Bookmark and Share

CTSI: Quality of Care and Health Outcomes


The CTSA will establish a multi-disciplinary quality research center offering an array of investigative and analytic capabilities designed to integrate and enhance the diverse and promising, but independent, quality research initiatives at UMB, and at partner institutions.

Several established and junior investigators at UMB are successfully engaged in research on quality of health care, here defined as measuring: “the ability of a (provider), hospital or health plan to provide services for individuals and populations that increase the likelihood of desired health outcomes, and are consistent with current professional knowledge”1Fink.  The UMB quality research community comprises multiple schools, a range of disciplines, and a diverse set of investigators.  The research accomplishments at UMB utilize a variety of research methods and analytical techniques, some of which will be highlighted below.  Each example of research on quality of health care reveals a distinct study design, data source, or statistical approach.  Continuation of work in each of these lines of investigation would benefit from a research center where there is a set of core research tools and analytical expertise to enable new and established investigators in this discipline to advance their work.

Data simulations in dialysis network quality indicators

Jeffrey C. Fink MD MS, Professor, Department of Medicine, School of Medicine; has examined the quality of dialysis practices over the last several years.  He has championed the modeling and tracking of center effects on key quality metrics in the end-stage renal disease (ESRD) Network 5 dialysis population (mid-Atlantic region) arguing that strong center effects are evidence of poor quality for a system for the practice in which the center effect is observed.  A key quality metric is the adequacy of dialysis as measured by the urea reduction rate (URR) in patients receiving hemodialysis in a given network In Figure 1, he demonstrated the extent to which dialysis facility practices influenced variations in URR.  The figure shows the distribution of actual URR values in the Network 5 sample (2004) versus a simulated sample of URR values with the same distribution characteristics (upper panel).  The lower panel shows the distribution of centers with their respective patients assigned their actual URR values versus the center distribution with simulated URR values derived from the same distribution. The analysis highlighted the broader degree of variation in achieved aggregate URR values when using actual data then what would be expected if URR values were distributed without reference to dialysis facility2Fink.  In other analyses he has demonstrated how the strong center effect in dialysis anemia management (using US Renal Data System data) and how center effects can be improved with Network-wide quality improvement initiatives.

Pharmaco-epidemiology and drug adherence in schizophrenia

Julie Kreyenbuhl, Pharm D, PhD, Associate Professor , Department of Psychiatry, School of Medicine has worked extensively on examining drug usage patterns among schizophrenics cared for in the VA health care system.  In a recent analysis of approximately 2000 patients initiated on antipsychotic treatment with one of five non-clozapine second-generation antipsychotics or either of the two most common first generation antipsychotics between 2004 and 2006.  The outcome of interest was duration of continuous antipsychotic possession from the index prescription, with possession interruption defined by 45 days between prescriptions.  She and her co-investigators demonstrated that 84% of patients discontinued therapy during follow-up (up to 33 months).  Figure 2 demonstrates the survival curves for time to discontinuation for all agents with risperidone having the greatest risk of discontinuation relative to olanzapine.   Several factors were identified as significant risk factors for discontinuation including homelessness, recent psychiatric hospitalization, and prescription of another antipsychotic (among others).  The investigators have concluded that current antipsychotic treatment only has limited acceptance by patients and further investigation in required to identify explanatory factors explaining poor antipsychotic drug adherence.

Cluster-randomized trial evaluating training methods to augment substance abuse screening and treatment

Robin Newhouse, PhD, RN, Chair & Professor, Organizational Systems and Adult Health, School of Nursing has recently completed a 7-center cluster-randomization trial comparing two implementation strategies (standard training vs Evidence-Based Behavioral Practice (EBBP) approach) to increase nurse Screening, Brief Intervention, and Referral to Treatment (SBIRT) for hospitalized patients at risk for unhealthy drug and/or alcohol use.  Both groups received a toolkit (algorithm, instruments, training materials, readiness ruler), a site visit, and monthly synchronous web meetings.   The EBBP intervention incorporated a conceptual model and review of the evidence supporting SBIRT through the five step process [1) Ask, 2) Acquire, 3) Appraise, 4) Apply, 5) Analyze & Adjust6Fink.  Results indicate that both groups performed equally as well with significant improvement in screening for drugs and alcohol. The number of brief interventions for patients at risk also improved ((Figure 4).  Dr Newhouse’s team will continue work on evaluation of evidence-based practice training models using the cluster-randomization method given her established alliance with regional network of hospitals.

Mixed-methods analyses in investigation of root causes to explain discharges against medical advice

Eberechukwu Onukwugha PhD, Assistant Professor, School of Pharmacy recently demonstrated that discharge against medical advice (DAMA) belie future health resource utilization as reflected in a CVD-related hospital readmission which was significantly greater than individuals who had a planned discharged for up to 6 months after index discharge7Fink. Ongoing research examines the costs associated with discharges against medical advice.   She has found that patients may leave against medical advice due to a variety of reasons, including perceptions of poor quality of hospital care.  Patients discharging from a high quality hospital are less likely to leave against medical advice compared to patients discharged from a low quality hospital as measured by JCAHO and H-CAHPS survey indicators of hospital quality8Fink.

Dr Onukwugha employed qualitative research to complement the empirical analyses by providing data regarding the patient’s reasons for leaving against medical advice.  In qualitative research conducted prior to and following the quantitative analyses, she explored patient and provider perspectives on patient’s reasons for leaving against medical advice.  Exploring the same issue from several perspectives and using mixed analytical methods provided the opportunity to compare patients’ perspectives with those of a variety of providers including physicians, nurses, and social workers.  Patients who left against medical advice and providers whose patients have left against medical advice participated in separate focus group interviews conducted at several Baltimore-area hospitals.  The qualitative research indicated that distance to the hospital was a relevant consideration for the patient in determining how likely they were to leave against medical advice.  The qualitative research also indicated that patients and providers did not always share the same perspective on DAMA.

Lastly, several reasons for DAMA emerged, some of which were related to the quality of care provided:

  1. drug seeking;
  2. inadequate pain management;
  3. obligations outside the hospital;
  4. long wait time;
  5. physician’s bedside manner was poor;
  6. teaching hospital setting was undesirable; and
  7. poor communication.

UMB quality research affiliates

Pharmaceutical Research Computing (PRC)

PRC is a research center within the Department of Pharmaceutical Health Services Research at the University of Maryland, Baltimore School of Pharmacy.  The mission of PRC is to provide research support for faculty, post-doctoral fellows, graduate students and other researchers by meeting their data warehousing, project management and statistical analysis needs.  PRC is entirely self-supported by revenues generated from the services we provide.  Since its establishment in 1998, PRC has worked on studies funded by federal government agencies (e.g., National Institutes of Health, Agency for Healthcare Research and Quality, the Food and Drug Administration), state government agencies (e.g., Pennsylvania Department of Public Welfare, Maryland Department of Health and Mental Hygiene, University of Massachusetts), foundations (e.g., Robert Wood Johnson Foundation, Commonwealth Foundation), pharmaceutical companies, and other public and private sectors.  PRC has extensive experience working with large administrative claims data (e.g., Medicare, Medicaid, HMO), secondary datasets (e.g., SEER-Medicare, Medicare Current Beneficiary Survey (MCBS), United States Renal Data System (USRDS), MarketScan, National Veterans Health Administration (VHA) data), electronic medical records (e.g., GE Centricity) and primary data including pharmacy and medical information (e.g., service evaluation, survey).  In addition, PRC has created a drug dictionary for both the MCBS community and institutional drug files.  Furthermore, the center has developed a web-based query system where authorized users can run queries and access results by file transfer protocol, as well as consultations on database design and data entry tools are also available.

The PRC personnel comprise a group of highly skilled professionals in the fields of information technology, statistics, computer programming, and pharmacy.  Besides a strong team of information technology specialists, programmers, and a statistician, who have expertise in data management and data analysis, we have a unique strength provided by the pharmacists on the team.  Their clinical expertise in pharmacotherapeutics, knowledge of reference files (e.g., drug dictionaries, ICD-9-CM, CPT, and HCPCS codes), and understanding in research methodology provide important contributions to study design, operationalization of variables, and data analysis.  Together, they provide quality research support for their client investigators.

Research HARBOR

Research HARBOR (Helping Advance Research By Organizing Resources) is described in detail elsewhere in this application has been designed to be an interactive, web-based portal that provides one-stop shopping for clinical and translational research support needs. Through this centralized hub, quality research investigators can access data warehouse capabilities, identify and utilize research support resources, tools and services, link with experts, access regulatory support, and identify educational and training opportunities.  The HARBOR Researchers Only space is the quintessential collection of data, tools, support services and educational materials available to support CTSA research. Investigators are often unaware or lacking in the necessary informatics skills to use the most efficient methods of data capture and management. Efficient solutions are often prohibitively expensive for an individual research project; whereas, a shared solution that is funded by a diverse array of stakeholders allows more cost-effective development and management of the solution. Further, it positions the community of researchers to be able to exploit the potential of data aggregated across sources, disciplines and levels of analysis (i.e., cellular, clinical trial, EHR, and community outcomes) to advance scientific discoveries more rapidly. The Research HARBOR Researchers Only space is organized around 5 “information docks”: Regulatory, Data Warehouse, Support Services, Research Tools, and Educational Resources.  These resources will enable researchers in quality of care (as described above) the opportunity to enhance their existing work and will serve as the home for the proposed Center for Quality.

Quality Research partners

Geisinger Health System (GHS)

GHS is an integrated health system that aspires to achieve the triple aim of providing world class quality care that also seeks to achieve high patient satisfaction at the most appropriate cost.  The strategy at GHS to achieve the triple aim has included building capacity to innovate, evaluate and learn as an organization.  For several years, GHS successfully managed a commitment to operational excellence as well as building the capacity to testing innovative models of care delivery.   GHS’s approach to effectively innovate in the context of a high-performing organization includes a number of activities and strategies including: Clinical Innovation and Decision Support: GHS’s division of Clinical Innovation and Decision Support seeks to design, implement and evaluate new models of care delivery to improve quality, efficiency and satisfaction with care.  Most of the work is designed to develop and leverage health information technology (HIT) coupled with key principles of clinical reengineering.  Clinical Reengineering: Concerned that existing models of care delivery will not be sufficient to meet the challenges of escalating costs and mediocre quality, GHS has designed and implemented a series of interventions and care delivery models that seek to move work away from physicians and provide that care through delegation to other care team members, automation through HIT, and activation of patients and families.  Digital Population Identification: In order to effectively reengineer care using HIT, GHS has develop methods to define and support the adoption of digital gold standard that identify what patients have what clinical conditions.  Accurate data located within the EHR forms the basis for effective decision support and ultimately supporting scalability and spread.

An example of GHS’s innovative approach to research and innovation is the ProvenCare line of clinical services, in which the HIT system is used to ensure that interdisciplinary, consensus best-practice care is provided to as many patients as possible (9). GHS initiated its first ProvenCare program in 2006 for elective coronary bypass surgery.  A 40-point checklist of surgical practices and care sub-processes proven to reduce complications was developed by GHS cardiovascular surgeons using recommendations of the American College of Cardiology and the American Heart Association.  The care process was redesigned and HIT support tools created to enable the completion and documentation of these processes.  This led to a substantial improvement in care-process performance:  all-process completion rates per patient increased from 59% to 95%, which has been sustained since 2007.  Clinical outcome also improved.

The ProvenCare model has since been extended to include other acute procedures, including total hip replacement, cataract surgery, bariatric surgery, percutaneous coronary intervention, and low-back surgery.  In implementing these innovations GHS has gained substantial experience with deploying clinical decision support tools within the EMR.  The GHS EMR includes 900 evidence-based inpatient order sets that are used to input 70% of orders.  Geisinger has also deployed over 100 EMR alerts and reminders with evidence of substantial improvement in the care of patients.  One example is the “diabetes bundle” of 9 best practice care processes for patients with diabetes.  Since the deployment of HIT-based support tools the all-or-none adherence to bundle best practice measures has increased over 300%.

This culture of innovation and commitment to HIT-supported, evidence-based, patient-focused care is ideal for developing and testing approaches to clinical practice. Key architects of GHS’s efforts in this area include, Jonathan Darer, MD MPH, Chief Innovation Officer, are co-investigators on the GHS eMERGE-II team.  In addition, GHS’s lead role in the Keystone Health Information Exchange (KeyHIE) and Keystone Beacon Community project provide means to export what is learned to non-Geisinger providers.  The Beacon project leverages KeyHIE (an existing health information exchange) to integrate evidence-based care redesign and care process integration across venues of care. GHS’s experience in this area provides a paradigm and institutional infrastructure for achieving the major goals of eMERGE Phase II.

To build on the proposed partnership in quality research between the UM CTSA and GHS a resource will be established to accommodate comprehensive, longitudinal data sets derived from mature EHRs such as that available at GHS, with the ability to preserve key phenotypic details with HIPAA compliant de-identification procedures.  The resource will include natural language processing applications to improve clinical operations and expand on quality research initiatives and provide capabilities to capture, store and examine patient self-reported data expanding on GHS’s online portal, deployed through mobile devices in outpatient clinics across the system and used to screen for depression, monitor asthma control, and track rheumatoid arthritis pain and functional outcomes.  The resource will also provide biostatistical and bioinformatics assets to develop and refine predictive modeling tools to identify high risk populations.

Center for Quality Research Advancement

The presentation of work conducted by several investigators at UMB and at GHS reveals the variety of topic areas, initiatives, and analytical methods pertinent to research in quality of health care.   Use of existing UMB resources through affiliates such as PRC and the Research HARBOR offer several elements necessary to assist researchers in their research progress and trajectories for growth.  The proposed partnership with GHS provides a platform for research validation and dissemination, and a fertile clinical environment to conduct further research initiatives.  In most cases a facility, which provides core services that would enhance growth in each discipline and expand on the promising findings of these and other investigators would be extremely useful and transformative to the broad quality research community at UMB.   The proposed center would be housed in the Research HARBOR and will aggregate several commonly-used research methods and capabilities which are instrumental to outcomes and quality research .   Key capabilities to be offered by the Center for Quality Research Advancement are as follows:

  1. Database warehousing and coordination:  As quality research often requires integration of data from multiple sources with harmonization of database configurations and variables, the Center will provide the capacity and resources to store disparate databases as expected in cluster randomization trials gathered from multiple centers or over time in quasi-experimental studies. Merging large data sets from health systems such as UMMS and GHS will require expert and effective data management resources to orchestrate joint analyses of such combined data sets.  The protocols and resources will be available allowing for necessary de-identification and standardization of heterogeneous data from independent data sources an EHR repositories.
  2. Statistical methodologies:  all quality and outcomes research utilize statistical methods to generate and test hypotheses.  The Center will offer both the personnel and tools needed for standard and novel statistical analyses.
  3. Computing resources: a key element to high-impact quality research is the use of research computing to conduct data simulations, decision analyses, Markov modeling and forecasting.  The Center will maintain and offer the processing capability and expertise to implement the necessary procedures to meet these computing needs and enhance the lines of inquiry of UMB quality researchers.
  4. Biomedical informatics: quality research initiatives within a successful clinical ecosystem such as GHS as described above, with the availability of a mature and functioning EHR require the availability of several biomedical informatics capabilities to implement a successful quality research program. The Center will including NLP and machine learning applications.  These tools organize and classify administrative and clinical health data for the purpose of high risk clinical population identification, identify of safety incidents of interest to a health system, determine key associations and health utilization patterns and prediction models for the purpose of personalizing treatment plans.
  5. Mixed-methods capabilities:  quality research, as demonstrated above, often call for the integration of both qualitative and quantitative methods and the proposed Center will offer the resources needed to conduct such coordinated analyses including space and equipment needed for focus groups and qualitative research and formalized approaches to mapping emerging themes from such activities with quantitative analyses.