The Need for Living Guidelines in a Learning Health System

Whew. As we head into another respiratory virus season, we are still catching our breath from last winter, which brought unprecedented patient volumes and staffing strains to children’s hospitals around the United States. The collision of the #tripledemic (a simultaneous increase in cases of influenza, coronavirus disease 2019 [COVID-19], and respiratory syncytial virus) with the misplaced priorities of the American health care system, including widespread shuttering of community pediatrics wards, imposed never-before-seen pressures on acute pediatric care referral centers.^1,2  Easing this burden was the standardization of care that is increasingly common and is often coupled with improved electronic health record (EHR) workflows. In its best form, EHR-integrated clinical decision support (CDS) can streamline routine tasks such as admitting an infant with bronchiolitis, weaning therapies for a child with status asthmaticus, or ordering antibiotics for an adolescent with community-acquired pneumonia. Standardized order sets are commonly designed to integrate best practices promoted by professional society disease-specific guidelines, with modifications based on local expertise and capacity.

Yet, the efficiencies realized by standardization only translate to patient benefit if both expert-developed guidelines and associated EHR-integrated CDS are up to date. For example, the American Academy of Pediatrics’ (AAP) bronchiolitis guidelines were last published in 2014 and the Infectious Diseases Society of America’s community-acquired pneumonia guidelines were last published in 2011.^3,4  The recently updated guidelines for the care of febrile infants relied substantially on literature that had been available for multiple years after the last version of the guidelines.⁵  In part, this delay reflects the slow accumulation of high-quality data, a lag that stands to be reduced as research techniques become more scalable and efficient with technological advances. Moreover, there have been important recent advances in clinical investigation aimed at addressing patient heterogeneity and better representing inherent clinical uncertainty. Additional literature has also emerged on important topics, such as the potential role of noninvasive respiratory support for children with bronchiolitis. However, incorporating new evidence into traditional guidelines commonly involves a resource-intensive process that aims to align a comprehensive set of recommendations with the current evidence base, rather than make targeted updates according to the availability of new evidence. As a result, many reported advances have not yet been included in the most recent version of guidelines, risking inconsistent implementation of these findings into effective, evidence-based CDS such as order sets. Slow adaptation of guidelines to new evidence and inconsistent incorporation into EHR workflows discourages the use of guidelines, promotes unnecessary variation, and erects barriers to providing children the best possible care. Living guidelines are an approach to evidence synthesis that addresses many of the shortcomings of the traditional guidelines process by catalyzing updates through continuous review and incorporation of new literature. The framework outlined by Cheyne et al is adapted in Fig 1 and illustrates the ongoing, cyclical reviews and updates that leverage information technology to facilitate the entire process.⁶  To appreciate the potential benefits of living guidelines to support high-quality pediatric care delivery, it is worth considering in more detail the current state of the AAP bronchiolitis guidelines.

Bronchiolitis guidelines were last updated 9 years ago, despite an expected 5-year update frequency. Many of the recommendations are still valid, including administering nasogastric or intravenous fluids to infants who cannot maintain hydration orally. However, other recommendations are absent. For example, no recommendations are offered regarding the role of high-flow nasal cannula or nasal continuous positive airway pressure in the support of infants with bronchiolitis, both of which have been widely adopted and studied.^7–9  Still other recommendations were controversial when released and warrant a reassessment of the data to determine whether revision is necessary. An example is the guidance that clinicians should not administer bronchodilators to infants and children with bronchiolitis, despite some published studies supporting administration.^10,11  New infectious causes of bronchiolitis have also emerged in the last decade, including COVID-19¹²  and enterovirus D68 (EV-D68).¹³  Children across a range of ages with EV-D68 have been reported to have asthma or reactive airway disease, and the emergence of both COVID-19 and EV-D68 as possible causes of bronchiolitis call into question whether the recommendation to not administer corticosteroids in the latest version of the guidelines adequately accounts for the variety of patient scenarios that are categorized as bronchiolitis.¹⁴  Even for respiratory syncytial virus, shifts in epidemiology have been reported, raising concerns that previous literature is less relevant.¹⁵ 

Pediatricians have long relied on the combination of collaboration and innovation to provide the best possible outcomes for their patients. So what are clinicians doing to bridge the current chasm between existing guidelines and more recent literature? For now, institutional committees at referral centers often fill the gap and provide systematic guidance to clinicians for high-volume diseases. For bronchiolitis, this has meant incorporating recommendations for unproven interventions like a trial of β-agonists for children with “severe” bronchiolitis or considering a fourth-generation cephalosporin as empirical therapy for community-acquired pneumonia for a child with possible sepsis. Some institutions can use their electronic record data to periodically audit the performance of care pathways to identify local guidelines in need of modification. The AAP has also endorsed learning quality improvement collaboratives, including recent efforts focused on more efficient utilization of high-flow nasal cannula among children with bronchiolitis.¹⁶  Collectively, these actions are exciting steps toward systematic learning embedded into care delivery; but without consistent professional society oversight and coordinated, multicenter collaboration, these efforts fall short. Left behind are potential for rapid evidence translation and large-scale learning afforded by modern EHRs, contemporary biostatistics, new methods of searching literature databases, and the use of now commonplace telecommunications.

One promising development is the new availability of EHRs for information sharing. As a result of the 21st Century Cures Act, which was passed in 2016 and contained key provisions promoting interoperability of data, EHR vendors are now required to map their databases to standards detailed by the US Core Data for Interoperability, with enforcement beginning in 2022.¹⁷  This has enabled some patients to access their own health data via third-party applications. Just as exciting, clinical investigators, quality improvement specialists, and an emerging group of learning health system experts who operate in both arenas capitalize on preharmonized data to assess care delivery across multiple centers with substantially less effort and resources than would have been required only a few years ago. The value of large-scale harmonization of clinical trial design and data collection was realized in the pandemic with the world’s first multiplatform randomized controlled studies.¹⁸  Similarly, US Core Data for Interoperability requirements can be expected to aid in care delivery data exchange and help to identify high-value practices, through both observational investigations and embedded clinical trials. Current recommendations for β-agonists and corticosteroids for bronchiolitis are informed by relatively small, observational, and underpowered studies. Integrating prospective investigation into EHR workflows while leveraging interoperability to scale data collection and analysis across multiple centers will help to overcome many of the limitations of our evidence base in pediatrics.

Large data sets also lend themselves well to emerging biostatistics techniques designed to better characterize the heterogeneity of treatment effect. The heterogeneity of treatment effect can be especially problematic in pediatric clinical investigations because even apparently straightforward diagnoses like bronchiolitis can occur at dramatically different stages of a patient’s molecular and anatomic development. Such individual differences within a population of patients enrolled in a study of a treatment of a disease can result in a summary result that does not represent the treatment effect for the typical enrolled patient. This type of differential treatment effect is widely recognized in the field of oncology, where molecular and genetic diagnostics are now routinely used to individualize chemotherapy regimens. For diseases in which the differences in molecular pathways and genetic risk factors as less defined, risk-based, and effect-based statistical techniques have emerged to help understand how some treatments may have differential effects for a single disease depending on individual patient characteristics.¹⁹  Relatedly, phenotyping and endotyping efforts are characterizing subpopulations within specific diagnoses that may respond differently to a given treatment.²⁰ 

Keeping up with the growing quantity of literature streaming from real-world data would have once been impossible. Fortunately, modern digital libraries and search methods are making it easier to efficiently search and summarize evidence. The Cochrane Community now has guidelines for the development and maintenance of living systematic reviews, which rely on “explicit, predefined methods describing search frequency…[and] when new evidence is incorporated into the review,” as well as processes for continuously surveilling for and flagging new evidence.²¹  These techniques are being looked to as the building blocks of living clinical guidelines.²²  Panels of experts are also able to be convened much more easily thanks to virtual meeting platforms. Systematic Delphi procedures and related methods can be cost-effectively and time-efficiently conducted in virtual sessions hosting content experts from around the world. Clinician-supervised generative artificial intelligence tools can be harnessed alongside virtual meetings to reduce the human workload associated with building and maintaining living systematic reviews and clinical guidelines. Many of these tools were leveraged by experts around the world to create living COVID-19 guidelines that provided essential guideposts for clinicians navigating a disease management landscape that changed at an unprecedented pace.²³ 

Together, these advances now provide the necessary tools to substantially revamp our field’s approach to guideline generation and evidence translation. Living guidelines woven into modern EHR workflows are a way of meeting the challenge of ensuring that every child receives top-notch care while building process resilience against resource constraints stemming from the shuttering of community pediatric wards. Creating living guidelines for high-volume diagnoses, including bronchiolitis and community-acquired pneumonia, would provide both community hospitals and pediatric acute care institutions alike with a synthesized evidence base to serve as a guidepost for ongoing updates and maintenance of the EHR CDS systems that are increasingly a cornerstone of consistent, high-value care.