Leveraging artificial intelligence to advance implementation science: potential opportunities and cautions.

BACKGROUND: The field of implementation science was developed to address the significant time delay between establishing an evidence-based practice and its widespread use. Although implementation science has contributed much toward bridging this gap, the evidence-to-practice chasm remains a challenge. There are some key aspects of implementation science in which advances are needed, including speed and assessing causality and mechanisms. The increasing availability of artificial intelligence applications offers opportunities to help address specific issues faced by the field of implementation science and expand its methods. MAIN TEXT: This paper discusses the many ways artificial intelligence can address key challenges in applying implementation science methods while also considering potential pitfalls to the use of artificial intelligence. We answer the questions of "why" the field of implementation science should consider artificial intelligence, for "what" (the purpose and methods), and the "what" (consequences and challenges). We describe specific ways artificial intelligence can address implementation science challenges related to (1) speed, (2) sustainability, (3) equity, (4) generalizability, (5) assessing context and context-outcome relationships, and (6) assessing causality and mechanisms. Examples are provided from global health systems, public health, and precision health that illustrate both potential advantages and hazards of integrating artificial intelligence applications into implementation science methods. We conclude by providing recommendations and resources for implementation researchers and practitioners to leverage artificial intelligence in their work responsibly. CONCLUSIONS: Artificial intelligence holds promise to advance implementation science methods ("why") and accelerate its goals of closing the evidence-to-practice gap ("purpose"). However, evaluation of artificial intelligence's potential unintended consequences must be considered and proactively monitored. Given the technical nature of artificial intelligence applications as well as their potential impact on the field, transdisciplinary collaboration is needed and may suggest the need for a subset of implementation scientists cross-trained in both fields to ensure artificial intelligence is used optimally and ethically.

Stakeholder perspectives on current determinants of ultrasound-guided thoracentesis in resource limited settings: a qualitative study.

OBJECTIVE: Preprocedure pleural fluid localization using bedside ultrasound has been shown to reduce complications related to thoracentesis and is now considered the standard of care. However, ultrasound-guided thoracentesis (USGT) has not been broadly adopted in many low-resource settings. With increasing affordability and portability of ultrasound equipment, barriers to USGT are changing. The aim of this multisite qualitative study is to understand the current barriers to USGT in two resource-limited settings. SETTING: We studied two geographically diverse settings, Harare, Zimbabwe, and Kathmandu, Nepal. PARTICIPANTS: 19 multilevel stakeholders including clinical trainees, attendings, clinical educators and hospital administrators were interviewed. There were no exclusion criteria. PRIMARY OUTCOME: To understand the current determinants of USGT adoption in these settings. RESULTS: Three main themes emerged from these interviews: (1) stakeholders perceived multiple advantages of USGT, (2) access to equipment and training were perceived as limited and (3) while an online training approach is feasible, stakeholders expressed scepticism that this was an appropriate modality for procedural training. CONCLUSION: Our data suggests that USGT implementation is desired by local stakeholders and that the development of an educational intervention, cocreated with local stakeholders, should be explored to ensure optimal contextual fit.

Use of the reach, effectiveness, adoption, implementation, and maintenance (RE-AIM) framework to guide iterative adaptations: Applications, lessons learned, and future directions.

Introduction: Implementation science frameworks have been used widely for planning and evaluation, but seldom to guide adaptations during program implementation. There is great potential for these frameworks to be used to inform conceptual and data-driven decisions about adaptations. Methods: We summarize recent applications using Iterative RE-AIM to capture and guide adaptations. Iterative RE-AIM can be repeated at multiple time points customized to each project and involves the following activities: identification of key implementation partners; rating importance of and progress on each RE-AIM dimension (reach, effectiveness, adoption, implementation, and maintenance); use of summary data on ratings to identify one or two RE-AIM dimensions for adaptations and implementation strategies; and evaluation of progress and impact of adaptations. We summarize recent and ongoing Iterative RE-AIM applications across multiple care coordination and pain management projects within the Veterans Health Administration, a hypertension control trial in Guatemala, a hospital-based lung ultrasound implementation pilot, and a colorectal cancer screening program in underserved communities. Results: Iterative RE-AIM appears feasible, helpful, and broadly applicable across diverse health care issues, interventions, contexts, and populations. In general, the RE-AIM dimension showing the largest gap between importance and progress has been Reach. The dimensions most frequently selected for improvement have been Reach and Implementation. We discuss commonalities, differences and lessons learned across these various applications of Iterative RE-AIM. Challenges include having objective real time data on which to make decisions, having key implementation staff available for all assessments, and rapidly scoring and providing actionable feedback. We discuss print and online resources and materials to support Iterative RE-AIM. Conclusions: The use of Iterative RE-AIM to guide and support understanding of adaptations has proven feasible across diverse projects and in multiple case studies, but there are still questions about its strengths, limitations, essential components, efficiency, comparative effectiveness, and delivery details. Future directions include investigating the optimal frequency and timing for iterative applications; adding contextual assessments; developing more continuous and rapid data on which to make adaptation decisions; identifying opportunities to enhance health equity; and determining the level of facilitation that is most cost-effective.

Using Iterative RE-AIM to enhance hospitalist adoption of lung ultrasound in the management of patients with COVID-19: an implementation pilot study.

BACKGROUND: Lung ultrasound (LUS) is a clinician-performed evidence-based imaging modality that has multiple advantages in the evaluation of dyspnea caused by multiple disease processes, including COVID-19. Despite these advantages, few hospitalists have been trained to perform LUS. The aim of this study was to increase adoption and implementation of LUS during the 2020 COVID-19 pandemic by using recurrent assessments of RE-AIM outcomes to iteratively revise our implementation strategies. METHODS: In an academic hospital, we implemented guidelines for the use of LUS in patients with COVID-19 in July 2020. Using a novel "RE-AIM dashboard," we used an iterative process of evaluating the high-priority outcomes of Reach, Adoption, and Implementation at twice monthly intervals to inform revisions of our implementation strategies for LUS delivery (i.e., Iterative RE-AIM process). Using a convergent mixed methods design, we integrated quantitative RE-AIM outcomes with qualitative hospitalist interview data to understand the dynamic determinants of LUS Reach, Adoption, and Implementation. RESULTS: Over the 1-year study period, 453 LUSs were performed in 298 of 12,567 eligible inpatients with COVID-19 (Reach = 2%). These 453 LUS were ordered by 43 out of 86 eligible hospitalists (LUS order adoption = 50%). However, the LUSs were performed/supervised by only 8 of these 86 hospitalists, 4 of whom were required to complete LUS credentialing as members of the hospitalist procedure service (proceduralist adoption 75% vs 1.2% non-procedural hospitalists adoption). Qualitative and quantitative data obtained to evaluate this Iterative RE-AIM process led to the deployment of six sequential implementation strategies and 3 key findings including (1) there were COVID-19-specific barriers to LUS adoption, (2) hospitalists were more willing to learn to make clinical decisions using LUS images than obtain the images themselves, and (3) mandating the credentialing of a strategically selected sub-group may be a successful strategy for improving Reach. CONCLUSIONS: Mandating use of a strategically selected subset of clinicians may be an effective strategy for improving Reach of LUS. Additionally, use of Iterative RE-AIM allowed for timely adjustments to implementation strategies, facilitating higher levels of LUS Adoption and Reach. Future studies should explore the replicability of these preliminary findings.

Comparison of four handheld point-of-care ultrasound devices by expert users.

BACKGROUND: Point-of-care ultrasound (POCUS) is rapidly becoming ubiquitous across healthcare specialties. This is due to several factors including its portability, immediacy of results to guide clinical decision-making, and lack of radiation exposure to patients. The recent growth of handheld ultrasound devices has improved access to ultrasound for many clinicians. Few studies have directly compared different handheld ultrasound devices among themselves or to cart-based ultrasound machines. We conducted a prospective observational study comparing four common handheld ultrasound devices for ease of use, image quality, and overall satisfaction. Twenty-four POCUS experts utilized four handheld devices (Butterfly iQ+™ by Butterfly Network Inc., Kosmos™ by EchoNous, Vscan Air™ by General Electric, and Lumify™ by Philips Healthcare) to obtain three ultrasound views on the same standardized patients using high- and low-frequency probes. RESULTS: Data were collected from 24 POCUS experts using all 4 handheld devices. No single ultrasound device was superior in all categories. For overall ease of use, the Vscan Air™ was rated highest, followed by the Lumify™. For overall image quality, Lumify™ was rated highest, followed by Kosmos™. The Lumify™ device was rated highest for overall satisfaction, while the Vscan Air™ was rated as the most likely to be purchased personally and carried in one's coat pocket. The top 5 characteristics of handheld ultrasound devices rated as being "very important" were image quality, ease of use, portability, total costs, and availability of different probes. CONCLUSIONS: In a comparison of four common handheld ultrasound devices in the United States, no single handheld ultrasound device was perceived to have all desired characteristics. POCUS experts rated the Lumify™ highest for image quality and Vscan Air™ highest for ease of use. Overall satisfaction was highest with the Lumify™ device, while the most likely to be purchased as a pocket device was the Vscan Air™. Image quality was felt to be the most important characteristic in evaluating handheld ultrasound devices.

Longitudinal analysis of serological responses of adults to Helicobacter pylori antigens.

Because Helicobacter pylori persist for decades in the human stomach, the aim of this study was to examine the long-term course of H. pylori-specific serum immunoglobulin G (IgG) responses with respect to subclass and antigenic target. We studied paired serum samples obtained in 1973 and in 1994 in Vammala, Finland, from 64 healthy H. pylori-positive adults and from other healthy control subjects. H. pylori serum immunoglobulin A, IgG, and IgG subclass responses were determined by antigen-specific enzyme-linked immunosorbent assays. H. pylori-specific IgG1 and IgG4 subtype responses from 47 subjects were similar in 1973 and 1994, but not when compared with unrelated persons. H. pylori-specific IgG1:IgG4 ratios among the participants varied >1000-fold; however, 57 (89.1%) of 64 subjects had an IgG1:IgG4 ratio >1.0, consistent with a predominant IgG1 (Th1) response. Furthermore, ratios in individual hosts were stable over the 21-year period (r = 0.56; P < .001). The immune response to heat shock protein HspA was unchanged in 49 (77%) of the 64 subjects tested; of the 15 whose serostatus changed, all seroconverted and were significantly younger than those whose status did not change. These findings indicate that H. pylori-specific antibody responses are host-specific with IgG1:IgG4 ratios stable over 21 years, IgG1 responses predominating, and HspA seroconversion with aging.

Multi-Level Stakeholder Perspectives on Determinants of Point of Care Ultrasound Implementation in a US Academic Medical Center.

There is growing interest from multiple specialties, including internal medicine, to incorporate diagnostic point of care ultrasound (POCUS) into standard clinical care. However, few internists currently use POCUS. The objective of this study was to understand the current determinants of POCUS adoption at both the health system and clinician level at a U.S. academic medical center from the perspective of multi-level stakeholders. We performed semi-structured interviews of multi-level stakeholders including hospitalists, subspecialists, and hospital leaders at an academic medical center in the U.S. Questions regarding the determinants of POCUS adoption were asked of study participants. Using the framework method, team-based analysis of interview transcripts were guided by the contextual domains of the Practical Robust Implementation and Sustainability Model (PRISM). Thirty-one stakeholders with diverse roles in POCUS adoption were interviewed. Analysis of interviews revealed three overarching themes that stakeholders considered important to adoption by clinicians and health systems: clinical impact, efficiency and cost. Subthemes included two that were deemed essential to high-fidelity implementation: the development of credentialing policies and robust quality assurance processes. These findings identify potential determinants of system and clinician level adoption that may be leveraged to achieve high-fidelity implementation of POCUS applications that result in improved patient outcomes.

Hospitalist Perceptions of Barriers to Lung Ultrasound Adoption in Diverse Hospital Environments.

Despite the many advantages of lung ultrasound (LUS) in the diagnosis and management of patients with dyspnea, its adoption among hospitalists has been slow. We performed semi-structured interviews of hospitals from four diverse health systems in the United States to understand determinants of adoption within a range of clinical settings. We used the diffusion of innovation theory to guide a framework analysis of the data. Of the 27 hospitalists invited, we performed 22 interviews from four hospitals of diverse types. Median years post-residency of interviewees was 10.5 [IQR:5-15]. Four main themes emerged: (1) There are important clinical advantages to LUS despite operator dependence, (2) LUS enhances patient and clinician experience, (3) Investment of clinician time to learn and perform LUS is a barrier to adoption but yields improved efficiency for the health system and (4) Mandated training and use may be necessary to achieve broad adoption as monetary incentives are less effective. Despite the perceived benefits of LUS for patients, clinicians and health systems, a significant barrier to broad LUS adoption is the experience of time scarcity by hospitalists. Future implementation strategies should focus on changes to the clinical environment that address clinician barriers to learning and adoption of new skills.

Discharge-ready volume status in acute decompensated heart failure: a survey of hospitalists.

Acute decompensated heart failure is the leading cause of hospitalization in older adults. Clinical practice guidelines recommend patients should be euvolemic at hospital discharge - yet accurate assessment of volume status is recognized to be exceptionally challenging. This conundrum led us to investigate how hospitalists are assessing volume status and discharge- readiness of patients hospitalized with heart failure. We collected audience response data during a didactic heart failure presentation at the 2019 Society of Hospital Medicine annual meeting. Respondents (n = 216), 76% of whom were practicing physician hospitalists caring for more than 20 acute heart failure patients per year, were presented six questions. Eighteen percent of respondents reported not being able to determine the completeness of decongestion on discharge and 32% reported that complete decongestion was not a treatment target. These findings suggest important differences between guideline recommendations and how hospitalists treat heart failure in current clinical practice.

Implications of Guideline Updates for the Management of Apparent Treatment Resistant Hypertension in the United States (A NCDR Research to Practice [R2P] Project).

The 2018 resistant hypertension scientific statement offers new treatment recommendations. To determine the implications of these changes, we sought to ascertain the prevalence of apparent treatment resistant hypertension (aTRH) and the therapies used to treat it in an US national ambulatory cardiovascular registry before these recent developments. Using the PINNACLE Registry from 2013 to 2014, we identified all patients receiving treatment for hypertension and then determined the proportion with aTRH as those who met the following criteria over ≥2 consecutive visits: (1) 3 blood pressure medication classes including a diuretic and blood pressure >140/90, OR (2) ≥4 blood pressure medications. Among those with aTRH, we examined past use of therapies now recommended in guidelines including: (1) first-line therapy with an angiotensin-converting enzyme inhibitor or angiotensin-II receptor blocker, calcium channel blocker and a thiazide diuretic, (2) use of chlorthalidone, and (3) use of a mineralocorticoid receptor antagonist (MRA) for those requiring a 4th medication. Of 84,624 patients on treatment for hypertension, 11,147 (13.1%) met criteria for prevalent aTRH. Among these patients: (1) Of those on 3 antihypertensive agents (n = 1,255), 315 (25%) were on the first-line regimen now recommended in guidelines, (2) 520 (6.7%) of the 7,930 patients on thiazides were using chlorthalidone, and (3) 3061 (27%) were using a MRA; another 4,523 (40.6%) were eligible for its addition. In conclusion, our findings of low historic use of therapies now recommended in guidelines suggest opportunities to improve care among patients with aTRH.

Stakeholder Perceptions of Point-of-Care Ultrasound Implementation in Resource-Limited Settings.

BACKGROUND: Nearly half of the world lacks access to diagnostic imaging. Point of care ultrasound (POCUS) is a versatile and relatively affordable imaging modality that offers promise as a means of bridging the radiology gap and improving care in low resource settings. METHODS: We performed semi-structured interviews of key stakeholders at two diverse hospitals where POCUS implementation programs had recently been conducted: one in a rural private hospital in Haiti and the other in a public referral hospital in Malawi. Questions regarding the clinical utility of POCUS, as well as barriers and facilitators of its implementation, were asked of study participants. Using the Framework Method, analysis of interview transcripts was guided by the WHO ASSURED criteria for point of care diagnostics. RESULTS: Fifteen stakeholders with diverse roles in POCUS implementation were interviewed. Interviewees from both sites considered POCUS a valuable diagnostic tool that improved clinical decisions. They perceived barriers to adequate training as one of the most important remaining barriers to POCUS implementation. CONCLUSIONS: In spite of the increasing affordability and portability of ultrasounds devices, there are still important barriers to the implementation of POCUS in resource-limited settings.

Predictors of Late Mortality in D-Transposition of the Great Arteries After Atrial Switch Repair: Systematic Review and Meta-Analysis.

Background Existing data on predictors of late mortality and prevention of sudden cardiac death after atrial switch repair surgery for D-transposition of the great arteries (D-TGA) are heterogeneous and limited by statistical power. Methods and Results We conducted a systematic review and meta-analysis of 29 observational studies, comprising 5035 patients, that reported mortality after atrial switch repair with a minimum follow-up of 10 years. We also examined 4 additional studies comprising 105 patients who reported rates of implantable cardioverter-defibrillator therapy in this population. Average survival dropped to 65% at 40 years after atrial switch repair, with sudden cardiac death accounting for 45% of all reported deaths. Mortality was significantly lower in cohorts that were more recent and operated on younger patients. Patient-level risk factors for late mortality were history of supraventricular tachycardia (odds ratio [OR] 3.8, 95% CI 1.4-10.7), Mustard procedure compared with Senning (OR 2.9, 95% CI 1.9-4.5) and complex D-TGA compared with simple D-TGA (OR 4.4, 95% CI 2.2-8.8). Significant risk factors for sudden cardiac death were history of supraventricular tachycardia (OR 4.7, 95% CI 2.2-9.8), Mustard procedure (OR 2.2, 95% CI 1.1-4.1), and complex D-TGA (OR 5.7, 95% CI 1.8-18.0). Out of a total 124 implantable cardioverter-defibrillator discharges over 330 patient-years in patients with implantable cardioverter-defibrillators for primary prevention, only 8% were appropriate. Conclusions Patient-level risk of both mortality and sudden cardiac death after atrial switch repair are significantly increased by history of supraventricular tachycardia, Mustard procedure, and complex D-TGA. This knowledge may help refine current selection practices for primary prevention implantable cardioverter-defibrillator implantation, given disproportionately high rates of inappropriate discharges.

Diagnostic Accuracy of Point-of-Care Lung Ultrasonography and Chest Radiography in Adults With Symptoms Suggestive of Acute Decompensated Heart Failure: A Systematic Review and Meta-analysis.

Importance: Standard tools used to diagnose pulmonary edema in acute decompensated heart failure (ADHF), including chest radiography (CXR), lack adequate sensitivity, which may delay appropriate diagnosis and treatment. Point-of-care lung ultrasonography (LUS) may be more accurate than CXR, but no meta-analysis of studies directly comparing the 2 tools was previously available. Objective: To compare the accuracy of LUS with the accuracy of CXR in the diagnosis of cardiogenic pulmonary edema in adult patients presenting with dyspnea. Data Sources: A comprehensive search of MEDLINE, Embase, and Cochrane Library databases and the gray literature was performed in May 2018. No language or year limits were applied. Study Selection: Study inclusion criteria were a prospective adult cohort of patients presenting to any clinical setting with dyspnea who underwent both LUS and CXR on initial assessment with imaging results compared with a reference standard ADHF diagnosis by a clinical expert after either a medical record review or a combination of echocardiography findings and brain-type natriuretic peptide criteria. Two reviewers independently assessed the studies for inclusion criteria, and disagreements were resolved with discussion. Data Extraction and Synthesis: Reporting adhered to the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy and the Preferred Reporting Items for Systematic Reviews and Meta-analyses guidelines. Two authors independently extracted data and assessed the risk of bias using a customized QUADAS-2 tool. The pooled sensitivity and specificity of LUS and CXR were determined using a hierarchical summary receiver operating characteristic approach. Main Outcomes and Measures: The comparative accuracy of LUS and CXR in diagnosing ADHF as measured by the differences between the 2 modalities in pooled sensitivity and specificity. Results: The literature search yielded 1377 nonduplicate titles that were screened, of which 43 articles (3.1%) underwent full-text review. Six studies met the inclusion criteria, representing a total of 1827 patients. Pooled estimates for LUS were 0.88 (95% Cl, 0.75-0.95) for sensitivity and 0.90 (95% Cl, 0.88-0.92) for specificity. Pooled estimates for CXR were 0.73 (95% CI, 0.70-0.76) for sensitivity and 0.90 (95% CI, 0.75-0.97) for specificity. The relative sensitivity ratio of LUS, compared with CXR, was 1.2 (95% CI, 1.08-1.34; P < .001), but no difference was found in specificity between tests (relative specificity ratio, 1.0; 95% CI, 0.90-1.11; P = .96). Conclusions and Relevance: The findings suggest that LUS is more sensitive than CXR in detecting pulmonary edema in ADHF; LUS should be considered as an adjunct imaging modality in the evaluation of patients with dyspnea at risk of ADHF.

Point-of-Care Ultrasound for Hospitalists: A Position Statement of the Society of Hospital Medicine.

Many hospitalists incorporate point-of-care ultrasound (POCUS) into their daily practice to answer specific diagnostic questions or to guide performance of invasive bedside procedures. However, standards for hospitalists in POCUS training and assessment are not yet established. Most internal medicine residency training programs, the major pipeline for incoming hospitalists, have only recently begun to incorporate POCUS in their curricula. The purpose of this document is to inform a broad audience on what POCUS is and how hospitalists are using it. This document is intended to provide guidance for the hospitalists who use POCUS and administrators who oversee its use. We discuss POCUS 1) applications, 2) training, 3) assessments, and 4) program management. Practicing hospitalists must continue to collaborate with their local credentialing bodies to outline requirements for POCUS use. Hospitalists should be integrally involved in decision-making processes surrounding POCUS program management.

Diagnosis of medullary sponge kidney by computed tomographic urography.

The diagnosis of medullary sponge kidney traditionally was established by means of intravenous pyelography. Beginning in the mid-1990s, intravenous pyelography rapidly was supplanted by computed axial tomography as the preferred imaging study for evaluating patients with renal stone disease. Conventional computed tomographic imaging has not been satisfactory for diagnosing medullary sponge kidney. The introduction of multidetector-row computed tomography in 1999 allowed radiologists to acquire images composed of elements allowing the creation of high-resolution 3-dimensional displays. Computed tomographic urography is an imaging technique that provides both cross-sectional displays and images of the contrast-filled renal collecting systems, ureters, and urinary bladder that are the equivalent of intravenous pyelography. We report a case of medullary sponge kidney diagnosed by means of 3-dimensional multidetector-row computed tomographic urography.