End user information needs for a SMART on FHIR-based automated transfer form to support the care of nursing home patients during emergency department visits

Abstract

Nursing home (NH) patients are extensive users of emergency department (ED) services. Problematically, poor information sharing and incomplete access to information complicates the delivery of care in EDs for NH patients. Paper-based transfer forms can support information sharing, but have significant limitations. Standards-based automated transfer-forms that leverage health information exchange data may address the limitations of paper-based forms and better support care delivery. This study developed a prototype SMART on FHIR automated transfer form for NH patients using priority data elements identified through individual interviews, a review of existing transfer forms, a targeted survey of end users, and a design workshop. Analyses were grounded in the 5 Rights of clinical decision support framework. The most valuable data elements included: emergency contact/healthcare proxy, current medication list, reason for transfer to the ED, baseline neurological state, and relevant diagnoses / medical history. The working prototype was successfully deployed within an Amazon Web Service environment.

Introduction

Nursing home (NH) patients are extensive users of emergency department (ED) services¹. These patients have higher rates of both ED overall visits and revisits than community dwelling older adults^2,3. Nearly a fifth of ED visits by NH patients are potentially preventable⁴. Moreover, in comparison to community dwelling older adults, NH patients are generally more medically complex with multiple chronic conditions⁵ and consequently undergo more imaging procedures in addition to being more likely to be hospitalized from the ED¹. Problematically, poor information sharing and incomplete access to information complicates the delivery of ED care for this high risk and high need patient population⁶.

Information that is incomplete or difficult to obtain is common for NH patients upon arrival in the ED^7–13. Even when information from the NH is available, it is often too sparse to be useful or, at the other extreme, includes large amounts of data that are of little value for clinical decision making¹⁴. Poor information sharing complicates work within the ED¹¹, delays care^15,16, and potentially places NH patients at greater risk for adverse events⁸. Some of the barriers to effective information sharing between NHs and EDs are due to the structure of the health care system. For example, less information is available to ED providers when the patient is transferred from low-resourced NHs¹⁷ or if the transfer occurs outside of normal business hours^18,19. Even phone calls to the NH may be of limited value; for example, the staff most knowledgeable about the resident’s transfer may not be available due to shift changes²⁰ or the afterhours provider on-call may not be familiar with the patient’s health issues¹⁴. Some information gaps are due to the unreliability of self-reports from patients¹¹; for example, due to cognitive impairment¹⁸ or delirium²¹.

Evidence suggests that standardized transfer forms documenting key information from the NH can increase the volume and improve the quality of information shared with ED providers^{10,19,22–24}, enhance provider satisfaction^25,26, and may be associated with more efficient care²⁷. However, paper-based transfer forms have significant limitations. Even when available, paper forms are not always utilized¹⁰ and fall into disuse over time²⁸, often due to staff turnover²⁹. Also, manually initiated transfer forms require NH staff education and training³⁰, take up NH staff time to complete³¹, and are inherently limited to the information accessible to the person completing the form. Paper-based forms may also simply get lost during transfers²⁰. Even if electronically shared, transfer forms that rely on manual data entry and submission by NH staff are not consistently used²⁵ and can place a time and workload burden on the NH²⁵. Similarly, electronic transfer forms, if reliant on end users to complete and send, fall into disuse over time²⁶. Changes in NH administration, which are frequent, can also disrupt usage and ultimately may require the transfer form process to be re-introduced²⁵. Importantly, the idea of a standardized form is a misnomer as a single ED can receive patients form multiple NHs, each of which may use a different type of transfer form³².

Advances in health information technology and health information exchange can address challenges arising from poor information sharing and paper-based transfer forms by automating processes (i.e. by not being human dependent) and improving the completeness of patient information by accessing information from multiple institutions simultaneously³³. In addition, the potentially negative impact of low levels of health information technology adoption among NHs on patient care can be mitigated through health information exchange which increases the breadth and amount of information available^34,35. While technical challenges hindered the development of earlier electronic transfer forms³⁶, interoperability standards have progressed significantly. In particular, the Fast Health Interoperability Resources (FHIR) standard from the Health Level Seven International (HL7) organization enables the consolidation and representation of numerous data elements within an electronic health record (EHR) environment^37,38. This study developed a SMART (Substitutable Medical Applications and Reusable Technologies) on FHIR prototype automated transfer form for NH patients who receive care in EDs.

Methods

We developed a prototype automated transfer form to meet ED providers’ information needs when delivering care to NH patients. Over a 10-month period beginning in April 2018, we identified priority data elements through a multistep process that included individual interviews with ED providers, a review of existing transfer forms, a targeted survey of end users, and a design workshop.

End user interviews & analysis

Key informants represented two different EDs that participated in HEALTHeLINK (Buffalo) or the Bronx Regional Health Information Organization (RHIO) information exchanges in New York State. We conducted 6 interviews with physicians and one interview with a nurse. Interviews followed a semi-structured interview guide, which covered the areas of information gathering, workflow, identification of information that is difficult to access and/or considered high priority, and information needs of ED clinicians. All interviews were conducted over the phone by two or more team members and were recorded with consent.

We analyzed the transcripts using a template analytical approach, which applies an a priori set of hierarchical codes ^39,40. Our template was based on the 5 Rights of clinical decision support⁴¹: important data elements (right information); the appropriate member of the care team (right person); a preferred point in the workflow (right time); accessing information in the EHR or through the health information exchange portal (right channel); and how end users receive the information (right format). Two coders applied the final coding scheme to all transcripts independently. We then conducted consensus coding and differences were resolved through joint readings and discussion. If consensus could not be reached by the two coders, a third team member was brought in to resolve differences.

Review of existing transfer forms

We identified five paper-based transfer forms specific to NH patients^42–46 and two clinical abstracts designed for ED end users ⁴⁷. From these forms we abstracted 220 (unduplicated) data elements.

End user survey

We then compiled, deduplicated, and collapsed (near) synonymous terms derived from the interviews and our review of paper-based forms. From the paper-based forms, we used elements that appeared on at least four of the seven total exemplars. We grouped the resultant elements into 5 categories to facilitate respondents’ completion of the survey: respondent demographics, patient identifiers, medical decision making, laboratory and imaging / radiology, allergies and medications, reasons for transfer and medical history, and NH information (Table 1). Survey respondents rated the value of each element when seeing a NH patient in the ED as: not at all, moderately, or extremely. Patient identifying information, such as names and medical record numbers, were assumed to be required information and therefore we excluded these elements from the ranking process. We distributed an anonymous REDCap^48,49 survey to a convenience sample of providers practicing in EDs located in New York State (using personal contacts and those with connections to HEALTHeLINK, the Bronx RHIO, and the New York eHealth Collaborative). A total of 10 individuals from 5 different hospitals responded to the anonymous survey.

Table 1.

Categories and data elements included in the end user survey

Category	Elements (example)
Patient identifiers	Name, medical record number, date of birth, race/ethnicity, gender, address, phone number, marital status
Medical decision making	Advanced directive, emergency contact / healthcare proxy, primary care provider
Laboratory & imaging	Complete blood count, blood type, creatinine, coagulation values, urine culture, imaging / radiology, other
Allergies & medications	Food allergies, medication allergies, current medication list, medication administration record
Reasons for transfer & medical history	Reason the patient was transferred to the ED, most recent vitals, baseline neurological state, current neurological state, relevant diagnoses / medical history, recent encounters (with dates), notes from recent encounters
Nursing home information	Phone number for staff familiar with patient, phone number for supervisor, phone number for front desk, facility type (nursing home, rehabilitation, etc.), nursing home care capabilities (e.g. can administer IV antibiotics)

Design workshop

We conducted a design workshop with six key informants from a Buffalo area health system (4 physicians, 1 registered nurse, 1 information systems specialist). The goal of the workshop was to inform the eventual design and implementation of the transfer form to maximize end user acceptance and usage. We provided all participants with a summary of our survey and qualitative findings on the most important data elements. The workshop commenced with a review of the overall project objectives and a nominal group technique to identify any key data elements that had not been previously mentioned. These included data both currently and not currently available electronically. Next, workshop participants broke out into two small groups each with a moderator. Each group was given examples of existing paper-based transfer forms, asked to develop a layout that best supported their decision making, and then create a paper mock-up. Groups reconvened and shared examples for group discussion. Finally, the workshop concluded with a moderated group discussion on the timing of usage within clinical workflows by staff type and how the example transfer form could be accessed through the EHR or the information exchange portal.

Prototype development & testing

We developed a SMART on FHIR prototype application designed to access a FHIR bundle and other standardized data sources to create and return a transfer form populated with data specific to the care of NH patients in the ED. The application generated an HTML page that could be viewed during a single session and/or printed. The prototype application was written in vanilla JavaScript and compliant with FHIR version 3. The prototype application was tested using data available from smarthealthit.org.

Results

Right information

Five data elements (Table 2) had a mean rating of 3.0, meaning that the entire convenience sample viewed them as extremely useful: emergency contact/healthcare proxy, current medication list, reason for transfer to the ED, baseline neurological state, and relevant diagnoses / medical history. The interviewees mentioned each of these elements, but also provided clarification, described current challenges in obtaining these elements, or illustrated their importance based on personal experience. For example, a desire for information on medical history was often bounded by its relevance to the patient’s current health problems and recency. A key informant commented, “I don't really care if they had their wisdom teeth out when they were 16 and they're 90, but just key problems that are important: coronary heart disease, hypertension, diabetes…” Also, key informants expressed a desire to know why the transfer occurred, e.g., "I want to know in simple language like why are they being sent to the emergency department. “The reason for transfer was noted as particularly important information for patients who were unable to provide the reason themselves: “ …(P)robably the most important part, particularly for patients who have some element of dementia, is an explanation of the events that led to the patient being transferred… (T)hat would be the first thing I would be looking for…” However, this information was not always available. One key informant noted, “I would say the reason for transfer, kind of really obvious, but sometimes you don't actually know that” and another recounted: “(S)ometimes I feel like you really don't get all the information… Why was this patient sent out at 5:00 in the morning, when we call the facility [and find out that the] facility doctor comes in at 7:00 am? What emergency… Why couldn't it have waited?” Emergency contact information was noted as important in understanding patient care preferences in the ED and for discharge planning. The design workshop participants echoed the importance of these factors, but also expanded the concept of current functional status to include ambulatory status.

Table 2.

Rankings of data elements.

Data element	Mean rating
Emergency Contact/Healthcare Proxy	3.0
Current Medication List	3.0
Reason patient transferred to ED	3.0
Baseline neurological state	3.0
Relevant diagnoses / medical history	3.0
Advanced directive	2.9
Medication allergies	2.9
Most recent vitals	2.9
Phone number for nursing home staff familiar with patient	2.9
Nursing home capabilities	2.8
Primary care provider	2.6
Current neurological state	2.6
Medication Administration Record (MAR)	2.4
Phone number for nursing home supervisor	2.4
Nursing home level of care	2.4
Recent healthcare encounter types and dates	2.3
Lab: CBC	2.2
Lab: Urine Cultures	2.2
Notes from recent encounters	2.2
Imaging/Radiology	2.11
Lab: Creatinine	2.1
Food Allergies	1.91
Lab: Coagulation Values	1.8
Phone number for nursing home front desk	1.81
Lab: other	1.61
Lab: Blood Type	1.2

¹Not specifically identified in qualitative interviews as important or desired factors.

An additional seven data elements (Table 2) were highly rated (mean >2.5) in the areas of medical decision making (advance directive, primary care provider contact information), current status (most recent vitals, current neurological state), allergies and medications, and NH information. Interviewees qualified the importance of medications by relevance and recency, and also for other data elements. For example, providers commented: “(N)ursing homes tend to send us a bunch of paper and most of the paper they just copy and send us the MAR [Medication Administration Record] that sometimes goes back weeks. It’s not helpful. It’s just a bunch of paper" and “We're often looking for needles in haystacks and people send us more hay. That doesn't help. We don't need to know every medication they've been on for the last year. We just need to know what medicines they're on right now and what medicine is they didn't take or didn't get.”

Two nonclinical data elements, primary care provider contact information and NH information, were also discussed in key informant interviews. One informant stated that these data elements could support further information gathering: “One of the things that’s frequently lacking and leads to frustration is a good contact number for somebody at the nursing home facility… Where you can actually speak to a nurse or care provider who is familiar with that patient and knows what was going on.” In addition, these data could influence care decisions. An ED physician reported: "It’s not real clear whether they're in there long term, assisted living, or rehab. Which is not that important from a medical standpoint, but if you're trying to admit [the patient]...what level of care they're at is important and it’s not always immediately obvious from the paperwork." These elements were also highlighted by design workshop participants.

Notably, laboratory tests and imaging results tended to rank at the bottom of the list of data elements. As one respondent noted, it was information the physicians already had access to: "Oftentimes with these patients, the lab work that they're sending is either so old that it’s not quite as useful or they actually just send back the lab work from their last visit with me at [hospital name]."

Right time, channel, format, and person

With some variation, key informants generally outlined a similar process for soliciting information about NH patients in the ED: 1) review triage notes (chief complaint) -> 2) review paper transfer forms (if available) -> 3) patient interview -> 3) review EHRs -> 4) additional steps if necessary, e.g., phone calls to the NH and search health information exchange. Within these steps, key informants stated a preference for their prioritized data elements as "earlier the better" or "before I go talk to the person".

Key informants noted several channels for obtaining relevant patient information. Based on the key informants’ experiences, current paper-based transfer forms were not the best channel due to information that was occasionally missing, incomplete, or that varied substantially. For example, one physician described the situation as follows: "(I)t’s so variable what you get from each site… Some places it'll just list like their medical problems. Maybe get a list of medications. Other places you won't get their medications, but you might get a very clear reason why they're there [that is] written out on a piece of paper. But it does not seem like there’s any standard way that information is gathered and it really depends on the staff. Some people will literally copy an entire chart on a patient. Other patients come over with like these sticky notes saying 'agitated'. So, it is very variable depending on the facility." Similarly, phone calls were not always productive: "Sometimes we're never able to get a hold of someone. We maybe don't meet expectations with the person who sent them to the emergency department so then they send them back the next day still without any communication of what we can do to help the patient."

Differences in data needs by job type (i.e. right person) were apparent in the interviews. While physicians did not need data elements such as insurance information and demographics, they did recognize their importance for other staff members. For example, a physician stated, "Having contact information for our social workers, because that’s usually when they get involved when our facility is refusing to take a patient back or there are concerns about the level of care they're receiving." Likewise, the sole nurse key informant had a case management role and reported that contact information helped place individuals into NHs.

Prototype transfer form design

In terms of format, the members of the design workshop expressed a preference for a single page, top down form, with different colors for the headers and information, and boxing or grouping of key information. Overall, the preference was for the top of the form to have "information about today", e.g. the information that helped explain the patient’s presence in the ED (such as the reason for the transfer) and that supported any immediate clinical decisions (such as advance directives). Information that was used in "later decision making", such as contact information and facility characteristics was suggested to be at the bottom of form. Participants wanted all data elements to include date/time stamps and to clearly designate when information was not available. Participants also suggested including a checkbox within the ED triage intake form to identify NH patients at arrival, which could trigger the report so that it was available prior to the patient interview (Right time).

The working prototype (Figure 1) included many of the highly rated data elements as well as additional elements identified in the design workshop. Not all the desired or represented elements are currently mapped as FHIR resources in HEALTHeLINK. While a desired field, the prototype cannot report the reason for the transfer to the ED since this data element currently has no FHIR specification and was not being sent to HEALTeLINK by any NH partners.

Figure 1.

Example of the prototype SMART on FHIR automated transfer form.

Limitations

This study has several limitations. Generalizability of the prototype in terms of end user acceptance may be limited. While we have the views of multiple health organizations and different staff types, the sample size was limited and interviews with these types of stakeholders in other states may have identified other key data elements. Similarly, providers working within different EHR environments may have had different preferences for the timing of the transfer form. Lastly, we did not study end user acceptance or uptake of the application in a live clinical environment. The next planned phase of the research is to pilot the transfer form in the ED and evaluate usage.

Conclusion

We developed a prototype SMART on FHIR transfer form for use in the ED to support the care of NH patients based on the information desired by end users and their preferences for its delivery. Notably, interviews, surveys, and the design workshop all indicated the importance of both clinical and nonclinical data in the care process. The clinical data elements to support care delivery identified in the study were consistent with prior research²⁸ and the nonclinical data elements were largely to facilitate communication with NHs in order to overcome poor information sharing.

A challenge facing the development of our prototype, and potentially other FHIR-based applications, is the electronic availability of all desired data elements. A prior effort to better share information from NHs during ED visits was undermined by the unavailability of standardized data³⁶. The use of an emerging standard is an advantage over prior efforts and the number of FHIR resources is expansive, but many potentially useful resources still remain at low maturity-levels or have not been specified. Opportunities exist to improve the content within the transfer form as more resources become available, or as more data types are shared by NHs through health information exchange. For the nonclinical information, some public directory resources, such as those from the Centers for Medicare and Medicaid Services, could be incorporated in future transfer forms. However, that information may not be specific enough to support care in the ED. An additional enhancement to the transfer form could be the ability to annotate or add very specific information such as NH staff contact information to supplement information in public directories.

Notably, our current transfer form prototype could not include all the data elements identified as important by key informant end users. However, identifying desired data elements that are currently unavailable provides an initial roadmap of new elements for EHR vendors, information exchanges, and NHs to pursue. As more data becomes available, they will be included in future iterations of the transfer form application.

The ED is a frequent site of care for NH patients, but care in this setting is fraught with communication challenges. The expansion of interoperable health information and standards may be leveraged to address the limitations of paper-based information sharing.

Acknowledgements

This work was supported by the New York eHealth Collaborative. We thank the support and cooperation of HEALTHeLINK and the Bronx RHIO.