The Analytics Showcase and AI Showcase feature data-related projects across a wide variety of clinical, financial, operational, technical, AI, and leadership topics.
Data-driven innovation often consists of a multitude of smaller, agile projects. Attendees will have the chance to explore a wide variety of different projects—tailored to many different team member roles.
ML models and methods, which are becoming more complex and less explainable, come with a risk of further encoding systemic biases into many real-world systems. Our study fills in existing data gaps by imputing race/ethnicity data via the BISG methodology and measuring clinical risk via evaluation of HCC performance in tandem with our Social Risk Index model (SRI) across race and ethnicity.
Results
• Closed data gap for race/ethnicity data by average of 33% across clients using BISG model.
• Members with low HCC risk and high SRI had an ED visit rate comparable to members with high HCC risk.
• Black and Hispanic members are overrepresented in high tiers of HCC risk model.
Both providers and consumers of healthcare benefit from finding mechanisms to pay for healthcare services. We have enhanced an existing predictive model to increase the likelihood that we focus outreach those people most likely to benefit. We needed to evaluate and possibly rebuild the predictive model that drove outreach both because the model had not been refreshed in a couple years and because COVID-19 altered utilization, coverage, and payment patterns.
Results
• We significantly improved predictive accuracy.
• The new predictive model is more efficient and easier to understand.
• Predictive models should be reviewed periodically for improvements.
As part of a state Medicaid contract, AlohaCare, a community-led non-profit health plan, needed to develop a comprehensive risk model to qualify members for specific clinical programs. By evaluating 9 separate models that included state criteria and health/financial/socials risks, the organization developed and deployed a predictive model directly into their care management processes that identified the most appropriate clinical program and care manager for each member.
Results
• Improved care program assignment per the risk model selection process.
• Expected and unexpected relationships found between risk models.
• Multi-dimensional risk models can ensure that profiles are comprehensive.
Blessing Health System, an integrated care delivery system in Western Illinois, wanted to evaluate the effectiveness of its Diabetes focused population health program -Be Well With Diabetes (BWWD). They sought to determine if there was a meaningful reduction in HbA1c levels, ED visits and hospitalizations within this population.
Results
• The HbA1C reduced by 0.8 in patients belonging to BWWD group compared to matched patients.
• BWWD patients had 65% lower odds of ED Visits.
• BWWD patients also had 44% lower odds of Hospitalization.
• The analysis showcases the use of advanced techniques to help evaluate population health interventions.
Castell, a value-based healthcare company in the intermountain west, required a solution to reduce the prevalence of duplicate patients from multiple payors in their data warehouse. An identity matching algorithm was developed using NLP and simulated data to identify duplicate records.
Results
• Allows Castell to regularly identify duplicate records based on patient demographics at a system scale.
• A patient matching model can be effectively trained using simulated data.
• Demonstrated upstream sources of person duplication in our data pipeline.
Healthcare organizations share the challenge of managing health equity. ChristianaCare launched multiple initiatives aiming to evaluate and close the gaps in health outcome disparities. To better support these initiatives, ChristianaCare developed a ML based platform to quantify, visualize, and interpret health equity on multiple health outcomes. The platform has been actively utilized to support the ongoing initiatives and organizational strategic goals.
Results
• ML based algorithm and user-oriented visualizations were developed to evaluate health equity.
• Key areas were identified as area of focus.
• Multidisciplinary clinical teams were engaged to utilize the tools in health equity interventions.
Where should we focus efforts to improve health equity? Too frequently, we depend on anecdote or manual data exploration. CHNw employed augmented intelligence (AI) to evaluate opportunities and successes across over 50 measures and 10 personal characteristics. AI produced quantitative measures of disparity, characteristic contribution, the number of people who could benefit from improvement, and subgroups to target. Leaders applied judgment and values to refine focus.
Results
• Health equity assessed across 51 HEDIS and MIPS measures & 10 personal characteristics.
• Largest opportunity was depression screening in teens & adults.
• Other opportunities in cancer screening & elderly pain assessment.
• Greatest equity achieved in chronic condition management & fall risk assessment.
HonorHealth, a healthcare and 1,404 bed hospital system based on Phoenix, AZ, sought to determine whether the NEWS2 score or derivatives should be deployed during COVID-19 to assist clinicians in identifying which non-ICU patients are most at risk for short term deterioration.
Results
• Neither NEWS2 nor its derivatives could meet the criteria for decision support during COVID-19.
• Organizations should measure and seek feedback before implementing point of care decision support.
HSHS, an integrated care delivery system in the Midwestern United States, sought to determine whether the perceived excess length of stay (LoS) associated with COVID-19 patients was unique and, if so, what was driving the excess LoS.
Results
• LoS did increase during COVID-19.
• HSHS did not have excess LoS compared to benchmarks, and may have been slightly favorable.
• Prior to implementing solutions, organizations should test the direction and magnitude of problems.
INTEGRIS Health, a not-for-profit integrated healthcare delivery system based in Oklahoma, wanted to move to a more satisfying experience for setting variable executive compensation measures and their targets. They leveraged AI to help select focus areas and performance targets that are data driven, reproducible, less prone to bias, and responsive to human value judgments.
Results
• The algorithms successfully identified focus areas and improvement trajectories.
• Modifications were made to the algorithms to balance leadership objectives.
• The algorithms have been applied uniformly across the system and individual hospitals.
Clinicians face an ever-increasing demand for clinical documentation. Clinician burnout is also a growing concern, which the pandemic has only exacerbated. We sought to understand if the combination of ambient listening technology and artificial intelligence could help clinicians with clinical documentation and help mitigate the adverse effects of increased documentation, pandemic-related staffing challenges, and ultimately help reduce clinician burnout.
Results
• Speech recognition technology has made recent and significant advances.
• AI has also matured and can be utilized to evaluate human conversation in near real-time.
• We are still in early trials but see promising results in “real world” clinical settings.
COVID-19 forced leaders to balance a precarious and volatile surge in patient volumes unmatched by prior experience along with an equally challenging reduction in nurse staffing due to COVID infection and staff resignations. Predictions of staffing needs based on daily average volumes from prior years have proved to be no longer accurate, so UnityPoint Health turned to real-time machine learning to address their nurse scheduling crisis.
Results
• Improved baseline prediction accuracy (Mean Absolute Percent Error) by 20-30%.
• Automated data refresh on a real-time basis to adjust to deviations in census trending.
• Visualized current and forecasted census across units in an intuitive format.
Like many others, the University of Utah Hospital experienced a significant increase in bed demand due to the COVID-19 pandemic. In an attempt to better manage capacity demands on the system, they developed a forecasting pipeline to estimate daily net new bed demand based on the scheduled surgical procedures/patients.
Results
• Capacity management decision making was improved (load leveling, ease of access to insights).
• Additional value may be gained via secondary use of data from predictive analytics projects.
• Rapid, iterative feedback during the development process aids in implementation success.
Breast cancer is the fourth leading cause of cancer mortality in the U.S.OSF HealthCare recognized early detection via mammography screening is critical for decreasing breast cancer death rates. It conducted a study using three intervention arms to assess how different patient engagement strategies affected screening rates among women of differing economic advantage (EA) levels, generating new knowledge about decreasing breast cancer screening inequity.
Results
• Targeting interventions only to lower EA patients reversed the disparity between the two groups.
• When interventions were equally applied, using more than one engagement strategy significantly increased screening rates for both groups.
• Future research is required to determine if interventions can improve screening rates for all without worsening disparity.
U.S. healthcare spending grew by $365B in 2020. Baylor Scott & White Health recognized that comprehensive care management is crucial for success in value-based care. It used propensity score matching to quantify the financial success of its care management activities. The organization identified a control group of patients not receiving services, allowing proper comparison of outcomes and quantifying expected savings for patients receiving care management support.
Results
• $30M in estimated annual savings over five years.
• 70% less utilization for commercial members receiving inpatient transition outreach.
• 63% less in annual spending for Medicare patients patients receiving outreach from the Proactive Care Team.
Medication errors account for more than 250,000 deaths in the U.S. every year. MultiCare Health System lacked actionable, integrated pharmacy data, making it challenging to improve safety. Leveraging its data platform, the organization integrated disparate data sources, turning data hidden in various sources into actionable insights in one tool that can simultaneously improve safety, inform decision-making, monitor the impact of interventions, and generate cost savings.
Results
• $10K annual cost savings, a result of eliminating 14 hours of manual data analysis a month.
• >500% increase in key performance indicator transparency to systemwide leadership.
• 100% increase in the availability of real-time smart pump interoperability data with baselines.
Clinicians receive a staggering number of interruptive alerts, leading to alert fatigue, burnout, and potential patient safety events. Carle identified that frequent EHR alerts were problematic—some were interruptive, detrimental to providers, and disruptive to patient care. It convened an improvement team that worked to understand the original rationale for a wide range of alerts and to engage stakeholders to devise alternative solutions.
Results
• 64% relative reduction in the number of total and interruptive alerts.
• 6.5K fewer interruptions per day.
• Collaborative improvement approach increased end-user engagement, ensuring the implementation of suitable workflows.
Globally, adverse patient care events remain one of the top ten leading causes of patient harm due to unsafe conditions in hospitals. LifeBridge Health decided to proactively lead the “Zero Harm” program even with added pressure on the health system during the pandemic. The program focuses on implementing and sustaining tactics that reduce the six most recurring types of patient harm in the hospital setting, such as patient falls, hospital-acquired infections, and pressure injuries
Results
• More than 7.5K events were reported, investigated, and closed.
• >10% decrease in patient falls, infections, and pressure injuries.
• Zero central line-associated bloodstream infections for four months.
Data Governance Program Advances Business Unit Engagement Unclear, inconsistent data governance leads to confusion and distrust of reporting and analytics, negatively impacting data-informed decision-making and limiting organizational effectiveness. ChristianaCare recognized that inconsistencies and lack of data clarity were impeding system improvement. The organization created a stable data backbone for consistent reporting, analytics, and decision-making by standardizing data governance and incorporating data stewards in the process
Results
• 24 sources of member data were standardized, resulting in more than 80K records reconciled.
• 300% growth in the analytic business glossary submissions from 2018 to 2021.
• Created five data standards with mapping for highly visible data.
• Doubled business unit participation in data governance.
Creating a Welcoming Analytic Front Door Healthcare has abundant, complex data. Froedtert Health and the Medical College of Wisconsin’s burdensome data request processes delayed access to data, impeded data quality, and created barriers to improvement. By implementing a single front door solution, users have self-service access to data explorers, learning resources, development guides and standards, and an analytics catalog with governed metrics, enabling improved data literacy and insight for improvement efforts
Results
• 50% increase in user application sessions per week.
• 75% increase in user-developed applications.
• 5% reduction in outstanding data requests.
• 2.7 point increase in the organization’s analytic maturity score.
Medical errors account for nearly 250K deaths in the U.S. annually. HonorHealth identified severe hypoglycemia and over-sedation from opioids as the top two adverse drug events, prioritizing both for improvement. It implemented the Hypoglycemia Event Analysis Tool and aNaloxone Event Analysis Tool to standardize event documentation, enabling team store view, identify, and remedy root causes. The organization exceeded the composite Patient Safety Index score improvement targets.
Results
• 50% decrease in severe hypoglycemic event rate per 1,000 patient days.
• 67% relative reduction in naloxone administration for opioid oversedation at the pilot site.
• $281K cost avoided by avoiding 17 severe hypoglycemic events and two opioid oversedation events.
Annually, 200 Hawaiians die by suicide. Native Hawaiian’s suicide risk is 2.5 times higher than Caucasians. Suicidality increased during the COVID era. The Queen’s Health System is the leading mental health provider in the Pacific. However, bed capacity limits access to new patients with essential needs. To improve safety and access, the team deployed a real-time suicide assessment and integrated treatment bundle for patients in the emergency department at risk for suicide.
• 87% of all admitted patients were screened for suicide risk.
• 81% percentage point increase in the number of patients receiving all suicide prevention bundle elements.
• >95% of patients at risk of suicide received timely post-discharge follow-up.
Hospitals spend $30M on supplies annually. The UnityPoint Health procurement team lacked an efficient way to view historical purchase order information, track total spending by contract, or ensure adherence to contract obligations, resulting infrequent requests for time-consuming ad hoc reports. It developed a self-service tool leveraging its data platform and supply management software, eliminated manual processes, and improved the timeliness and accuracy of supply data.
Results
• $1M increase in vendor rebates and cost savings.
• Improved the organization’s ability to meet contract compliance and use lower-cost vendors.
• Five-day reduction in report turnaround time—reports are now available in minutes.
Healthcare data are abundant but require translation for decision-making. UnityPoint Health complemented its system-level analytics with business partners to engage with customers to ensure problems are fully understood. Teams use standard tools to create meaningful, easy-to-understand visualizations and ensure delivery of the robust analytics required to improve performance. The organization translated data into action and developed a data-informed decision-making culture.
Results
• $200M in crucial decisions supported by robust analytics.
• Developed thousands of data analyses and reports for improved leadership decision-making.
• Hundreds of improvement initiatives are or have been supported.
• Increased analytics and improvement team satisfaction and effectiveness.
Healthcare organizations face pressures to deliver high-quality care while controlling costs. New Hanover Regional Medical Center lacked a systematic approach for identifying opportunities, implementing evidence-based best practices, and prioritizing efforts to decrease unwarranted care variation. It established a synergistic improvement framework by leveraging data-analytics, improvement science, and adaptive leadership to achieve data-informed, exceptional, cost-effective outcomes.
Results
• $5.4M savings from 20 initiatives, positively impacting 37K patients.
• $1.4M labor costs avoided by utilizing automated, advanced analytics.
• >1.5K units of blood avoided.
• 28.6% reduction in the mortality rate for patients with sepsis.
The COVID-19 pandemic magnified health equity gaps. Mount Sinai needed to accelerate work already in progress to incorporate patient data such as race, ethnicity, and language into reports and dashboards, enabling analysis and identification of health disparities to target for improvement. The health system established focus groups and deployed change management strategies to communicate the importance of accurate data collection and its use for improving health outcomes.
Results
• Race, ethnicity, and language (REL) data collected for 90% of patient population.
• Development of a core equity toolkit, the result of improved REL data capture.
• $20M grants received to support health equity initiatives.
Aligning clinical practices with current evidence is challenging, often resulting in an evidence-practice gap. WakeMed developed a global, scalable dashboard to monitor adherence to its early recovery after surgery (ERAS)program and best practices. The organization leveraged analytics to illustrate that patients receiving the ERAS protocol had better outcomes and engaged providers in reviewing the data, supporting program expansion, further improving outcomes, decreasing costs, and closing the evidence-practice gap.
Results
• 11 ERAS populations integrated into an outcome and process-based analytic tool.
• Visualizations support the measurement of on-pathway vs. off-pathway cohorts by various clinical outcomes.
• Enabled comparisons of outcomes by equity dimensions and provider, including control charts and patient-level details.
• 54% relative reduction in rehospitalization for cesarean-section.
Social determinants of health (SDoH) contribute to health disparities and inequities. Nemours Children’s Health recognized that many of its patients were impacted by SDoH but lacked the insight to identify patients’ unique needs and provide the necessary support to improve health outcomes. By pairing analytics with a systemwide screening tool, the organization gained insights into its patient populations’ needs and identified improvement opportunities.
Results
• 13.7% of patients self-reported at least one social need.
• 35.8% of patients who reported needs requested resources.
• Top three needs identified included reliable/affordable internet, neighborhood safety, and food insecurity.
Over 12M Americans will have atrial fibrillation(AF) by 2030, substantially increasing stroke risk. Community Health Network recognized that only 64 percent of patients with AF were receiving appropriate anticoagulation therapy. The organization created best practice work flows, provided education, improved visibility of anticoagulation status, and monitored care gaps, resulting in an increased number of patients receiving anticoagulation therapy and decreased risk of stroke.
Results
• 13% increase in the number of patients appropriately anticoagulated.
• 547 patients appropriately anticoagulated after the AF best practice alert was triggered.
• 27% AF event monitor detection for patients post-cryptogenic stroke.
Understanding the patient care experience is crucial for healthcare organizations to succeed. The Permanente Medical Group from the South Sacramento Medical Center looked through the lens of Behavioral Science to build a robust framework for its care experience program, leading to improved data and effective improvement plans. The program yielded new measurements, more in-depth results, better data integrity, culture enhancements, and resources for continuous improvement.
Results
• 11% increase in Net Promoter Score.
• 20% increase in patient fulfillment score with a 10% increase in response rate.
• Improved qualitative and quantitative understanding of care experience data and interpretation of results.
• Effective training program developed using relevant patient experience data.
Centers for Medicare & Medicaid Services regulations require payers to implement Application Programming Interfaces to improve the electronic exchange of health care data and improve patient access to information. Carle and Health Alliance needed a rapid solution to address the new rule. It rapidly implemented a Fast Healthcare Interoperability Resources (FHIR) solution enabling better patient access to data and meeting regulation timelines for implementation.
Results
• 250K members gained meaningful access to personal health information using a personally selected third-party application.
• 1K analyst hours avoided by rapid deployment of FHIR solution.
• Penalty for non-compliance avoided—$100K cost avoidance.
An estimated 80 percent of a person’s health is related to factors beyond medical care, referred to as social determinants of health (SDoH). MetroHealth’s Institute for H.O.P.E.™ sought to understand the impact of SDoH on the community. It initiated a screening process for all adult patients at the point of care. The organization created an interactive dashboard by integrating data from multiple sources and surfaced opportunities for system changes and improvements.
Results
• Successfully linked patients needing assistance with health-related social needs to community organizations for services.
• Developed programs to address social needs and created a system of referrals and research.
• Over $6M in grants received for programs and research.
Each year, preventable patient safety events in the U.S.cost over $100B. Henry Ford Hospital’s patient safety indicator (PSI) score was above the national average composite, and clinical review related to PSIs was lacking. The organization leveraged analytics to facilitate interdisciplinary review of all PSIs, resulting in improved data accuracy, integrity, reporting, identification of improvement opportunities, and an increase in CMS Overall Hospital Quality Star Rating.
Results
• Statistically significant improvement in PSI score (p-value=0.0065).
• Improvement in PSI-90 composite score:
o 2019 PSI-90 composite: 1.44
o 2020 PSI-90 composite: 1.03
o 2021 PSI-90 composite: 0.75
• Expected increase in Overall Hospital Quality Star Rating and Hospital Safety Grade.
Many lab tests, often accompanied by painful needlesticks that increase infection risk, are unnecessary, can lead to anemia, and do not improve patient outcomes. Intermountain Healthcare used SaVia’s technology to eliminate unnecessary therapeutic interventions and needlesticks for NICU patients.
Results
• $1.2M annual cost avoidance.
• 23% reduction in needlesticks—96K needlesticks eliminated.
• Multiday reduction in NICU length of stay.
• Increased caregiver and parental satisfaction.
Vendors promote medical device benefits, but real-world outcomes data and the impact of new devices are difficult to surface. Baylor Scott and White Health wanted to evaluate new medical devices to assess and illustrate product impacts. The organization effectively evaluates emerging and disruptive medical devices by leveraging an analytic approach to visualize device ordering and utilization, population insights to trend patient outcomes, and billing data to ensure coding accuracy.
Results
• Identified $3.3M increased spending over four years.
• Visualized facility and physician utilization of Impella device by version.
• Surfaced Impella patient outcomes (LOS, readmission rate, and mortality rate).
• Detected $1M opportunity in Impella coding accuracy.
Readmission cost Medicare and Medicaid more than $15B annually. Cone Health recognized an opportunity to reduce readmissions by leveraging analytics to identify patients at extreme and high risk of readmission. As a result, the organization redesigned social work and nursing workflows, developed tailored, personalized care plans to improve the patient transition from hospital to home, and enabled care teams to intervene before unnecessary readmissions, substantially reducing readmission risk.
Results
• 20% reduction in the number of patients with extreme or high readmission risk.
• Two-day reduction in the average length of stay for patients with high readmission risk.
• 74% of patients with high readmission risk referred to preferred short-term nursing facilities.
• 90% of patients with high readmission risk received interventions by care coordinators.
We take pride in providing you with relevant, useful content. May we use cookies to track what you read? We take your privacy very seriously. Please see our privacy policy for details and any questions.