Talk to the Veterans Crisis Line now
An official website of the United States government
CRE 12-321
Cognitive Support Informatics for Nurse Medication Stewardship
Frank A Drews, MS PhD VA Salt Lake City Health Care System, Salt Lake City, UT Salt Lake City, UT Funding Period: November 2013 - April 2018 |
|
BACKGROUND/RATIONALE:
Acute change in mental status (AMSC) is a common clinical problem. Coordination between physicians and nurses are critical in prevention, monitoring, and patient safety. Quality of care requires effective coordination between clinical roles. Improving cognitive support for clinical team coordination is the goal of this work. OBJECTIVE(S): AIM 1: Characterize nurses' medication management information needs, documentation and communication processes associated with identifying and monitoring AMSC in inpatient settings. AIM 2: Create a predictive model that integrates AMSC predictors and text descriptors with staffing data to support therapeutic decision-making, resource planning and collaboration AIM 3: Design and test three cognitive support interventions: 1) Info-button decision-support provided through BCMA; 2) alerts imbedded in white board display; 3) display of documented mental status changes from the narrative electronic notes. AIM 4: Implement and assess the impact of a comprehensive program for AMSC. METHODS: AIM 1: Study 1: A national phone survey of 58 Nurse Managers of inpatient medical wards were randomly selected within all VA VISN (regional areas) to assess ward-level documentation practices and AMSC policies. Study 2: Ethnographic observations were conducted in 3 inpatient floors (total=28) on the flow of mental status information, clinical decisions for patient care, and clinician's handoffs for older adults. Study 3:Identification of mental models, information needs and decision tasks of 11 physicians regarding care of patients with AMSC using Cognitive Task Analyses and qualitative analysis. Study 4: Randomly selected 30 patients > 65 years of age from 1 year of inpatient admissions to a SLC VA medical ward and evaluated documentation practices across nurses and physicians for AMSC. Study 5: Three expert usability experts evaluated the VA eMAR/BCMA system using Nielson's heuristic taxonomy. AIM 2 Study 1: Developed and tested a quantitative predictive model for daily risk of delirium from structured data predicting VINCI inpatient patients with orders for sitters and restraints. Study 2: Validated the risk model in Study 1 using CAM (Confusion Assessment Method) assessed on 125 patients in a medical ward in SLC VA. Study 3: Developed and tested an NLP model using topic-modeling to mental status terms associated with AMSC using progress notes in VINCI patient data. Study 4: Validated NLP model using EPRP reviewers for 2017 charts assessing the VA delirium-based performance measure as gold standard. AIM 3: Study 1: Designed a CPRS clinical reminder that included physician assessment of risk at admission, delirium risk score, and an order set for increased nurse monitoring. Study 2: Designed a dashboard for physicians that provides decision support in patients at risk for delirium and tested for usability, usefulness and impact on physician decision-making. Study 3: An initial pilot survey of 17 physician's expectations and satisfaction regarding communication from nurses on mental status changes was conducted with SLC resident teams. Study 4: Design decision support for nurses in CPRS to support nurse-physician communication that integrates physician notes, predictive risk scores, and nurses' assessments. AIM 4: Study 1: The full implementation of the program is ongoing and under the direction of the physician and nursing staff in SLC. They will direct the evaluation of the work. Study 2: The implementation of an NLP automated program to support national EPRP performance assessment of delirium has been turned over to the directors of EPRP and the VA Office of Quality Assessment. FINDINGS/RESULTS: AIM 1 Study 1: Four themes from the interviews emerged: 1) "Fuzzy Concepts"- mental status terminology is often colloquial, vague, and imprecise; 2) "Grey Data" - data about mental status is often hidden or hard to find; 3) "Context is Critical"- mental status assessment and documentation depends on the patient situation; and 4) "Competing Goals"- nurses will document the mental status related to behavioral issues over other types of symptoms. Structured data found: 1) Assessment for mental status limited to orientation (93%), and few identify risk (5%); 2) Preventative interventions were rarely used (10%). Sitters was the predominant intervention (88%); 3) formal stewardship, such as tracking ACMS, was rare (5%); and 4) formal CAM monitoring is rarely ordered (5%). Study 2: Key deficits in communication were noted, specifically issues with communicating risk (80% of patients at risk were not identified nor was there increased monitoring) poor communication (delays in AMSC in 75% of patients with changes), no causal attribution was noted for 73% of patients with AMSC. Study 3: 4 thematic areas emerged - all related to uncertainty: 1) Unavailable baseline information, 2) causal attribution is ambiguous; 3) information sources lack credibility, and 4) high perceived effort. Study 4: 11/30 of patients had some documentation referring to AMSC. Physicians and nurses agreed on 10/11. ICD9 codes only identified 3 of the 30. Nurses focused largely on orientation while physicians had significantly more terms related to AMSC. Study 5: 99 usability problems were identified with 15 rated as catastrophic. Situational awareness was affected at all levels for nurses caring for patients in the inpatient setting. AIM 2- Study 1: Accuracy (C statistic) maximized at 33 (C=73.1%). PPV was low (7.35%). Study 2: C statistic for the ROC curve = 0.801. Study 3: Three different topic-modeling methods (LDA and 2 ICD-based method) were tested. Keyword search method was highly specific but insufficiently sensitive (F-measure = 0.442). All 3-topic models had better recall but worse precision. LCD-2 had an F-measure of 0.677. 5/1000 topics found related to AMSC. Study 4: a: NLP more accurate than EPRP reviewers with "presence of delirium" had an AUC=89.3% and "assessment of delirium" had an AUC=93%. Results b: An NLP pipeline ontology, called "POETenceph" was developed to rank clinical notes on the evidence for delirium using our realist ontology of encephalopathy, POETenceph correctly classified 65% of the documents. AIM 3: Study 1: The assessment was inserted in physician's H&P as a structured variable and is in use at present. Study 2: A randomized trial compared regular CPRS access with dashboard display of delirium information using vignettes found that medical residents using the dashboard were significantly more likely to: 1) identify more risk factors (4.54 vs. 3.25, p = 0.123); 2) identify baseline mental status (93 % vs. 75% p=0.007); 3) have a more appropriate plan overall (p < 0.05) and express more confidence in their decision (p =. 0.04). Medication and baseline information was rated as the most useful. Study 3: An initial pilot survey of 17 physician's expectations regarding communication from nurses on mental status identified five areas identified as most important: a) Lab values (glucose, BMP), b) Vital Signs, c) relevant medications, d) Onset/timing, brief situation and baseline, e) symptoms, and f) oxygen saturation. On ratings scales of 1 (low) to 7 (high), they reported they had to often request more information (M=4.8), rated the appropriateness overall as high (M=5.6), as was satisfaction (M=5.8). Study 4: Design decision support for nurses in CPRS to support nurse-physician communication that integrates physician notes, predictive risk scores, and nurses' assessments. Results: Initial evaluation showed initial moderate ratings of satisfaction. Aim 4 Study 1: The full implementation of the program is ongoing and under the direction of the physician and nursing staff in SLC. They will direct the evaluation of the work. Study 2: The implementation of an NLP automated program to support national EPRP performance assessment of delirium has been turned over to the directors of EPRP and the VA Office of Quality Assessment. IMPACT: This study will significantly improve the care of patients with delirium by improving communication about risk and early identification of possible causes. We have developed an alert, documentation reminder dialogues, created a dashboard, and created an NLP module for EPRP performance measures. External Links for this ProjectNIH ReporterGrant Number: I01HX001154-01Link: https://reporter.nih.gov/project-details/8495033 Dimensions for VA Dimensions for VA is a web-based tool available to VA staff that enables detailed searches of published research and research projects. Learn more about Dimensions for VA. VA staff not currently on the VA network can access Dimensions by registering for an account using their VA email address. Search Dimensions for this project PUBLICATIONS:Journal Articles
DRA:
Aging, Older Veterans' Health and Care, Health Systems
DRE: Diagnosis, Treatment - Observational, Technology Development and Assessment Keywords: none MeSH Terms: none |
