The clinical informatics solutions are team-based operations. Where a team is listed, the presenter is noted. iHealth 2017 presentations cover the following topics:
- Analytics and the Learning Health System
- Bridging Analytics, Bedside care, and Education
- Care Coordination
- Clinical Decision Support
- Data Sciences
- Health Policy and Payment Reform
- Interoperability and Informatics Infrastructure
- Population Health
- Safety and Quality
Wednesday, May 3
Querying Electronic Health Data for Population Health Activities using PopMedNetTM
Jessica Malenfant, Harvard Pilgrim Health Care Institute (Presenter)
Kyle Erickson, Harvard Pilgrim Health Care Institute
Zachary Wyner, Harvard Pilgrim Health Care Institute
Chayim Herzig-Marx, Harvard Pilgrim Health Care Institute
Dean Corriveau, Lincoln Peak Partners
Jeffrey Brown, Harvard Pilgrim Health Care Institute
This presentation is an overview of new informatics infrastructure developed in the query platform, PopMedNetTM (PMN) that can provide framework for the Learning Health System. Specifically, the query interface in PMN was re-architected to incorporate current technologies and scalable query functionality that will address technology differences across collaborators in distributed health data networks. PMN was also enhanced to incorporate customizable electronic workflows to facilitate the entire query cycle to supplement the new MDQ infrastructure.
Defining Patient Sub-Populations with SNOMED
Vaishnavi Kannan, University of Texas Southwestern Health System (Presenter)
DuWayne Willett, University of Texas Southwestern Health System
With the move to value-based care, defining patient subpopulations becomes imperative. For constructing codesets, SNOMED offers an attractive alternative to ICD-10, as SNOMED has higher clinical fidelity, is a hierarchical ontology, and is widely implemented in EHRs. In 2015, we constructed 43 specialty patient registries, most employing 1 or more diagnosis groupers for inclusion criteria. Developing a publicly-available curated source of SNOMED-based condition groupers would accelerate population health efforts and "smart" EHR feature configuration across organizations.
Enabling Prospective Analysis of Care Redesign with reCAP, The REal-Time Care Analysis Platform
Colin Walsh, Vanderbilt University Medical Center
Michael Ripperger, Vanderbilt University (Presenter)
To guide interventions preventing low-value service utilization, healthcare leaders may rely on retrospective analyses that are not readily reproducible, cannot differentiate system from individual factors, and may not capture explanatory temporal patterns. Developed in parallel with a quality improvement care-pilot, reCAP, the REal-time Care Analysis Platform, is an interactive analytic platform permitting temporal/aggregate, patient/cohort/program-level views. This study evaluates via user feedback the impact of novel data source integration, interoperability-enabled design, and clinical role customization.
S03: Presentations - Telemedicine
Telemedicine to Manage HIV Patients within Pennsylvania Prisons
Jennifer Aldridge, Lewis Katz School of Medicine, Temple University (Presenter)
Kelly Lattanzi, Lewis Katz School of Medicine, Temple University
Mark Weiner, Lewis Katz School of Medicine, Temple University
Ellen Tedaldi, Lewis Katz School of Medicine, Temple University
The Pennsylvania State Prison System includes HIV positive inmates scattered across 26 sites. While prisons are staffed with nurses, and patients receive medical evaluations and care by staff physicians, the provision of specialized HIV care is challenging, given the number of sites and geographic distance between the prisons and the specialists’ locations. This presentation describes the architecture and function of a telemedicine program that supports the care of hundreds of HIV positive prisoners every year
Feasibility, Acceptability and Impact of a Pediatric Teledermatology Mobile Health Application
Alexander Fiks, University of Pennsylvania School of Medicine/The Children's Hospital of Philadelphia
Patrick McMahon, The Children's Hospital of Philadelphia/University of Pennsylvania School of Medicine
Lindsay Berrigan, The Children's Hospital of Philadelphia (Presenter)
Emily Sykes, The Children's Hospital of Philadelphia
Rachel Gruver, The Children's Hospital of Philadelphia
Katherine Halkyard, The Children's Hospital of Philadelphia
Flaura Winston, The Children's Hospital of Philadelphia
Linda Fleisher, The Children's Hospital of Philadelphia
Research has not evaluated the use of teledermatology to facilitate the routine dermatologic care of children. Following an initial implementation in a large health care system, we found through open and closed-ended survey questions and electronic logs of actual use that a mobile teledermatology application proved feasible/usable, acceptable and impacted clinical care. Such applications may ultimately improve pediatric dermatology care.
Analysis of Abnormal Heart Rates Recorded when Patients were Invited to Upload Personal Fitness Device Data to the Electronic Health Record
Joshua Pevnick, Cedars-Sinai Health System
Yaron Elad, Cedars-Sinai Health System
Richard Riggs, Cedars-Sinai Health System
Pamela Roberts, Cedars-Sinai Health System
Ray Duncan, Cedars-Sinai Health System
Inviting patients to upload personal fitness device data (PFDD) to an electronic health record resulted in 151 patients with 5693 heart rates (HRs)<40 and 61 patients with 4444 HRs>200. This represented <0.2% of uploaded HRs. Four patients had >80% of abnormal HRs. In six concerning cases, chart review did not suggest device error. Three cases corroborated PFDD, but had already been recognized. PFDD may have offered earlier detection, but did not improve these patients’ outcomes.
S05: Presentations - Data Sciences
Effects of Digoxin and Diltiazem on Mortality among Atrial Fibrillation Patients
Shrie Raam Sathyanarayanan, Oklahoma State University
Atrial fibrillation (AF) is a sustained dysrhythmia frequently encountered in clinical practice. Digoxin and diltiazem are two commonly used rate control drugs for AF. This study compared mortality rates of AF patients prescribed these drugs and used logistic regression to assess the predictive power of drug, age, race, gender, length of stay, and payer (Medicaid/Medicare). Digoxin-associated mortality was significantly higher than for diltiazem [5.23% vs 3.98%, p<0.001], and the model had high accuracy [AUC 0.730].
Estimating Data Requirements to Detect Pediatric Critical Decompensation
Melissa Aczon, Children's Hospital Los Angeles (Presenter)
David Ledbetter, Children's Hospital Los Angeles
Sareen Shah, Children's Hospital Los Angeles
We developed a new early warning tool using machine learning techniques that are able to analyze time-series data. This tool enables calculation of the trajectory of risk over time by integrating dynamic data (continuously changing values) instead of relying solely on static data (values at one point in time). We assessed our model’s ability to predict critical decompensation of floor patients as a function of data available for training.
Mining Care Pathways of Multimorbid Patients from Electronic Health Records
Yiye Zhang, Weill Cornell Medical College/Cornell University (Presenter)
Rema Padman, Carnegie Mellon University
This preliminary study introduces an analytics method to mine care pathways of multimorbid patients from electronic health records and cost data. We categorized care pathways of patients who were initially diagnosed with chronic kidney disease, hypertension and diabetes from 2009 to 2011 by clinical complexity and medical spending. Analysis and visualization show significant variations in care and costs as conditions progress that may reveal potential inconsistencies in care needs and service utilization in future studies.
S06: Presentations - Clinical Decision Support
Development of a Web-based Decision Support Tool for Operationalizing and Optimizing Management of Hyperbilirubinemia in Preterm Infants
Yassar Arain, Stanford University (Presenter)
Jonathan Palma, Stanford University/Stanford University
Premie BiliRecs is a novel electronic clinical decision support tool for the management of hyperbilirubinemia in moderately preterm infants less than 35 weeks gestational age. It serves to operationalize and automate current expert consensus-based guidelines, and to aid in the generation of new practice-based evidence to inform future guidelines. Preliminary data collection and analysis seeks to show standardization in treatment of hyperbilirbubinemia in moderately preterm infants.
Design and Implementation of a Real Time Acuity Detection System in a Pediatric ICU
Eric Shelov, The Children's Hospital of Philadelphia (Presenter)
Naveen Muthu, The Children's Hospital of Philadelphia
Maya Dewan, Cincinnati Children's Medical Center
The Pediatric Intensive Care Unit (PICU) at the Children’s Hospital of Philadelphia recently designed and implemented a real-time acuity detection system. The goal of this system, based on retrospectively validated clinical criteria, is to detect patients at risk for a “code” event or cardiac arrest. After successful implementation, several PDSA cycles with the system have led to sustained improvement in bedside preparedness for high risk patients in the PICU.
Extracting the Amount of Alcohol Consumption from Clinical Narratives using Natural Language Processing for use in Clinical Decision Support Tools
Branden Hickey, Mayo Clinic
Rajeev Chaudhry, Mayo Clinic (Presenter)
Ravikumar Komandur Elayavilli, Mayo Clinic
Maya Kessler, Mayo Clinic
Charity Maynard, Mayo Clinic
Timothy Miksch, Mayo Clinic
Steve Peters, Mayo Clinic
Marianne Scheitel, Mayo Clinic
Jane Shellum, Mayo Clinic
Low success in extracting the discrete amount of alcohol consumption per week or day from a patient’s electronic health record has potential to decrease accuracy for several clinical decision support tools. At Mayo Clinic, we utilized natural language processing (NLP) to detect mentions of alcohol consumption by looking for alcohol type, units, and frequency from the patient’s social history note section. In this work, we analyzed the performance of the NLP algorithm to detect amount of alcohol consumption with the goal to improve specificity for 1-year risk of major bleeding (HAS-BLED) and 10-year probability of a major osteoporotic fracture (FRAX) calculators.
S09: Presentations - Interoperability and Informatics Infrastructure
FHIR®-based Predictive Analytics: A Breast Cancer Pilot
Deven Atnoor, Sysbiochem, LLC (Presenter)
Grant Wood, Intermountain Healthcare
Brett Johnson, Department of Veteran's Affairs
Kevin Hughes, Mass General Hospital
Intermountain Health, Massachusetts General Hospital and Sysbiochem have built a set of services for deploying clinical predictive models using the HL7-FHIR standard. These services have enabled the stakeholders a way to integrate family history data with clinical data from EHR and genomic results using HL7-FHIR standards for submitting to a risk prediction application dynamically, and obtaining the prediction results as a FHIR message back into the calling system. This presentation provides a detailed overview of the work done for building the integration workflow, lessons learnt and results obtained. We will also provide a demonstration of the end-to-end workflow.
Leveraging Search Patterns in Electronic Health Records to Make Information Retrieval More Efficient
Titus Schleyer, Indiana University–Purdue University Indianapolis/Regenstrief Institute
Xia Ning, Indiana University–Purdue University Indianapolis
Martin Doug, Indiana University–Purdue University Indianapolis/Regenstrief Institute
When clinicians search the electronic health record with regard to a particular patient problem, the EHR does not make suggestions for potentially useful information. In this study, we analyzed 2,926,932 searches for information about ~4.5m patients that clinicians performed in Indiana’s major HIE from 9/2012 to 7/2016. We found strong evidence that clinicians’ search terms about similar patients cluster in distinct patterns, making it possible to recommend information items for future searches.
S11: Presentations - Health Policy and Payment Reform
Decision Modeling of Clinical Content for Collaborative Documentation Purposes in the Management of Chronic Conditions
Hari Nandigam, Partners Healthcare Systems (Presenter)
Perry Mar, Partners HealthCare System/Brigham and Women’s Hospital/Harvard Medical School
Kira Tsivkim, Partners HealthCare System
Dina Iskhakova, Partners HealthCare System
Objective: This paper is focused on modeling of the decision entity for interprofessional care planning with focus on patients with chronic diseases. Methods: We developed an information model of decision concept and determined its relationship upstream with patients’ health problem and goals and downstream with interventions and assessment based on expert validated clinical scenarios. We used Altova UModel for translation into UML model and then mapped decision entity and valuesets to CTS-2 schemas. Results: In the management of chronic conditions, determining target concepts and their attributes such as role, act, rationale, clinical guidelines, and anticipatory guidance are critical for decision modeling. Discussion: Proper modeling of decision entities is required for structured documentation to accommodate guidelines and logic for the decision, especially in highly collaborative management process. Conclusion: By representing decisions through CTS2 model we addressed two essential aspects: knowledge representation and interchange. CTS2 is also considered a standard for a shared semantic model, allowing applications to access services without knowing the internal database structure or data model. Clinical guidelines change constantly, and anticipatory guidance, rationales and corresponding attributes for decisions are often patient or situation specific and with time dependencies. We conclude that because of this flexibility it also allows updating content without breaking applications that utilize the services.
An Informatics Framework for Value-based Care at NewYork-Presbyterian
Gilad Kuperman, New York Presbyterian
The goal of value-based care (VBC) is to increase efficiency in the health care system. Health care payers are advancing VBC through payment models that create incentives for providers to restructure care delivery. NewYork-Presbyterian (NYP) is an integrated delivery network in the New York metropolitan area that includes (i) NYP Hospital, a 2400-bed academic center, (ii) NYP Physician Services, a network of 1000 physicians, (iii) the NYP Regional Hospital Network, 3 community hospitals with a total of 1000 beds, and (iv) the NYP Division of Community and Population Health, which provides over 1 million ambulatory encounters annually. NYP is also affiliated with Columbia and Cornell medical schools– each of which has 1000 faculty physicians. Over the past five years, in response to evolving federal and state payment programs, NYP has established a number of value-based care delivery programs. Governmental payment models have unfolded unevenly and the information systems infrastructure to support the redesigned models of care has emerged on a program-by-program basis. As VBC has matured at NYP, cross-program needs are now better understood and NYP increasingly is shifting from a tactical approach for the support of VBC to a strategic approach.
Re-Identification Risk in HIPAA Safe Harbor De-Identified Datasets: Using non-HIPAA Identifiers / the Case of the MVA Attack
Peter Elkin, University at Buffalo (Presenter)
Victor Janmey, University at Buffalo
Daniel Schlegel, University at Buffalo
Christopher Crowner, University at Buffalo
We present a re-identification attack, the MVA attack, that uses indirect (non-HIPAA) identifiers to target a vulnerable subset of records de-identified to the HIPAA Safe Harbor standard, those involving motor vehicle accidents (MVAs). The attack is demonstrated through a case report involving re-identification of a patient from a de-identified dataset. Remediation strategies to prevent this type of attack are also discussed.
Rema Padman, Carnegie Mellon University (Presenter)
Alexandra Chouldechova, Carnegie Mellon University
IncharaDiwakar, Carnegie Mellon University
Mark Clements, Children's Mercy Hospitals & Clinics
Extracting predictive features from asynchronous multivariate data streams for early prediction of diabetic ketoacidosis in pediatric type 1 diabetes patients is a challenging problem. We propose a smoothing technique for segmenting measurements collected at irregularly spaced time points into trend and value abstractions, and show that using these abstracted temporal features in building prediction models improves predictive accuracy over models that solely consider summary statistics such as averages and counts.
Identifying Patterns of Co-occurring Medical Conditions through Topic Models of Electronic Health Records
Moumita Bhattacharya, University of Delaware
Claudine Jurkovitz, Christiana Care Health System
Hagit Shatkay, University of Delaware
Multiple adverse health conditions co-occurring in a patient are typically associated with poor prognosis and increased office or hospital visits. Developing methods to identify patterns of co-occurring conditions can assist in diagnosis. We aim to identify patterns of association among conditions by applying a machine learning method, namely topic modeling, to EHRs. We show that conditions that are highly probable to be associated with the same topic, indeed tend to co-occur in patients.
Developing and Implementing a Reliable and Validated Solution for Big Data Text Analytics (NLP, Text Mining) at a Tertiary Pediatric Hospital.
Luis Ahumada, The Children's Hospital of Philadelphia
William Nieczpiel, The Children's Hospital of Philadelphia
John Martin, The Children's Hospital of Philadelphia
By definition, big data in healthcare refers to electronic health data sets so large and complex that they are difficult to manage with traditional software and/or hardware; nor can they be easily managed with traditional or common data management tools and methods. Our approach takes advantage of modern technology infrastructure and the novel techniques in data science to manage and provide access to clinician notes and extract insights for making better informed decisions in healthcare. The traditional method of extracting data from clinical notes is done through manual chart review.
S15: Presentations - Care Coordination
Integrating mHealth Medication Reconciliation and Symptom Reporting for Patient-centered Ambulatory Care
Lisa Grossman, Columbia University
Rui Sim, Columbia University
Ruth Masterson Creber, Columbia University
Appropriate symptom monitoring and medication reconciliation is essential to patient-centered care in heart failure. We present an electronic platform, MyClinic ©, for symptom monitoring and medication reconciliation in the ambulatory care setting. We discuss the usability of MyClinic ©, as well as the specific utilities. This highly usable, symptom-focused electronic application may afford better value to patients and providers than traditional intake applications.
Developing an Electronic Care Plan to Improve Longitudinal Care Coordination in Chronic Kidney Disease
Theresa Cullen, Regenstrief Institute
Jenna Norton, National Institute of Diabetes and Digestive and Kidney Diseases (Presente)
Andrew Narva, National Institute of Diabetes and Digestive and Kidney Diseases
Chronic kidney disease (CKD) is a complex condition requiring coordination across multiple settings. Development of an electronic care plan for CKD has the potential to improve CKD outcomes by supporting digital transfer of patient information across settings, among providers, and between providers and patients. Several challenges and lessons learned have been identified in the CKD e-care plan development process that may be relevant to the creation of e-care plans for other chronic diseases.
Reducing Readmission and Post-hospital Mortality through Data beyond Your Borders
Philip Smith, Remote Patient Monitoring, Inc.
Readmission and Post-hospital Mortality are difficult problems for today’s hospitals as they represent potential failures of hand-offs during critical transitions of care. However, since these often occur into the community rather than solely within the borders of the health system, they are often not captured within the hospital’s enterprise data warehouse or other traditional means. Often the complete data is not available to the hospital until the results are compiled and reported by the Centers for Medicare and Medicaid (CMS) several years later when penalties are assigned (CMS will be reporting 2016 penalties in mid-2017, for discharges Oct 1, 2011 – Sept. 30, 2014.). It is difficult to improve performance when data and information are not available in a timely fashion. The author has done extensive work in developing clinical close dashboards for other hospital measures and now has a process to obtain data using a patient-centric technique that overcomes the health system’s and hospital’s limitations. This novel process will permit hospitals to compile data at 30, 60 and 90 days following hospital discharge, provide pre-admission status for the prior 90 days, and follow joint replacement patients for up to 180 days. This new process will allow timeliness of data to validate performance improvement trials, determine optimal follow up intervals and improve predictive analytics. In addition, de-identified information will be utilized to create a readmission and post-hospital mortality benchmark well in advance of the current CMS benchmarks. Moreover, the process will allow discrete data for numerous interventions through leveraging clinical decision support from within the electronic health record. The process is being refined through a pilot at a large, multi-state health system in Q4 2016.
S20: Presentations - Safety and Quality
Reduced Rates of Hospital-Associated Clostridium Difficile Infection Associated with a Clinical Surveillance System
Stephen Ross, Wolters Kluwer Health/University of Colorado
Sue Ie, Wolters Kluwer Health
Sharad Manaktala, Wolters Kluwer Health
Justin Clark, Wolters Kluwer Health
Using data from 64 affiliated hospitals, we evaluated the effect of implementation of a clinical surveillance system on the standardized infection ratio (SIR) for hospital-associated Clostridium difficile infection (CDI). Using a generalized linear regression model with mixed effects, implementation of the surveillance system was associated with a decrease in CDI SIR of .230 (p=.038). This supports the clinical value of “active surveillance systems” recommended in recent CMS requirements for hospital antimicrobial stewardship programs.
Making Acute Care More Patient-Centered: Implementing a Learning Lab
Theresa Fuller, Brigham and Women's Hospital (Presenter)
Jenzel Espares, Brigham and Women's Hospital
Megan Duckworth, Brigham and Women's Hospital
Brittany Couture, Brigham and Women's Hospital
Jeffrey Schnipper, Brigham and Women's Hospital/Harvard University
Patricia Dykes, Brigham and Women's Hospital/Harvard University
Sarah Collins, Brigham and Women's Hospital/Harvard University/Partners HealthCare
Anuj Dalal, Brigham and Women's Hospital/Harvard University
Ronen Rozenblum, Brigham and Women's Hospital/Harvard University
Kumiko Schnock, Brigham and Women's Hospital
Esteban Gershanik, Brigham and Women's Hospital/Partners HealthCare/Harvard University
Stuart Lipsitz, Brigham and Women's Hospital/Harvard University
Alexandra Businger, Brigham and Women's Hospital
Pamela Neri, Partners HealthCare
James Benneyan, Northeastern University
David Bates, Brigham and Women's Hospital/Harvard University/Partners HealthCare
Learning laboratories allow for healthcare initiatives to employ the development processes and iterative refinement of design features and subsystems normally found in large-scale engineering ventures. While there are clear benefits to this approach, the nature of implementing a large scale project as a learning lab and the inherent constraints and subtleties must also be approached thoughtfully. This presentation explores the early experience of implementing a learning lab, its limitations, and the lessons learned.
First Contact Provider (FCP): Improving Inpatient Critical Results Reporting
Anisha Chandiramani, University of Chicago Medicine (Presenter)
Janet Gervasio, University of Chicago Medicine
Michelle Johnson, University of Chicago Medicine
Jessica Kolek, University of Chicago Medicine
Steven. Zibrat, University of Chicago Medicine
Dana Edelson, University of Chicago Medicine
We created a new field in EPIC, called the First Contact Provider (FCP), which produces a consistent and clear display of the paging contact information for every admitted patient’s Licensed Independent Practitioner (LIP). FCP implementation was widely adopted by admitting providers throughout UCMC and resulted in significant improvement in critical lab reporting to LIPs.
S22: Presentations - Bridging Analytics, Bedside care, and Education
Granular ED Throughput Data: Helping Providers Find Agency in Process Improvement
Tom Spiegel, University of Chicago (Presenter)
Jacob Moore, University of Chicago
Motivation: Can provider-specific, granular throughput data inspire provider agency in operations improvement? Methods: Periodic, innovative modeling isolated patient bed-time directly attributable to delayed provider decision making. Providers were informed and encouraged to critically evaluate ways to improve their practice. Findings: Emergency Department (ED) providers improved their post workup times, thereby decreasing total ED length-of-stay for discharged patients. Implications: Granular, actionable clinical data can foster provider agency in ED operations improvement.
Manual, Automated, or Derived Measures: The Value of Variability in the Meaningful Use of Vital Sign Data
Keith Feldman, University of Notre Dame (Presenter)
Annie Rohan, Stony Brook University
Nitesh Chawla, University of Notre Dame
Traditionally, vital sign data has been manually collected and documented. However technological advancements now provide for the automated collection of vital signs via electronic medical records. Limited knowledge exists as to whether data documented automatically are comparable, inferior, or superior to manually collected data for clinical applications. This study provides an overview of differences in manually versus automatically collected heart rate data, and explores the potential applications of automated vitals data utilizing measures of variability.
S23: Presentations - Workflow
Prototyping The Future of CPOE: Starting With Indications
Aaron Nathan, Brigham and Women's Hospital (Presenter)
Adam Wright, Brigham and Women's Hospital/Harvard Medical School
Gordon Schiff, Brigham and Women's Hospital/Harvard Medical School
Lynn Volk, Partners Healthcare
Pamela Neri, Partners Healthcare
Kevin Kron, Partners Healthcare
Sara Myers, Brigham and Women's Hospital
Mary Amato, MCPHS University
Alejandra Salazar, Brigham and Women's Hospital
Enrique Seoane-Vasquez, MCPHS University
Tewodros Eguale, MCPHS University
Sarah McCord, MCPHS University
Rosa Rodriquez-Monguio, University of Massachusetts
Hannah Dym, Brigham and Women's Hospital
Employing user-centered design methods, a multidisciplinary team at Brigham and Women's Hospital designed an innovative CPOE system driven by indications. AHRQ has funded this project to meet the demands of modern patient-centered care, and improve patient safety. Commercial vendors limit indication association to after-the-fact additions, however, this prototype will showcase how starting with the indication drives CDS providing drug of choice suggestions based on clinical guidelines, formulary coverage, and patient-specific factors.
Design and Application of a Bariatric Ambulatory Workflow for the Pre-operative Evaluation of Bariatric Surgical Candidates
Tatyan Clarke, Temple University School of Medicine (Presenter)
Mark Weiner, Temple University School of Medicine
David Fleece, Temple University School of Medicine
In preparation for the surgical management of morbid obesity, Bariatric patients undergo a complex, often lengthy, multidisciplinary evaluation and risk-stratification to determine their candidacy for these moderate-risk operative interventions. This presentation will review the creation and application of a Bariatric Ambulatory Workflow to collect and audit elements of the preoperative work-up of potential Bariatric patients and its application to monitoring program efficacy.
Leveraging Advanced Physician Documentation for Improving Quality, Accuracy, and Compliance
Kang Hsu, St. Joseph's Hospital/MEDITECH
Dr. Kang Hsu will discuss a multiple year effort underway at St. Joseph's Health system to leverage their EHR’s physician documentation tool to improve their documentation compliance rate and case mix index. Through a collaborative effort between the medical staff, Clinical Documentation Improvement team, HIM department, IT, and administration, they have developed a process to electronically communicate with the medical staff via physician documentation regarding documentation improvement and compliance items. As a result, they have seen a trend of improved documentation, quality, accurate reflection of the true patient case mix index, and increased hospital reimbursement.