Skip to main content
Sign In
 

D2V Research

Pilot Projects



New Projects:​​

 Data-Informed Management of Intracranial Aneurysms

Laura Wiley, PhD
Christopher Roark, MD

Approximately 1 in 50 individuals in the United States harbor an intracranial aneurysm with a small percentage of aneurysm rupturing each year. The consequence of aneurysm rupture is severe, with a 50% six-month mortality rate. Only 1/3 of survivors will return to their pre-rupture state of health. In the last thirty years, the quality and availability of advanced imaging have both grown exponentially and minimally invasive endovascular therapy has revolutionized the field. Due to these events, we are treating more incidental aneurysms than ever, however the expansion of preventative treatment has not lowered the population aneurysm rupture rate. Thus, there is a need to more selectively treat high risk aneurysm rather than treat more aneurysms. The PHASES score was developed as a meta-analysis of prospective studies of aneurysm rupture risk, but roughly 80% of the patients studied were from homogeneous populations (Japan and Finland). This score has not been validated in a large-scale study using a more heterogeneous population. We hypothesize that electronic health records (EHRs) can be used to provide a highly-detailed picture of patients with ruptured and unruptured aneurysms and provide a more diverse view of rupture risk. In this study, we will develop a continuously updating repository of all aneurysms treated at UCHealth since the institution of the Epic EHR. We will use then us this dataset to validate the PHASES score in the diverse UCHealth population. In future studies, we will collaborate with peer institutions to develop a more accurate risk prediction tool.​7

 Collection of Patients' Disability Status Data to Inform Patient-Centered Care

​Megan Morris, PhD, MPH, CCC-SLP

Approximately 19% of the U.S. population lives with a disability. A growing body of evidence reveals that people with disabilities are medically complex, have high rates of chronic conditions and experience significant disparities in the receipt and access to high quality healthcare services. To address these disparities, healthcare organizations need to systematically collect data on patients’ disability status. Patients’ disability status needs to be collected at the time of registration or scheduling so that the information can be used by the care team to inform patient-centered care. Despite policy requirements for collection, no standards exist on how to document patients’ disability status in a patient-centered manner.

The proposed study will develop and implement two strategies to increase and promote accurate and patient-centered collection of patients’ disability status: (1) training for frontline staff members who will collect the data and (2) a prompt that precedes the disability status questions and explains to patients why organizations are collecting the information. Using the UCHealth System as a “laboratory”, we will implement the study with PatientLine, the department that oversees patient registration and scheduling for all UCHealth outpatient primary care clinics. Using a step-wedge type approach, we will first implement the training for the staff members and then after two months, will implement the inclusion of the prompt. Guided by the RE-AIM framework, we will collect both quantitative and qualitative data on the frequency that patients’ disability status is asked and recorded, staff members’ and patients’ perceptions of implementing the interventions, and the time and resources required to implement the strategies.

Aligned with the D2V mission, we engaged key stakeholders in the development of the proposed study and will plan to continue to engage them throughout the study to ensure that the intervention and outcomes are aligned with patients’ and the healthcare systems’ values.​
8

 Data-Driven Innovations to Improve Discharge for the Medically Complex Hospitalized Patient

​Marisha Burden, MD, FACP, SFHM

Millions of people are hospitalized in the United States annually and many of those patients are medically complex requiring a team-based approach for their medical problems, complex planning for discharge and the transitions thereafter. Planning for these transitions ideally begins at admission but because of the variability of conditions and practices, this is not always the case.

Late discharges, in the afternoon or evening, adversely affects patient flow throughout the hospital which, in turn, results in delays in care, more medication errors, increased mortality, longer lengths of stay, higher costs, and lower patient satisfaction. Patients who are possible and definite discharges, as characterized by Burden et al., face different barriers to discharge with possible discharges typically awaiting assessment by other providers or awaiting tests. Orders for definite discharges are often delayed because providers are caring for other patients thus requiring different interventions for the different phenotypes. Typical solutions to improving the flow of these complex patients rely on one-way communication mechanisms and data entry that is not fluid (i.e. orders whereby providers estimate day of discharge in advance).

We propose a multi-faceted study, bridging patient-centered care and existing Epic functionality with data, to align team efforts around a communication tool within the electronic medical record that will improve communication, triaging of pending tests, and signal to care teams when action is needed to help facilitate discharge. The goals of this study are to engage patients, families, caregivers and care team members from a variety of disciplines to (1) assess functionality of existing tools and processes for documenting discharge readiness and directing discharge planning work (2) utilize this aforementioned work and our previous work on barriers to discharge to implement an Epic tool and (3) implement a multidisciplinary discharge team to prioritize definite discharges and facilitate earlier discharge from the hospital.​
9

 A Virtual Gluten-Free Diet Education Program for Pediatric Patients with Celiac Disease

​Pooja Mehta, MD, MSCI

Celiac disease (CD) is a chronic autoimmune disease with a rising incidence in the pediatric population. Children with celiac disease are medically complex – often with comorbidities such as Type 1 Diabetes and Down syndrome – and require the care of multidisciplinary teams. Strict adherence to a gluten-free diet remains the only treatment for CD yet there is no standardized method of educating children and families regarding this suddenly evoked lifestyle change. The current gold standard is attending an educational class with a registered dietitian but this practice is imperfect. First, classes are only offered in large tertiary care centers. This may preclude the attendance of children who live at a distance and particularly children from rural areas. Second, educational classes do not provide tailored information specific to individual circumstances. Third, dietitian-run educational classes do not incorporate psychological and behavioral interventions that have been shown to improve adherence. 

This proposal aims to overcome these barriers by creating a virtual gluten-free diet program that incorporates telehealth education sessions, digital uploads of an individual’s home environment, and virtual group sessions with a psychologist. To ensure the program is both effective and pragmatic, we will collaborate with D2V’s Stakeholder Engagement and Governance Core and the Patient
Systems and Value Core. By engaging stakeholders in the design process, creating interventions that incorporate information technology, and measuring patient outcomes, this novel proposal aligns with D2V’s mission of developing, implementing, and disseminating person-centered, high-value healthcare. The results of this pilot study will provide data to support a subsequent grant to AHRQ or the NIH. In this larger grant, we will conduct a nationwide study of a virtual program to treat nutrition-based gastrointestinal diseases.​
10

 Developing Processes to Use Actionable Data to Improve Care for Medically Complex Patients with Type 2 Diabetes

​Amy Huebschmann, MD, MS

This proposal seeks to improve healthcare value for complex patients by advancing innovations in data and health systems science. The focus of the proposal is on medically complex patients with type 2 diabetes (MCPs with T2D), defined as patients with uncontrolled blood glucose levels (HbA1c >9%) and at least one additional chronic condition, such as hypertension or arthritis. MCPs with T2D have dramatically worse health outcomes and higher medical costs than patients with normal glucose levels. It is promising that interventions to improve glycemic control have improved both health outcomes and costs, but these interventions must address glucose in the context of patients’ behavioral health risks and social/economic needs. Recently, evidence-based platforms have been developed to allow patients and primary care providers (PCPs) to share engaging and actionable displays of glucose data (Glooko platform) and data on behavioral health risks and social/economic needs (My Own Health Report (MOHR) platform, but these are not currently integrated with the University of Colorado Health (UCHealth) system primary care processes or health records. 

We propose a pilot study to engage with multiple stakeholders and potential end-users of these data—patients and PCPs/clinic staff—in order to identify and pilot-test the best method(s) of sharing and acting on Glooko and MOHR data. By engaging with MCPs with T2D and their providers to improve care and by collaborating with industry (Glooko, MOHR), this project addresses the D2V goals to consider multiple stakeholder perspectives on how best to utilize data to improve patient outcomes and experiences. Improving outcomes for MCPs with T2Ds also has the potential to reduce costs of care. This study would also provide pilot data for a future R01 proposal to NIH PA-18-007 (Dissemination and Implementation in Health) or AHRQ
PA-17-246 (Health Information Technology to Improve Health Care Quality and Outcomes).​
11


Ongoing Projects:

 The Value of Post-Acute Care to Patients and Family Caregivers Following Hospitalization

​Cari Levy, MD
Mark Moss, MD

As family caregivers are often the key to keeping older adults who have health problems at home, this study aims to determine how easy it will be to recruit caregivers of patients that have been hospitalized for a study and keep them in the study, so we can better research the experience of family caregivers. In addition, the study will look at how difficult family caregivers find different questionnaires that ask about their well being. Lastly, this study will explore, through interviews, what health and social services patients and families feel they need to recover after a hospitalization and their experience of the quality of services they receive that after a hospitalization in both the home setting or in a nursing home. This study is important as there is limited knowledge regarding what services patients and families need most following a hospitalization to recover, particularly as current research does show that most older adults would prefer to be cared for at home rather than in nursing homes.
6

 Expanding the Reach of Palliative Care: pilot study of a psychosocial intervention in Latino patients with advanced cancer

​David Bekelman, MD, PhD
Stacy Fischer, MD

People with serious illnesses have poor quality of life and are often depressed. Latino patients with advanced cancer rarely receive palliative care services that are demonstrated to improve quality of life. We developed a psychosocial intervention for patients with serious illness to help them adjust to symptoms and functional limitations that accompany serious illness. This psychosocial intervention is being tested in patients with heart failure and COPD. This proposal will adapt and pilot test this psychosocial intervention to Latinos with advanced cancer. This project is significant because it addresses common problems (depression and quality of life) in a large but underserved patient population (Latinos with advanced cancer). This project is innovative because (1) it examines the feasibility of several approaches to extend the reach of psychosocial and palliative care (trained lay persons to serve as translators for Spanish-speaking patients; video-conferencing to conduct the psychosocial intervention at a distance); and (2) conducts pilot testing in multiple sites in the UC Health System.​​​​3

 Record Linkage

Toan Ong, PhD
Michael Kahn, MD, PhD
Lisa Schilling, MD


Effective record linkage methods can be classified according to whether the linking variables have been encrypted (privacy preserving) or not (clear-text), and whether the decision to link records is based on a perfect match (deterministic) or an approximate match (probabilistic). While clear-text record linkage methods are well established, privacy preserving (encrypted) record linkage (PPRL) methods are new and perform poorly when confronted with data quality problems such as incorrect and missing data. Our team has developed novel PPRL methods to probabilistically link data under different data quality constraints. Because linked data are critical to many D2V research agendas, we will implement multiple scalable record linkage methods and test them using existing clinical and financial data sources. With ~ 4.2M patients in the
 
Health Data Compass (HDC) database, we will develop targeted cohorts such as adults with hospital-acquired surgical infections and children with traumatic brain injury to test the linkage performance features (sensitivity, specificity, PPV). We will also design, implement and test geo-spatial-based linkage techniques using state population/census data and public health exposure data from Colorado Department of Public Health & Environment (CDPHE). ​​

​​​​9

 Early Identification and Intervention for Frail Hospitalized Patients

Tyson Oberndorfer, MD, MS
David Kao, MD
Mark Gritz, PhD
John Welton, PhD, RN

Hospitalization and acute illness is a period of high risk for older patients, with a large proportion developing hospital acquired conditions and hospitalization-associated disability. The resulting increases in length of stay, loss of independence leading to post-acute institutionalization, and excess mortality exacts a high price from individual patients and families, health care providers, and society. Interventions like University of Colorado Hospital’s Acute Care for Elders Services can ameliorate some of the effects of acute illness and hospitalization, however, their reach is limited by bed availability. 

The proposed project will develop an approach to identifying at risk elders using nursing assessment data available on hospital admission to create and validate an automatically generated frailty index available in EPIC, identifying at risk patients throughout the health system at high risk for poor outcomes. An early intervention program will be created in which telehealth geriatric consultation and nursing protocols will be implemented to improve clinical outcomes for the frail older population across UCHealth hospitals.      

12

 Developing Causal Inference Methods for Nested (or Clustered) Data with Application to Common Pediatric Surgical Conditions

Debashis Ghosh, PhD

In healthcare learning systems, two potentially important questions of interest are to understand the surgical learning curve for a given procedure as well as to determine effects of surgeons on patient outcomes. The thrust of this proposal is that causal inference methods based on the surgeon-patient dyad could yield new findings above and beyond what has been given in the literature on volume/outcome relationships as well as on benchmarking physicians. In this grant, we propose the following three aims:

Aim 1: Develop a causal inference framework for quantification of surgical learning effects and expand the methodology to allow for clustering within hospitals;

Aim 2: Apply the methodology in Specific Aim 1 to data on tonsillectomy and spine fusion procedures from the Pediatric Health Information System (PHIS).

Aim 3: Determine if the results found in Specific Aim 2 generalize to the Premier database​
15

 Big Data to Nursing Value

John Welton, PhD, RN, FAAN

The primary purpose for funding is to develop partnerships and expertise at University of Colorado Anschutz to create the infrastructure necessary to store and standardize multiple hospitals EHR data and engage a multidisciplinary team to develop new research opportunities and future funding. Our ultimate goal is to precisely measure nursing costs per patient and to relate nurse assignment to patients’ outcomes, thus measuring the value of nursing care. Funding will allow assembly of a team to test different data extraction and data analysis methods using modern data science techniques. These efforts will lead to developing methods for many hospitals in the US to provide data within a consortium in the future and to position the Anschutz campus as leader in this research area.

Aim 1: Develop infrastructure to create a data science capability at the University of Colorado to aggregate large clinical and operational data focused on the nursing care of patients in multiple hospitals across the country.

Aim 2: Develop and test new analysis approaches to accurately measure nursing costs at a micro level linking nursing care staffing and assignment nurse patterns with short (daily) and hospital level outcomes to measure the value of nursing care.​
16

 Methods for Assessing Sustainability of Health Care Delivery Innovations: Application to Palliative Care for Neurologic Disorders

R. Brett McQueen, PhD

In the United States, more than 50% of all health care costs are attributable to 15% of individuals with life-limiting conditions and functional limitations. The disproportionately higher costs in this population reflects the complexity of care, with many patients forced to rely on acute hospital care given a lack of alternative options. In response to poor health outcomes and inefficient care for patients with life-limiting conditions, there is broad support to invest in innovative and sustainable care models. Sustainability requires improvement in patient outcomes along with positive financial returns for providers and health systems delivering the innovation. Providers are currently unable to assess sustainability of innovative care models, such as neurology palliative care (PC), because the link between cost and health outcomes is missing in health systems.

Aim 1: Develop process maps for each step of care delivery for patients with Parkinson's Disease (PD) at the University of Colorado neurology PC and comparator clinics. We will develop process maps to identify resources utilized at each step of care delivery for patients including facility usage, personnel, and equipment.

Aim 2: Estimate the incremental costs of expanding PC services for PD patients into other clinics. We will link process maps with financial data from CU Medicine to calculate average costs of implementing PC services for PD patients into clinic settings not currently offering PC services.

Aim 3: Quantify opportunities to improve returns on investing in PC services. Using costs and returns from investing in neurology PC services, we will identify sustainability scenarios that do not sacrifice quality of care and health outcomes for PD patients.​
17


Completed Projects:

 Exploring and Improving CHORDS for Asthma Research

Heather Hoch, MD 
Art Davidson, MD 
Stan Szefler, MD 

Over eight percent of children in the US carried a diagnosis of asthma in 2013, and 58% of those children had an asthma exacerbation in that year. Asthma exacerbations/hospitalizations remain a significant source of healthcare expenditures. Attempts have been made to develop risk scores in an attempt to predict those children most likely to suffer an asthma exacerbation. One previous attempt found that risk factors might differ by season of the year, with fall being the most prevalent season for asthma exacerbations. Our previous work sought to validate these risk factors (in the form of a risk index called the saEPI [seasonal asthma exacerbation predictive index]) in a small set of children with documented atopy and severe asthma, and though we validated this risk index, we also found that there may be room for improvement in the index’s reliability (manuscript recently accepted for publication, JACI). The limitation of this evaluation, however, was that the index has only been studied in children with severe asthma and atopy. With this evaluation utilizing the Colorado Health Observation Regional Data Service (CHORDS)​ data set, we aim to first evaluate whether or not the data exist in the database to fully evaluate this question, and second, to utilize the data available to test the risk index. Additionally, we hope to evaluate the repeatability of seasonal patterns of exacerbations in individual patients.​
1

 Preventing Surgical Infections

William Henderson, PhD, MPH 
Karl Hammermeister, MD 
Robert Meguid, MD, MPH 
Kathryn Colborn, PhD 


The National Surgical Quality Improvement Program (NSQIP) is an audit and feedback program of risk-adjusted surgical outcomes. It was originally developed in the VA healthcare system in 1991, and subsequently adopted by the American College of Surgeons for non-VA hospitals in 2005. Although currently deployed in >700 hospitals, NSQIP is limited by its sampling frame, its dependence on labor-intensive manual chart abstraction, and its lack of patient input. One objective of NSQIP is to estimate post-surgery infection rates. Many of these infections occur after patients have been discharged from the hospital. Identification of these infections is done through manual chart review, and because of the labor involved, only about 15% of patients are assessed. The objective of this project is to develop an electronic method of identifying surgical infections through supervised learning. Using a dataset of more than 6,000 chart reviewed records of surgical patients from 2013-2016, the team will develop a model or combination of models that predicts observed infections well using patients’ clinical information, such as ICD-9 codes, CPT codes, drug prescriptions and demographic data. Various machine learning algorithms and ensemble methods will be compared with respect to their squared error loss, sensitivity, specificity and other relevant statistics in order to identify the best technique for identifying infections using electronic medical record data. Data that will be used for this project include the All Payer Claims Database, EPIC and NSQIP. Dr. Henderson is a biostatistician in the Department of Biostatistics in the Colorado School of Public Health. Dr. Hammermeister is a cardiologist in the School of Medicine. They have worked on the NSQIP since 1991. Dr. Meguid is a thoracic surgeon in the Department of Surgery in the School of Medicine, and the Surgical Director of the Surgical Outcomes and Applied Research Program. Dr. Colborn is a biostatistician in the Division of Health Care Policy and Research in the School of Medicine. ​​​​

2

 Pediatric Traumatic Brain Injury in Colorado

Tell Bennett, MD, MS

Traumatic brain injury (TBI) is the leading cause of death and acquired disability in U.S. children. Little high-quality evidence is available to guide acute care decisions. Unfortunately, many randomized trials of interventions in children and adults with TBI have failed, at least in part because of variation in care between centers. This project will 1) create a single dataset that contains information about the entire continuum of care for children with TBI in Colorado. To accomplish this, we will leverage D2V informatics and record linkage capabilities to link the Colorado All Payer Claims Database (APCD) with the trauma registries and electronic medical records (EMRs) of the two pediatric trauma centers in Colorado, Childrens Hospital Colorado (CHCO) and Denver Health (DH). Using the linked dataset, we will 2) characterize the medical needs and functional outcomes of children who survive their TBI and are discharged from the hospital back to their community and 3) use observational causal inference techniques to assess the effectiveness of various treatment strategies.​​

3

 Strategies to Improve Value in Noninvasive Cardiovascular Testing

​Vinay Kini, MD, MSPH

Aim 1: Identify hospitals in Colorado (n=86) that perform high- and low-value NCT at outlier levels, and describe patient-, provider-, and site-level characteristics associated with their use.

Aim 2: Establish a Stakeholder Panel to 1) provide context regarding patterns of high- and low-value NCT identified in Aim 1, and 2) inform development of an interview guide.

This research will 1) use a novel data source to generate new insights into the effect of insurance and payer mix on healthcare value, and 2) engage patient, provider, and health system stakeholders to begin identification of transferable strategies to achieve high value healthcare. This pilot study will generate preliminary data for future funding applications including an AHRQ K08 award and an American Heart Association Scientist Development Grant.

The study is consistent with D2V’s mission of implementing high value healthcare by advancing innovations in data science. By drawing upon the expertise of D2V’s Analytics and Stakeholder Engagement Cores, we will leverage a novel data source and engage key stakeholders to identify and successfully implement strategies to maximize value in NCT.​​
4

 The Feasibility, Safety, and Value of a Virtual Cardiac Implantable Electronic Device Wound Check

Lucas Marzec, MD
Larry Allen, MD
Daniel Matlock, MD, MPH
Mark Gritz, PhD
Gary Grunwald, PhD
Megan Morris, PhD
Heidi Wald, MD, MSPH

Infection is the most common complication following implantation of a cardiac implantable 
electronic device (CIED). Standard of care is for patients to be seen in-person by a nurse within 
2 weeks following CIED implantation to evaluate the surgical site. These in-person CIED wound 
check visits burden patients, providers, and UCHealth. A virtual nurse visit may be a safe 
alternative that reduces the burden on patients, providers, and the UCH system. However, the 
safety, effectiveness, and cost savings of a virtual wound check visits is unknown. We will 
conduct a single-center prospective cohort study of 75 patients implanted with a CIED at the UC 
Hospital to assess the safety and value of a virtual nurse wound check visit compared to an in-person 
nurse wound check visit.​

Aim 1: In collaboration with the D2V Stakeholder Engagement and Governance Core, establish a multi-stakeholder panel to guide refinement of our current virtual CIED wound check protocol.

Aim 2: Assess the safety of a virtual CIED wound check visit in comparison to an in-person nurse CIED wound visit.

Aim 3: Assess the value of the virtual CIED wound check visits compared to in-person wound check visits, through the use of semi-structured patient interviews, nurse, physician, and administrator focus groups, and time-driven activity-based costing​.​
5

 Linking Research Prioritization with Health Outcomes: A Value-based Approach

Jonathan Campbell, PhD
R. Brett McQueen, PhD

Research question: “Are quantitative health outcome methods practical and actionable for decisions related to the prioritization of future research?" To address this question, we will use the ICER framework as a launching pad to generate new VOI outputs from previous modeling analyses that inform optimal future clinical trial design for a rheumatoid arthritis case.

Aim 1: Estimate value of information outputs to inform future clinical trial design. Using a previously developed value-based decision model in moderate-to-severe rheumatoid arthritis,5 we will estimate specific characteristics to inform the design of future rheumatoid arthritis clinical trials including optimal sample size, therapeutic comparators, endpoints, and follow-up times.

Aim 2: Evaluate stakeholder perceptions for research prioritization with and without value of information analyses.
Sub-aim 2.1 Educate stakeholders on the use of VOI methods.
Sub-aim 2.2 Conduct rank experiments with stakeholders to evaluate the additional benefit of VOI methods for the prioritization of future research.
Sub-aim 2.3 Evaluate and disseminate stakeholders’ perceptions of VOI analyses on the prioritization of future research.​​
​​6