This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Research Protocols, is properly cited. The complete bibliographic information, a link to the original publication on http://www.researchprotocols.org, as well as this copyright and license information must be included.
Durable ventricular assist device (VAD) therapy is reserved for patients with advanced heart failure who have a poor estimated 1-year survival. However, despite highly protocolized management processes, patients are at a unique risk for developing a health care–associated infection (HAI). Few studies have examined optimal strategies for HAI prevention after durable VAD implantation, despite variability in rates across centers and their impact on short- and long-term outcomes.
The objective of this study is to develop recommendations for preventing the most significant HAIs after durable VAD implantation. The study has 3 specific aims: (1) identify determinants of center-level variability in HAI rates, (2) develop comprehensive understanding of barriers and facilitators for achieving low center-level HAI rates, and (3) develop and disseminate a best practices toolkit for preventing HAIs that accommodates various center contexts.
This is a sequential mixed methods study starting with a cross-sectional assessment of current practices. To address aim 1, we will conduct (1) a systematic review of HAI prevention studies and (2) in-depth quantitative analyses using administrative claims, in-depth clinical data, and organizational surveys of VAD centers. For aim 2, we will apply a mixed methods patient tracer assessment framework to conduct semistructured interviews, field observations, and document analysis informed by findings from aim 1 at 5 high-performing (ie, low HAIs) and 5 low-performing (ie, high HAI) centers, which will be examined using a mixed methods case series analysis. For aim 3, we will build upon the findings from the previous aims to develop and field test an HAI preventive toolkit, acquire stakeholder input at an annual cardiac surgical conference, disseminate the final version to VAD centers nationwide, and conduct follow-up surveys to assess the toolkit’s adoption.
The project was funded by the Agency for Healthcare Research and Quality in 2018 and enrollment for the overall project is ongoing. Data analysis is currently under way and the first results are expected to be submitted for publication in 2019.
This mixed methods study seeks to quantitatively assess the determinants of HAIs across clinical centers and qualitatively identify the context-specific facilitators and barriers for attaining low HAI rates. The mixed data findings will be used to develop and disseminate a stakeholder-acceptable toolkit of evidence-based HAI prevention recommendations that will accommodate the specific contexts and needs of VAD centers.
PRR1-10.2196/14701
Heart failure affects nearly 5.7 million Americans and is a contributing cause of 1 in 9 deaths [
Patients with VADs are at a heightened risk for device-related and nondevice-related health care–associated infection (HAI) despite a highly protocolized perioperative and postoperative course. The nature of durable VAD therapy requires an uninterrupted external power source connected to the patient through a driveline (ie, a percutaneous lead to provide power and control to the implantable pump), which serves as a potential source for the development of HAIs given its connectivity between the patient and the external environment [
The majority of HAIs occur within 90 days after a VAD implantation, with a decreased but ongoing risk beyond the 90-day period [
Large variation exists in both the rate of HAIs and, more notably, the adoption of preventive strategies (eg, checklists, effective teamwork, and unit and center leadership) across clinical centers [
Conceptual model for health care–associated infection development and impact. VAD: ventricular assist device.
Few studies have focused on understanding both the optimal strategies for HAI prevention after VAD implantation as well as the approaches for enhancing the local adoption of these practices. Relying on administrative claims or clinical data is insufficient for identifying novel HAI practices, let alone understanding the characteristics of center- and unit-level strategies and contexts contributing to HAIs. To address this research gap, this mixed methods study seeks to identify recommendations for preventing the most clinically significant HAIs after VAD implantation. First, we plan to characterize key determinants of HAIs using quantitative approaches. Subsequently, we plan to identify context-specific promoters and barriers to preventing HAIs across low- and high-performing centers using qualitative approaches. More importantly, a multidisciplinary study team will use both findings to develop an action-oriented modular toolkit that provides evidenced HAI prevention recommendations, which accommodates the specific needs of individual centers. This project will contribute to research knowledge and interventions aimed at preventing HAIs by encouraging field-wide adoption of evidence-based practices.
As shown in
Overall study design. HAI: health care–associated infection; INTERMACS: Interagency Registry for Mechanically Assisted Circulatory Support.
To address this aim, we will first conduct a systematic review of interventional HAI prevention studies. Second, we will supplement the clinical dataset with (1) data from an electronically distributed survey of VAD centers to identify HAI prevention strategies not already tracked through INTERMACS and (2) center-specific measures of provider social network configurations using a merged INTERMACS-Medicare dataset [
Given that Medicare is the largest US health care payer, Center for Medicare and Medicaid Services (CMS) is the sole national data source of exhaustive claims for noninstitutional and institutional providers. The CMS files contain data about (1) beneficiary (eg, age, diagnoses, and type of benefits); (2) institutional admissions; (3) provider services; (4) outpatient, hospice, home health agency, and skilled nursing facility services; and (5) prescription drugs (eg, heart failure medications). Each file contains data about the date and location of services and payments. We will characterize the social networks of providers (eg, nurse practitioners, physician assistants, and physicians) as a measure of provider teamwork. Following precedent from prior research on physician social networks, we will restrict these networks to specific provider specialties (eg, anesthesiology and surgery) that are most likely to be directly involved with the care of VAD patients [
Data from this annual survey, to be merged with CMS files, contain center-specific structure and organizational measures (eg, size, urban or rural location, teaching status, affiliation with networks, training programs, ownership, staffing levels, and types of surgical services provided).
INTERMACS is a multicenter, Joint Commission–recognized, nationally audited database of FDA-approved VADs. INTERMACS contains extensive data regarding (1) preimplant details (eg, age and comorbid diseases); (2) operative details (eg, device type and operative duration); (3) patient status until death, cardiac transplantation, or device removal; (4) adverse events (eg, device malfunctions); and (5) functional status, neurocognition, and health-related quality of life. INTERMACS monitors device-related (driveline, exit cannula, pump pocket, and pump interior) and nondevice-related (positive blood cultures, central line associated sepsis, pulmonary, urinary tract, mediastinum, peripheral wound, gastrointestinal) HAIs.
The survey, informed by a systematic review of the literature and prior work, will identify potential HAI determinants that are not available through existing data sources [
Preliminary survey drafts will be developed based on a conceptual framework and by our investigative team of clinical and research experts, with input from survey and decision science experts. We will test the draft survey instrument with key stakeholders at selected centers with low and high HAI rates to ensure readability, face validity, and completeness of response options. Interviews will be conducted by Web-based teleconference in real time as the draft survey is completed to solicit qualitative feedback on survey content. The survey will be revised based on this feedback.
We will use evidence-based strategies to maximize survey response rate and minimize nonresponse bias, especially as a function of a center’s HAI rate. For example, the email invitation will be personalized, and the survey length will be limited. In addition, nonrespondents (VAD coordinator, surgeon, and cardiologist) will be contacted by the study team as necessary. Our team will provide respondents with a small gift card after completing the survey.
We project analyzing data among 9339 VAD patients receiving durable VAD from 2008 to 2017. We will administer surveys to potential respondents (VAD coordinator, surgeon, and cardiologist) from an estimated 153 VAD centers.
First, we will conduct a systematic review of the published literature to identify established HAI preventive practices [
Our primary endpoint is any HAI 90 days after device implant. We will account for HAI competing risks (eg, death and device removal) and other censoring (follow-up less than 90 days) using time-to-event analysis. Furthermore, we will use 2 approaches for modeling HAI risk. First, we will use the Cox proportional hazards model. We will adjust for surgery year and season and investigate interactions by key biological variables (eg, patient age, sex, and race). We will account for patient demographics (eg, age and sex), disease characteristics (eg, pulmonary disease), and surgical history (eg, prior cardiac surgery) and study the effect of intraoperative care (eg, cardiopulmonary bypass duration) and postoperative care (eg, duration of intubation). We will use the Lasso variable selection method to assess variables for model inclusion [
We will also conduct exploratory analyses for device-related and nondevice-related HAIs and their main components (to address heterogeneity of risk factors across HAI subtypes). We will quantify the importance of a group of factors in explaining center variability in HAI by fitting models with and without such factors and comparing how well the expected numbers from different models track with the observed numbers. Primary analyses will include all centers, and for the purpose of sensitivity analysis, we will also exclude low-volume centers. We will explore different center volume thresholds, trying to strike a balance between 2 considerations: stable estimate of HAIs rate (quantified by standard error) and number of centers remaining for analysis. We will use simple mean imputation, median imputation, or multiple imputation to account for anticipated missing covariates depending on their pattern and mechanism [
Provider social networks represent a measure of teamwork and communication that are associated with the quality of care [
Claims data do not map perfectly to true care patterns; therefore, our network measures may include false negatives and false positives. To assess for these biases, we will run several sensitivity analyses, motivated by prior work on missing data and thresholding techniques in the study of social networks [
The approaches discussed above consider the time-to-event endpoint (ie, time to first HAI), which counts multiple HAIs from a single patient only once. Alternatively, we will consider the recurrent event endpoint, where multiple HAIs from a single patient are counted multiple times. We will use a counting process, Anderson-Gill model, to rerun the analysis using a recurrent event endpoint to help distinguish the 2 outcome scenarios [
A mixed methods patient tracer assessment framework will be applied (adapted from the Joint Commission’s tracer methodology) in aim 2 to examine center-level resources and local barriers and facilitators for achieving low HAI rates [
A priori, we will use HAI rates derived from aim 1 to intentionally sample 5 high-performing centers with low HAI rates and 5 low-performing centers with high HAI rates for site visits. However, the sampling process will be further refined based on other criteria pending the nature of findings emerging from aim 1 and center-specific practices (eg, measures of provider teamwork).
We will conduct 2-day site visits at 5 high-performing and 5 low-performing centers. The mixed methods patient tracer assessment procedure focuses on the patient’s trajectory from the index hospitalization to 90 days. Before site visits, we will conduct advanced analytics to examine performance features, measures, and other quantitative data that will inform qualitative data collection. During the center visits, the site-visiting team will systematically trace the patient’s movement through the health system and investigate HAI risk factors or preventive strategies at each
The site-visiting team will follow a patient’s trajectory and conduct semistructured interviews with stakeholders impacting VAD patient care, including administrators (eg, chief medical or nursing officer, quality or safety officer, physician service chief, clinical unit manager, and hospital epidemiologist), physicians (eg, attending surgeon, intensivist, hospitalist, and medical consultants—pulmonologist and cardiologist), advanced practice providers (eg, nurse practitioner and physician assistant), and nurses (eg, ward and intensive care unit). Informed consent will be obtained from informants before starting each interview. Each interview, which will be conducted in private offices or conference rooms, will last for 40 min to 60 min. Stakeholder interviews will focus on answering questions from the advanced analytics, understanding perceptions of the center’s resources and local strategies for HAI prevention, and eliciting key features that would enhance local toolkit adoption. Interviews will continue until reaching informational redundancy or saturation (ie, no new information is being identified) at each center. Following each interview, the study team will provide a gift certificate to each interviewee to acknowledge his or her contribution.
For each site visit, the site-visiting team will conduct direct observations of the clinical work environment and behavior within each
During each site visit, we will collect relevant protocol documents and interview participants at the relevant transition point about the use of the documents, variations, or other volunteered information. Other relevant documents, patient education documents, postsurgical order sets, and variations by surgeons will be collected.
Preliminary analysis will begin with a research team debrief at the end of each day, which will reflect team members’ field notes, impressions, and observations. These recorded conversations will provide an additional source of data for defining facility- and unit-level contextual characteristics. Debriefing sessions during day 1 will also identify areas of focus and inquiry for day 2. On day 2, we will explore questions raised on day 1 to develop an expanded understanding. At the end of day 2, the most current version of the toolkit will be presented to the sites for feedback.
For each site, we will develop a case study following a structured outline that will have a sufficiently flexible format to accommodate site variation [
All quantitative (aim 1 findings about the specific sites) and qualitative data (aim 2 findings) will be integrated into the case study for each site. Related findings from both data sources will be matched to provide a comprehensive understanding of each center’s HAI prevention strategies. Multiple case series analyses will be performed. As appropriate for related findings, joint display analysis (ie, the process of iteratively creating, interpreting, and restructuring tables and figures that integrate the quantitative and qualitative findings) will be used to illuminate similarities and differences in HAI prevention strategies across centers and draw overall conclusions [
Member checking is the process of providing qualitative and mixed methods findings back to study participants to elucidate their input on the overall interpretation [
The study team will create a HAI preventive toolkit of evidence-based recommendations that may be customized to the context of each center. The development of the toolkit will be informed by the findings from aims 1 and 2. We will field test our prototype toolkit at the same 10 centers that participated in aim 2. This process will identify approaches for optimizing adoption (ie, fit and suitability for everyday use) as well as inform local context needs and any necessary modifications (eg, content, design, and options). The team will host a dedicated 90-min session at an annual cardiac surgical conference (eg, the International Society for Heart and Lung Transplantation) to elicit further user feedback about the toolkit’s acceptability. We will use the feedback to develop a final toolkit, which will be distributed to US VAD centers. Finally, a follow-up survey will be electronically distributed to US VAD centers to assess national adoption rates of the toolkit.
We will develop an initial version of a printed HAI prevention toolkit of evidence-based recommendations based on the Agency for Healthcare Research and Quality (AHRQ) guidelines [
Development of the health care–associated infection preventive toolkit. HAI: health care–associated infection; INTERMACS: Interagency Registry for Mechanically Assisted Circulatory Support; VAD: ventricular assist device.
The prototype toolkit will be field tested at the same 10 centers that participated in aim 2. Study team members will conduct audio-recorded interviews of VAD clinical stakeholders to elicit user feedback regarding the toolkit’s usability, acceptability, and likelihood of significant effectiveness for preventing HAIs using a
A moderated expert panel will be convened at an annual cardiac surgical conference (eg, the International Society for Heart and Lung Transplantation) to gather stakeholder input on the refined toolkit. During this session, we will share the rationale and intended use of the toolkit and highlight the toolkit’s usability, acceptability, and likelihood of significant effectiveness. Attendees will use an audience response system to provide Likert-scaled responses to each question and will be invited to offer additional qualitative feedback concerning the reasons underlying their quantitative survey responses. Panelists (representing cardiac surgery, heart failure cardiology, infectious disease, and epidemiology) will be invited to respond to attendees’ remarks. The entire session will be double audiotaped (with deidentified transcripts created thereafter). Analysis of the audience response system data and the themes emerging from the discussion for each item will be integrated to draw overall conclusions and inform final revisions of the toolkit. Thereafter, we will distribute the final version to US VAD centers.
A follow-up survey intended to assess toolkit adoption rates among US VAD centers will be developed and electronically distributed using the same approach as the center survey described in aim 1. The survey will be pretested to assess face and content validity, comprehensibility, time to completion, and ambiguity. The survey is anticipated to solicit information using a mixture of multiple-choice responses (eg, awareness of the toolkit, roles of those involved with local adoption, identified surgeon champion, types of resource support provided to the adoption team, and frequency of team meetings) and open-ended responses (eg, method used to implement the toolkit, perceived barriers and facilitators for implementation, ongoing quality improvement initiatives to enhance local adoption, and perceived effectiveness of the toolkit).
The project was funded by the AHRQ in 2018 and enrollment for the overall project is ongoing. We are conducting a systematic review of interventional HAI prevention studies and developing the survey concerning HAI determinants across US VAD centers. We anticipate that survey data collection will begin in November 2019. Findings from aims 1 and 2 will be used to develop a toolkit of evidence-based HAI prevention practices that may be adopted to the local contexts and across VAD centers. The first results are expected to be submitted for publication in 2019.
Although current emphasis is placed on surgical technical competence and checklists, further improvements in patient safety and outcomes may only be achieved with greater attention to optimizing the organization of clinical practice to reliably deliver safe and effective care. Our mixed methods study has several strengths. First, we will employ network analytic tools to assess whether provider teamwork is a determinant of variation in center-level HAI rates. This analytic method will enable us to account for differences in collaboration and communication across provider teams that would not be captured through traditional patient risk factors. Second, we will employ a novel patient tracer mixed methods assessment in which center-specific outcomes inform our qualitative investigation as we follow a hypothetical patient through each critical transition (or stop) of the patient’s care trajectory. Using this novel technique, we will identify potentially modifiable contexts, communication, and practices that could be missed if solely relying on quantitative approaches. Third, we will enhance our toolkit’s adoption by field testing a prototype during site visits to assess end-user usability and adoption, incorporating broad provider community input at an annual cardiac surgical conference (eg, the International Society for Heart and Lung Transplantation) before national dissemination and conducting follow-up surveys to assess the uptake of the toolkit.
Although unlikely, there are a few unanticipated challenges with this study. First, it is possible that we will not identify determinants of HAI across centers. There is a possibility that we will not find distinct HAI determinants across centers. However, given the documented center variability in anticoagulation practices and pump implant techniques, we anticipate ample variability related to HAI prevention practices [
This study seeks to elucidate determinants of HAI across clinical centers using quantitative approaches and identify context-specific facilitators and barriers for attaining low HAI rates using qualitative approaches. We will use these findings to develop and disseminate a stakeholder-acceptable toolkit of evidence-based HAI prevention recommendations that will accommodate the specific needs of VAD centers and address AHRQ patient safety guidelines [
Agency for Healthcare Research and Quality
Center for Medicare and Medicaid Services
Food and Drug Administration
health care–associated infection
Interagency Registry for Mechanically Assisted Circulatory Support
National Heart, Lung, and Blood Institute
National Institutes of Health
ventricular assist device
The INTERMACS dataset was provided by Dr. Marissa Miller, National Heart, Lung, and Blood Institute (NHLBI) of the National Institutes of Health (NIH). Data in INTERMACS used for this study were previously collected under contract with the NIH (contract number HHSN268201100025C). This project was supported by grant number R01HS026003 from the AHRQ, US Department of Health and Human Services. Opinions expressed in this manuscript do not represent those of INTERMACS, NHLBI, CMS, US FDA or AHRQ.
None declared.