Evaluation of PHI Hunter in Natural Language Processing Research.

OBJECTIVES: We introduce and evaluate a new, easily accessible tool using a common statistical analysis and business analytics software suite, SAS, which can be programmed to remove specific protected health information (PHI) from a text document. Removal of PHI is important because the quantity of text documents used for research with natural language processing (NLP) is increasing. When using existing data for research, an investigator must remove all PHI not needed for the research to comply with human subjects' right to privacy. This process is similar, but not identical, to de-identification of a given set of documents. MATERIALS AND METHODS: PHI Hunter removes PHI from free-form text. It is a set of rules to identify and remove patterns in text. PHI Hunter was applied to 473 Department of Veterans Affairs (VA) text documents randomly drawn from a research corpus stored as unstructured text in VA files. RESULTS: PHI Hunter performed well with PHI in the form of identification numbers such as Social Security numbers, phone numbers, and medical record numbers. The most commonly missed PHI items were names and locations. Incorrect removal of information occurred with text that looked like identification numbers. DISCUSSION: PHI Hunter fills a niche role that is related to but not equal to the role of de-identification tools. It gives research staff a tool to reasonably increase patient privacy. It performs well for highly sensitive PHI categories that are rarely used in research, but still shows possible areas for improvement. More development for patterns of text and linked demographic tables from electronic health records (EHRs) would improve the program so that more precise identifiable information can be removed. CONCLUSIONS: PHI Hunter is an accessible tool that can flexibly remove PHI not needed for research. If it can be tailored to the specific data set via linked demographic tables, its performance will improve in each new document set.

The cost of adverse drug events in ambulatory care.

BACKGROUND: Many justifications for ePrescribing predict savings achieved by reducing the number of adverse drug events (ADEs) in the ambulatory setting however, there is little evidence from which to estimate the size of these savings. Estimating the cost of ADEs in the ambulatory setting would improve the reliability of these predictions. METHODS: We identified patients with potential ADEs in a primary care practice setting and characterized the patient's age along with charge and utilization indicators for 6 weeks pre- and post-event. We then used linear regression to determine charges attributable to an ADE. RESULTS: Charges were higher for patients following an ambulatory visit who were determined to have ADEs. This occurred in a linear fashion: 2 ADEs ($4,976); 1 ADE ($2,337); and no ADEs ($1,943). The charge attributable to a single ADE is $643 (2001 US dollars) or $926 (cost adjusted to 2006 US dollars). CONCLUSIONS: Patients with ADEs incur greater charges. The charges attributable to an ambulatory ADE are a significant cost to the health care delivery system on the order of $8 billion annually.

Common comorbidity scales were similar in their ability to predict health care costs and mortality.

OBJECTIVE: To compare the ability of commonly used measures of medical comorbidity (ambulatory care groups [ACGs], Charlson comorbidity index, chronic disease score, number of prescribed medications, and number of chronic diseases) to predict mortality and health care costs over 1 year. STUDY DESIGN AND SETTING: A prospective cohort study of community-dwelling older adults (n=3,496) attending a large primary care practice. RESULTS: For predicting health care charges, the number of medications had the highest predictive validity (R(2)=13.6%) after adjusting for demographics. ACGs (R(2)=16.4%) and the number of medications (15.0%) had the highest predictive validity for predicting ambulatory visits. ACGs and the Charlson comorbidity index (area under the receiver operator characteristic [ROC] curve=0.695-0.767) performed better than medication-based measures (area under the ROC curve=0.662-0.679) for predicting mortality. There is relatively little difference, however, in the predictive validity across these scales. CONCLUSION: In an outpatient setting, a simple count of medications may be the most efficient comorbidity measure for predicting utilization and health-care charges over the ensuing year. In contrast, diagnosis-based measures have greater predictive validity for 1-year mortality. Current comorbidity measures, however, have only poor to moderate predictive validity for costs or mortality over 1 year.

A tiered approach is more cost effective than traditional pharmacist-based review for classifying computer-detected signals as adverse drug events.

OBJECTIVE: To develop a cost-efficient method for identifying adverse drug events (ADEs) and medication errors (MEs) identified using outpatient electronic medical records within ambulatory settings. DESIGN: Comparison of sensitivity and cost of "traditional" pharmacist based approach to identifying ADEs and MEs during a 4 month period with a tiered approach. RESULTS: The proportion of computer generated signals analyzed identified as ADEs were similar using the two approaches while the number of MEs was nearly double with tiered reviews suggesting the same or better sensitivity. Traditional pharmacist review cost $68.70 US dollars to detect an ADE and tiered approach cost only $42.40. CONCLUSION: Tiered review of ADEs and MEs by personnel with increasing clinical capability is more cost-efficient than pharmacist review.