Redesigning intensive care unit flow using variability management to improve access and safety

Novel Dynamic Prediction of Daily Patient Discharge in Acute and Critical Care

Determining when a patient can be discharged from a care setting is critical to optimize the utilization and delivery of timely care. Furthermore, timely discharge can lead to better clinical outcomes by effectively mitigating the prolonged length of stay in a care environment. This paper presents a novel algorithm for the prediction of likelihood of patient discharge within the next 24 or 48 hours from acute or critical care environments on a daily basis. Continuous patient monitoring and health data obtained from acute hospital at home environment (n=303 patients) and a critical care unit environment (n=9,520 patients) are retrospectively used to train, validate and test numerous machine learning models for dynamic daily predictions of patients discharge. In the acute hospital at home environment, the area under the receiver operating characteristic (AUROC) curve performance of a top XGBoost model was 0.816 ± 0.025 and 0.758 ± 0.029 for daily discharge prediction within 24 hours and 48 hours respectively. Similar independent prediction models from the critical care environment resulted in relatively a lower AUROC for likewise predicting daily patient discharge. Overall, the results demonstrate the efficacy and utility of our novel algorithm for dynamic predictions of daily patient discharge in both acute- and critical care healthcare settings.

Enhancing the value to users of machine learning-based clinical decision support tools: A framework for iterative, collaborative development and implementation

BACKGROUND

Health care organizations are integrating a variety of machine learning (ML)-based clinical decision support (CDS) tools into their operations, but practitioners lack clear guidance regarding how to implement these tools so that they assist end users in their work.

PURPOSE

We designed this study to identify how health care organizations can facilitate collaborative development of ML-based CDS tools to enhance their value for health care delivery in real-world settings.

METHODOLOGY/APPROACH

We utilized qualitative methods, including 37 interviews in a large, multispecialty health system that developed and implemented two operational ML-based CDS tools in two of its hospital sites. We performed thematic analyses to inform presentation of an explanatory framework and recommendations.

RESULTS

We found that ML-based CDS tool development and implementation into clinical workflows proceeded in four phases: iterative solution coidentification, iterative coengagement, iterative coapplication, and iterative corefinement. Each phase is characterized by a collaborative back-and-forth process between the technology's developers and users, through which both users' activities and the technology itself are transformed.

CONCLUSION

Health care organizations that anticipate iterative collaboration to be an integral aspect of their ML-based CDS tools' development and implementation process may have more success in deploying ML-based CDS tools that assist end users in their work than organizations that expect a traditional technology innovation process.

PRACTICE IMPLICATIONS

Managers developing and implementing ML-based CDS tools should frame the work as a collaborative learning opportunity for both users and the technology itself and should solicit constructive feedback from users on potential changes to the technology, in addition to potential changes to user workflows, in an ongoing, iterative manner.

Huddles and their effectiveness at the frontlines of clinical care: a scoping review

BACKGROUND

Brief, stand-up meetings known as huddles may improve clinical care, but knowledge about huddle implementation and effectiveness at the frontlines is fragmented and setting specific. This work provides a comprehensive overview of huddles used in diverse health care settings, examines the empirical support for huddle effectiveness, and identifies knowledge gaps and opportunities for future research.

METHODS

A scoping review was completed by searching the databases PubMed, EBSCOhost, ProQuest, and OvidSP for studies published in English from inception to May 31, 2019. Eligible studies described huddles that (1) took place in a clinical or medical setting providing health care patient services, (2) included frontline staff members, (3) were used to improve care quality, and (4) were studied empirically. Two reviewers independently screened abstracts and full texts; seven reviewers independently abstracted data from full texts.

RESULTS

Of 2,185 identified studies, 158 met inclusion criteria. The majority (67.7%) of studies described huddles used to improve team communication, collaboration, and/or coordination. Huddles positively impacted team process outcomes in 67.7% of studies, including improvements in efficiency, process-based functioning, and communication across clinical roles (64.4%); situational awareness and staff perceptions of safety and safety climate (44.6%); and staff satisfaction and engagement (29.7%). Almost half of studies (44.3%) reported huddles positively impacting clinical care outcomes such as patients receiving timely and/or evidence-based assessments and care (31.4%); decreased medical errors and adverse drug events (24.3%); and decreased rates of other negative outcomes (20.0%).

DISCUSSION

Huddles involving frontline staff are an increasingly prevalent practice across diverse health care settings. Huddles are generally interdisciplinary and aimed at improving team communication, collaboration, and/or coordination. Data from the scoping review point to the effectiveness of huddles at improving work and team process outcomes and indicate the positive impact of huddles can extend beyond processes to include improvements in clinical outcomes.

STUDY REGISTRATION

This scoping review was registered with the Open Science Framework on 18 January 2019 ( https://osf.io/bdj2x/ ).

Sample sizes for surveillance of S. aureus transmission to monitor effectiveness and provide feedback on intraoperative infection control including for COVID-19

Reductions in perioperative surgical site infections are obtained by a multifaceted approach including patient decolonization, hand hygiene, and hub disinfection, and environmental cleaning. Associated surveillance of S. aureus transmission quantifies the effectiveness of the basic measures to prevent the transmission to patients and clinicians of pathogenic bacteria and viruses, including Coronavirus Disease 2019 (COVID-19). To measure transmission, the observational units are pairs of successive surgical cases in the same operating room on the same day. We evaluated appropriate sample sizes and strategies for measuring transmission.

There was absence of serial correlation among observed counts of transmitted isolates within each of several periods (all P ≥.18). Similarly, observing transmission within or between cases of a pair did not increase the probability that the next sampled pair of cases also had observed transmission (all P ≥.23).

Most pairs of cases had no detected transmitted isolates. Also, although transmission (yes/no) was associated with surgical site infection (P =.004), among cases with transmission, there was no detected dose response between counts of transmitted isolates and probability of infection (P =.25).

The first of a fixed series of tests is to use the binomial test to compare the proportion of pairs of cases with S. aureus transmission to an acceptable threshold. An appropriate sample size for this screening is N =25 pairs. If significant, more samples are obtained while additional measures are implemented to reduce transmission and infections. Subsequent sampling is done to evaluate effectiveness. The two independent binomial proportions are compared using Boschloo's exact test. The total sample size for the 1^st and 2^nd stage is N =100 pairs.

Because S. aureus transmission is invisible without testing, when choosing what population(s) to screen for surveillance, another endpoint needs to be used (e.g., infections). Only 10/298 combinations of specialty and operating room were relatively common (≥1.0% of cases) and had expected incidence ≥0.20 infections per 8 hours of sampled cases. The 10 combinations encompassed ≅17% of cases, showing the value of targeting surveillance of transmission to a few combinations of specialties and rooms.

In conclusion, we created a sampling protocol and appropriate sample sizes for using S. aureus transmission within and between pairs of successive cases in the same operating room, the purpose being to monitor the quality of prevention of intraoperative spread of pathogenic bacteria and viruses.

How improving access times had unforeseen consequences: a case study in a Dutch hospital

OBJECTIVES

To investigate the consequences of increasing capacity to reduce access times, and to explore how patient waiting times and use of physical capacity were influenced by variability.

DESIGN

A retrospective case study that combines both primary and secondary data. Secondary data were retrieved from a hospital database to establish inflow and outflow of patients, utilisation of resources and available capacity, realised access times and the weekly number of new patients seen over 1 year. Primary data consisted of field notes, onsite visits and observations, and semistructured interviews.

SETTING

A secondary care facility, that is, a rheumatology department, in a large Dutch hospital.

PARTICIPANTS

Analyses are based on secondary patient data from the hospital database, and the responses of the interviews with physicians, nurses and Lean Six Sigma project leaders.

RESULTS

The study shows that artificial variability was increased by managerial decisions to add capacity and to allow an increased inflow of new patients. This, in turn, resulted in undesirable and significant fluctuations in access times. We argue that we witnessed a new multiplier effect that typifies the fluctuations.

CONCLUSIONS

Adding capacity resources to reduce access times might appear an obvious and effective solution. However, the outcomes were less straightforward than expected, and even led to new artificial variability. The study reveals a phenomenon that is specific to service environments, and especially healthcare, and has detrimental consequences for access times.

Lean readiness within emergency departments: a conceptual framework

Purpose

The purpose of this paper is to develop a framework to assess the lean readiness within emergency departments (EDs) and identify the key quality practices deemed essential for lean system (LS) implementation.

Design/methodology/approach

An extensive review of the lean healthcare literature was conducted, including LS implementation within the healthcare sector (both generally and in EDs), best ED quality practices, essential factors for LS implementation within healthcare and lean readiness assessment frameworks. The authors identified six main categories from a literature review (top management and leadership, human resources, patient relations, supplier relations, processes and continuous improvement (CI)), and validated these based on experts’ opinion.

Findings

Several factors were identified as crucial for EDs, including top management and leadership, human resources, patient relations, supplier relations, processes and CI.

Research limitations/implications

The framework has not yet been tested, which prevents the author from declaring it fit for EDs.

Practical implications

This framework will help ED managers determine the factors that will enable/hinder the implementation of LSs within their premises.

Originality/value

To the author’s knowledge, this is the first lean readiness assessment framework for EDs and one of the few lean readiness assessment frameworks in the literature.

Improving decision making in acute healthcare through implementation of an intensive care unit (ICU) intervention in Australia: a multimethod study

OBJECTIVE

To evaluate the implementation of an intensive care unit (ICU) intervention designed to establish rules for making ICU decisions about postsurgery beds.

DESIGN

Preintervention/postintervention case study using a multimethod approach, involving two phases of staff interviews, process mapping and collection of administrative data.

SETTING

ICU in a 700-bed regional tertiary care hospital in Australia.

PARTICIPANTS

31 interview participants. Phases 1 and 2 participants drawn from three groups of staff: bedside nursing staff in the ICU, ICU specialist doctors and senior management staff involved in oversight of ICU operations. Phase 2 included an additional participant group: staff from surgery and emergency departments.

INTERVENTION

Implementation of an ICU escalation plan and introduction of a multidisciplinary morning meeting to determine ICU bed status in accordance with the plan.

MAIN OUTCOME MEASURES

Interview data consisted of preintervention staff perceptions of ICU workplace cohesiveness with bed pressure, and postintervention staff perceptions of the escalation plan and ICU performance. Administrative data consisted of bed status (red, amber or green), monthly number of planned elective surgeries requiring an ICU bed and monthly number of elective surgeries cancelled due to unavailability of ICU beds.

RESULTS

Improved internal communication, decision making and cohesion within the ICU and better coordination between ICU and other hospital departments. Significant reduction in elective surgeries cancelled due to unavailability of ICU beds, χ²₁=24.9, p<0.0001.

CONCLUSIONS

By establishing rules for decision making around ICU bed allocation, the intervention improved internal professional relationships within the ICU as well as between the ICU and external departments and reduced the number of elective surgeries cancelled.

Hospital-Level Changes in Adult ICU Bed Supply in the United States

OBJECTIVES

Although the number of intensive care beds in the United States is increasing, little is known about the hospitals responsible for this growth. We sought to better characterize national growth in intensive care beds by identifying hospital-level factors associated with increasing numbers of intensive care beds over time.

DESIGN

We performed a repeated-measures time series analysis of hospital-level intensive care bed supply using data from Centers for Medicare and Medicaid Services.

SETTING

All United States acute care hospitals with adult intensive care beds over the years 1996-2011.

PATIENTS

None.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

We described the number of beds, teaching status, ownership, intensive care occupancy, and urbanicity for each hospital in each year of the study. We then examined the relationship between increasing intensive care beds and these characteristics, controlling for other factors. The study included 4,457 hospitals and 55,865 hospital-years. Overall, the majority of intensive care bed growth occurred in teaching hospitals (net, +13,471 beds; 72.1% of total growth), hospitals with 250 or more beds (net, +18,327 beds; 91.8% of total growth), and hospitals in the highest quartile of occupancy (net, +10,157 beds; 54.0% of total growth). In a longitudinal multivariable model, larger hospital size, teaching status, and high intensive care occupancy were associated with subsequent-year growth. Furthermore, the effects of hospital size and teaching status were modified by occupancy: the greatest odds of increasing ICU beds were in hospitals with 500 or more beds in the highest quartile of occupancy (adjusted odds ratio, 18.9; 95% CI, 14.0-25.5; p < 0.01) and large teaching hospitals in the highest quartile of occupancy (adjusted odds ratio, 7.3; 95% CI, 5.3-9.9; p < 0.01).

CONCLUSIONS

Increasingly, intensive care bed expansion in the United States is occurring in larger hospitals and teaching centers, particularly following a year with high ICU occupancy.

Real-time prediction of inpatient length of stay for discharge prioritization

OBJECTIVE

Hospitals are challenged to provide timely patient care while maintaining high resource utilization. This has prompted hospital initiatives to increase patient flow and minimize nonvalue added care time. Real-time demand capacity management (RTDC) is one such initiative whereby clinicians convene each morning to predict patients able to leave the same day and prioritize their remaining tasks for early discharge. Our objective is to automate and improve these discharge predictions by applying supervised machine learning methods to readily available health information.

MATERIALS AND METHODS

The authors use supervised machine learning methods to predict patients' likelihood of discharge by 2 p.m. and by midnight each day for an inpatient medical unit. Using data collected over 8000 patient stays and 20 000 patient days, the predictive performance of the model is compared to clinicians using sensitivity, specificity, Youden's Index (i.e., sensitivity + specificity - 1), and aggregate accuracy measures.

RESULTS

The model compared to clinician predictions demonstrated significantly higher sensitivity (P < .01), lower specificity (P < .01), and a comparable Youden Index (P > .10). Early discharges were less predictable than midnight discharges. The model was more accurate than clinicians in predicting the total number of daily discharges and capable of ranking patients closest to future discharge.

CONCLUSIONS

There is potential to use readily available health information to predict daily patient discharges with accuracies comparable to clinician predictions. This approach may be used to automate and support daily RTDC predictions aimed at improving patient flow.

Solving the worldwide emergency department crowding problem - what can we learn from an Israeli ED?

ED crowding is a prevalent and important issue facing hospitals in Israel and around the world, including North and South America, Europe, Australia, Asia and Africa. ED crowding is associated with poorer quality of care and poorer health outcomes, along with extended waits for care. Crowding is caused by a periodic mismatch between the supply of ED and hospital resources and the demand for patient care. In a recent article in the Israel Journal of Health Policy Research, Bashkin et al. present an Ishikawa diagram describing several factors related to longer length of stay (LOS), and higher levels of ED crowding, including management, process, environmental, human factors, and resource issues. Several solutions exist to reduce ED crowding, which involve addressing several of the issues identified by Bashkin et al. This includes reducing the demand for and variation in care, and better matching the supply of resources to demands in care in real time. However, what is needed to reduce crowding is an institutional imperative from senior leadership, implemented by engaged ED and hospital leadership with multi-disciplinary cross-unit collaboration, sufficient resources to implement effective interventions, access to data, and a sustained commitment over time. This may move the culture of a hospital to facilitate improved flow within and across units and ultimately improve quality and safety over the long-term.

Relationship between ICU bed availability, ICU readmission, and cardiac arrest in the general wards

OBJECTIVE

The decision to admit a patient to the ICU is complex, reflecting patient factors and available resources. Previous work has shown that ICU census does not impact mortality of patients admitted to the ICU. However, the effect of ICU bed availability on patients outside the ICU is unknown. We sought to determine the association between ICU bed availability, ICU readmissions, and ward cardiac arrests.

DESIGN

In this observational study using data collected between 2009 and 2011, rates of ICU readmission and ward cardiac arrest were determined per 12-hour shift. The relationship between these rates and the number of available ICU beds at the start of each shift (accounting for census and nursing capacity) was investigated. Grouped logistic regression was used to adjust for potential confounders.

SETTING

Five specialized adult ICUs comprising 63 adult ICU beds in an academic medical center.

PATIENTS

Any patient admitted to a non-ICU inpatient unit was counted in the ward census and considered at risk for ward cardiac arrest. Patients discharged from an ICU were considered at risk for ICU readmission.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Data were available for 2,086 of 2,190 shifts. The odds of ICU readmission increased with each decrease in the overall number of available ICU beds (odds ratio = 1.06; 95% CI, 1.00-1.12; p = 0.03), with a similar but not statistically significant association demonstrated in ward cardiac arrest rate (odds ratio = 1.06; 95% CI, 0.98-1.14; p = 0.16). In subgroup analysis, the odds of ward cardiac arrest increased with each decrease in the number of medical ICU beds available (odds ratio = 1.26; 95% CI, 1.06-1.49; p = 0.01).

CONCLUSIONS

Reduced ICU bed availability is associated with increased rates of ICU readmission and ward cardiac arrest. This suggests that systemic factors are associated with patient outcomes, and flexible critical care resources may be needed when demand is high.

A daily huddle facilitates patient transports from a neonatal intensive care unit

To improve hospital access for expectant women and newborns in the state of Maryland, a quality improvement team reviewed the patient flow characteristics of our neonatal intensive care unit. We identified inefficiencies in patient discharges, including delays in patient transports. Several patient transport delays were caused by late preparation and delivery of the patient transfer summary.

Baseline data collection revealed that transfer summaries were prepared on-time by the resident or nurse practitioner only 41% of the time on average, while the same transfer summaries were signed on-time by the neonatologist 5% of the time on average. Our aim was to improve the rate of on-time transfer summaries to 50% over a four month time period.

We performed two PDSA cycles based on feedback from our quality improvement team. In the first cycle, we instituted a daily huddle to increase opportunities for communication about patient transports. In the second cycle, we increased computer access for residents and nurse practitioners preparing the transfer summaries. The on-time summary preparation by residents/nurse practitioners improved to an average of 72% over a nine month period. The same summaries were signed on-time by a neonatologist 26% of the time on average over a nine month period.

In conclusion, institution of a daily huddle combined with augmented computer resources significantly increased the percentage of on-time transfer summaries. Current data show a trend toward improved ability to accept patient referrals. Further data collection and analysis is needed to determine the impact of these interventions on access to hospital care for expectant women and newborns in our state.

The safe patient flow initiative: a collaborative quality improvement journey at Yale-New Haven Hospital

BACKGROUND

Yale-New Haven Hospital (YNHH) began a successful journey to achieve safe patient flow in fiscal year (FY) 2008 (October 1, 2007-September 30, 2008). The 966-bed (now 1,541-bed) academic medical center faced several challenges, including overcrowding in the Adult Emergency Department (ED); delays in the postanesthesia care unit, which affected the flow of patients through the operating rooms; pinched capacity during the central part of the day; and a lack of interdependent institutionwide coordination of patients.

METHODS

The Safe Patient Flow Steering Committee oversaw improvement efforts, most of which were implemented in FY 2009 (October 2008-September 2009), through a cascade of operational meetings. Process changes were made in various departments, such as the Adult ED, Physicians/Providers, and the Bed Management Department. Organizationwide method changes involved standardizing the discharge process, using status boards for visual control, and improving accuracy and timeliness of data entry.

RESULTS

Between FY 2008 and FY 2011, YNHH experienced an 84% improvement in discharges by 11:00 A.M. The average length of stay decreased from 5.23 to 5.05 days, thereby accommodating an additional 45 inpatients on a daily basis, contributing to YNHH's positive operating margin amid increasing volume and overall decreasing inpatient length of stay.

CONCLUSIONS

YNHH improved clinical, operational, and financial outcomes by embracing five key components of demand capacity management: real-time communication, inter/intradepartmental and interdisciplinary collaboration, staff empowerment, standardization of best practices, and institutional memory.

Quality improvement initiative to reduce serious safety events and improve patient safety culture

BACKGROUND AND OBJECTIVE

Many thousands of patients die every year in the United States as a result of serious and largely preventable safety events or medical errors. Safety events are common in hospitalized children. We conducted a quality improvement initiative to implement cultural and system changes with the goal of reducing serious safety events (SSEs) by 80% within 4 years at our large, urban pediatric hospital.

METHODS

A multidisciplinary SSE reduction team reviewed the safety literature, examined recent SSEs, interviewed internal leaders, and visited other leading organizations. Senior hospital leaders provided oversight, monitored progress, and helped to overcome barriers. Interventions focused on: (1) error prevention; (2) restructuring patient safety governance; (3) a new root cause analysis process and a common cause database; (4) a highly visible lessons learned program; and (5) specific tactical interventions for high-risk areas. Our outcome measures were the rate of SSEs and the change in patient safety culture.

RESULTS

SSEs per 10000 adjusted patient-days decreased from a mean of 0.9 at baseline to 0.3 (P < .0001). The days between SSEs increased from a mean of 19.4 at baseline to 55.2 (P < .0001). After a worsening of patient safety culture outcomes in the first year of intervention, significant improvements were observed between 2007 and 2009.

CONCLUSIONS

Our multifaceted approach was associated with a significant and sustained reduction of SSEs and improvements in patient safety culture. Multisite studies are needed to better understand contextual factors and the significance of specific interventions.

Managing ICU throughput and understanding ICU census

PURPOSE OF REVIEW

Traditionally, hospitals have coped with chronically high ICU census by building more ICU beds, but this strategy is unlikely to be tenable under future financial models. Therefore, ICUs need additional tools to manage census, inflow, and throughput.

RECENT FINDINGS

Higher ICU census, without compensatory surges in nursing capacity, is associated with several adverse effects on patients and providers, but its relationship to mortality is uncertain. Providers also discharge patients more aggressively during times of high census. Little's Law (L = λ W), a cornerstone of queuing theory, provides an eminently practical basis for managing ICU census and throughput. One target for improving throughput is minimizing process steps that are without value to the patient, e.g., waiting for a bed at ICU discharge. Larger gains in ICU throughput can be found in ICU quality improvement. For example, spontaneous breathing trials, daily wake-ups, and early physical/occupational therapy programmes are all likely to improve throughput by reducing ICU length of stay. The magnitude of these interventions' effects on ICU census can be startling.

SUMMARY

ICUs should actively manage throughput and census. Operations management tools such as Little's Law can provide practical guidance about the relationship between census, throughput, and patient demand. Standard ICU quality improvement techniques can meaningfully affect both ICU census and throughput.

Evaluating large-scale health programmes at a district level in resource-limited countries

Recent experience in evaluating large-scale global health programmes has highlighted the need to consider contextual differences between sites implementing the same intervention. Traditional randomized controlled trials are ill-suited for this purpose, as they are designed to identify whether an intervention works, not how, when and why it works. In this paper we review several evaluation designs that attempt to account for contextual factors that contribute to intervention effectiveness. Using these designs as a base, we propose a set of principles that may help to capture information on context. Finally, we propose a tool, called a driver diagram, traditionally used in implementation that would allow evaluators to systematically monitor changing dynamics in project implementation and identify contextual variation across sites. We describe an implementation-related example from South Africa to underline the strengths of the tool. If used across multiple sites and multiple projects, the resulting driver diagrams could be pooled together to form a generalized theory for how, when and why a widely-used intervention works. Mechanisms similar to the driver diagram are urgently needed to complement existing evaluations of large-scale implementation efforts.

Using real-time demand capacity management to improve hospitalwide patient flow

BACKGROUND

The Joint Commission's accreditation standard on managing patient flow, effective January 2005, served as a call to action for hospitals, yet many hospitals still lack the processes and structures to admit or transfer patients to an inpatient bed on a timely basis. In 2007 the University of Pittsburgh Medical Center (UPMC) at Shadyside, a 526-bed tertiary care hospital, began testing and implementing real-time demand capacity management (RTDC) at an initial pilot site. The hospital had identified improved patient flow as a strategic goal in 2002, but a series of patient flow projects failed to result in improvement. IMPLEMENTING RTDC: Standard processes for the four RTDC steps-Predicting Capacity, Predicting Demand, Developing a Plan, and Evaluating a Plan--and standard structures for unit bed huddles and the hospital bed meetings were developed. The neurosurgery (NS) service line's ICU and stepdown unit were designated as the first pilot sites, but work was quickly spread to other units.

RESULTS

Improvements were achieved and have been sustained through early 2011 for all measures, including (1) the unit-based reliability of discharge predictions; (2) overnight holds in the postanesthesia care unit, a problem eliminated two months after RTDC work began; (3) the percentage of patients who left without being seen (LWBS), routinely < 0.5% by May 2008; (5) the emergency department median length of stay for admitted patients, routinely < 4 hours after March 2008; and (6) aggregate length of stay (ALOS), generally maintained at < 5.75 days.

CONCLUSIONS

RTDC represents a promising approach to improving hospitalwide patient flow. Its four steps, integrated into current bed management processes, are not an add-on to the work needing to be accomplished everyday.