Detection of allergic transfusion-related adverse events from electronic medical records.

BACKGROUND: Transfusion-related adverse events can be unrecognized and unreported. As part of the US Food and Drug Administration's Center for Biologics Evaluation and Research Biologics Effectiveness and Safety initiative, we explored whether machine learning methods, such as natural language processing (NLP), can identify and report transfusion allergic reactions (ARs) from electronic health records (EHRs). STUDY DESIGN AND METHODS: In a 4-year period, all 146 reported transfusion ARs were pulled from a database of 86,764 transfusions in an academic health system, along with a random sample of 605 transfusions without reported ARs. Structured and unstructured EHR data were retrieved, including demographics, new symptoms, medications, and lab results. In unstructured data, evidence from clinicians' notes, test results, and prescriptions fields identified transfusion ARs, which were used to extract NLP features. Clinician reviews of selected validation cases assessed and confirmed model performance. RESULTS: Clinician reviews of selected validation cases yielded a sensitivity of 67.9% and a specificity of 97.5% at a threshold of 0.9, with a positive predictive value (PPV) of 84%, estimated to 4.5% when extrapolated to match transfusion AR incidence in the full transfusion dataset. A higher threshold achieved sensitivity of 43% with specificity/PPV of 100% in our validation set. Essential features predicting ARs were recognized transfusion reactions, administration of antihistamines or glucocorticoids, and skin symptoms (e.g., hives and itching). Removal of NLP features decreased model performance. DISCUSSION: NLP algorithms can identify transfusion reactions from the EHR with a reasonable level of precision for subsequent clinician review and confirmation.

The Food and Drug Administration Biologics Effectiveness and Safety Initiative Facilitates Detection of Vaccine Administrations From Unstructured Data in Medical Records Through Natural Language Processing.

Introduction: The Food and Drug Administration Center for Biologics Evaluation and Research conducts post-market surveillance of biologic products to ensure their safety and effectiveness. Studies have found that common vaccine exposures may be missing from structured data elements of electronic health records (EHRs), instead being captured in clinical notes. This impacts monitoring of adverse events following immunizations (AEFIs). For example, COVID-19 vaccines have been regularly administered outside of traditional medical settings. We developed a natural language processing (NLP) algorithm to mine unstructured clinical notes for vaccinations not captured in structured EHR data. Methods: A random sample of 1,000 influenza vaccine administrations, representing 995 unique patients, was extracted from a large U.S. EHR database. NLP techniques were used to detect administrations from the clinical notes in the training dataset [80% (N = 797) of patients]. The algorithm was applied to the validation dataset [20% (N = 198) of patients] to assess performance. Full medical charts for 28 randomly selected administration events in the validation dataset were reviewed by clinicians. The NLP algorithm was then applied across the entire dataset (N = 995) to quantify the number of additional events identified. Results: A total of 3,199 administrations were identified in the structured data and clinical notes combined. Of these, 2,740 (85.7%) were identified in the structured data, while the NLP algorithm identified 1,183 (37.0%) administrations in clinical notes; 459 were not also captured in the structured data. This represents a 16.8% increase in the identification of vaccine administrations compared to using structured data alone. The validation of 28 vaccine administrations confirmed 27 (96.4%) as "definite" vaccine administrations; 18 (64.3%) had evidence of a vaccination event in the structured data, while 10 (35.7%) were found solely in the unstructured notes. Discussion: We demonstrated the utility of an NLP algorithm to identify vaccine administrations not captured in structured EHR data. NLP techniques have the potential to improve detection of vaccine administrations not otherwise reported without increasing the analysis burden on physicians or practitioners. Future applications could include refining estimates of vaccine coverage and detecting other exposures, population characteristics, and outcomes not reliably captured in structured EHR data.

A Randomized Trial to Evaluate the Efficacy of a Web-Based HIV Behavioral Intervention for High-Risk African American Women.

The aim of this study was to develop and test a cost-effective, scalable HIV behavioral intervention for African American women. Eighty-three African American women were recruited from a community health center and randomly assigned to either the web-based Safe Sistah program or to a delayed HIV education control condition. The primary outcome was self-reported condom use. Secondary measures assessed other aspects of the gender-focused training included in Safe Sistah. Participants completed self-report assessments prior to randomization, 1- and 4-months after their program experience. Across the entire study period, women in the experimental condition significantly increased their condom use relative to controls (F = 5.126, p = 0.027). Significant effects were also found for sexual communication, sex refusal, condom use after alcohol consumption, and HIV prevention knowledge. These findings indicate that this web-based program could be an important component in reducing the HIV disparities among African American women.

Test of a web-based program to improve adherence to HIV medications.

We evaluated the effectiveness of a web-based version of the Life-Steps intervention combined with modules for stress reduction and mood management, designed to improve medication adherence among HIV infected individuals. 168 HIV+ adults were randomized into either the Life-Steps program or a waitlist control condition. All participants completed a baseline assessment and provided a 2-week electronic pill (MEMS) cap baseline reading. Follow up data collection was conducted at 3, 6 and 9 months. Patients in the web-based Life-Steps condition had significantly higher antiretroviral medication adherence rates than patients in the control group over the nine-month period as measured by the MEMS cap. In addition, analysis of viral load data indicated that the program also resulted in a significant decrease in viral load. These findings indicate that a web-based Life-Steps program can be a useful and implementable tool for helping patients living with HIV maintain medication adherence.

Community-powered problem solving.

Traditionally, companies have managed their constituencies with specific processes: marketing to customers, procuring from vendors, developing HR policies for employees, and so on. The problem is, such processes focus on repeatability and compliance, so they can lead to stagnation. Inviting your constituencies to collectively help you solve problems and exploit opportunities--"co-creation"--is a better approach. It allows you to continually tap the skills and insights of huge numbers of stakeholders and develop new ways to produce value for all. The idea is to provide stakeholders with platforms (physical and digital forums) on which they can interact, get them to start exploring new experiences and connections, and let the system grow organically. A co-creation initiative by a unit of Becton, Dickinson and Company demonstrates how this works. A global leader in syringes, BD set out to deepen its ties with hospital customers and help them reduce the incidence of infections from unsafe injection and syringe disposal practices. The effort began with a cross-functional internal team, brought in the hospital procurement and supply managers BD had relationships with, and then reached out to hospitals' infection-prevention and occupational health leaders. Eventually product designers, nurses, sustainability staffers, and even hospital CFOs were using the platform, contributing data that generated new best practices and reduced infections.

A web-based mental health program: reaching parents at work.

OBJECTIVE: The purpose of the project was to test a web-based program providing working parents with the knowledge and skills necessary for prevention and early intervention of mental health problems in youth. METHOD: Study sample consisted of 99 parents who were randomized into either an experimental (program use) or waitlist control condition. Analysis of covariance and paired t-tests were used to compare response outcomes for the two groups. RESULTS: Parents receiving the intervention had greater knowledge of youth mental health issues and greater self-efficacy in handling these issues compared to controls. CONCLUSIONS: Findings suggest that referring parents to a multimedia web-based program can improve parents' knowledge of children's' mental health and their confidence in addressing mental health issues. Web-based programs can be offered to individuals in multiple settings, including the workplace, thereby reaching large numbers of parents.

A web-based approach to managing stress and mood disorders in the workforce.

OBJECTIVE: To evaluate the effectiveness of a web-based multimedia health promotion program for the workplace, designed to help reduce stress and to prevent depression, anxiety, and substance abuse. METHODS: Using a randomized controlled trial design, 309 working adults were randomly assigned to the web-based condition or to a wait-list control condition. All participants were assessed on multiple self-reported outcomes at pretest and posttest. RESULTS: Relative to controls, the web-based group reduced their stress, increased their knowledge of depression and anxiety, developed more positive attitudes toward treatment, and adopted a more healthy approach to alcohol consumption. CONCLUSIONS: We found that a brief and easily adaptable web-based stress management program can simultaneously reduce worker stress and address stigmatized behavioral health problems by embedding this prevention material into a more positive stress management framework.

A field test of a web-based workplace health promotion program to improve dietary practices, reduce stress, and increase physical activity: randomized controlled trial.

BACKGROUND: Most work sites engage in some form of health promotion programming designed to improve worker health and reduce health care costs. Although these programs have typically been delivered through combinations of seminars and print materials, workplace health promotion programs are increasingly being delivered through the Internet. OBJECTIVE: The purpose of this research was to evaluate the effectiveness of a Web-based multimedia health promotion program for the workplace, designed to improve dietary practices, reduce stress, and increase physical activity. METHODS: Using a randomized controlled trial design with pretest-posttest comparisons within each group, 419 employees of a human resources company were randomly assigned to the Web-based condition or to a condition that provided print materials on the same topics. All subjects were assessed at pretest and posttest through an online questionnaire containing multiple measures of health behavior and attitudes. The test period was 3 months. Questionnaire data were analyzed mainly by analysis of covariance and t tests. RESULTS: Retention rates were good for both groups-85% for the Web-based group and 87% for the print group. Subjects using the Web-based program performed significantly better than the print group on Attitudes Toward a Healthful Diet (F(1,415) = 7.104, P = .008) and Dietary Stage of Change (F(1,408) = 6.487, P = .01), but there were no significant group differences on the five other dietary measures. Both groups also showed improvement from pretest to posttest on most dietary measures, as indicated by significant t tests. Within the Web-based group, dosage analyses showed significant effects of the number of times the subject accessed the program on measures of Dietary Self-Efficacy (F(2,203) = 5.270, P = .003), Attitudes Toward a Healthful Diet (F(2,204) = 2.585, P = .045), and Dietary Stage of Change (F(2,200) = 4.627, P = .005). No significant differences were found between the two groups on measures of stress or physical activity, although t tests of pretest-posttest changes indicated that both groups improved on several of these measures. The Web-based group gave significantly higher ratings to the program materials than the print group on all health topics and in their overall evaluation (F(1,410) = 9.808, P = .002). CONCLUSIONS: The Web-based program was more effective than print materials in producing improvements in the areas of diet and nutrition but was not more effective in reducing stress or increasing physical activity. The higher ratings given to the Web-based program suggest that workers preferred it to the print materials. Both groups showed numerous pretest-posttest improvements in all health topics, although such improvements might be attributable in part to a Hawthorne effect. Results suggest that a multimedia Web-based program can be a promising means of delivering health promotion material to the workforce, particularly in the area of diet and nutrition.