A structured approach to recording AIDS-defining illnesses in Kenya: A SNOMED CT based solution.

INTRODUCTION: Several studies conducted in sub-Saharan Africa (SSA) have shown that routine clinical data in HIV clinics often have errors. Lack of structured and coded documentation of diagnosis of AIDS defining illnesses (ADIs) can compromise data quality and decisions made on clinical care. METHODS: We used a structured framework to derive a reference set of concepts and terms used to describe ADIs. The four sources used were: (i) CDC/Accenture list of opportunistic infections, (ii) SNOMED Clinical Terms (SNOMED CT), (iii) Focus Group Discussion (FGD) among clinicians and nurses attending to patients at a referral provincial hospital in western Kenya, and (iv) chart abstraction from the Maternal Child Health (MCH) and HIV clinics at the same hospital. Using the January 2014 release of SNOMED CT, concepts were retrieved that matched terms abstracted from approach iii & iv, and the content coverage assessed. Post-coordination matching was applied when needed. RESULTS: The final reference set had 1054 unique ADI concepts which were described by 1860 unique terms. Content coverage of SNOMED CT was high (99.9% with pre-coordinated concepts; 100% with post-coordination). The resulting reference set for ADIs was implemented as the interface terminology on OpenMRS data entry forms. CONCLUSION: Different sources demonstrate complementarity in the collection of concepts and terms for an interface terminology. SNOMED CT provides a high coverage in the domain of ADIs. Further work is needed to evaluate the effect of the interface terminology on data quality and quality of care.

Supporting patient screening to identify suitable clinical trials.

To support the efficient execution of post-genomic multi-centric clinical trials in breast cancer we propose a solution that streamlines the assessment of the eligibility of patients for available trials. The assessment of the eligibility of a patient for a trial requires evaluating whether each eligibility criterion is satisfied and is often a time consuming and manual task. The main focus in the literature has been on proposing different methods for modelling and formalizing the eligibility criteria. However the current adoption of these approaches in clinical care is limited. Less effort has been dedicated to the automatic matching of criteria to the patient data managed in clinical care. We address both aspects and propose a scalable, efficient and pragmatic patient screening solution enabling automatic evaluation of eligibility of patients for a relevant set of trials. This covers the flexible formalization of criteria and of other relevant trial metadata and the efficient management of these representations.

Learning regular expressions for clinical text classification.

OBJECTIVES: Natural language processing (NLP) applications typically use regular expressions that have been developed manually by human experts. Our goal is to automate both the creation and utilization of regular expressions in text classification. METHODS: We designed a novel regular expression discovery (RED) algorithm and implemented two text classifiers based on RED. The RED+ALIGN classifier combines RED with an alignment algorithm, and RED+SVM combines RED with a support vector machine (SVM) classifier. Two clinical datasets were used for testing and evaluation: the SMOKE dataset, containing 1091 text snippets describing smoking status; and the PAIN dataset, containing 702 snippets describing pain status. We performed 10-fold cross-validation to calculate accuracy, precision, recall, and F-measure metrics. In the evaluation, an SVM classifier was trained as the control. RESULTS: The two RED classifiers achieved 80.9-83.0% in overall accuracy on the two datasets, which is 1.3-3% higher than SVM's accuracy (p<0.001). Similarly, small but consistent improvements have been observed in precision, recall, and F-measure when RED classifiers are compared with SVM alone. More significantly, RED+ALIGN correctly classified many instances that were misclassified by the SVM classifier (8.1-10.3% of the total instances and 43.8-53.0% of SVM's misclassifications). CONCLUSIONS: Machine-generated regular expressions can be effectively used in clinical text classification. The regular expression-based classifier can be combined with other classifiers, like SVM, to improve classification performance.

Photodocumentation as an emergency department documentation tool in soft tissue infection: a randomized trial.

OBJECTIVES: Current documentation methods for patients with skin and soft tissue infections receiving outpatient parenteral anti-infective therapy (OPAT) include written descriptions and drawings of the infection that may inadequately communicate clinical status. We undertook a study to determine whether photodocumentation (PD) improves the duration of outpatient treatment of skin and soft tissue infections. METHODS: A single-blinded, prospective, randomized trial was conducted in the emergency departments of a community hospital and an academic tertiary centre. Participants included consecutive patients age ≥ 14 years presenting with noninvasive skin and soft tissue infections requiring OPAT. Patients in the intervention arm were treated with standard of care plus PD at each emergency physician assessment. Control subjects received care provided at the discretion of the treating physician and non-photographic documentation. The primary outcome was duration of therapy measured in half-days. The required sample size to detect a difference of one half-day was 253 patients per group (α  =  0.05). Secondary outcomes included (1) completion and therapeutic failure rates, (2) patient satisfaction, and (3) physician and nurse satisfaction. RESULTS: Enrolment was slower and follow-up rates lower than anticipated, and the trial was terminated when funds were exhausted. A total of 468 subjects with similar age and gender characteristics were enrolled, with 244 receiving the intervention and 224 in the control arm. The mean OPAT duration was similar in the two groups (3.6 days v. 3.5 days, p  =  0.73). No differences in the rate for completion and therapeutic failure were observed (71% v. 68% and < 1% for both, respectively). Survey response rates varied significantly: patients, 65%; nurses, 17%; and physicians, 87%. Physicians endorsed more comfort with their assessment and OPAT judgment with PD (65% and 64%, respectively). Physicians cited too much time lost with technological challenges, which would affect implementation in a busy ED. CONCLUSIONS: PD as an intervention is acceptable to patients and has reasonable endorsement by the majority of physicians. This trial had significant limitations that threatened the integrity of the study, so the results are inconclusive.

The usability-error ontology.

Clinical Systems have become standard partners with clinicians in the care of patients. As these systems become integral parts of the clinical workflow, they have the potential to help improve patient outcomes, however they have also in some cases have led to adverse events and has resulted in patients coming to harm. Often the root cause analysis of these adverse events can be traced back to Usability Errors in the Health Information Technology (HIT) or its interaction with users. Interoperability of the documentation of HIT related Usability Errors in a consistent fashion can improve our ability to do systematic reviews and meta-analyses. In an effort to support improved and more interoperable data capture regarding Usability Errors, we have created the Usability Error Ontology (UEO) as a classification method for representing knowledge regarding Usability Errors. We expect the UEO will grow over time to support an increasing number of HIT system types. In this manuscript, we present this Ontology of Usability Error Types and specifically address Computerized Physician Order Entry (CPOE), Electronic Health Records (EHR) and Revenue Cycle HIT systems.

Full-text automated detection of surgical site infections secondary to neurosurgery in Rennes, France.

The surveillance of Surgical Site Infections (SSI) contributes to the management of risk in French hospitals. Manual identification of infections is costly, time-consuming and limits the promotion of preventive procedures by the dedicated teams. The introduction of alternative methods using automated detection strategies is promising to improve this surveillance. The present study describes an automated detection strategy for SSI in neurosurgery, based on textual analysis of medical reports stored in a clinical data warehouse. The method consists firstly, of enrichment and concept extraction from full-text reports using NOMINDEX, and secondly, text similarity measurement using a vector space model. The text detection was compared to the conventional strategy based on self-declaration and to the automated detection using the diagnosis-related group database. The text-mining approach showed the best detection accuracy, with recall and precision equal to 92% and 40% respectively, and confirmed the interest of reusing full-text medical reports to perform automated detection of SSI.

Engineering natural language processing solutions for structured information from clinical text: extracting sentinel events from palliative care consult letters.

Despite a trend to formalize and codify medical information, natural language communications still play a prominent role in health care workflows, in particular when it comes to hand-overs between providers. Natural language processing (NLP) attempts to bridge the gap between informal, natural language information and coded, machine-interpretable data. This paper reports on a study that applies an advanced NLP method for the extraction of sentinel events in palliative care consult letters. Sentinel events are of interest to predict survival and trajectory for patients with acute palliative conditions. Our NLP method combines several novel characteristics, e.g., the consideration of topological knowledge structures sourced from an ontological terminology system (SNOMED CT). The method has been applied to the extraction of different types of sentinel events, including simple facts, temporal conditions, quantities, and degrees. A random selection of 215 anonymized consult letters was used for the study. The results of the NLP extraction were evaluated by comparison with coded sentinel event data captured independently by clinicians. The average accuracy of the automated extraction was 73.6%.

Evaluating documentation of dietetic care in Swedish medical records.

An adequate documentation in medical records is essential for patient safety and high quality care. The aim of this study was to evaluate documentation by dietitians in Swedish medical records. A retrospective audit of147 dietetic notes in electronic medical records was performed. The audit focused at documentation of essential parts of the dietetic care, as well as other quality aspects such as lingual clarity and structure of the documentation. The nutrition intervention showed to be the most documented part of dietetic care. However, the audit showed that several important parts of nutrition care were poorly documented, for instance nearly half of the audited records had no clear nutrition problem documented, and in most of the records, the goal of nutrition intervention was missing. The study shows that Swedish dietitians need to improve documentation in medical records, as a suggestion by implementing a more structured documentation model.

SNOMED CT's RF2: Is the future bright?

SNOMED CT's new RF2 format is said to come with features for better configuration management of the SNOMED vocabulary, thereby accommodating evolving requirements without the need for further fundamental change in the foreseeable future. Although the available documentation is not yet convincing enough to support this claim, the newly introduced Model Component hierarchy and associated reference set mechanism seem to hold real promise of being able to deal successfully with a number of ontological issues that have been discussed in the recent literature. Backed up by a study of the old and new format and of the relevant literature and documentation, three recommendations are presented that would free SNOMED CT from use-mention confusions, unclear referencing of real-world entities and uninformative reasons for change in a way that does not force SNOMED CT to take a specific philosophical or ontological position.

Validation of electronic medical database in patients with atrial fibrillation in community hospitals.

BACKGROUND: The electronic medical database (EMD) has been increasingly used for clinical research as it reflects a real-world practice with large and heterogeneous samples. However, few studies have reported on the validity of EMD from community hospitals for research purposes. OBJECTIVE: To assess the validity of EMD based on data from patients with atrial fibrillation (AF) receiving care from community hospitals in Phitsanulok Province, Thailand. MATERIAL AND METHOD: The validity of EMD was determined using hand-written out-patient medical records (OPMRs) as a criterion standard. One hundred ninety three records of patient with ICD-10 of AF (148) were retrieved from the EMD of two community hospitals between August 2007 and July 2008. For each patient, data of a randomly selected visit from the EMD was matched to data of the same visit from OPMRs, abstracted by a standardized data collection form. The EMD was cross-validated with OPMRs based on patient's diagnosis of AE co-morbidities (risk factors for stroke) and bleeding events. All data were tabulated in a 2 x 2 format to calculate sensitivity, specificity and the Cohen's Kappa. RESULTS: Out of 193 AF patients retrieved from the EMD, 169 (87.56%) were documented as having a diagnosis of AF in OPMRs. The EMD data on risk factors for stroke showed moderate to high sensitivity (range: 66.67-100%) and high specificity (range: 98.77-100%). The agreement between the two databases was considered good to very good (calculated kappa range: 0.7942-0.9681). The specificity based on major bleeding was 100%; however, sensitivity and the Cohens Kappa could not be determined as the major bleeding diagnosis was found in neither the EMD nor the OPMRs. CONCLUSION: The EMD of AF patients from community hospitals in Phitsanulok was valid and in good agreement with the OPMRs. The EMD from community hospitals appeared suitable for health research in patients with AF

Advancing oral medicine through informatics and information technology: a proposed framework and strategy.

The implementation of information technology in healthcare is a significant focus for many nations around the world. However, information technology support for clinical care, research and education in oral medicine is currently poorly developed. This situation hampers our ability to transform oral medicine into a 'learning healthcare discipline' in which the divide between clinical practice and research is diminished and, ultimately, eliminated. This paper reviews the needs of and requirements for information technology support of oral medicine and proposes an agenda designed to meet those needs. For oral medicine, this agenda includes analyzing and reviewing current clinical and documentation practices, working toward progressively standardizing clinical data, and helping define requirements for oral medicine systems. IT professionals can contribute by conducting baseline studies about the use of electronic systems, helping develop controlled vocabularies and ontologies, and designing, implementing, and evaluating novel systems centered on the needs of clinicians, researchers and educators. Successfully advancing IT support for oral medicine will require close coordination and collaboration among oral medicine professionals, information technology professionals, system vendors, and funding agencies. If current barriers and obstacles are overcome, practice and research in oral medicine stand ready to derive significant benefits from the application of information technology.

Clinical text classification under the Open and Closed Topic Assumptions.

This paper investigates multi-topic aspects in automatic classification of clinical free text in comparison with general text. In this paper, we facilitate two different views on multi-topics: the Closed Topic Assumption (CTA) and the Open Topic Assumption (OTA). Experimental results show that the characteristics of multi-topic assignments in the Computational Medicine Centre (CMC) Medical NLP Challenge Data is strongly OTA-oriented but general text Reuters-21578 is characterised in the middle of the OTA and CTA spectrum.

Evaluation of an automated inferencing engine generating ICD-9CM codes for physician and hospital billing.

Recent Medicare changes to Severity Diagnosis Related Groups (MS-DRGs) for inpatients have made the appropriate and timely coding of services provided by hospitals and physicians a challenge, and require education for clinicians and coders. Clinical departments have limited funds to hire dedicated personnel to code and prepare payor submissions. Automating the process can assist in accurate data collection and reimbursement.

Topic maps for exploring nosological, lexical, semantic and HL7 structures for clinical data.

A topic map is implemented for learning about clinical data associated with a hospital stay for patients diagnosed with chronic kidney disease, diabetes and hypertension. The question posed is: how might a topic map help bridge perspectival differences among communities of practice and help make commensurable the different classifications they use? The knowledge layer of the topic map was generated from existing ontological relationships in nosological, lexical, semantic and HL7 boundary objects. Discharge summaries, patient charts and clinical data warehouse entries rectified the clinical knowledge used in practice. These clinical data were normalized to HL7 Clinical Document Architecture (CDA) markup standard and stored in the Clinical Document Repository. Each CDA entry was given a subject identifier and linked with the topic map. The ability of topic maps to function as the infostructure ;glue' is assessed using dimensions of semantic interoperability and commensurability.

Evaluating healthcare quality using natural language processing.

Consistent monitoring for quality indicators as adverse events or missed screening opportunities remains a difficult proposition for most healthcare organizations. Much of the clinical data needed for quality reports is imbedded in narrative reports in the electronic health record. Narrative data most often require costly retrieval by manual data extraction. NUD*IST, a qualitative research computer program, was used as an automated natural Language processing tool to extract and code data for analysis of screening and treatment for breast cancer. The study method demonstrated acceptable Levels of precision and recall compared to large-scale natural Language processing programs.