Health information exchange (HIE) addresses these shortcomings by providing clinicians who are not part of the same organizational entity with the capability to electronically retrieve and share health-related information across regional, community, or organizational boundaries.8-10 Studies have shown that HIE can lead to numerous benefits such as reduced utilization of healthcare services,11 reduced diagnostic testing,12 prescription safety,13 public health support, and cost savings.14-16 Healthcare organizations can engage in HIE using different technological architectures for point-to-point communication (ie, the direct exchange of data between electronic health record [EHR] systems from different vendors) between providers, and under various interorganizational collaborations such as services provided through a single intermediary health information organization.16,17
HIE models vary across the United States, but often involve physicians, hospitals, and other healthcare stakeholders in the community joining a local HIE effort run by a regional health information organization (RHIO)18 or another model for exchange. These include enterprise HIE efforts (ie, an HIE system privately owned, funded, and developed by a single enterprise or organization)19 and vendor-mediated exchange.20,21 In general, RHIOs have emerged as the leading model to facilitate the electronic exchange of information between healthcare organizations.22 A common feature for all HIE models is that information exchange is enabled through the electronic transmissionof data represented by 2 exchange methods—“push” and “pull”.17,23 Push takes place when clinical data are electronically deposited into a recipient’s system after a sender initiates transfer. Pull is initiated when a user proactively uses a HIE system to retrieve aggregated patient health data stemming from multiple sources across a community.
Existing studies have evaluated the use of pull-based HIE systems,24-27 and 1 quantitative study examined RHIO-managed HIEs with both push and pull capabilities,17 however, little qualitative research exists that examines the experience of healthcare professionals using multiple HIE models and push/pull methods at their place of work.
With the myriad options organizations have for HIE, our goal was to understand the dimensions of the information retrieval pathways used by healthcare professionals to engage in HIE. We sought to assess how healthcare professionals met their information needs in the context of laboratory test results, radiological images and reports, and medication histories. Access to these types of data has the greatest potential for impact on patient health, and holds the greatest promise for cost savings,28 improved diagnosing and monitoring of health and diseases,29 detection of diseases at an early stage,30 and reduced frequency of repeat diagnostic testing.31
Study Design and Setting
We undertook a qualitative multicase study approach in New York state. The state provides a relevant setting to evaluate HIE; it has invested more than $440 million in health information technology and HIE adoption over the last 7 years.32 We consulted with the New York eHealth Collaborative, a private-public nonprofit organization charged with facilitating statewide HIE by coordinating the creation of a network to connect healthcare providers, to identify 3 RHIOs serving distinctly diverse communities; this selection process would maximize variation and allow for comparison.5 The RHIOs that matched our criteria for selection agreed to participate in the study and cooperated in identifying care organizations for our study sites. The communities varied in terms of population size and consented patients, and in the number of users (clinicians and staff) with access to their local RHIO’s HIE (Table 1). Each RHIO implemented a different commercial HIE platform; the exchange architectures also differed. Community A and Community C used a federated model, and Community B used a centralized model. In a federated model, an organization locally stores and retains control over the patient information, and it responds when another organization, also a member of the same RHIO HIE effort, requests information. In a centralized model, patient data is stored in a central repository maintained by the RHIO after being collected from the organizations participating in the local RHIO effort.
We provided a brief project overview and obtained verbal and written informed consent from each care organization's clinicians and staff who agreed to participate in this study. The consent form was provided by the Institutional Review Board of Weill Cornell Medical College, which approved this study.
Data was collected during 2-day site visits at each participating care organization. To increase the internal validity of our study, we used 2 sources for evidence: observation and interviews.34 All data collection began in May 2013 and ended in June 2013. RHIO staff helped identify and secure the cooperation of key informants who had experience utilizing various HIE models. We used a snowball sampling procedure in which we first interviewed a key informant; they were then asked to identify and help recruit other potential informants who were familiar with HIE and could provide additional insight about their information retrieval experiences.35 Because these other potential informants came from the key informants' social networks, we believed this would reflect the perspectives of a diverse community of healthcare professionals.36 We also observed participants and recorded the field notes of healthcare professionals engaging in HIE activities.37 Data gathering at each site ended when the point of data saturation was reached—that is, when the interviews and observations did not produce any new information.38 All interviews lasted an average of 30 minutes and were conducted using a semi-structured protocol. This allowed us to delve deeper into a topic concerning information retrieval approaches, to pose additional questions based on subject responses, and to thoroughly understand the answers that were given. All interviews were audio recorded and transcribed.
From the interviews and observations, it was possible to obtain a narrative of HIE models and exchange methods used in each community. For this analysis, we undertook an iterative, open-coding approach to the interview and observational data.39 To begin, we identified broad themes around the important concepts of information retrieval pathways for the types of information sought, based on our research objective. We completed the initial coding as close as possible to the date of data collection. During a second cycle of coding, we identified new concepts and relationships that were not adequately represented by the initial categories. This open coding was undertaken independently by authors PK and JRV, who were responsible for all phases of the coding. Collaboratively, PK and JRV reviewed and revised the coding schemes to consolidate redundant codes.
RESULTS We interviewed a total of 48 individuals across 3 emergency departments (EDs), 7 primary care practices (PCPs), and 2 local public health departments (LHDs) (Table 1). We observed healthcare practitioner activities for more than 40 hours across all 3 communities. The PCPs included a federally qualified health center, a community health center, solo practices, and large group practices. Each care organization was equipped with an EHR) and had access to their RHIO’s HIE system. The LHDs provided clinical services but did not have an EHR. LHDs were included in this study because HIE can provide them with access to previously hard-to-obtain clinical and demographic data for the purposes of disease surveillance, disaster response, and healthcare service delivery. We obtained the perspective of 3 broad types of professionals: clinicians (physicians, nurses, physician assistants, scribes, pharmacists, and medical assistants); case managers and public health disease investigation staff; and administrative staff (directors, information technology specialists, medical records personnel, quality managers, clerks, and front-desk staff). We used these larger groupings to protect confidentiality.Information Retrieval Pathways