shared task on speech recognition.

  • Image of pregnant woman, drawn by friend.  Gifted to our project for use, no attribution required.

    Balancing the Interests of Patient Data Protection and Medication Safety Monitoring in a Public-Private Partnership


    Obtaining data without the intervention of a health care provider represents an opportunity to expand understanding of the safety of medications used in difficult-to-study situations, like the first trimester of pregnancy when women may not present for medical care. While it is widely agreed that personal data, and in particular medical data, needs to be protected from unauthorized use, data protection requirements for population-based studies vary substantially by country. For public-private partnerships, the complexities are enhanced. The objective of this viewpoint paper is to illustrate the challenges related to data protection based on our experiences when performing relatively straightforward direct-to-patient noninterventional research via the Internet or telephone in four European countries. Pregnant women were invited to participate via the Internet or using an automated telephone response system in Denmark, the Netherlands, Poland, and the United Kingdom. Information was sought on medications, other factors that may cause birth defects, and pregnancy outcome. Issues relating to legal controllership of data were most problematic; assuring compliance with data protection requirements took about two years. There were also inconsistencies in the willingness to accept nonwritten informed consent. Nonetheless, enrollment and data collection have been completed, and analysis is in progress. Using direct reporting from consumers to study the safety of medicinal products allows researchers to address a myriad of research questions relating to everyday clinical practice, including treatment heterogeneity in population subgroups not traditionally included in clinical trials, like pregnant women, children, and the elderly. Nonetheless, there are a variety of administrative barriers relating to data protection and informed consent, particularly within the structure of a public-private partnership.

  • This image was created by the authors using deidentified patient data. It represents a classifier picking out text.

    Prioritization of Free-Text Clinical Documents: A Novel Use of a Bayesian Classifier


    Background: The amount of incoming data into physicians’ offices is increasing, thereby making it difficult to process information efficiently and accurately to maximize positive patient outcomes. Current manual processes of screening for individual terms within long free-text documents are tedious and error-prone. This paper explores the use of statistical methods and computer systems to assist clinical data management. Objective: The objective of this study was to verify and validate the use of a naive Bayesian classifier as a means of properly prioritizing important clinical data, specifically that of free-text radiology reports. Methods: There were one hundred reports that were first used to train the algorithm based on physicians’ categorization of clinical reports as high-priority or low-priority. Then, the algorithm was used to evaluate 354 reports. Additional beautification procedures such as section extraction, text preprocessing, and negation detection were performed. Results: The algorithm evaluated the 354 reports with discrimination between high-priority and low-priority reports, resulting in a bimodal probability distribution. In all scenarios tested, the false negative rates were below 1.1% and the recall rates ranged from 95.65% to 98.91%. In the case of 50% prior probability and 80% threshold probability, the accuracy of this Bayesian classifier was 93.50%, with a positive predictive value (precision) of 80.54%. It also showed a sensitivity (recall) of 98.91% and a F-measure of 88.78%. Conclusions: The results showed that the algorithm could be trained to detect abnormal radiology results by accurately screening clinical reports. Such a technique can potentially be used to enable automatic flagging of critical results. In addition to accuracy, the algorithm was able to minimize false negatives, which is important for clinical applications. We conclude that a Bayesian statistical classifier, by flagging reports with abnormal findings, can assist a physician in reviewing radiology reports more efficiently. This higher level of prioritization allows physicians to address important radiologic findings in a timelier manner and may also aid in minimizing errors of omission.

  • HINTS Logo (National Cancer Institute).

    The Role of Health Care Experience and Consumer Information Efficacy in Shaping Privacy and Security Perceptions of Medical Records: National Consumer Survey...


    Background: Providers’ adoption of electronic health records (EHRs) is increasing and consumers have expressed concerns about the potential effects of EHRs on privacy and security. Yet, we lack a comprehensive understanding regarding factors that affect individuals’ perceptions regarding the privacy and security of their medical information. Objective: The aim of this study was to describe national perceptions regarding the privacy and security of medical records and identify a comprehensive set of factors associated with these perceptions. Methods: Using a nationally representative 2011-2012 survey, we reported on adults’ perceptions regarding privacy and security of medical records and sharing of health information between providers, and whether adults withheld information from a health care provider due to privacy or security concerns. We used multivariable models to examine the association between these outcomes and sociodemographic characteristics, health and health care experience, information efficacy, and technology-related variables. Results: Approximately one-quarter of American adults (weighted n=235,217,323; unweighted n=3959) indicated they were very confident (n=989) and approximately half indicated they were somewhat confident (n=1597) in the privacy of their medical records; we found similar results regarding adults’ confidence in the security of medical records (very confident: n=828; somewhat confident: n=1742). In all, 12.33% (520/3904) withheld information from a health care provider and 59.06% (2100/3459) expressed concerns about the security of both faxed and electronic health information. Adjusting for other characteristics, adults who reported higher quality of care had significantly greater confidence in the privacy and security of their medical records and were less likely to withhold information from their health care provider due to privacy or security concerns. Adults with higher information efficacy had significantly greater confidence in the privacy and security of medical records and less concern about sharing of health information by both fax and electronic means. Individuals’ perceptions of whether their providers use an EHR was not associated with any privacy or security outcomes. Conclusions: Although most adults are confident in the privacy and security of their medical records, many express concerns regarding sharing of information between providers; a minority report withholding information from their providers due to privacy and security concerns. Whether individuals thought their provider was using an EHR was not associated with negative privacy/security perceptions or withholding, suggesting the transition to EHRs is not associated with negative perceptions regarding the privacy and security of medical information. However, monitoring to see how this evolves will be important. Given that positive health care experiences and higher information efficacy were associated with more favorable perceptions of privacy and security, efforts should continue to encourage providers to secure medical records, provide patients with a “meaningful choice” in how their data are shared, and enable individuals to access information they need to manage their care.

  • Implementation details of the CIMIDx framework.

    CIMIDx: Prototype for a Cloud-Based System to Support Intelligent Medical Image Diagnosis With Efficiency


    Background: The Internet has greatly enhanced health care, helping patients stay up-to-date on medical issues and general knowledge. Many cancer patients use the Internet for cancer diagnosis and related information. Recently, cloud computing has emerged as a new way of delivering health services but currently, there is no generic and fully automated cloud-based self-management intervention for breast cancer patients, as practical guidelines are lacking. Objective: We investigated the prevalence and predictors of cloud use for medical diagnosis among women with breast cancer to gain insight into meaningful usage parameters to evaluate the use of generic, fully automated cloud-based self-intervention, by assessing how breast cancer survivors use a generic self-management model. The goal of this study was implemented and evaluated with a new prototype called “CIMIDx”, based on representative association rules that support the diagnosis of medical images (mammograms). Methods: The proposed Cloud-Based System Support Intelligent Medical Image Diagnosis (CIMIDx) prototype includes two modules. The first is the design and development of the CIMIDx training and test cloud services. Deployed in the cloud, the prototype can be used for diagnosis and screening mammography by assessing the cancers detected, tumor sizes, histology, and stage of classification accuracy. To analyze the prototype’s classification accuracy, we conducted an experiment with data provided by clients. Second, by monitoring cloud server requests, the CIMIDx usage statistics were recorded for the cloud-based self-intervention groups. We conducted an evaluation of the CIMIDx cloud service usage, in which browsing functionalities were evaluated from the end-user’s perspective. Results: We performed several experiments to validate the CIMIDx prototype for breast health issues. The first set of experiments evaluated the diagnostic performance of the CIMIDx framework. We collected medical information from 150 breast cancer survivors from hospitals and health centers. The CIMIDx prototype achieved high sensitivity of up to 99.29%, and accuracy of up to 98%. The second set of experiments evaluated CIMIDx use for breast health issues, using t tests and Pearson chi-square tests to assess differences, and binary logistic regression to estimate the odds ratio (OR) for the predictors’ use of CIMIDx. For the prototype usage statistics for the same 150 breast cancer survivors, we interviewed 114 (76.0%), through self-report questionnaires from CIMIDx blogs. The frequency of log-ins/person ranged from 0 to 30, total duration/person from 0 to 1500 minutes (25 hours). The 114 participants continued logging in to all phases, resulting in an intervention adherence rate of 44.3% (95% CI 33.2-55.9). The overall performance of the prototype for the good category, reported usefulness of the prototype (P=.77), overall satisfaction of the prototype (P=.31), ease of navigation (P=.89), user friendliness evaluation (P=.31), and overall satisfaction (P=.31). Positive evaluations given by 100 participants via a Web-based questionnaire supported our hypothesis. Conclusions: The present study shows that women felt favorably about the use of a generic fully automated cloud-based self- management prototype. The study also demonstrated that the CIMIDx prototype resulted in the detection of more cancers in screening and diagnosing patients, with an increased accuracy rate.

  • This is a royalty free image (
License: CC0 Public Domain.

    Effects of Individual Health Topic Familiarity on Activity Patterns During Health Information Searches


    Background: Non-medical professionals (consumers) are increasingly using the Internet to support their health information needs. However, the cognitive effort required to perform health information searches is affected by the consumer’s familiarity with health topics. Consumers may have different levels of familiarity with individual health topics. This variation in familiarity may cause misunderstandings because the information presented by search engines may not be understood correctly by the consumers. Objective: As a first step toward the improvement of the health information search process, we aimed to examine the effects of health topic familiarity on health information search behaviors by identifying the common search activity patterns exhibited by groups of consumers with different levels of familiarity. Methods: Each participant completed a health terminology familiarity questionnaire and health information search tasks. The responses to the familiarity questionnaire were used to grade the familiarity of participants with predefined health topics. The search task data were transcribed into a sequence of search activities using a coding scheme. A computational model was constructed from the sequence data using a Markov chain model to identify the common search patterns in each familiarity group. Results: Forty participants were classified into L1 (not familiar), L2 (somewhat familiar), and L3 (familiar) groups based on their questionnaire responses. They had different levels of familiarity with four health topics. The video data obtained from all of the participants were transcribed into 4595 search activities (mean 28.7, SD 23.27 per session). The most frequent search activities and transitions in all the familiarity groups were related to evaluations of the relevancy of selected web pages in the retrieval results. However, the next most frequent transitions differed in each group and a chi-squared test confirmed this finding (P<.001). Next, according to the results of a perplexity evaluation, the health information search patterns were best represented as a 5-gram sequence pattern. The most common patterns in group L1 were frequent query modifications, with relatively low search efficiency, and accessing and evaluating selected results from a health website. Group L2 performed frequent query modifications, but with better search efficiency, and accessed and evaluated selected results from a health website. Finally, the members of group L3 successfully discovered relevant results from the first query submission, performed verification by accessing several health websites after they discovered relevant results, and directly accessed consumer health information websites. Conclusions: Familiarity with health topics affects health information search behaviors. Our analysis of state transitions in search activities detected unique behaviors and common search activity patterns in each familiarity group during health information searches.

  • This is a royalty free image by hyena reality (

    Real-Time Patient Survey Data During Routine Clinical Activities for Rapid-Cycle Quality Improvement


    Background: Surveying patients is increasingly important for evaluating and improving health care delivery, but practical survey strategies during routine care activities have not been available. Objective: We examined the feasibility of conducting routine patient surveys in a primary care clinic using commercially available technology (Web-based survey creation, deployment on tablet computers, cloud-based management of survey data) to expedite and enhance several steps in data collection and management for rapid quality improvement cycles. Methods: We used a Web-based data management tool (survey creation, deployment on tablet computers, real-time data accumulation and display of survey results) to conduct four patient surveys during routine clinic sessions over a one-month period. Each survey consisted of three questions and focused on a specific patient care domain (dental care, waiting room experience, care access/continuity, Internet connectivity). Results: Of the 727 available patients during clinic survey days, 316 patients (43.4%) attempted the survey, and 293 (40.3%) completed the survey. For the four 3-question surveys, the average time per survey was overall 40.4 seconds, with a range of 5.4 to 20.3 seconds for individual questions. Yes/No questions took less time than multiple choice questions (average 9.6 seconds versus 14.0). Average response time showed no clear pattern by order of questions or by proctor strategy, but monotonically increased with number of words in the question (<20 words, 21-30 words, >30 words)—8.0, 11.8, 16.8, seconds, respectively. Conclusions: This technology-enabled data management system helped capture patient opinions, accelerate turnaround of survey data, with minimal impact on a busy primary care clinic. This new model of patient survey data management is feasible and sustainable in a busy office setting, supports and engages clinicians in the quality improvement process, and harmonizes with the vision of a learning health care system.

  • Screen capture of an application prototype to manage the PHQ-9 object.

    Information Persistence Services Designed to Support Home Care


    Background: Due to the challenges faced by health and social care systems, in particular those related to actual demographic trends, home care emerges as a potentially cost-effective solution to answer the needs of citizens, and to allow the reallocation of resources to alternatives to hospitalization or institutionalization. Objective: Home care services require cooperation between different actors, including health and social caregivers, care receivers, and their informal caregivers (eg, relatives or friends), across time, space, and organizational boundaries. Therefore, it is foreseeable that eHealth services can contribute to their improvement. The aim of this study is to evaluate information persistence services based on the Reference Information Model (RIM) of the Health Level Seven (HL7) version 3 to support formal caregivers, both health and social care providers, and informal caregivers in the context of home care services. Methods: A pilot study was set up involving two Portuguese institutions that provide home care services for the elderly. Defining of information requirements was performed according to a comprehensive process. This included a review of the literature, observations of work activities, interviews with caregivers, care receivers and their relatives, analysis of paper documentation related to care receivers’ histories, health conditions and care plans, and brainstorming groups involving specialized professionals. Following this, information objects were implemented and validated. Results: The methodological approach, as well as the information persistence services, proved to be robust and adequate to specify, implement, and validate different types of information objects related to home care services for the elderly. This study also reinforces the application of the RIM of the HL7 version 3 beyond the strict scope of health care, allowing the persistence of not only health care information, but also information related to social assistance activities. Conclusions: This study contributes to the ongoing efforts related to the development of eHealth applications to improve the cooperation among formal health care and social caregivers, as well as care receivers and their informal caregivers.

  • Development and evaluation of a clinical decision support system (CDSS) for Sepsis. 
(A) Synopsis of the EHR database, distribution of vitals, states, actions. Likelihood functions were used to calculate state-specific transition and observation probabilities. (B) The state-action diagram describes underlying patient states, possible transitions and beliefs based on the values of the observed variables. There are two absorbing states, “Death” and “Dismissal”. The training of the POMDP model that is applied on the state-action space performs value iteration updates on a sample set of beliefs, effectively using a Monte Carlo approach for sampling together with dynamic programming for the calculation of the value iteration. (C) Belief (ie, probability distribution of the patient states) is updated based on the action taken and the new observations. At each time step patient vitals are observed and the action that corresponds to the optimal policy is taken. A new set of observations (vitals, tests) will lead to an updated belief that may lead to a new action to be undertaken. The update is asynchronous, as it is calculated on-the-fly as new information arrives.  (D) Evaluation of the CDSS framework was performed through 5-fold cross validation and data size sensitivity analysis.

    From Data to Optimal Decision Making: A Data-Driven, Probabilistic Machine Learning Approach to Decision Support for Patients With Sepsis


    Background: A tantalizing question in medical informatics is how to construct knowledge from heterogeneous datasets, and as an extension, inform clinical decisions. The emergence of large-scale data integration in electronic health records (EHR) presents tremendous opportunities. However, our ability to efficiently extract informed decision support is limited due to the complexity of the clinical states and decision process, missing data and lack of analytical tools to advice based on statistical relationships. Objective: Development and assessment of a data-driven method that infers the probability distribution of the current state of patients with sepsis, likely trajectories, optimal actions related to antibiotic administration, prediction of mortality and length-of-stay. Methods: We present a data-driven, probabilistic framework for clinical decision support in sepsis-related cases. We first define states, actions, observations and rewards based on clinical practice, expert knowledge and data representations in an EHR dataset of 1492 patients. We then use Partially Observable Markov Decision Process (POMDP) model to derive the optimal policy based on individual patient trajectories and we evaluate the performance of the model-derived policies in a separate test set. Policy decisions were focused on the type of antibiotic combinations to administer. Multi-class and discriminative classifiers were used to predict mortality and length of stay. Results: Data-derived antibiotic administration policies led to a favorable patient outcome in 49% of the cases, versus 37% when the alternative policies were followed (P=1.3e-13). Sensitivity analysis on the model parameters and missing data argue for a highly robust decision support tool that withstands parameter variation and data uncertainty. When the optimal policy was followed, 387 patients (25.9%) have 90% of their transitions to better states and 503 patients (33.7%) patients had 90% of their transitions to worse states (P=4.0e-06), while in the non-policy cases, these numbers are 192 (12.9%) and 764 (51.2%) patients (P=4.6e-117), respectively. Furthermore, the percentage of transitions within a trajectory that lead to a better or better/same state are significantly higher by following the policy than for non-policy cases (605 vs 344 patients, P=8.6e-25). Mortality was predicted with an AUC of 0.7 and 0.82 accuracy in the general case and similar performance was obtained for the inference of the length-of-stay (AUC of 0.69 to 0.73 with accuracies from 0.69 to 0.82). Conclusions: A data-driven model was able to suggest favorable actions, predict mortality and length of stay with high accuracy. This work provides a solid basis for a scalable probabilistic clinical decision support framework for sepsis treatment that can be expanded to other clinically relevant states and actions, as well as a data-driven model that can be adopted in other clinical areas with sufficient training data.

Citing this Article

Right click to copy or hit: ctrl+c (cmd+c on mac)

Latest Submissions Open for Peer-Review:

View All Open Peer Review Articles
  • Putting Meaning into Meaningful Use: A Roadmap to successful Integration of Evidence at the Point of Care

    Date Submitted: Apr 18, 2015

    Open Peer Review Period: Apr 20, 2015 - Jun 15, 2015

    Pressures to contain healthcare costs, personalize patient care, use of big data, and enhance healthcare quality have highlighted the need for integration of evidence at the point of care.(1-7) But th...

    Pressures to contain healthcare costs, personalize patient care, use of big data, and enhance healthcare quality have highlighted the need for integration of evidence at the point of care.(1-7) But the application of EBM has great promise in the era of electronic health records (EHRs) and health technology. Most successful integration of evidence into EHRs has been complex decision tools that trigger at a critical point of the clinical visit and include patient specific recommendations. Incorporation of complex CDS tools into the EMR is equally complex and continues to challenge health service researchers and implementation scientists. Poor adoption and sustainably of EBM guidelines and CDS tools at the point of care has persisted and continue to document low rates of usage. The barriers cited by physicians include efficiency, perception of usefulness, information content, user interface, and over-triggering. Building on the traditional EHR implementation frameworks we review keys strategies for successful CDSs: 1) the quality of the evidence, 2) the potential to reduce unnecessary care, 3) ease of integrating evidence at the point of care, 4) the evidence’s consistency with clinician perceptions and preferences, 5) incorporating bundled sets or automated documentation and 6) shared decision making tools.