Combining International Standards to Develop Clinical Decision Support for Parent Smoking Cessation in Pediatrics

Introduction

Smoking causes severe health consequences, and there is no safe level of secondhand smoke (SHS) exposure [1,2]. SHS exposure among children increases the risk of sudden infant death syndrome, chronic respiratory diseases such as asthma, and lung cancer in adulthood [2]. Over 40% of children in the United States are regularly exposed to SHS, most often by a parent [3]. When a parent quits smoking, they significantly lower their chances of developing lung and other related cancers and lengthen their life expectancy [4]. Parental smoking cessation eliminates the majority of children’s SHS exposure and decreases the odds that their children will become tobacco users themselves [5-7]. Pediatricians are well positioned to help parents quit smoking by implementing interventions that connect parents with cost-effective, evidence-based tobacco cessation treatments [8,9]. Furthermore, for low-income households, where the smoking rate can be 2 times the general population level, pediatricians are often parents’ primary source of interaction with the health care system [10,11]. Thus, integrating tobacco cessation treatment within pediatric care holds great promise for mitigating the harm of tobacco within families.

While pediatric health care systems are uniquely positioned to help parents and household members (henceforth referred to as parents) quit smoking, evidence-based treatments are significantly underused [12]. Cost-effective cessation strategies exist, especially when behavioral and pharmacological interventions are combined (eg, nicotine replacement therapy [NRT] plus the Quitline) [13]. Compared to controls, behavioral interventions, such as telephone- or text-messaging–based counseling, and NRTs significantly increase smoking cessation rates and can easily be disseminated [14,15]. The 2021 US Preventive Services Task Force, Healthy People 2030, the Department of Health and Human Services, and the surgeon general all strongly recommend that clinicians ask all adults about tobacco use and provide behavioral interventions and pharmacotherapy for cessation [13,16-18]. To best address tobacco use among parents, it is ideal to develop scalable solutions that are coordinated across health systems, community partners, and national services.

Fortunately, advances in health interoperability standards, such as Fast Healthcare Interoperability Resources (FHIR), SMART on FHIR (hereafter referred to as SMART), and CDS Hooks, have created opportunities to more easily support these efforts [19-21]. FHIR is a health information exchange standard that allows clients (eg, clinical decision support [CDS] services) to obtain granular access to data, such as encounters, allergies, or medications, using a standardized application programming interface (API). SMART is a standard that allows authorized apps to be opened directly from an electronic health record (EHR) or patient portal so that clinicians and patients can seamlessly access external apps within their workflow. CDS Hooks is an event-driven framework in which specific actions in the clinical workflow, such as ordering a medication and trigger messages, known as “hooks,” are sent to subscribing services. These systems can then return recommendations as a “card” to the user directly within their EHR workflow. Systems built on these standards, known as service-oriented systems, are separate from the host system (eg, EHR) but connected using a standard API [22]. This allows for a modular, scalable approach that can extend care beyond the individual patient.

Several researchers have already begun to explore using these standards for developing CDS systems. Investigators in one health system demonstrated the utility of the SMART framework within the EHR using both clinician- and patient-facing apps to address clinical needs such as patient-specific medication instructions [23]. Dolin et al [24] developed a prototype pharmacogenomics CDS service that interfaced with a commercial EHR using the CDS Hooks standard, and Jung et al [25] used CDS Hooks to implement a national service for drug allergy interaction checking. Other researchers have sought to combine standards within the same system. Theiss et al [26] developed a prototype app that combined several health information technology standards in a nonproduction setting, and Morgan et al [27] combined contextual CDS Hooks prompts to direct clinicians to SMART apps within their hospital’s emergency department.

These studies highlighted the utility of each standard; however, the apps were designed mainly to support clinicians, not patients or families, and were deployed in nonpediatric settings. Lastly, while some use cases have connected to external systems (eg, Value Set Authority Center or RXNorm) to retrieve information, the apps did not interface with community health care partners. In this viewpoint, we describe our experience in developing and implementing a parent tobacco treatment platform (PTTP) within a pediatric institution that leverages multiple international standards to support interoperability, with the overarching goal of providing a model for how such work can be approached. We highlight three key areas where critical decisions shaped the design and trajectory of the system: (1) international standards, (2) community partnerships, and (3) family-centered care. By sharing these lessons, we aim to guide other health IT developers, physicians, health care leadership, and standards development organizations in designing, implementing, and scaling CDS interventions.

How We Designed the PTTP

Project Setting

The PTTP was developed as part of a National Institutes of Health funded research initiative focused on connecting the parents (including caregivers and other household members) of our patients with smoking cessation services within the Children’s Hospital of Philadelphia (CHOP) Care Network and Pediatric Research Consortium [28], which includes a combination of suburban, urban, and semirural practices that use a common EHR (Epic Systems Inc). The tool was initially deployed in 5 clinics and has since expanded across all CHOP’s primary care and pulmonary medicine networks (35 sites). Development (including user analysis and design) of the CDS system occurred between January 2020 and November 2023. To support this work, a multidisciplinary team was formed that consisted of physicians, a software engineer and EHR-integration expert, a human-computer interaction specialist, and project managers. The project was approved by the CHOP Institutional Review Board (IRB 20‐018146).

System Overview

On the basis of a rigorous human-centered design evaluation, which consisted of interviews with clinicians, parents, and front-desk staff, formative testing of prototypes, and iterative feedback sessions, we identified that the system needed to support 2 high-level workflows (Figure 1) [29,30].

‎

First, the system needed to screen, identify, and automatically connect parents to the requested treatment options. Although some health care providers (HCP) already routinely ask parents about their smoking status, our team found that the process is inconsistent. Furthermore, parents felt more comfortable disclosing their smoking status through a survey that used nonjudgmental language [29]. Additionally, to engage parents during a potential activation window, where they are energized to engage in treatment, all treatment connections needed to be processed immediately upon user responses. From these priorities, we hypothesized that an event-driven architecture, responding to previsit questionnaires in the patient portal, would meet the needs of our project. This architecture includes the following high-level layers: (1) producer (source of events), (2) broker (sender of events), and (3) subscriber (listener of events; Figure 2) [31]. Questionnaires can be completed up to 7 days ahead of the visit within the patient’s personal health portal (MyChart; Epic System), during check-in, or by the HCP during the visit. The questionnaire is configurable within our vendor EHR and includes questions about smoking use; if the respondent indicates that they smoke, it provides a motivational message to encourage treatment engagement and three evidence-based treatment options: (1) NRT, (2) Quitline enrollment, and (3) SmokefreeTXT enrollment. The Quitline is a free, state-specific, telephone-based smoking cessation service that provides counseling, referrals to other programs, and access to free medication. SmokefreeTXT is a free mobile text messaging service offered by the National Cancer Institute that provides daily tips and 24-hour support for quitting smoking. Treatment connections are processed immediately upon completion of the questionnaire.

‎

Second, the system should incorporate a component embedded within the clinician workflow in the EHR. HCPs were interested in understanding the smoking status of the household, given its impact on the child. To reduce the burden on HCPs, who have to juggle many competing priorities in a limited amount of time, the clinician-facing component needed to not only summarize smoking use in the household but also quickly allow HCPs to connect parents to smoking cessation treatment and support other clinical tasks within the EHR, such as documentation and billing. We hypothesized that a client-side rendering architecture would serve the needs of this component [32]. This architecture includes the following layers: (1) presentation (eg, the user-facing app), (2) application (eg, clinical logic), (3) service (eg, connection to database and treatment partners), and (4) database (storage of app data; Figure 3).

‎

Interoperability Standards Evaluation

We mapped the architectural needs of our system to interoperability standards across all versions supported by Health Level 7 (HL7) International (draft standard for trial use, standard for trial use, and release). To reduce potential dissemination barriers of our system to other organizations, we prioritized standards based on modern internet protocols (eg, FHIR) and limited them to only those well supported by our commercial EHR vendor. Refer to Table 1 for the list of standards that were evaluated.

While HL7 version 2 messaging is the most common event-driven standard used in health care, its implementation often uses a transmission control protocol, which is at a lower level of the Open Systems Interconnection model compared to more modern standards, such as FHIR [36]. As a result, it is generally not a preferred health information exchange standard for new developments. Evaluations occurred at the following technology layers, where standards were available: (1) EHR-integration framework (eg, SMART, CDS Hooks, and web messaging), (2) data model (eg, FHIR), (3) logic (eg, clinical quality language [CQL]), and (4) community information exchange (eg, direct).

Interoperability Standards Selection

The smoking cessation system (Multimedia Appendix 1) has 2 high-level components, a parent-facing and a clinician-facing system, with services shared between them (Figure 4). Both components are described below, with a separate section dedicated to shared services. Table 2 provides an overview of the use of standards across each component.

‎

Patient-Facing Component

We first evaluated technologies that would meet the needs of the parent-facing CDS, which was delivered through a previsit questionnaire. Given the existing support for the direct messaging protocol within our EHR and the Pennsylvania Quitline, we initially attempted to facilitate connections using this method. However, as CHOP is a pediatric hospital and does not maintain charts for adults, the Direct protocol did not readily support transmitting information about the parent through the child’s chart. Additionally, other treatment partners did not support the direct protocol. Therefore, a separate service was required to create a connection between the parent and our community treatment partners.

We then evaluated the CDS Hooks standard. While CDS Hooks services are traditionally used to deliver information (eg, recommendations) in the form of a card within a user’s (eg, clinician) workflow, we were interested in using the standard to perform asynchronous tasks. Unfortunately, the current version of the standard does not support a hook based on the completion of a questionnaire. However, our EHR’s implementation of CDS Hooks provides a richer set of triggers, including the completion of a previsit questionnaire. While the use of widely available standards was preferred, other options would have required significant customization and project-specific code. As a result of this and early demonstrations related to adding a hook for responding to patient-completed questionnaires [37], we chose to move forward with the CDS Hooks. Immediately after submitting the questionnaire, a Patient_View hook is immediately triggered from the EHR to our system. The system authenticates the message, evaluates questionnaire responses retrieved via FHIR, and connects parents to the requested treatment options.

After determining the trigger mechanism, we explored CQL as a standards-based method for evaluating the patient’s data to determine whether treatment options were requested. While the CQL rule performed as expected, part of the EHR configuration required to send a CDS Hooks message after completion of a previsit questionnaire included a native EHR rule, which already evaluated questionnaire responses. Given this and the additional infrastructure required to support the CQL standard, such as the rules engine, which is not readily supported by our EHR, the CQL rule was considered redundant and removed. In its place, a combination of EHR-based rules within the questionnaire configuration and logic statements in the CDS Hooks service was used.

Clinician-Facing Component

Next, we evaluated standards-based methods for developing and implementing the clinician-facing component. Given the recent federal regulation mandating support for SMART [38], we chose this method to embed this component directly within the HCP’s EHR workflow, requiring no additional clicks to navigate to or open. The app was written in React (JavaScript) and supports multiple HCP tasks within the EHR. For parents who smoke but do not select a treatment during the questionnaire, the app allows HCPs to enroll the parent in treatment if they change their mind. The system also allows HCPs to add SHS to the problem list with a single click and includes a section with prompts on how to discuss tobacco use with parents and treatment information. Furthermore, the app automates several routine tasks, including populating text within the encounter note and after-visit summary and defaulting visit diagnoses in order sets. When a parent accepts treatment and counseling is performed by the HCP, the system also provides information on how to appropriately bill for the visit.

To facilitate a tight integration between the app and the EHR, we took advantage of our EHR vendor’s web-messaging framework that enables navigation to different areas of the chart, such as viewing questionnaire responses. Although similar to SMART Web Messaging, our EHR’s framework also allows embedded apps to “listen” to events originating from the EHR, such as when a patient’s medications or problem list are updated. For example, to ensure that SHS is not added to the patient’s chart multiple times, our system listens for changes to the patient’s problem list and updates itself accordingly.

Community Partner Connections

Both the parent- and clinician-facing components use a set of shared services to connect parents to smoking cessation treatment. At the time of our project, limited options existed to support information exchange between the EHR and community partners. Unfortunately, due to varying technical resources across our treatment partners, as well as the lack of support for family-centered care within our EHR, standards-based methods for connecting parents from the child’s chart were not possible. As a result, a separate service was required to connect parents to each partner, as detailed in the subsequent sections. All services are secured using the OAuth 2.0 token infrastructure.

NRT Treatment

Since the smoking cessation system was embedded within the child’s chart, traditional methods to prescribe medication (eg, E-Prescribe) were not available. To connect parents to NRT, we contracted a third-party pharmacy that delivered medication to the parents’ homes. However, as a smaller organization, the pharmacy lacked the resources to develop a process for ingesting parent information using more modern techniques, such as an API. Consequently, we developed a service to send parent information and a system-generated PDF of a prescription using encrypted email.

Text Message–Based Counseling

We partnered with SmokefreeTXT to provide text-based counseling to parents. SmokefreeTXT is a text message program provided by the National Cancer Institute that provides a web form that individuals can complete to enroll [39]. We initially attempted to integrate this web form directly into our previsit questionnaire, but due to network restrictions within our organization, this was not possible. Additionally, because the questionnaire includes other treatment options beyond SmokefreeTXT, the use of the web form may have required users to enter their information multiple times if they were interested in other treatment options. At the time of our project, no automated mechanism existed to connect individuals to the program, so we engaged directly with the SmokefreeTXT operating partner to request the development of an API. The API initially only accepted the participant’s phone number; however, a primary goal of our project was to remove as many potential barriers to treatment engagement as possible. As a result, to streamline the enrollment process, which required the user to respond with additional information after enrolling (eg, target quit date and current age), a second API was enabled to programmatically update these values, using information already provided in the questionnaire.

Telephone-Based Counseling

Because CHOP provides care across multiple states, we needed to partner with 2 separate telephone counseling organizations in Pennsylvania and New Jersey, each with different connection requirements. Quitline selection is determined by the location of the participating clinic. The Pennsylvania Quitline supports the use of the Direct Messaging protocol. Unfortunately, as mentioned above, Direct Messaging does not account for the family-centered nature of care and can only readily be used if the connection originates from the patient’s chart. In an attempt to establish real-time connections, we evaluated the development of an API similar to that used for SmokefreeTXT; however, this was not possible at the time due to limited resources. Additionally, the Pennsylvania Quitline had already established a robust infrastructure where enrollments are processed through a daily file exchange using the Secure File Transfer Protocol. To build the file, a script queries our clinical data warehouse to identify the participants who are requested to be enrolled. This file is then retrieved by the Quitline partner, who is given remote access to the directory, and processed by their system. For our New Jersey partner, due to resource constraints, connections are sent using encrypted email in a process similar to that of our pharmacy partner (described above).

Impact on Family-Centered Care

Parents want their child’s physician to discuss smoking cessation treatment with them [13]. However, in pediatric settings, appropriate tobacco treatments are rarely, if ever, provided to parents who smoke [12]. This is largely due to workflow limitations in the pediatric EHR, especially the inability to account for the family aspect of child health care [40,41]. EHR vendors often develop features for the physician-adult patient dyad. However, as noted above, they provide limited support to connect family relationships, making it challenging to leverage existing technologies to support teamwork across the patient’s entire network (eg, care team and community groups), particularly in pediatric settings [42].

Our system focused directly on the family-centered nature of a child’s care. Therefore, over a 1-year period, our system was used in 194,946 visits and identified 7847 (4.03%) parents who smoked. Of those who smoked, 37.64% (n=2954) selected at least one treatment option, 29.27% (n=2297) selected NRT, 29.46% (n=2312) selected SmokefreeTXT, and 21.8% (n=1711) selected the Quitline (Table 3). Performance of the app was excellent, with 99% of all treatment connections handled through the previsit questionnaire and CDS Hooks service without clinician involvement. The remaining 1% of the connections were facilitated using the clinician-facing tool. Interaction (eg, clicks) within the clinician-facing system was limited, occurring in only 774 (<1%) encounters. Given the low level of interaction with the system by HCPs, routine check-ins with practice management were established to ensure that any issues with usability were identified and corrected. No issues were identified during the project period. The uptime of the system (across all components) was 99.99%, with most outages related to network faults that occurred for less than 1 minute. We identified multiple failures in the web-messaging framework; however, these only limited navigation to different areas of the chart from within the app (eg, viewing the questionnaire) and did not impact parent connections. The median (IQR) response time across all system services was 60 (14,393) milliseconds.

Discussion and Future Directions

By combining multiple international health data standards and reusing app services across system components (parent and clinician facing), we were able to develop the EHR-integrated PTTP. Our system connected parents (including caregivers and other household members) with smoking cessation treatment at scale, serving hundreds of thousands of families across a diverse patient and practice population. This represents a significant increase compared to prior work [43-45]. Furthermore, by developing our system in a modular way, we helped to lower the barriers to implementation in another health system on the same EHR, which is using components of our platform [46]. However, our work also showed that there are limits to using international interoperability standards to support family-centered care in pediatrics.

First, some standards, such as CDS Hooks, are limited in scope and support only a fraction of the potentially actionable events that occur within a health care system. Engaging with patients and families outside of the clinical encounter is a growing area of interest in the health care community [37]. One method of doing this is through previsit questionnaires, which can be completed in the patient’s health portal. Using our EHR vendor’s enhanced set of CDS Hooks triggers, our system was able to immediately respond to parents completing a previsit questionnaire. Compared to adult settings, where only 7.6% of patients received a medication associated with tobacco use treatment [47], our system connected 29% of the parents. Additionally, 99% of these connections occurred automatically, without requiring any work from the visit HCP. While CDS Hooks was originally intended to support more synchronous workflows, such as validating orders at the time of ordering, our team envisions a much larger role for the standard to support asynchronous workflows that do not require input from the health team. This could help organizations across EHR platforms change how they engage with their patients and even alter the dynamics of an individual visit.

Additionally, variability in health care workflows and implementation of standards across EHR vendors creates a need to adapt even standards-adherent systems for implementation in different health systems [46,48,49]. This is especially true for standards that are flexible to institutional workflows, such as SMART, which, depending on the EHR, can integrate into the EHR at different stages of the visit or locations in the chart. Furthermore, health care is a complex adaptive system, and a single workflow may not work across organizations [50]. Therefore, even for EHRs that support SMART, individual implementations may differ across care domains (eg, inpatient vs primary care) or organizations. Given these limitations, future research and implementation efforts should evaluate whether a combination of standards-based and native EHR technologies would increase portability while reducing maintenance.

Second, standards are not widely adopted outside of health care organizations, such as hospitals or clinics. Recent regulations require EHR vendors to support health interoperability technologies for certification with the Office of the National Coordinator for Health Information Technology [38]. However, smaller, health care–adjacent organizations similar to our community partners may not use commercial EHRs and often lack the resources to support integrations using international standards. Additionally, standards supporting EHR integration with community partners, such as the Direct protocol, do not support connections for others within the patient’s network (eg, parents). As a result, our system needed to support 3 different methods of information exchange (eg, encrypted email, API, and secure file transfer protocol), and even those that shared a protocol required different information within the message. To fully realize the potential of our system to support interoperability at the community partner level, additional work is needed to create a standardized method for exchanging information. One promising standard-based method to support this is the FHIR ServiceRequest profile, which has been part of the United States Core Interoperability implementation guide since version 2 was released in April 2022.

Lastly, our work showed that the concept of the EHR as a hub, where the EHR provides the necessary capabilities to support teamwork across the patient’s entire network (eg, care team and community groups), is not well supported, particularly in pediatrics [51]. For example, if a parent has 2 children, ideally, our system would only request information (via the previsit questionnaire) once. However, as patient records rarely contain family-based associations [42], the parent is presented with the questionnaire multiple times, which may result in parental frustration and potentially conflicting information across their children’s charts.

Performance of the PTTP was acceptable, with near-perfect uptime across all components. However, we identified issues that warranted attention, including network failures and communication with our EHR’s web-messaging framework. While a growing body of work has identified how CDS can fail, the focus has largely been on centralized computing models where the data and the computation are on the same machine [52-54]. Architectures taking advantage of newer HL7 standards create new complexities and failure modes that are not recognized with existing frameworks [55]. To ensure the ongoing safety and effectiveness of service-oriented CDS, more research into these failure modes is necessary.

Conclusions

We found that it is feasible to develop and integrate CDS using a hybrid architecture that combines several health IT standards such as SMART, FHIR, and CDS Hooks. These types of systems are likely to increase over time; however, additional work is necessary to both expand existing standards to integrate with the EHR and streamline information exchange processes for sharing data, when appropriate, with third-party systems about a patient’s care network (eg, parents) whose behaviors impact the patient. Similarly, improved EHR support for linking patient records of family relationships will help to align health data across families.