Applied Clinical Informatics最新文献

Nursing documentation burden is a growing point of concern in the United States health care system. Documentation in the electronic health record (EHR) is a contributor to perceptions of burden. Efficiency tools like flowsheet macros are one development intended to ease the burden of documentation.This study aimed to evaluate whether flowsheet macros, a documentation efficiency tool in the EHR that consolidates documentation into a single click, reduces the time spent on documentation activities and the EHR overall.Nurses in the health system were encouraged to create and utilize flowsheet macros for their documentation. Flowsheet documentation and time in system data for nurses' first and last shifts in the evaluation period were extracted from the EHR. Linear regression with control variables was utilized to understand if the utilization of flowsheet macros for documentation reduced the time spent in flowsheets or the EHR.The results of linear regression showed a significant, negative relationship between flowsheet macros use and time in flowsheets (adjusted odds ratio [AOR] = -0.291, 95% confidence interval [CI] = -0.342 to -0.240, p < 0.001). Flowsheet macros use and time in system also had a significant, negative relationship (AOR = -0.269, CI = -0.390 to -0.147, p ≤ 0.001). Subgroups for department specialties showed time savings in flowsheet activities for medical surgical, critical care, and obstetrics units, however, a significant relationship was not found in emergency and rehabilitation units.Utilization of flowsheet macros was associated with a decrease in the amount of time a nurse spends in both flowsheets and the EHR. Adoption and timesavings varied by the department setting, suggesting flowsheet macros may not be applicable to all patient types or conditions. Future research should investigate if the time savings from this tool yield benefits in perceptions of nurse documentation burden.

The CONCERN Early Warning System (CONCERN EWS) is an artificial intelligence-based clinical decision support system (AI-CDSS) for the prediction of clinical deterioration, leveraging signals from nursing documentation patterns. While a recent multisite randomized controlled trial (RCT) demonstrated its effectiveness in reducing inpatient mortality and length of stay, evaluating implementation outcomes is essential to ensure equitable results across patient populations.This study aims to (1) assess whether clinicians' usage of the CONCERN EWS, as measured by CONCERN Detailed Prediction Screen launches, varied by patient demographic characteristics, including sex, race, ethnicity, and primary language; (2) evaluate whether CONCERN EWS's effectiveness in reducing the risk of in-hospital mortality varied across patient demographic groups.We conducted a retrospective observational analysis of electronic health record log files and clinical outcomes from a multisite, pragmatic, cluster-RCT involving four hospitals across two health care systems. Equity in usage was assessed by comparing CONCERN Detailed Prediction Screen launches across demographic groups, and effectiveness was examined by comparing the risk of in-hospital mortality between intervention and usual care groups using Cox proportional hazards models adjusted for patient characteristics.Clinicians' CONCERN Detailed Prediction Screen launches did not significantly differ by patients' demographic characteristics, suggesting equitable usage. The CONCERN EWS was significantly associated with reduced risk of in-hospital mortality overall (adjusted hazard ratio [HR] = 0.644, 95% CI: 0.532-0.778, p < 0.0001), with consistent effectiveness across most groups. Notably, patients whose primary language was not English experienced a greater reduction of mortality risk compared to patients whose primary language was English (adjusted HR = 0.419, 95% CI: 0.287-0.610, p = 0.0082).This study presents a case of evaluating equity in AI-CDSS usage and effectiveness, contributing to the limited literature. While findings suggest equitable engagement and effectiveness, ongoing evaluations are needed to understand the observed variability and ensure responsible implementation.

Cancer staging is integral to ensuring cancer patients receive appropriate risk-adapted therapy. Discrete cancer staging using a structured staging form helps ensure accurate staging, provides a single source of truth for staging information, and allows for reporting to regulatory authorities. Our institution created pediatric oncology specific discrete staging forms that have been shared with the broader Epic community. By November 2023, baseline utilization of the staging form for patients with leukemia or lymphoma was 43%, and the override rate for our existing alert was 99.9%.Improve discrete documentation of cancer stage for patients with leukemia or lymphoma within 60 days following initiation of chemotherapy to >80% by July 2024 as measured by signed staging form.Model for improving plan-do-study-act (PDSA) cycles was implemented, and statistical process control charts were used to evaluate impact. The first intervention was educational training to oncology providers. The second PDSA cycle involved sharing monthly individual completion data with the primary oncologist regarding their personal patient metrics. The third PDSA cycle involved removing the interruptive alert.Within 6 months, documentation of primary oncologist improved from 86 to 100%, and initiation of staging form improved from 57 to 90%. Completion of signed cancer staging form reached 80%. Patients marked as not needing staging increased from 5 to 17%.Completion of a digital cancer staging form is important for continuity of care, and to facilitate reporting to regulatory authorities, though frequent interruptive alerts were an ineffective method for improving documentation. Education and data sharing increased staging completion to near target, with ongoing efforts to reach the goal of 80%.

Accurate discharge summaries are essential for effective communication between hospital and outpatient providers but generating them is labor-intensive. Large language models (LLMs), such as GPT-4, have shown promise in automating this process, potentially reducing clinician workload and improving documentation quality. A recent study using GPT-4 to generate discharge summaries via concatenated clinical notes found that while the summaries were concise and coherent, they often lacked comprehensiveness and contained errors. To address this, we evaluated a structured prompting strategy, summarize-then-prompt, which first generates concise summaries of individual clinical notes before combining them to create a more focused input for the LLM.The objective of this study was to assess the effectiveness of a novel prompting strategy, summarize-then-prompt, in generating discharge summaries that are more complete, accurate, and concise in comparison to the approach that simply concatenates clinical notes.We conducted a retrospective study comparing two prompting strategies: direct concatenation (M1) and summarize-then-prompt (M2). A random sample of 50 hospital stays was selected from a large hospital system. Three attending physicians independently evaluated the generated hospital course summaries for completeness, correctness, and conciseness using a 5-point Likert scale.The summarize-then-prompt strategy outperformed direct concatenation strategy in both completeness (4.28 ± 0.63 vs. 4.01 ± 0.69, p < 0.001) and correctness (4.37 ± 0.54 vs. 4.17 ± 0.57, p = 0.002) of the summarization of the hospital course. However, the two strategies showed no significant difference in conciseness (p = 0.308).Summarizing individual notes before concatenation improves LLM-generated discharge summaries, enhancing their completeness and accuracy without sacrificing conciseness. This approach may facilitate the integration of LLMs into clinical workflows, offering a promising strategy for automating discharge summary generation and could reduce clinician burden.

The goal of national interoperability is to improve care quality and decrease administrative burden and costs. Patients, providers, and other stakeholders are increasingly concerned that indiscriminate sharing of data may have deleterious, permanent consequences, as well as fail to provide granular control over the sharing of individual health data. Data segmentation and consent standards to date have been limited in scope and implementation, which has hindered efforts to scale data sharing preferences. Shift, an independent expert stakeholder task force, has been convened to mature standards, terminologies, and consensus-driven implementation guidance, which are prerequisites for more robust policy drivers needed to support nationwide sensitive data segmentation and consent capabilities. This paper describes Shift's framework and processes as means to advance equitable interoperability.

The global aging population is rapidly increasing, and the prevalence of age-related cognitive conditions, such as mild cognitive impairment (MCI), is becoming more common. This condition, which represents intermediate stages between normal aging and dementia, underscores the importance of early detection and timely intervention to address the growing demand for health services. Traditional cognitive assessments have limitations, such as the consistency of results, prompting the need for innovative technology-based solutions.This study aimed to examine how technology-based mobility data collection methods and machine learning algorithms are used to detect MCI in adults.A systematic scoping review was conducted to identify papers that analyzed mobility-related data using machine learning algorithms, focusing on adults aged 18 or older with MCI. Seven databases were searched: MEDLINE, EMBASE, IEEE Xplore, PsycINFO, Scopus, Web of Science, and ACM Digital Library, yielding 2,901 papers.Twenty-four papers met the inclusion criteria, highlighting 116 mobility indicators used to classify or indicate MCI. Wearable devices were the most common data collection method, with mobile applications being the least utilized. The most frequently reported mobility indicator for walking was walking speed. For driving, indicators included the number of hard braking events, the number of night trips, and speed. Logistic regression, random forest, and neural networks were the most used machine learning algorithms. Overall, the mean accuracy, sensitivity, and specificity of all the algorithms were 86.1% (standard deviation [SD] = 6.7%), 84% (SD = 6.5%), and 72.8% (SD = 12%), respectively. The mean area under the curve and the harmonic mean of precision and recall scores (F1) were 0.77 (SD = 0.08) and 0.83 (SD = 0.16), respectively.This review highlights the use of technology-based methods, particularly wearable devices, in assessing mobility and applying machine learning algorithms to detect MCI. However, a notable gap in research on mobile app-based mobility monitoring suggests a promising direction for future studies.

Electronic health records (EHRs) contain valuable patient information, yet certain aspects of care remain infrequently documented and difficult to extract. Identifying these rarely documented elements requires advanced informatics approaches to uncover clinical documentation patterns that would otherwise remain inaccessible for research and quality improvement.This study developed and validated an informatics approach using natural language processing (NLP) to detect and characterize rarely documented elements in EHRs, using spiritual care documentation as an exemplar case.Using EHR data from a Midwestern US hospital (2010-2023), we fine-tuned Spiritual-BERT, an NLP model based on Bio-Clinical-BERT. The model was trained on 80% of a manually annotated, gold-standard corpus of EHR notes, and its performance was validated using the remaining 20% of the corpus, alongside 150 synthetic notes generated by GPT-4 and curated by clinical experts. We applied Spiritual-BERT to identify spiritual care documentation and analyzed patterns across diverse patient populations, provider roles, and clinical services.Spiritual-BERT demonstrated high accuracy in capturing spiritual care documentation (F1-scores: 0.938 internal validation, 0.832 external validation). Analysis of nearly 3.6 million EHR notes from 14,729 older adults revealed that 2% of clinical notes contained spiritual care references, while 73% of patients had spiritual care documented in at least one note. Significant variations were observed across provider types: chaplains documented spiritual care in 99.4% of their notes, compared to 1.7% for nurses and 1.2% for physicians. Documentation patterns also varied based on ethnicity, language, and medical diagnosis.This study demonstrates how advanced NLP techniques can effectively identify and characterize rarely documented elements in EHRs that would be challenging to detect through traditional methods. This approach revealed distinct documentation patterns across provider types, clinical settings, and patient characteristics, with promise for analyzing other under-documented clinical information.

Approximately 10% of patients have a documented penicillin "allergy"; however, up to 95% have subsequent negative testing. These patients may receive suboptimal antibiotics, leading to longer hospitalizations and higher costs, rates of resistant and nosocomial infections, and all-cause mortality. To mitigate these risks in children, we implemented an inpatient penicillin allergy delabeling protocol and integrated it into the electronic health record (EHR) through a mixed methods approach of clinical decision support (CDS).We describe our protocol implementation across three sequential phases: "Pilot," "Active Antimicrobial Stewardship Program (ASP)," and "Mixed CDS." We highlight several potential pitfalls that may have contributed to poor clinician adoption.Patients were risk-stratified as nonallergic, low-risk, or high-risk based on history. Process measures included: evaluation rate, oral challenge rate for low-risk, and allergy referral rate for high- or low-risk when oral challenge was deferred. The primary outcome measure was the penicillin allergy delabeling rate among low-risk or nonallergic. Balancing measures included the rate of epinephrine or antihistamine administrations.The pilot and ASP phases used clinician education and an order set, but were mostly manual processes. The mixed CDS phase introduced interruptive alerts, dynamic text in note templates, and patient list columns to guide clinicians, but little education was provided. The mixed CDS phase had the lowest evaluation rate compared with the pilot and active ASP phases (6.4 vs. 25 vs. 15%). However, when the evaluation was performed, the mixed CDS phase had the highest oral challenge rate (33 vs. 26 vs. 13%) and delabeling rate (43 vs. 33 vs. 27%). No adverse events occurred.CDS tools improve clinician decision-making and optimize patient care. However, relying on CDS for complex clinical evaluations can lead to failure when clinicians cannot find the tool or appreciate its importance. Person-to-person communication can be vital in establishing a process and educating intended users for successful CDS implementation.

In the emergency department-initiated buprenorphine for opioid use disorder (EMBED) trial, a clinical decision support (CDS) tool had no effect on rates of buprenorphine initiation in emergency department (ED) patients with opioid use disorder. The Agency for Healthcare Research and Quality (AHRQ) recently released a CDS Performance Measure Inventory to guide data-driven CDS development and evaluation. Through partner co-design, we tailored AHRQ inventory measures to evaluate EMBED CDS performance and drive improvements.Relevant AHRQ inventory measures were selected and adapted using a partner co-design approach grounded in consensus methodology, with three iterative, multidisciplinary partner working group sessions involving stakeholders from various roles and institutions; meetings were followed by postmeeting surveys. The co-design process was divided into conceptualization, specification, and evaluation phases building on the Centers for Medicare and Medicaid Services' measure life cycle framework. Final measures were evaluated in three EDs in a single health system from January 1, 2023, to December 31, 2024.The partner working group included 25 members. During conceptualization, 13 initial candidate metrics were narrowed to 6 priority categories. These were further specified and validated as the following measures, presented with preliminary values based on the use of the current (i.e., preoptimization) EMBED CDS: eligible encounters with CDS engagement, 5.0% (95% confidence interval: 4.3-5.8%); teamwork on ED initiation of buprenorphine, 39.9% (32.5-47.3%); proportion of eligible users who used EMBED, 58.3% (50.9-65.8%); time spent on EMBED, 29.0 seconds (20.4-37.7 seconds); proportion of buprenorphine orders placed through EMBED, 6.5% (3.4-9.6%); and task completion, 13.8% (8.9-18.7%) for buprenorphine order/prescription.A measurement science framework informed by partner co-design was a feasible approach to develop measures to guide CDS improvement. Subsequent research could adapt this approach to evaluate other CDS applications.