What works in medication reconciliation: an on-treatment and site analysis of the MARQUIS2 study.

BACKGROUND: The second Multicenter Medication Reconciliation Quality Improvement Study demonstrated a marked reduction in medication discrepancies per patient. The aim of the current analysis was to determine the association of patient exposure to each system-level intervention and receipt of each patient-level intervention on these results. METHODS: This study was conducted at 17 North American Hospitals, the study period was 18 months per site, and sites typically adopted interventions after 2-5 months of preintervention data collection. We conducted an on-treatment analysis (ie, an evaluation of outcomes based on patient exposure) of system-level interventions, both at the category level and at the individual component level, based on monthly surveys of implementation site leads at each site (response rate 65%). We then conducted a similar analysis of patient-level interventions, as determined by study pharmacist review of documented activities in the medical record. We analysed the association of each intervention on the adjusted number of medication discrepancies per patient in admission and discharge orders, based on a random sample of up to 22 patients per month per site, using mixed-effects Poisson regression with hospital site as a random effect. We then used a generalised linear mixed-effects model (GLMM) decision tree to determine which patient-level interventions explained the most variance in discrepancy rates. RESULTS: Among 4947 patients, patient exposure to seven of the eight system-level component categories was associated with modest but significant reductions in discrepancy rates (adjusted rate ratios (ARR) 0.75-0.97), as were 15 of the 17 individual system-level intervention components, including hiring, reallocating and training personnel to take a best possible medication history (BPMH) and training personnel to perform discharge medication reconciliation and patient counselling. Receipt of five of seven patient-level interventions was independently associated with large reductions in discrepancy rates, including receipt of a BPMH in the emergency department (ED) by a trained clinician (ARR 0.40, 95% CI 0.37 to 0.43), admission medication reconciliation by a trained clinician (ARR 0.57, 95% CI 0.50 to 0.64) and discharge medication reconciliation by a trained clinician (ARR 0.64, 95% CI 0.57 to 0.73). In GLMM decision tree analyses, patients who received both a BPMH in the ED and discharge medication reconciliation by a trained clinician experienced the lowest discrepancy rates (0.08 per medication per patient). CONCLUSION AND RELEVANCE: Patient-level interventions most associated with reductions in discrepancies were receipt of a BPMH of admitted patients in the ED and admission and discharge medication reconciliation by a trained clinician. System-level interventions were associated with modest reduction in discrepancies for the average patient but are likely important to support patient-level interventions and may reach more patients. These findings can be used to help hospitals and health systems prioritise interventions to improve medication safety during care transitions.

Prospective, multi-site study of patient outcomes after implementation of the TREWS machine learning-based early warning system for sepsis.

Early recognition and treatment of sepsis are linked to improved patient outcomes. Machine learning-based early warning systems may reduce the time to recognition, but few systems have undergone clinical evaluation. In this prospective, multi-site cohort study, we examined the association between patient outcomes and provider interaction with a deployed sepsis alert system called the Targeted Real-time Early Warning System (TREWS). During the study, 590,736 patients were monitored by TREWS across five hospitals. We focused our analysis on 6,877 patients with sepsis who were identified by the alert before initiation of antibiotic therapy. Adjusting for patient presentation and severity, patients in this group whose alert was confirmed by a provider within 3 h of the alert had a reduced in-hospital mortality rate (3.3%, confidence interval (CI) 1.7, 5.1%, adjusted absolute reduction, and 18.7%, CI 9.4, 27.0%, adjusted relative reduction), organ failure and length of stay compared with patients whose alert was not confirmed by a provider within 3 h. Improvements in mortality rate (4.5%, CI 0.8, 8.3%, adjusted absolute reduction) and organ failure were larger among those patients who were additionally flagged as high risk. Our findings indicate that early warning systems have the potential to identify sepsis patients early and improve patient outcomes and that sepsis patients who would benefit the most from early treatment can be identified and prioritized at the time of the alert.

Factors driving provider adoption of the TREWS machine learning-based early warning system and its effects on sepsis treatment timing.

Machine learning-based clinical decision support tools for sepsis create opportunities to identify at-risk patients and initiate treatments at early time points, which is critical for improving sepsis outcomes. In view of the increasing use of such systems, better understanding of how they are adopted and used by healthcare providers is needed. Here, we analyzed provider interactions with a sepsis early detection tool (Targeted Real-time Early Warning System), which was deployed at five hospitals over a 2-year period. Among 9,805 retrospectively identified sepsis cases, the early detection tool achieved high sensitivity (82% of sepsis cases were identified) and a high rate of adoption: 89% of all alerts by the system were evaluated by a physician or advanced practice provider and 38% of evaluated alerts were confirmed by a provider. Adjusting for patient presentation and severity, patients with sepsis whose alert was confirmed by a provider within 3 h had a 1.85-h (95% CI 1.66-2.00) reduction in median time to first antibiotic order compared to patients with sepsis whose alert was either dismissed, confirmed more than 3 h after the alert or never addressed in the system. Finally, we found that emergency department providers and providers who had previous interactions with an alert were more likely to interact with alerts, as well as to confirm alerts on retrospectively identified patients with sepsis. Beyond efforts to improve the performance of early warning systems, efforts to improve adoption are essential to their clinical impact and should focus on understanding providers' knowledge of, experience with and attitudes toward such systems.

Human-machine teaming is key to AI adoption: clinicians' experiences with a deployed machine learning system.

While a growing number of machine learning (ML) systems have been deployed in clinical settings with the promise of improving patient care, many have struggled to gain adoption and realize this promise. Based on a qualitative analysis of coded interviews with clinicians who use an ML-based system for sepsis, we found that, rather than viewing the system as a surrogate for their clinical judgment, clinicians perceived themselves as partnering with the technology. Our findings suggest that, even without a deep understanding of machine learning, clinicians can build trust with an ML system through experience, expert endorsement and validation, and systems designed to accommodate clinicians' autonomy and support them across their entire workflow.

Effects of a refined evidence-based toolkit and mentored implementation on medication reconciliation at 18 hospitals: results of the MARQUIS2 study.

BACKGROUND: The first Multicenter Medication Reconciliation Quality Improvement (QI) Study (MARQUIS1) demonstrated that mentored implementation of a medication reconciliation best practices toolkit decreased total unintentional medication discrepancies in five hospitals, but results varied by site. The objective of this study was to determine the effects of a refined toolkit on a larger group of hospitals. METHODS: We conducted a pragmatic quality improvement study (MARQUIS2) at 18 North American hospitals or hospital systems from 2016 to 2018. Incorporating lessons learnt from MARQUIS1, we implemented a refined toolkit, offering 17 system-level and 6 patient-level interventions. One of eight physician mentors coached each site via monthly calls and performed one to two site visits. The primary outcome was number of unintentional medication discrepancies in admission or discharge orders per patient. Time series analysis used multivariable Poisson regression. RESULTS: A total of 4947 patients were sampled, including 1229 patients preimplementation and 3718 patients postimplementation. Both the number of system-level interventions adopted per site and the proportion of patients receiving patient-level interventions increased over time. During the intervention, patients experienced a steady decline in their medication discrepancy rate from 2.85 discrepancies per patient to 0.98 discrepancies per patient. An interrupted time series analysis of the 17 sites with sufficient data for analysis showed the intervention was associated with a 5% relative decrease in discrepancies per month over baseline temporal trends (adjusted incidence rate ratio: 0.95, 95% CI 0.93 to 0.97, p<0.001). Receipt of patient-level interventions was associated with decreased discrepancy rates, and these associations increased over time as sites adopted more system-level interventions. CONCLUSION: A multicentre medication reconciliation QI initiative using mentored implementation of a refined best practices toolkit, including patient-level and system-level interventions, was associated with a substantial decrease in unintentional medication discrepancies over time. Future efforts should focus on sustainability and spread.

Design of MARQUIS2: study protocol for a mentored implementation study of an evidence-based toolkit to improve patient safety through medication reconciliation.

BACKGROUND: The first Multi-center Medication Reconciliation Quality Improvement Study (MARQUIS1) demonstrated that implementation of a medication reconciliation best practices toolkit decreased total unintentional medication discrepancies in five hospitals. We sought to implement the MARQUIS toolkit in more diverse hospitals, incorporating lessons learned from MARQUIS1. METHODS: MARQUIS2 is a pragmatic, mentored implementation QI study which collected clinical and implementation outcomes. Sites implemented a revised toolkit, which included interventions from these domains: 1) best possible medication history (BPMH)-taking; 2) discharge medication reconciliation and patient/caregiver counseling; 3) identifying and defining clinician roles and responsibilities; 4) risk stratification; 5) health information technology improvements; 6) improved access to medication sources; 7) identification and correction of real-time discrepancies; and, 8) stakeholder engagement. Eight hospitalists mentored the sites via one site visit and monthly phone calls over the 18-month intervention period. Each site's local QI team assessed opportunities to improve, implemented at least one of the 17 toolkit components, and accessed a variety of resources (e.g. implementation manual, webinars, and workshops). Outcomes to be assessed will include unintentional medication discrepancies per patient. DISCUSSION: A mentored multi-center medication reconciliation QI initiative using a best practices toolkit was successfully implemented across 18 medical centers. The 18 participating sites varied in size, teaching status, location, and electronic health record (EHR) platform. We introduce barriers to implementation and lessons learned from MARQUIS1, such as the importance of utilizing dedicated, trained medication history takers, simple EHR solutions, clarifying roles and responsibilities, and the input of patients and families when improving medication reconciliation.

What happens to the medication regimens of older adults during and after an acute hospitalization?

BACKGROUND: Hospitalized elderly patients are at risk for medication errors and nonadherence when discharged home. OBJECTIVE: To describe how older patients' medications change during and after hospitalizations and what patients ultimately take after discharge. METHODS: We conducted an observational cohort study of 95 patients aged 65 years and older admitted to Johns Hopkins Bayview Medical Center in 2007. Inclusion criteria included admissions longer than 24 hours and discharge to home. Medication lists from three periods were recorded: prehospitalization, day of discharge, and 3 days after discharge. In comparing lists, we characterized: new and discontinued medications, changes in dosage, and changes in frequency. RESULTS: Before admission, patients were taking a total of 701 medications (mean, 7 per patient). Upon discharge, 192 new medicines were started (2.0 per patient), 76 discontinued (0.8 per patient), 67 changed in frequency, (0.7 per patient), and 45 changed in dosage (0.5 per patient). Antibiotics and antihypertensives were the most commonly prescribed new medications. Antihypertensives were also most likely to be discontinued. At day 3 after discharge, patients were adherent with 98% (763/778) of medications. However, 25% of antihypertensives and 88% analgesics discontinued by hospitalists on the day of discharge were reinitiated by patients upon their return home. CONCLUSIONS: During hospitalizations, medications of older adults change substantially. Despite clear medication reconciliation efforts in the hospital environment, medication errors occur upon discharge to home. Because current standards are yielding suboptimal results, alternate methodologies for promoting medication adherence should also be considered, developed, and studied for effectiveness.

A quality improvement intervention to facilitate the transition of older adults from three hospitals back to their homes.

OBJECTIVES: To study the feasibility and effectiveness of a discharge planning intervention. DESIGN: Quasi-experimental pre-post study design. SETTING: General medicine wards at three hospitals: an academic medical center, a community teaching hospital, and a community-based nonteaching hospital. PARTICIPANTS: All patients aged 65 and older admitted to the hospitalist services. INTERVENTION: The intervention toolkit had five core elements: admission form with geriatric cues, facsimile to the primary care provider, interdisciplinary worksheet to identify barriers to discharge, pharmacist-physician collaborative medication reconciliation, and predischarge planning appointments. MEASUREMENTS: Thirty-day readmission and return to emergency department rates and patient satisfaction with discharge. Odds ratios were determined, and site effects were examined accordig to interaction terms and Breslow Day statistics. RESULTS: Two hundred thirty-seven patients were followed during the preintervention period, and 185 were exposed to the intervention. Patients characteristics were similar across the two time periods. The proportion of patients with high-quality transitions home, measured according to Coleman's Care Transition Measures, increased from 68% to 89% (odds ratio (OR)=3.49, 95% confidence interval (CI)=2.06-5.92). Return to the emergency department within 3 days of discharge was lower in the intervention period (10% vs 3%, OR=0.25, 95% CI=0.10-0.62). At 30 days, there was a lower rate of readmission (22% vs 14%, OR=0.59, 95% CI=0.34-0.97) and fewer visits to the emergency department (21% vs 14%, OR=0.61, 95% CI=0.36-1.03) (P=.06). CONCLUSION: When hospitalized elderly patients are treated with consideration of their specific needs, healthcare outcomes can be improved.

A curriculum to teach internal medicine residents to perform house calls for older adults.

Physician house calls are an important but underused mode of delivering health care to a growing population of homebound elderly patients. One major barrier to internal medicine physicians making house calls is a lack of training in this setting. This article describes a needs assessment survey of residents from nine internal medicine residency programs for a house call curriculum that combines a longitudinal clinical experience with Internet-based learning. Implementation of the curriculum was begun in July 2006, and data will be collected and results evaluated for at least 2 years. Several educational outcomes from the intervention are anticipated, including increased learner knowledge of house call medicine, improved learner confidence in making house calls, and program director satisfaction with the curriculum. This early work lays the foundation for determining the effect of a carefully designed curriculum on the number of practicing internists with the skills, knowledge, and attitudes necessary to meet the growing need for physician house calls.