Expanding Health Information Exchange Improves Identification of Frequent Emergency Department Users.

STUDY OBJECTIVE: Frequent emergency department (ED) users are of interest to policymakers and hospitals. The objective of this study is to examine the effect of health information exchange size on the identification of frequent ED users. METHODS: We retrospectively analyzed data from Healthix, a health information exchange in New York that previously included 10 hospitals and then grew to 31 hospitals. We divided patients into 3 cohorts: high-frequency ED users with 4 or more visits in any 30-day period, medium-frequency ED users with 4 or more visits in any year, and infrequent ED users with fewer than 4 visits in any year. For both the smaller (10-hospital) and larger (31-hospital) health information exchanges, we compared the identification rate of frequent ED users that was based on hospital-specific data with the corresponding rates that were based on health information exchange data. RESULTS: The smaller health information exchange (n=1,696,279 unique ED patients) identified 11.4% more high-frequency users (33,467 versus 30,057) and 9.5% more medium-frequency users (109,497 versus 100,014) than the hospital-specific data. The larger health information exchange (n=3,684,999) identified 19.6% more high-frequency patients (52,727 versus 44,079) and 18.2% more medium-frequency patients (222,574 versus 192,541) than the hospital-specific data. Expanding from the smaller health information exchange to the larger one, we found an absolute increase of 8.2% and 8.7% identified high- and medium-frequency users, respectively. CONCLUSION: Increasing health information exchange size more accurately reflects how patients access EDs and ultimately improves not only the total number of identified frequent ED users but also their identification rate.

Mapping of HIE CT terms to LOINC®: analysis of content-dependent coverage and coverage improvement through new term creation.

Objective: We describe and evaluate the mapping of computerized tomography (CT) terms from 40 hospitals participating in a health information exchange (HIE) to a standard terminology. Methods: Proprietary CT exam terms and corresponding exam frequency data were obtained from 40 participant HIE sites that transmitted radiology data to the HIE from January 2013 through October 2015. These terms were mapped to the Logical Observations Identifiers Names and Codes (LOINC®) terminology using the Regenstrief LOINC mapping assistant (RELMA) beginning in January 2016. Terms without initial LOINC match were submitted to LOINC as new term requests on an ongoing basis. After new LOINC terms were created, proprietary terms without an initial match were reviewed and mapped to these new LOINC terms where appropriate. Content type and token coverage were calculated for the LOINC version at the time of initial mapping (v2.54) and for the most recently released version at the time of our analysis (v2.63). Descriptive analysis was performed to assess for significant differences in content-dependent coverage between the 2 versions. Results: LOINC's content type and token coverages of HIE CT exam terms for version 2.54 were 83% and 95%, respectively. Two-hundred-fifteen new LOINC CT terms were created in the interval between the releases of version 2.54 and 2.63, and content type and token coverages, respectively, increased to 93% and 99% (P < .001). Conclusion: LOINC's content type coverage of proprietary CT terms across 40 HIE sites was 83% but improved significantly to 93% following new term creation.

Bouncing Back Elsewhere: Multilevel Analysis of Return Visits to the Same or a Different Hospital After Initial Emergency Department Presentation.

STUDY OBJECTIVE: Analyses of 72-hour emergency department (ED) return visits are frequently used for quality assurance purposes and have been proposed as a means of measuring provider performance. These analyses have traditionally examined only patients returning to the same hospital as the initial visit. We use a health information exchange network to describe differences between ED visits resulting in 72-hour revisits to the same hospital and those resulting in revisits to a different site. METHODS: We examined data from a 31-hospital health information exchange of all ED visits during a 5-year period to identify 72-hour return visits and collected available encounter, patient, and hospital variables. Next, we used multilevel analysis of encounter-level, patient-level, and hospital-level data to describe differences between initial ED visits resulting in different-site and same-site return visits. RESULTS: We identified 12,621,159 patient visits to the 31 study EDs, including 841,259 same-site and 107,713 different-site return visits within 72 hours of initial ED presentation. We calculated odds ratios (ORs) and 95% confidence intervals (CIs) for the initial-visit characteristics' predictive relationship that any return visit would be at a different site: daytime visit (OR 1.10; 95% CI 1.07 to 1.12), patient-hospital county concordance (OR 1.40; 95% CI 1.36 to 1.44), male sex (OR 1.27; 95% CI 1.24 to 1.30), aged 65 years or older (OR 0.55; 95% CI 0.53 to 0.57), sites with an ED residency (OR 0.41; 95% CI 0.40 to 0.43), sites at an academic hospital (OR 1.12; 95% CI 1.08 to 1.15), sites with high density of surrounding EDs (OR 1.73; 95% CI 1.68 to 1.77), and sites with a high frequency of same-site return visits (OR 0.10; 95% CI 0.10 to 0.11). CONCLUSION: This analysis describes how ED encounters with early revisits to the same hospital differ from those with revisits to a second hospital. These findings challenge the use of single-site return-visit frequency as a quality measure, and, more constructively, describe how hospitals can use health information exchange to more accurately identify early ED return visits and to support programs related to these revisits.

Standard Anatomic Terminologies: Comparison for Use in a Health Information Exchange-Based Prior Computed Tomography (CT) Alerting System.

BACKGROUND: A health information exchange (HIE)-based prior computed tomography (CT) alerting system may reduce avoidable CT imaging by notifying ordering clinicians of prior relevant studies when a study is ordered. For maximal effectiveness, a system would alert not only for prior same CTs (exams mapped to the same code from an exam name terminology) but also for similar CTs (exams mapped to different exam name terminology codes but in the same anatomic region) and anatomically proximate CTs (exams in adjacent anatomic regions). Notification of previous same studies across an HIE requires mapping of local site CT codes to a standard terminology for exam names (such as Logical Observation Identifiers Names and Codes [LOINC]) to show that two studies with different local codes and descriptions are equivalent. Notifying of prior similar or proximate CTs requires an additional mapping of exam codes to anatomic regions, ideally coded by an anatomic terminology. Several anatomic terminologies exist, but no prior studies have evaluated how well they would support an alerting use case. OBJECTIVE: The aim of this study was to evaluate the fitness of five existing standard anatomic terminologies to support similar or proximate alerts of an HIE-based prior CT alerting system. METHODS: We compared five standard anatomic terminologies (Foundational Model of Anatomy, Systematized Nomenclature of Medicine Clinical Terms, RadLex, LOINC, and LOINC/Radiological Society of North America [RSNA] Radiology Playbook) to an anatomic framework created specifically for our use case (Simple ANatomic Ontology for Proximity or Similarity [SANOPS]), to determine whether the existing terminologies could support our use case without modification. On the basis of an assessment of optimal terminology features for our purpose, we developed an ordinal anatomic terminology utility classification. We mapped samples of 100 random and the 100 most frequent LOINC CT codes to anatomic regions in each terminology, assigned utility classes for each mapping, and statistically compared each terminology's utility class rankings. We also constructed seven hypothetical alerting scenarios to illustrate the terminologies' differences. RESULTS: Both RadLex and the LOINC/RSNA Radiology Playbook anatomic terminologies ranked significantly better (P<.001) than the other standard terminologies for the 100 most frequent CTs, but no terminology ranked significantly better than any other for 100 random CTs. Hypothetical scenarios illustrated instances where no standard terminology would support appropriate proximate or similar alerts, without modification. CONCLUSIONS: LOINC/RSNA Radiology Playbook and RadLex's anatomic terminologies appear well suited to support proximate or similar alerts for commonly ordered CTs, but for less commonly ordered tests, modification of the existing terminologies with concepts and relations from SANOPS would likely be required. Our findings suggest SANOPS may serve as a framework for enhancing anatomic terminologies in support of other similar use cases.

Patient crossover and potentially avoidable repeat computed tomography exams across a health information exchange.

OBJECTIVE: The purpose of this study was to measure the number of repeat computed tomography (CT) scans performed across an established health information exchange (HIE) in New York City. The long-term objective is to build an HIE-based duplicate CT alerting system to reduce potentially avoidable duplicate CTs. METHODS: This retrospective cohort analysis was based on HIE CT study records performed between March 2009 and July 2012. The number of CTs performed, the total number of patients receiving CTs, and the hospital locations where CTs were performed for each unique patient were calculated. Using a previously described process established by one of the authors, hospital-specific proprietary CT codes were mapped to the Logical Observation Identifiers Names and Codes (LOINC®) standard terminology for inter-site comparison. The number of locations where there was a repeated CT performed with the same LOINC code was then calculated for each unique patient. RESULTS: There were 717 231 CTs performed on 349 321 patients. Of these patients, 339 821 had all of their imaging studies performed at a single location, accounting for 668 938 CTs. Of these, 9500 patients had 48 293 CTs performed at more than one location. Of these, 6284 patients had 24 978 CTs with the same LOINC code performed at multiple locations. The median time between studies with the same LOINC code was 232 days (range of 0 to 1227); however, 1327 were performed within 7 days and 5000 within 30 days. CONCLUSIONS: A small proportion (3%) of our cohort had CTs performed at more than one location, however this represents a large number of scans (48 293). A noteworthy portion of these CTs (51.7%) shared the same LOINC code and may represent potentially avoidable studies, especially those done within a short time frame. This represents an addressable issue, and future HIE-based alerts could be utilized to reduce potentially avoidable CT scans.

Increased Identification of Emergency Department 72-hour Returns Using Multihospital Health Information Exchange.

OBJECTIVES: Emergency departments (EDs) commonly analyze cases of patients returning within 72 hours of initial ED discharge as potential opportunities for quality improvement. In this study, we tested the use of a health information exchange (HIE) to improve identification of 72-hour return visits compared to individual hospitals' site-specific data. METHODS: We collected deidentified patient data over a 5-year study period from Healthix, an HIE in the New York metropolitan area. We measured site-specific 72-hour ED returns and compared these data to those obtained from a regional 31-site HIE (Healthix) and to those from a smaller, antecedent 11-site HIE. Although only ED visits were counted as index visits, either ED or inpatient revisits within 72 hours of the index visit were considered as early returns. RESULTS: A total of 12,669,657 patient encounters were analyzed across the 31 HIE EDs, including 6,352,829 encounters from the antecedent 11-site HIE. Site-specific 72-hour return visit rates ranged from 1.1% to 15.2% (median = 5.8%) among the individual 31 sites. When the larger HIE was used to identify return visits to any site, individual EDs had a 72-hour return frequency of 1.8% to 15.5% (median = 6.8%). HIE increased the identification ability of 72-hour ED return analyses by a mean of 11.16% (95% confidence interval = 11.10% to 11.22%) compared with site-specific (no HIE) analyses. CONCLUSION: This analysis demonstrates incremental improvements in our ability to identify early ED returns using increasing levels of HIE data aggregation. Although intuitive, this has not been previously described using HIE. ED quality measurement and patient safety efforts may be aided by using HIE in 72-hour return analyses.

Validating Health Information Exchange (HIE) Data For Quality Measurement Across Four Hospitals.

Health information exchange (HIE) provides an essential enhancement to electronic health records (EHR), allowing information to follow patients across provider organizations. There is also an opportunity to improve public health surveillance, quality measurement, and research through secondary use of HIE data, but data quality presents potential barriers. Our objective was to validate the secondary use of HIE data for two emergency department (ED) quality measures: identification of frequent ED users and early (72-hour) ED returns. We compared concordance of various demographic and encounter data from an HIE for four hospitals to data provided by the hospitals from their EHRs over a two year period, and then compared measurement of our two quality measures using both HIE and EHR data. We found that, following data cleaning, there was no significant difference in the total counts for frequent ED users or early ED returns for any of the four hospitals (p<0.001).

The visit-data warehouse: enabling novel secondary use of health information exchange data.

INTRODUCTION/OBJECTIVES: Health Information Exchange (HIE) efforts face challenges with data quality and performance, and this becomes especially problematic when data is leveraged for uses beyond primary clinical use. We describe a secondary data infrastructure focusing on patient-encounter, nonclinical data that was built on top of a functioning HIE platform to support novel secondary data uses and prevent potentially negative impacts these uses might have otherwise had on HIE system performance. BACKGROUND: HIE efforts have generally formed for the primary clinical use of individual clinical providers searching for data on individual patients under their care, but many secondary uses have been proposed and are being piloted to support care management, quality improvement, and public health. DESCRIPTION OF THE HIE AND BASE INFRASTRUCTURE: This infrastructure review describes a module built into the Healthix HIE. Healthix, based in the New York metropolitan region, comprises 107 participating organizations with 29,946 acute-care beds in 383 facilities, and includes more than 9.2 million unique patients. The primary infrastructure is based on the InterSystems proprietary Caché data model distributed across servers in multiple locations, and uses a master patient index to link individual patients' records across multiple sites. We built a parallel platform, the "visit data warehouse," of patient encounter data (demographics, date, time, and type of visit) using a relational database model to allow accessibility using standard database tools and flexibility for developing secondary data use cases. These four secondary use cases include the following: (1) tracking encounter-based metrics in a newly established geriatric emergency department (ED), (2) creating a dashboard to provide a visual display as well as a tabular output of near-real-time de-identified encounter data from the data warehouse, (3) tracking frequent ED users as part of a regional-approach to case management intervention, and (4) improving an existing quality improvement program that analyzes patients with return visits to EDs within 72 hours of discharge. RESULTS/LESSONS LEARNED: Setting up a separate, near-real-time, encounters-based relational database to complement an HIE built on a hierarchical database is feasible, and may be necessary to support many secondary uses of HIE data. As of November 2014, the visit-data warehouse (VDW) built by Healthix is undergoing technical validation testing and updates on an hourly basis. We had to address data integrity issues with both nonstandard and missing HL7 messages because of varied HL7 implementation across the HIE. Also, given our HIEs federated structure, some sites expressed concerns regarding data centralization for the VDW. An established and stable HIE governance structure was critical in overcoming this initial reluctance. CONCLUSIONS: As secondary use of HIE data becomes more prevalent, it may be increasingly necessary to build separate infrastructure to support secondary use without compromising performance. More research is needed to determine optimal ways of building such infrastructure and validating its use for secondary purposes.