- Email: [email protected]
Health information exchange system usage patterns in three communities: Practice sites, users, patients, and data
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
journal homepage: www.ijmijournal.com
Health information exchange system usage patterns in three communities: Practice sites, users, patients, and data Thomas R. Campion Jr. a,b,c,d,e,∗ , Alison M. Edwards a,b,d,e , Stephen B. Johnson a,b,d , Rainu Kaushal a,b,c,d,e,f,g , with the HITEC investigators a
Division of Quality and Medical Informatics, Weill Cornell Medical College, New York, NY, United States Department of Public Health, Weill Cornell Medical College, New York, NY, United States c Department of Pediatrics, Weill Cornell Medical College, New York, NY, United States d Center for Healthcare Informatics and Policy, Weill Cornell Medical College, New York, NY, United States e Health Information Technology Evaluation Collaborative (HITEC), New York, NY, United States f Department of Medicine, Weill Cornell Medical College, New York, NY, United States g Komansky Center for Children’s Health at NewYork-Presbyterian Hospital, New York, NY, United States b
a r t i c l e
i n f o
a b s t r a c t
Article history:
Objectives: Public and private organizations are implementing systems for query-based
Received 22 October 2012
health information exchange (HIE), the electronic aggregation of patient data from multi-
Received in revised form
ple institutions. However, existing studies of query-based HIE system usage have addressed
30 April 2013
a limited number of settings. Our goal was to quantify the breadth and depth of usage
Accepted 2 May 2013
of a query-based HIE system implemented across multiple communities with diverse care settings and patient populations.
Keywords:
Methods: We performed a cross-sectional study in three communities in New York State
Health information technology
using system access log files from January 2009 to May 2011 to measure usage patterns of a
Electronic health records
query-based HIE web portal system with respect to practice sites, users, patients, and data.
Evaluation studies
Results: System access occurred from 60% (n = 200) of practice sites registered to use the
Community
system in Community A, 59% (n = 156) in Community B, and 82% (n = 28) in Community
Health information exchange
C. In Communities A and B, users were primarily non-clinical staff in outpatient settings, while in Community C inpatient physicians were the main users. Across communities, proportions of patients whose data were accessed varied with 5% (n = 11,263) in Community A, 60% (n = 212,586) in Community B, and 1% (n = 1107) in Community C. In Community B, users updated patient consent through the HIE portal, whereas in the other communities, users updated patient consent through a separate system. Across communities, users most frequently accessed only patient summary data displayed by default followed by detailed laboratory and radiology data. Conclusions: This study is among the first to illustrate large-scale usage of a query-based HIE system implemented across multiple communities. Patient summary data displayed by default may be an important feature of query-based HIE systems. User role, practice site type, and patient consent workflow may affect patterns of query-based HIE web portal system usage in the communities studied and elsewhere. © 2013 Elsevier Ireland Ltd. All rights reserved.
∗
Corresponding author at: 402 East 61st Street, DV-308, New York, NY 10065, United States. Tel.: +1 646 962 2345; fax: +1 646 962 0105. E-mail address: [email protected] (T.R. Campion Jr.). 1386-5056/$ – see front matter © 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ijmedinf.2013.05.001
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
1.
Introduction
Electronic transfer of patient data across healthcare organizations has the potential to improve patient care coordination, facilitate public health efforts, and reduce healthcare costs [1,2]. Currently public and private organizations are implementing multiple forms of clinical data exchange [3] including directed exchange that transmits patient data from one provider to another and query-based health information exchange (HIE) that aggregates patient data from multiple healthcare institutions [4]. The two forms of clinical data exchange address different but complementary patient data needs [5]. For example, directed exchange automates paper-based point-to-point processes such as patient referrals and delivery of laboratory results, while query-based HIE assembles comprehensive patient data from numerous organizations across a community. Query-based exchange using Cross-Enterprise Document Sharing (XDS), a specification developed by Integrating the Healthcare Enterprise (IHE), has been successfully deployed in multiple communities across Europe [6,7] and Asia [8,9]. In the United States today, federal policies have aligned usage of clinical data exchange with financial incentives. Novel healthcare models such as accountable care organizations rely increasingly on query-based HIE to gather patient population data from multiple institutions so clinicians can coordinate care for quality- and cost-based reimbursement [10]. Evidence of query-based HIE’s impact is also beginning to appear in the literature. Usage of query-based exchange systems facilitated by a private insurer [11] and a non-profit regional health information organization (RHIO) [12] have been associated with cost reductions in emergency department settings in two communities, but it is unknown whether findings generalize to other communities and care settings. Additionally, optimal approaches for the configuration and usage of query-based HIE systems are unknown. As a first step in evaluating community-based health information technologies, practitioners and researchers must understand how clinicians and staff use electronic systems [13–16]. For query-based HIE, quantitative measures of usage are critical for understanding the breadth and depth of system adoption, implementation, and resistance in communities, as availability of systems does not guarantee their usage or the quality of their usage [16]. However, existing quantitative studies of query-based exchange usage are limited to emergency department and safety net settings in two communities [17,18]. Furthermore, the two communities implemented different query-based exchange systems, limiting the ability to compare system usage patterns across communities. Broader study of query-based exchange system usage patterns across multiple communities, care settings, patient populations, and software systems is necessary to identify optimal approaches for HIE that improve healthcare decision making and outcomes [13–15]. In this study, our goal was to quantify the breadth and depth of usage of a commercially available system for query-based exchange implemented in three communities by measuring usage patterns with respect to practice sites, users, patients, and data. Understanding usage patterns of a
811
query-based exchange system implemented on a large-scale can inform efforts by organizations implementing HIE, vendors developing software for HIE, and policymakers defining rules for HIE to increase HIE usage in support of patient care and public health.
2.
Methods
We performed a cross-sectional study using system access log files to measure patterns of query-based exchange usage. The Institutional Review Boards of Weill Cornell Medical College and academic research institutions with which community organizations had affiliations approved this study.
2.1.
Setting
This study examined query-based exchange usage in three separate communities in New York State—here called Community A, Community B, and Community C—funded through the Healthcare Efficiency and Affordability Law for New Yorkers Capital Grant Program (HEAL NY). HEAL NY, which started in 2006 and represents more than an $800 million investment of public-private funds in EHRs and HIE, aims to develop a health information network for New York State by linking together community-based RHIOs that adhere to common standards and policies. RHIOs’ roles included convening and governing community stakeholders, promoting collaboration and data sharing, and implementing technology for HIE. As of 2012, twelve non-profit RHIOs provided HIE across New York State in compliance with state requirements using a variety of commercial products. The three non-profit RHIOs investigated in this study implemented the same commercial HIE system, which we describe in Section 2.3. As shown in Table 1, the RHIOs studied provided HIE for communities of about one million adults. Across the three communities, the numbers of encounter types and physicians varied. Of note, patients in Community A and Community B rarely received care outside the catchment area of their local RHIO, while patients in Community C may have sought care either within the catchment area of the local RHIO or in neighboring areas served by other RHIOs. Whereas providers in Community A and Community B practiced in settings ranging from large health systems to solo practices, providers in Community C practiced almost exclusively in large health systems. Years of RHIO operations differed across the communities. Notably, in 2009, Community C replaced a previous HIE product with its current system, which delayed the start of HIE data collection from community healthcare institutions.
2.2.
Patient consent
In New York State, patient data can be made available to a RHIO without consent, but patients must provide affirmative consent for that data to be accessed. Patient data can be accessed without affirmative consent in emergency situations according to a “break the glass” policy, and patients can also opt out of this type of access when choosing to not participate in HIE. Within the framework of this guidance,
812
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
Table 1 – Characteristics of communities engaged in HIE. Community A
Community B
Community C
Patient population, n (%) Female Male Age of patient population 18–24, n (%) 25–44 45–64 65 and older
1,207,007 641,222 (53) 565,784 (47)
968,600 506,820 (52) 461,780 (48)
982,800 555,282 (57) 427,518 (43)
143,747 (12) 451,498 (37) 364,554 (30) 247,208 (20)
121,131 (13) 375,427 (39) 298,450 (31) 173,591 (18)
148,400 (15) 429,800 (44) 263,200 (27) 141,400 (14)
Annual inpatient admissions Annual outpatient visits Annual emergency department visits Physicians
203,000 4,336,000 683,000 3600
157,000 3,523,000 264,000 3000
200,000 4,500,000 600,000 4000
Year RHIO founded Year RHIO data feeds started
2004 2007
2006 2007
2006 2010
Population and annual patient volume estimates are stated for 2010. Percentages may not sum to 100% due to rounding. HIE was available for adults and children in each community, but this study investigated HIE usage for adults only; adults represented 70–75% of the total population in each community. RHIO: regional health information organization.
the communities operationalized their consent procedures somewhat differently. In Community A, patients were invited to provide a single affirmative consent to all covered entities belonging to the RHIO, and were approached to provide additional consent when a new covered entity joined the RHIO. In Community B and Community C, patients were approached to provide multiple affirmative consents to individual covered entities to access their data. In the communities studied, the percentage of patients who provided affirmative consent for HIE participation ranged from 90% to 97%. Workflows for updating patient consent status differed across communities. In Community A, users updated consent status using a patient credentialing application linked to but separate from the HIE system. Users then needed to launch the HIE system to access a patient’s clinical data. In Community B, users updated consent status by logging into the HIE system and attempting to access a patient’s record. If the system detected that a patient had not provided consent to a user’s practice site, a popup window appeared requiring the user to specify the patient’s consent status. If the user indicated that the patient provided affirmative consent to the practice site, the system then displayed the patient’s clinical data. Users in Community B also had the option of updating consent using a patient credentialing system linked to but separate from the HIE system as well as an EHR form that transmitted consent status to the HIE. In Community C, users updated consent status using an EHR form only. Additionally, toward the beginning of operations in Community A and Community B, HIE administrators updated consent manually by processing paper forms received from practice sites.
2.3.
System description
The three communities implemented the same commercial HIE system, Axolotl Virtual Health Record (now known as Optum Virtual Health Record) [19], a secure stand-alone web portal that enabled registered users to query data sources from across the community for patients in their care. The system employed a federated architecture with a master patient
index, record locator service, and user directory to transmit patient data. Healthcare institutions in each community made their data available for access via HIE. Users in practice sites were able to access patient data via query-based exchange regardless of being affiliated with a particular healthcare institution that provided data. The system enabled users to look-up patient data from but not send patient data to the HIE. Fig. 1 depicts system workflow. To access the system, users in Community A logged in using a username, password, and secure token, while users in Community B and Community C logged in using a username and password. Users then searched for patients using a combination of demographic and administrative identifiers including last name, first name, date of birth, gender, medical record number, and insurance policy number. After a user selected a patient from search results, the system checked the patient’s consent status and displayed data accordingly. For a patient who had not provided affirmative consent to the user’s practice site, the system displayed a message stating it would not show patient data unless the patient provided affirmative consent. If at this time the patient decided to participate, a user would update the patient’s status to indicate affirmative consent as described in the previous section. For a patient who provided affirmative consent to the user’s practice site, the system displayed a default landing page—configured as a recent result summary report in Community A and Community B and a patient information report in Community C. A recent result summary presented links to the five most recent patient data reports for each data report category (e.g. laboratory, radiology), while a patient information report displayed patient contact information and the names of entities to which a patient had affirmatively consented. The system’s user interface presented tabs at the top of the screen corresponding to categories of patient data reports (e.g. laboratory, radiology). Upon clicking a category tab, a user was presented with a list of links to individual data reports of that category containing each report’s name, date, and responsible physician. After clicking a link from the list, a user accessed the details of an individual data
813
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
report. HIE users demonstrated two forms of patient data access—detailed access, or access of individual data reports for a patient, and default only access, or access of only the default landing page for a patient. Default only access may have provided users with useful clinical information from listings of available reports for a patient (e.g. indication that all laboratory results were outdated) without needing to access individual data reports.
2.4.
2.5.
We analyzed system usage patterns with respect to practice sites, users, patients, and data to determine where, by whom, for whom, and for what HIE system access occurred. Because Community A and Community B had 27 months of system availability and Community C had 9 months of system availability, we conducted analysis of the first 9 months of usage for all communities and the first 27 months of usage for only Community A and Community B. For practice sites, we determined the number and types of settings (e.g. inpatient, outpatient, emergency) from which system access occurred as well as the degree of access in each practice site as measured by sessions, or instances in which a user accessed a patient’s data on a given day. This measure enabled counting accesses on different days by a single user as well as multiple users accessing data for a single patient across one or more days. For users, we determined the number, roles, and types of practice sites of those who accessed the system. For patients, we determined the number of affirmatively consented patients, number of affirmatively consented patients whose data were accessed, the length of time from initial consent to first access of data for patients, and the number of patients whose data were accessed from multiple practice sites. We also differentiated between detailed access and default only access of data for patients. For data, we determined the number and types of data reports accessed by user roles, number and types of data reports accessed over time, and the number and types of data reports accessed by practice sites.
Data collection
For all affirmatively consented patients who were at least 18 years old at the time of consent, we obtained de-identified system access log data for the first 27 months of system availability in Community A and Community B (January 1, 2009 through March 31, 2011) and first nine months of system availability in Community C (September 1, 2010 through May 31, 2011). System access log data specified the patient, user, practice site, date, time, and category of each data report accessed. From the system access log data, we also obtained patient characteristics including gender, initial consent date, and age at initial consent. Additionally, using each community’s customer relationship management records, we obtained user characteristics—professional role (e.g. provider, nurse, staff, other clinician), primary practice site, and physician specialty—and practice site designation (e.g. inpatient, outpatient, emergency). We defined staff as non-clinical workers (e.g. medical receptionists), nurses as registered nurses, providers as clinicians with prescribing privileges (e.g. physicians including medical doctors and doctors of osteopathic medicine as well as physician extenders including nurse practitioners and physician assistants), and other clinicians as allied health professionals (e.g. pharmacists, physical therapists, registered dieticians).
Y Update in credenaling system
Data analysis
2.6.
Statistical analysis
We determined descriptive statistics including frequencies and percentages for categorical variables, mean and
Community A
Community B
Community C
Login to HIE
Login to HIE
Login to HIE
Search for paent
Search for paent
Search for paent Y
N
Check consent Y
Update in HIE popup window Y
N
Check consent Y
Display recent summary
Display recent summary
Display detailed report
Display detailed report
N Update in EHR
Check consent Y Display paent informaon Display detailed report
Fig. 1 – HIE system workflow across the three communities. The system displayed a default landing page to users for each patient who provided affirmative consent. Users then had the option of accessing individual detailed reports (e.g. laboratory, radiology, transcribed, admission-discharge-transfer). HIE data sources varied across communities and included hospitals, freestanding laboratories, freestanding radiology centers, homecare services, community health centers, and long-term care facilities.
814
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
Table 2 – Characteristics of practice sites from which system access and sessions occurred. Months 1–9
Months 1–27
Community A
Community B
Community C
Community A
Community B
Sites where system access occurred, N Outpatient, n (%) Inpatient Emergency Other
59 57 (97%) 1 (2%) 1 (2%) 0 (0%)
80 62 (78%) 4 (5%) 2 (3%) 12 (15%)
28 16 (57%) 8 (29%) 4 (14%) –
200 190 (95%) 6 (3%) 2 (1%) 2 (1%)
156 122 (78%) 10 (6%) 7 (5%) 17 (11%)
Sites where system sessions occurred, N Outpatient, n (%) Inpatient Emergency Other
1964 1903 (97%) 17 (1%) 44 (2%) –
27,675 22,611 (82%) 2439 (9%) 1645 (6%) 980 (4%)
1965 376 (19%) 1531 (78%) 58 (3%) –
15,272 14,465 (95%) 77 (1%) 304 (2%) 426 (3%)
319,089 259,135 (81%) 22,493 (7%) 35,201 (11%) 2260 (1%)
System access occurred from a practice site if any user from the practice site accessed the system. System sessions, or instances in which a user accessed a patient’s data on a given day from a practice site, accounted for the frequency of access from a practice site. Other practice sites included public health settings, long-term care, care and disease management, and home care settings.
standard deviation (SD) for normally distributed continuous data, median and interquartile ranges (IQR) for non-normally distributed continuous data, and minimum and maximum values for non-normally distributed continuous data that were highly skewed. We used SAS 9.2 and Microsoft Excel 2007 to perform calculations and manage data.
3.
Results
3.1.
Practice sites
Of practice sites registered to use the system in each community, access occurred in 18% (n = 59) of sites in Community A, 30% (n = 80) in Community B, and 82% (n = 28) in Community C during the first nine months of system availability (Table 2). After 27 months, access had occurred from 60% (n = 200) of sites in Community A and 59% (n = 156) in Community B. In each community, the majority of practice sites from which system access occurred were outpatient. In Community A and Community B, the majority of system sessions occurred from outpatient sites, while in Community C inpatient sites accounted for the majority of sessions.
3.2.
Users
The number, roles, and primary practice site of users who accessed the system varied across the three communities (Table 3). Community A registered the greatest number of users to access the system (n = 3658) followed by Community B (n = 3461) and Community C (n = 118). After 27 months, the number of users accessing the system in Community A and Community B had more than tripled. Over time more than half of users in Community A and Community B were nurses and staff, while after nine months more than two thirds of users in Community C were providers (e.g. physicians and physician extenders). In all three communities, the majority of users practiced primarily in outpatient practice sites. Of note, the proportion of users primarily practicing in inpatient sites was substantially greater in Community C after nine months compared to Community A and Community B over time.
3.3.
Patients
As shown in Table 4, during the first nine months of system availability, the number of patients who affirmatively consented to HIE participation and the proportion of patients whose data were accessed varied widely across the three communities. After 27 months in Community A and Community B, the number of affirmatively consented patients had increased more than six-fold while the proportion of patients whose data were accessed remained relatively similar. In each of the three communities, patients whose data were accessed were older than those whose data were not accessed. Additionally, patients whose data were accessed tended to be older than the overall population (Table 1). Of patients whose data were accessed after nine months, 60% (n = 936) in Community A, 7% (n = 1724) in Community B, and 27% (n = 296) in Community C had detailed access rather than default only access. After 27 months, the proportion of patients with detailed access rather than default only access remained relatively similar in Community A (53%, n = 5936) and had almost doubled in Community B (13%, n = 28,036). During the first nine months of system availability, the majority of patients whose data were accessed had their data accessed for the first time on the same day as consent in Community A (77%, n = 1096) and Community B (96%, n = 23,867). However, in Community C 17% (n = 196) of patients had their data accessed for the first time on the same day as consent, and the median (IQR) number of days from consent to first access was 17 (1–84). After 27 months, first time patient data access occurred on the same day as consent for 55% (n = 6003) of patients in Community A and 86% (n = 183,771) of patients in Community B. Over time the majority of patients in Community A and Community B had their data accessed in an outpatient setting, while after nine months the majority of patients in Community C had their data accessed in an inpatient setting (Table 5). During the first nine months, the number of patients whose data were accessed from two or more practice sites was 33 (0.1%) in Community A, 843 (1.8%) in Community B, and 12 (0.01%) in Community C. After 27 months, the number of patients whose data were accessed from two
815
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
Table 3 – Characteristics of users who accessed patient data. Months 1–9
Users by role, n Nurses and staff, n (%) Nurses, n (%) Staff, n (%) Providers, n (%) Physicians Internal medicine Emergency medicine Family medicine Pediatrics Obstetrics and gynecology Other specialty Oncology/hematology Radiology Physician extenders Other clinicians, n (%) Users by practice site type Outpatient, n (%) Inpatient Emergency Other
Months 1–27
Community A
Community B
Community C
Community A
Community B
155 110 (71) NA NA 39 (25) 32 (21) 5 3 10 4 1 9 0 0 7 (4) 6 (4)
399 240 (60) 35 (9) 205 (51) 104 (26) 76 (19) 15 3 16 1 3 24 4 10 28 (7) 55 (14)
88 22 (25) 20 (23) 2 (2) 60 (68) 54 (61) 20 14 8 7 4 1 0 0 6 (7) 6 (7)
672 487 (72) NA NA 134 (20) 105 (16) 13 11 27 11 6 32 5 0 29 (4) 51 (8)
1318 717 (54) 134 (10) 583 (44) 398 (30) 295 (22) 77 35 53 5 9 70 17 29 103 (8) 202 (15)
147 (95) 2 (1) 4 (3) 2 (1)
334 (84) 21 (5) 19 (5) 25 (6)
43 (49) 29 (33) 16 (18) –
637 (95) 8 (1) 17 (3) 10 (1)
1131 (86) 44 (3) 91 (7) 51 (4)
Available data sources in Community A did not distinguish between nurse and staff users. Physician extenders included nurse practitioners and physician assistants. Other practice sites included public health, long-term care, care and disease management, and home care settings.
or more practice sites increased to 315 (0.1%) in Community A and 41,210 (11.6%) in Community B.
3.4.
Data
As depicted in Fig. 2, data report access increased over time in all three communities. After nine months, the number of data reports accessed ranged from 5111 in Community A and 31,065 in Community B to 7690 in Community C. After 27 months, the number of data reports accessed increased more than six-fold to 32,784 in Community A and more than ten-fold to 382,920 in Community B. In Community B and Community C, users most frequently accessed only the default report, while in Community A users most frequently accessed a detailed report. Of detailed data reports accessed in Community A and Community B, users accessed laboratory reports most frequently followed by radiology, transcribed (e.g. discharge summary, history and physical, consultation), and admission-dischargetransfer (ADT) reports, which were structured documents generated by individual hospitals’ patient tracking systems. As shown in Table 6, in Community B staff accessed data reports more frequently than other user roles, and in Community A, where data sources did not distinguish between nurse and staff users, nurses and staff accessed data reports more frequently than other user roles. However, in Community C, providers accessed detailed data reports most frequently while nurses accessed the majority of default reports.
4.
Discussion
This study is the first to our knowledge to quantify patterns of usage of a commercial query-based health information
exchange system implemented on a large, multi-community scale. Across the three communities, between 59% and 82% of practice sites registered to use the system actually accessed the system. In two communities, outpatient staff primarily accessed patient data, while in one community inpatient providers were the main users. Consistent with previous studies [18,20], patients whose data were accessed tended to be older than those whose data were not accessed, as older patients may have more complex conditions that motivate users to access HIE [17,18,20]. Users primarily accessed summaries of recent results as well as detailed reports of laboratory and radiology studies. Over time, system usage increased in each community but with varying proportions of patients whose data were accessed. While the proportions of patients whose data were accessed in Community C (1%) and Community A (5%) are comparable to patient-level access rates reported in the literature [16], Community B’s exceptionally high 53–60% may be due to consent workflow procedures. The predominant consent workflow in Community B involved users logging into the HIE, searching for a patient, adjusting consent status in a popup window, and then landing on the recent result summary by default. Of patients who provided affirmative consent in Community B during the first nine months of system availability, 96% had their data accessed for the first time on the same day as consent, suggesting that users sought access to HIE data almost immediately upon being granted permission. In contrast, during the same time period, users in Community A updated consent through a separate credentialing application and accessed patient data on the same day as consent for 77% of patients while users in Community C updated consent through an EHR form and accessed data on the same day as consent for 18% of patients. Furthermore, after 27 months,
816
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
Table 4 – Characteristics of patients who affirmatively consented to have their data accessed.
Months 1–9 Community A n (%) Female Male Age at consent 18–24, n (%) 25–44 45–64 65 and older Community B n (%) Female Male Age at consent 18–24, n (%) 25–44 45–64 65 and older Community C n (%) Female Male Age at consent 18–24, n (%) 25–44 45–64 65 and older Months 1–27 Community A n (%) Female Male Age at consent 18–24, n (%) 25–44 45–64 65 and older Community B n (%) Female Male Age at consent 18–24, n (%) 25–44 45–64 65 and older
Unique patients affirmatively consented
Unique patients whose data were accessed
Unique patients whose data were not accessed
31,356 17,245 (55) 14,099 (45)
1571 (5) 963 (61) 608 (39)
29,785 (95) 16,282 (55) 13,491 (45)
3014 (10) 8045 (26) 11,439 (36) 8853 (28)
61 (4) 284 (18) 582 (37) 643 (41)
2953 (10) 7761 (26) 10,857 (36) 8210 (28)
46,948 35,223 (75) 11,723 (25)
24,765 (53) 17,500 (71) 7263 (29)
22,183 (47) 17,723 (80) 4460 (20)
2569 (5) 10,960 (23) 19,653 (42) 13,766 (29)
1672 (7) 5964 (24) 9326 (38) 7803 (32)
897 (4) 4996 (23) 10,327 (47) 5963 (27)
129,692 84,023 (65) 45,665(35)
1107 (1) 677 (61) 430 (39)
128,585 (99) 83,346 (65) 45,235 (35)
17,186 (13) 44,747 (35) 43,313 (33) 24,446 (19)
17 (2) 69 (6) 428 (39) 593 (54)
17,169 (13) 44,678 (35) 42,885 (33) 23,853 (19)
219,647 141,725 (65) 77,890 (35)
11,263 (5) 6697 (59) 4566 (41)
208,384 (95) 135,028 (65) 73,324 (35)
16,688 (8) 55,278 (25) 89,285 (41) 58,389 (27)
424 (4) 2001 (18) 4418 (39) 4419 (39)
16,264 (8) 53,277 (26) 84,867 (41) 53,970 (26)
354,161 238,116 (67) 115,919 (33)
212,586 (60) 142,834 (67) 69,648 (33)
141,575 (40) 95,282 (67) 46,271 (33)
27,201 (8) 88,085 (25) 148,006 (42) 90,866 (26)
14,070 (7) 49,645 (23) 89,735 (42) 59,136 (28)
13,131 (9) 38,440 (27) 58,271 (41) 31,730 (22)
Patients whose data were accessed tended to be older than those whose data were not accessed.
Table 5 – Unique patients for whom data were accessed stratified by practice site type. Months 1–9
Unique patients, N Outpatient, n (%) Inpatient Emergency Other
Months 1–27
Community A
Community B
Community C
Community A
Community B
1571 1515 (96%) 14 (1%) 44 (3%) 0 (0%)
24,765 20,677 (82%) 2235 (9%) 1466 (6%) 918 (4%)
1107 278 (25%) 778 (70%) 57 (5%) –
11,263 10,697 (95%) 64 (1%) 277 (2%) 271 (2%)
212,586 188,418 (83%) 18,860 (8%) 19,020 (8%) 1982 (1%)
Patients can be counted more than once if accessed in more than one type of practice site. As a result, sum of n’s may exceed total N.
817
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
Table 6 – Data reports accessed by users. Months 1–9
Laboratory reports, N Staff Nurses Providers Other clinicians Radiology reports Staff Nurses Providers Other clinicians Transcribed reports Nurses Providers Other clinicians ADT Staff Nurses Providers Other clinicians Recent result summary Staff Nurses Providers Other clinicians Patient information Staff Nurses Providers Other clinicians
Months 1–27
Community A
Community B
Community C
Community A
Community B
2491 1454 (58%) – 888 (36%) 149 (6%) 1152 872 (76%) – 255 (22%) 25 (2%) 221 120 (54%) – 96 (43%) 5 (2%) 68 48 (71%) – 20 (29%) – 1179* 730 (62%) – 387 (33%) 62 (5%) NA NA NA NA NA
3221 1785 (55%) 136 (4%) 918 (29%) 382 (12%) 1521 896 (59%) 41 (3%) 362 (24%) 222 (15%) 261 177 (68%) 16 (6%) 54 (21%) 14 (5%) NA NA NA NA NA 25,655* 23,330 (91%) 81 (<1%) 329 (1%) 1915 (7%) NA NA NA NA NA
300 – 6 (2%) 204 (68%) 90 (30%) 196 – 4 (2%) 190 (97%) 2 (1%) 44 0 (0%) 2 (5%) 40 (91%) 2 (5%) 278 1 (<1%) 58 (21%) 131 (47%) 88 (32%) 1739 0 (0%) 221 (13%) 1030 (59%) 488 (28%) 5133* 130 (3%) 4571 (89%) 345 (7%) 87 (2%)
14,663 11,076 (76%) – 2813 (19%) 774 (5%) 6168 4858 (79%) – 929 (15%) 381 (6%) 2144 1656 (77%) – 372 (17%) 116 (5%) 68 48 (71%) – 20 (29%) – 9741* 7090 (73%) – 1284 (13%) 1367 (14%) NA NA NA NA NA
58,713 30,223 (51%) 5388 (9%) 14,902 (25%) 8199 (14%) 29,846 18,613 (62%) 1954 (7%) 5938 (20%) 3341 (11%) 7210 3053 (42%) 834 (12%) 2426 (34%) 895 (12%) 1354 528 (39%) 164 (12%) 385 (28%) 277 (20%) 281,124* 267,793 (95%) 2627 (1%) 2391 (1%) 8313 (3%) NA NA NA NA NA
Community A data sources did not distinguish between nurses and staff. As denoted by *, in Community A and Community B the recent result summary was the default report whereas in Community C the patient information report was the default report. Transcribed reports included discharge summary, history and physical, consultation, etc. ADT: admission-discharge-transfer.
the proportion of patients with detailed access—beyond the default landing page—in Community B (7.9%) was substantially greater than in Community A (2.7%). In Community B, the practice of logging into the HIE to update consent may have become engrained in clinical culture and created a habit for users to access HIE data in the course of care. Substantial increases in the numbers of reports accessed after nine months versus after 27 months in Community B, especially when compared to Community A, support this hypothesis. Updating patient consent status directly through the querybased HIE portal may be inconvenient and time consuming with respect to other clinical tasks, but the tradeoff of workflow efficiency and regular HIE system usage may warrant consideration in other HIE portal implementations requiring affirmative patient consent. Future investigations should use qualitative methods to understand HIE usage with respect to consent workflow, system usability, and utility of clinical data for different practice types and communities. Across the three communities, clinicians and staff accessed a variety of data reports using the system, but they accessed only the default landing page with greatest frequency. This is consistent with a study of query-based HIE portal usage for medically indigent patients where users most frequently accessed only the default, or “gateway,” screen listing a patient’s most recent encounters [21]. Together these
findings emphasize the importance of data displayed on the default landing page of query-based portals, as the data may provide clinical utility to users without needing to display a detailed report, and consideration of levels of access, such as default only and detailed, when examining usage of querybased HIE portals [21]. Nurses and staff often prepare HIE data for physicians to use by generating printouts [22], and researchers have suggested that HIE designs incorporate user- and role-specific display of data to accommodate differing information needs [21,22]. Additionally, another study noted that regular HIE usage by nurses and staff was associated with increased overall levels of HIE access in practice sites [17]. Findings from our investigation support these observations, as staff in outpatient settings predominantly accessed the HIE system in Community A and Community B, the communities with the highest degree of access, while providers in inpatient settings mostly accessed the HIE in Community C. Although outpatient and inpatient clinical workflows diverge and may explain the difference in system usage between Community C and the two other communities, the majority of users in Community C (49%) practiced primarily in an outpatient setting, suggesting an opportunity to increase overall system usage through increased nurse and staff usage in outpatient settings. Additionally, further study is required to understand inpatient
818
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
Fig. 2 – Data reports accessed over time in Communities A, B, and C. Access of data reports through the system increased over time in the three communities studied. Default data reports accessed are presented on the secondary axis due to scaling issues resulting from their frequent access in Community B. Users accessed 1354 ADT reports in Community B (not shown). “Summary Only” and “Patient Info Only” refer to default only access as described in Section 2.
usage of HIE systems and whether nurse and staff usage in inpatient areas is related to overall levels of HIE access. To increase system usage, researchers and practitioners should engage all members of the care team and explore role- and practice setting-specific presentation of HIE data. From the current investigation and previous studies of query-based exchange, common observations—including older patients having HIE data accessed more frequently than younger patients [18,20], nurse and staff system usage being related to overall increased HIE usage [17,21], and importance of default patient data displays in HIE [17,21]—may
indicate emergence of theory for query-based exchange that can increase usefulness of HIE systems. Novel observations from this study, such as the potential relationship between opt-in consent workflow and increased system usage, may also contribute to nascent query-based exchange theory. Researchers have documented a dearth of theory in biomedical informatics research [23,24], and additional study in query-based HIE may help fill this gap in the literature. Our investigation may be overly conservative in measuring HIE usage. We examined access of individual static data reports, such as laboratory results and transcribed notes, and excluded aggregated dynamic data views, including just-intime insurance eligibility verification and medication history lookup, that were in early stages of development at the time of the study. Because users must first access a default landing page prior to accessing dynamic transaction features, our measures of access captured any use of these features by proxy. Furthermore, we examined usage of one web portal system for query-based HIE, and particularities of the system and its implementation, including but not limited to user interface design, workflow integration, organizational commitment, and training, may have affected user behavior [25]. Finally, we investigated usage of HIE in communities funded by New York State’s unique health information technology program, which may not exist in other settings. However, in contrast to previous approaches, the commercial query-based exchange system implementations studied were facilitated by community organizations without involvement of academic institutions with substantial informatics resources [17,26] and targeted all patients community-wide rather than only the medically indigent [27], which may enable findings to generalize to other settings. Patient privacy concerns [28,29], healthcare system structure [30,31], and patient population characteristics [32] influence HIE infrastructure and implementation decisions, and findings from this study may only be useful to settings where opt-in patient consent, communitywide data sharing, and query-based exchange match social norms. This investigation illustrates usage of a commercial querybased exchange system across multiple communities and care settings. Consistent with other studies, increased HIE usage by staff members appeared to be related to higher levels of system usage by all clinicians. Additionally, users most frequently accessed only the system’s default landing page as opposed to detailed clinical reports. Workflow for updating patient consent may have contributed to higher cultural acceptance of HIE system usage in one community, which may be useful for increasing usage of HIE portal systems in other settings requiring opt-in consent. User role, practice site type, and consent workflow may affect patterns of query-based HIE system usage in the communities studied and elsewhere.
Authors’ contributions TRC conceptualized the manuscript, guided data collection, authored the manuscript, and incorporated co-authors’ feedback into the manuscript. AME performed statistical calculations and provided critical manuscript feedback. SBJ
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
819
references Summary points What was already known • Usage of query-based health information exchange (HIE) portal systems has been associated with healthcare cost reductions. • Studies of query-based HIE portal system usage have addressed only emergency department and safety net settings in two communities. Additionally, the two communities implemented different query-based HIE systems, limiting the ability to compare system usage patterns across communities. What this study added to our knowledge • This study is among the first to quantify patterns of usage of a commercial query-based health information exchange system implemented on a large, multi-community scale. • Consent workflow requiring users to login to a querybased HIE portal system, as opposed to a separate system, may have contributed to higher overall HIE usage in one community, where 60% of patients had their data accessed. • Across the three communities studied, users most frequently accessed only default patient data, emphasizing the importance of data initially presented to users in HIE systems, followed by detailed laboratory reports, indicating a need for more research of HIE system usage in multiple communities and settings.
provided critical manuscript feedback. RK conceptualized the manuscript and provided critical manuscript feedback.
Conflicts of interest The authors declare that they have no conflicts of interest in the research.
Acknowledgments The study was conducted as part of the Health Information Technology Evaluation Collaborative (HITEC), the multidisciplinary academic consortium charged with evaluating the effects of New York State’s investment in health information technology. Authors of the manuscript include members (TRC and AME) and the director (RK) of HITEC. This study was conducted with funding from the New York State Department of Health (contract number C023699). The authors would like to thank the community organizations for participating in the study; Lisa M, Kern, M.D., M.P.H. for conceptual feedback; Kenneth S. Boockvar, M.D., M.S. for data collection assistance; Renny V. Thomas, M.P.H. for data collection assistance; and Jessica S. Ancker, M.P.H., Ph.D. for manuscript assistance.
[1] J.R. Vest, L.D. Gamm, Health information exchange: persistent challenges and new strategies, J. Am. Med. Inform. Assoc.: JAMIA 17 (2010) 288–294. [2] T. Mäenpää, T. Suominen, P. Asikainen, et al., The outcomes of regional healthcare information systems in health care: a review of the research literature, Int. J. Med. Inf. 78 (2009) 757–771. [3] L. Lenert, D. Sundwall, M.E. Lenert, Shifts in the architecture of the Nationwide Health Information Network, J Am Med Inform Assoc: JAMIA 19 (2012) 98–502. [4] C. Williams, F. Mostashari, K. Mertz, et al., From the Office of the National Coordinator: the strategy for advancing the exchange of health information, Health Aff. (Project Hope) 31 (2012) 527–536. [5] G.J. Kuperman, Health-information exchange: why are we doing it, and what are we doing? J. Am. Med. Inform. Assoc.: JAMIA 18 (2011) 678–682. [6] O. Heinze, M. Birkle, L. Köster, et al., Architecture of a consent management suite and integration into IHE-based Regional Health Information Networks, BMC Med. Inform. Decision Making 11 (2011) 58. [7] T. Schabetsberger, F. Wozak, B. Katt, et al., Implementation of a secure and interoperable generic e-Health infrastructure for shared electronic health records based on IHE integration profiles, Stud. Health Technol. Inform. 160 (2010) 889–893. [8] J. Zhang, K. Zhang, Y. Yang, et al., Grid-based implementation of XDS-I as part of image-enabled EHR for regional healthcare in Shanghai, Int. J. Comput. Assisted Radiol. Surg. 6 (2011) 273–284. [9] S. Liu, B. Zhou, G. Xie, et al., Beyond regional health information exchange in China: a practical and industrial-strength approach, AMIA. Annual Symposium Proceedings/AMIA Symposium AMIA Symposium 2011 (2011) 824–833. [10] S. Devore, R.W. Champion, Driving population health through accountable care organizations, Health Aff. (Millwood) 30 (2011) 41–50. [11] A. Tzeel, V. Lawnicki, K. Pemble, The business case for payer support of a community-based health information exchange: a Humana pilot evaluating its effectiveness in cost control for plan, Am. Health Drug Benefits 4 (2011) 207–216. [12] M.E. Frisse, K.B. Johnson, H. Nian, et al., The financial impact of health information exchange on emergency department care, J. Am. Med. Inform. Assoc.: JAMIA 19 (2012) 328–333. [13] J.S. Ancker, L.M. Kern, E. Abramson, et al., The Triangle Model for evaluating the effect of health information technology on healthcare quality and safety, J. Am. Med. Inform. Assoc.: JAMIA 19 (2012) 61–65. [14] C.P. Friedman, “Smallball” evaluation: a prescription for studying community-based information interventions, J. Med. Library Assoc.: JMLA 93 (2005) S43–S48. [15] K.B. Johnson, C. Gadd, Playing smallball: approaches to evaluating pilot health information exchange systems, J. Biomed. Inform. 40 (2007) S21–S26. [16] J.R. Vest, J. Jasperson, What should we measure? Conceptualizing usage in health information exchange, J. Am. Med. Inform. Assoc.: JAMIA 17 (2010) 302–307. [17] K.B. Johnson, K.M. Unertl, Q. Chen, et al., Health information exchange usage in emergency departments and clinics: the who, what, and why, J. Am. Med. Inform. Assoc.: JAMIA 18 (2011) 690–697. [18] J.R. Vest, L.D. Gamm, R.L. Ohsfeldt, et al., Factors associated with health information exchange system usage in a
820
[19]
[20] [21]
[22]
[23]
[24]
[25]
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
safety-net ambulatory care clinic setting, J. Med. Syst. 36 (2012) 2455–2461. Virtual Health Record – OptumInsight.com. http://www. optuminsight.com/physicians/solutions/health-information -exchange/optum-virtual-health-record/overview/ (accessed 20.04.13). J.R. Vest, Health information exchange and healthcare utilization, J. Med. Syst. 33 (2009) 223–231. J.R. Vest, Jasperson ‘jon Sean. How are health professionals using health information exchange systems? Measuring usage for evaluation and system improvement, J. Med. Syst. 36 (2012) 3195–3204. K.M. Unertl, K.B. Johnson, N.M. Lorenzi, Health information exchange technology on the front lines of healthcare: workflow factors and patterns of use, J. Am. Med. Inform. Assoc.: JAMIA 19 (2012) 392–400. B. Kaplan, Evaluating informatics applications—some alternative approaches: theory, social interactionism, and call for methodological pluralism, Int. J. Med. Inf. 64 (2001) 39–56. M. Chiasson, M. Reddy, B. Kaplan, et al., Expanding multi-disciplinary approaches to healthcare information technologies: what does information systems offer medical informatics? Int. J. Med. Inf. 76 (Suppl. 1) (2007) S89–S97. C.P. Friedman, J.C. Wyatt, Evaluation Methods in Biomedical Informatics, Springer, 2010.
[26] K.B. Johnson, C.S. Gadd, D. Aronsky, et al., The MidSouth eHealth Alliance: use and impact in the first year, AMIA. Annual Symposium proceedings/AMIA Symposium AMIA Symposium (2008) 333–337. [27] J.R. Vest, H. Zhao, J. Jasperson, et al., Factors motivating and affecting health information exchange usage, J. Am. Med. Inform. Assoc.: JAMIA 18 (2011) 143–149. [28] A. Geissbuhler, S. Spahni, A. Assimacopoulos, et al., Design of a patient-centered, multi-institutional healthcare information network using peer-to-peer communication in a highly distributed architecture, Stud. Health Technol. Inform. 107 (2004) 1048–1052. [29] A. Geissbuhler, Lessons learned implementing a regional health information exchange in Geneva as a pilot for the Swiss national eHealth strategy, Int. J. Med. Inf. (2013), doi:10.1016/j.ijmedinf.2012.11.002. [30] H. Park, S.-I. Lee, Y. Kim, et al., Patients’ perceptions of a health information exchange: a pilot program in South Korea, Int. J. Med. Inf. 82 (2013) 98–107. [31] T. Bowden, E. Coiera, Comparing New Zealand’s “Middle Out” health information technology strategy with other OECD nations, Int. J. Med. Inf. (2012), doi:10.1016/j.ijmedinf.2012.12.002. [32] W.T. Steward, K.A. Koester, S.P. Collins, et al., The essential role of reconfiguration capabilities in the implementation of HIV-related health information exchanges, Int. J. Med. Inf. 81 (2012) e10–e20.
journal homepage: www.ijmijournal.com
Health information exchange system usage patterns in three communities: Practice sites, users, patients, and data Thomas R. Campion Jr. a,b,c,d,e,∗ , Alison M. Edwards a,b,d,e , Stephen B. Johnson a,b,d , Rainu Kaushal a,b,c,d,e,f,g , with the HITEC investigators a
Division of Quality and Medical Informatics, Weill Cornell Medical College, New York, NY, United States Department of Public Health, Weill Cornell Medical College, New York, NY, United States c Department of Pediatrics, Weill Cornell Medical College, New York, NY, United States d Center for Healthcare Informatics and Policy, Weill Cornell Medical College, New York, NY, United States e Health Information Technology Evaluation Collaborative (HITEC), New York, NY, United States f Department of Medicine, Weill Cornell Medical College, New York, NY, United States g Komansky Center for Children’s Health at NewYork-Presbyterian Hospital, New York, NY, United States b
a r t i c l e
i n f o
a b s t r a c t
Article history:
Objectives: Public and private organizations are implementing systems for query-based
Received 22 October 2012
health information exchange (HIE), the electronic aggregation of patient data from multi-
Received in revised form
ple institutions. However, existing studies of query-based HIE system usage have addressed
30 April 2013
a limited number of settings. Our goal was to quantify the breadth and depth of usage
Accepted 2 May 2013
of a query-based HIE system implemented across multiple communities with diverse care settings and patient populations.
Keywords:
Methods: We performed a cross-sectional study in three communities in New York State
Health information technology
using system access log files from January 2009 to May 2011 to measure usage patterns of a
Electronic health records
query-based HIE web portal system with respect to practice sites, users, patients, and data.
Evaluation studies
Results: System access occurred from 60% (n = 200) of practice sites registered to use the
Community
system in Community A, 59% (n = 156) in Community B, and 82% (n = 28) in Community
Health information exchange
C. In Communities A and B, users were primarily non-clinical staff in outpatient settings, while in Community C inpatient physicians were the main users. Across communities, proportions of patients whose data were accessed varied with 5% (n = 11,263) in Community A, 60% (n = 212,586) in Community B, and 1% (n = 1107) in Community C. In Community B, users updated patient consent through the HIE portal, whereas in the other communities, users updated patient consent through a separate system. Across communities, users most frequently accessed only patient summary data displayed by default followed by detailed laboratory and radiology data. Conclusions: This study is among the first to illustrate large-scale usage of a query-based HIE system implemented across multiple communities. Patient summary data displayed by default may be an important feature of query-based HIE systems. User role, practice site type, and patient consent workflow may affect patterns of query-based HIE web portal system usage in the communities studied and elsewhere. © 2013 Elsevier Ireland Ltd. All rights reserved.
∗
Corresponding author at: 402 East 61st Street, DV-308, New York, NY 10065, United States. Tel.: +1 646 962 2345; fax: +1 646 962 0105. E-mail address: [email protected] (T.R. Campion Jr.). 1386-5056/$ – see front matter © 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ijmedinf.2013.05.001
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
1.
Introduction
Electronic transfer of patient data across healthcare organizations has the potential to improve patient care coordination, facilitate public health efforts, and reduce healthcare costs [1,2]. Currently public and private organizations are implementing multiple forms of clinical data exchange [3] including directed exchange that transmits patient data from one provider to another and query-based health information exchange (HIE) that aggregates patient data from multiple healthcare institutions [4]. The two forms of clinical data exchange address different but complementary patient data needs [5]. For example, directed exchange automates paper-based point-to-point processes such as patient referrals and delivery of laboratory results, while query-based HIE assembles comprehensive patient data from numerous organizations across a community. Query-based exchange using Cross-Enterprise Document Sharing (XDS), a specification developed by Integrating the Healthcare Enterprise (IHE), has been successfully deployed in multiple communities across Europe [6,7] and Asia [8,9]. In the United States today, federal policies have aligned usage of clinical data exchange with financial incentives. Novel healthcare models such as accountable care organizations rely increasingly on query-based HIE to gather patient population data from multiple institutions so clinicians can coordinate care for quality- and cost-based reimbursement [10]. Evidence of query-based HIE’s impact is also beginning to appear in the literature. Usage of query-based exchange systems facilitated by a private insurer [11] and a non-profit regional health information organization (RHIO) [12] have been associated with cost reductions in emergency department settings in two communities, but it is unknown whether findings generalize to other communities and care settings. Additionally, optimal approaches for the configuration and usage of query-based HIE systems are unknown. As a first step in evaluating community-based health information technologies, practitioners and researchers must understand how clinicians and staff use electronic systems [13–16]. For query-based HIE, quantitative measures of usage are critical for understanding the breadth and depth of system adoption, implementation, and resistance in communities, as availability of systems does not guarantee their usage or the quality of their usage [16]. However, existing quantitative studies of query-based exchange usage are limited to emergency department and safety net settings in two communities [17,18]. Furthermore, the two communities implemented different query-based exchange systems, limiting the ability to compare system usage patterns across communities. Broader study of query-based exchange system usage patterns across multiple communities, care settings, patient populations, and software systems is necessary to identify optimal approaches for HIE that improve healthcare decision making and outcomes [13–15]. In this study, our goal was to quantify the breadth and depth of usage of a commercially available system for query-based exchange implemented in three communities by measuring usage patterns with respect to practice sites, users, patients, and data. Understanding usage patterns of a
811
query-based exchange system implemented on a large-scale can inform efforts by organizations implementing HIE, vendors developing software for HIE, and policymakers defining rules for HIE to increase HIE usage in support of patient care and public health.
2.
Methods
We performed a cross-sectional study using system access log files to measure patterns of query-based exchange usage. The Institutional Review Boards of Weill Cornell Medical College and academic research institutions with which community organizations had affiliations approved this study.
2.1.
Setting
This study examined query-based exchange usage in three separate communities in New York State—here called Community A, Community B, and Community C—funded through the Healthcare Efficiency and Affordability Law for New Yorkers Capital Grant Program (HEAL NY). HEAL NY, which started in 2006 and represents more than an $800 million investment of public-private funds in EHRs and HIE, aims to develop a health information network for New York State by linking together community-based RHIOs that adhere to common standards and policies. RHIOs’ roles included convening and governing community stakeholders, promoting collaboration and data sharing, and implementing technology for HIE. As of 2012, twelve non-profit RHIOs provided HIE across New York State in compliance with state requirements using a variety of commercial products. The three non-profit RHIOs investigated in this study implemented the same commercial HIE system, which we describe in Section 2.3. As shown in Table 1, the RHIOs studied provided HIE for communities of about one million adults. Across the three communities, the numbers of encounter types and physicians varied. Of note, patients in Community A and Community B rarely received care outside the catchment area of their local RHIO, while patients in Community C may have sought care either within the catchment area of the local RHIO or in neighboring areas served by other RHIOs. Whereas providers in Community A and Community B practiced in settings ranging from large health systems to solo practices, providers in Community C practiced almost exclusively in large health systems. Years of RHIO operations differed across the communities. Notably, in 2009, Community C replaced a previous HIE product with its current system, which delayed the start of HIE data collection from community healthcare institutions.
2.2.
Patient consent
In New York State, patient data can be made available to a RHIO without consent, but patients must provide affirmative consent for that data to be accessed. Patient data can be accessed without affirmative consent in emergency situations according to a “break the glass” policy, and patients can also opt out of this type of access when choosing to not participate in HIE. Within the framework of this guidance,
812
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
Table 1 – Characteristics of communities engaged in HIE. Community A
Community B
Community C
Patient population, n (%) Female Male Age of patient population 18–24, n (%) 25–44 45–64 65 and older
1,207,007 641,222 (53) 565,784 (47)
968,600 506,820 (52) 461,780 (48)
982,800 555,282 (57) 427,518 (43)
143,747 (12) 451,498 (37) 364,554 (30) 247,208 (20)
121,131 (13) 375,427 (39) 298,450 (31) 173,591 (18)
148,400 (15) 429,800 (44) 263,200 (27) 141,400 (14)
Annual inpatient admissions Annual outpatient visits Annual emergency department visits Physicians
203,000 4,336,000 683,000 3600
157,000 3,523,000 264,000 3000
200,000 4,500,000 600,000 4000
Year RHIO founded Year RHIO data feeds started
2004 2007
2006 2007
2006 2010
Population and annual patient volume estimates are stated for 2010. Percentages may not sum to 100% due to rounding. HIE was available for adults and children in each community, but this study investigated HIE usage for adults only; adults represented 70–75% of the total population in each community. RHIO: regional health information organization.
the communities operationalized their consent procedures somewhat differently. In Community A, patients were invited to provide a single affirmative consent to all covered entities belonging to the RHIO, and were approached to provide additional consent when a new covered entity joined the RHIO. In Community B and Community C, patients were approached to provide multiple affirmative consents to individual covered entities to access their data. In the communities studied, the percentage of patients who provided affirmative consent for HIE participation ranged from 90% to 97%. Workflows for updating patient consent status differed across communities. In Community A, users updated consent status using a patient credentialing application linked to but separate from the HIE system. Users then needed to launch the HIE system to access a patient’s clinical data. In Community B, users updated consent status by logging into the HIE system and attempting to access a patient’s record. If the system detected that a patient had not provided consent to a user’s practice site, a popup window appeared requiring the user to specify the patient’s consent status. If the user indicated that the patient provided affirmative consent to the practice site, the system then displayed the patient’s clinical data. Users in Community B also had the option of updating consent using a patient credentialing system linked to but separate from the HIE system as well as an EHR form that transmitted consent status to the HIE. In Community C, users updated consent status using an EHR form only. Additionally, toward the beginning of operations in Community A and Community B, HIE administrators updated consent manually by processing paper forms received from practice sites.
2.3.
System description
The three communities implemented the same commercial HIE system, Axolotl Virtual Health Record (now known as Optum Virtual Health Record) [19], a secure stand-alone web portal that enabled registered users to query data sources from across the community for patients in their care. The system employed a federated architecture with a master patient
index, record locator service, and user directory to transmit patient data. Healthcare institutions in each community made their data available for access via HIE. Users in practice sites were able to access patient data via query-based exchange regardless of being affiliated with a particular healthcare institution that provided data. The system enabled users to look-up patient data from but not send patient data to the HIE. Fig. 1 depicts system workflow. To access the system, users in Community A logged in using a username, password, and secure token, while users in Community B and Community C logged in using a username and password. Users then searched for patients using a combination of demographic and administrative identifiers including last name, first name, date of birth, gender, medical record number, and insurance policy number. After a user selected a patient from search results, the system checked the patient’s consent status and displayed data accordingly. For a patient who had not provided affirmative consent to the user’s practice site, the system displayed a message stating it would not show patient data unless the patient provided affirmative consent. If at this time the patient decided to participate, a user would update the patient’s status to indicate affirmative consent as described in the previous section. For a patient who provided affirmative consent to the user’s practice site, the system displayed a default landing page—configured as a recent result summary report in Community A and Community B and a patient information report in Community C. A recent result summary presented links to the five most recent patient data reports for each data report category (e.g. laboratory, radiology), while a patient information report displayed patient contact information and the names of entities to which a patient had affirmatively consented. The system’s user interface presented tabs at the top of the screen corresponding to categories of patient data reports (e.g. laboratory, radiology). Upon clicking a category tab, a user was presented with a list of links to individual data reports of that category containing each report’s name, date, and responsible physician. After clicking a link from the list, a user accessed the details of an individual data
813
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
report. HIE users demonstrated two forms of patient data access—detailed access, or access of individual data reports for a patient, and default only access, or access of only the default landing page for a patient. Default only access may have provided users with useful clinical information from listings of available reports for a patient (e.g. indication that all laboratory results were outdated) without needing to access individual data reports.
2.4.
2.5.
We analyzed system usage patterns with respect to practice sites, users, patients, and data to determine where, by whom, for whom, and for what HIE system access occurred. Because Community A and Community B had 27 months of system availability and Community C had 9 months of system availability, we conducted analysis of the first 9 months of usage for all communities and the first 27 months of usage for only Community A and Community B. For practice sites, we determined the number and types of settings (e.g. inpatient, outpatient, emergency) from which system access occurred as well as the degree of access in each practice site as measured by sessions, or instances in which a user accessed a patient’s data on a given day. This measure enabled counting accesses on different days by a single user as well as multiple users accessing data for a single patient across one or more days. For users, we determined the number, roles, and types of practice sites of those who accessed the system. For patients, we determined the number of affirmatively consented patients, number of affirmatively consented patients whose data were accessed, the length of time from initial consent to first access of data for patients, and the number of patients whose data were accessed from multiple practice sites. We also differentiated between detailed access and default only access of data for patients. For data, we determined the number and types of data reports accessed by user roles, number and types of data reports accessed over time, and the number and types of data reports accessed by practice sites.
Data collection
For all affirmatively consented patients who were at least 18 years old at the time of consent, we obtained de-identified system access log data for the first 27 months of system availability in Community A and Community B (January 1, 2009 through March 31, 2011) and first nine months of system availability in Community C (September 1, 2010 through May 31, 2011). System access log data specified the patient, user, practice site, date, time, and category of each data report accessed. From the system access log data, we also obtained patient characteristics including gender, initial consent date, and age at initial consent. Additionally, using each community’s customer relationship management records, we obtained user characteristics—professional role (e.g. provider, nurse, staff, other clinician), primary practice site, and physician specialty—and practice site designation (e.g. inpatient, outpatient, emergency). We defined staff as non-clinical workers (e.g. medical receptionists), nurses as registered nurses, providers as clinicians with prescribing privileges (e.g. physicians including medical doctors and doctors of osteopathic medicine as well as physician extenders including nurse practitioners and physician assistants), and other clinicians as allied health professionals (e.g. pharmacists, physical therapists, registered dieticians).
Y Update in credenaling system
Data analysis
2.6.
Statistical analysis
We determined descriptive statistics including frequencies and percentages for categorical variables, mean and
Community A
Community B
Community C
Login to HIE
Login to HIE
Login to HIE
Search for paent
Search for paent
Search for paent Y
N
Check consent Y
Update in HIE popup window Y
N
Check consent Y
Display recent summary
Display recent summary
Display detailed report
Display detailed report
N Update in EHR
Check consent Y Display paent informaon Display detailed report
Fig. 1 – HIE system workflow across the three communities. The system displayed a default landing page to users for each patient who provided affirmative consent. Users then had the option of accessing individual detailed reports (e.g. laboratory, radiology, transcribed, admission-discharge-transfer). HIE data sources varied across communities and included hospitals, freestanding laboratories, freestanding radiology centers, homecare services, community health centers, and long-term care facilities.
814
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
Table 2 – Characteristics of practice sites from which system access and sessions occurred. Months 1–9
Months 1–27
Community A
Community B
Community C
Community A
Community B
Sites where system access occurred, N Outpatient, n (%) Inpatient Emergency Other
59 57 (97%) 1 (2%) 1 (2%) 0 (0%)
80 62 (78%) 4 (5%) 2 (3%) 12 (15%)
28 16 (57%) 8 (29%) 4 (14%) –
200 190 (95%) 6 (3%) 2 (1%) 2 (1%)
156 122 (78%) 10 (6%) 7 (5%) 17 (11%)
Sites where system sessions occurred, N Outpatient, n (%) Inpatient Emergency Other
1964 1903 (97%) 17 (1%) 44 (2%) –
27,675 22,611 (82%) 2439 (9%) 1645 (6%) 980 (4%)
1965 376 (19%) 1531 (78%) 58 (3%) –
15,272 14,465 (95%) 77 (1%) 304 (2%) 426 (3%)
319,089 259,135 (81%) 22,493 (7%) 35,201 (11%) 2260 (1%)
System access occurred from a practice site if any user from the practice site accessed the system. System sessions, or instances in which a user accessed a patient’s data on a given day from a practice site, accounted for the frequency of access from a practice site. Other practice sites included public health settings, long-term care, care and disease management, and home care settings.
standard deviation (SD) for normally distributed continuous data, median and interquartile ranges (IQR) for non-normally distributed continuous data, and minimum and maximum values for non-normally distributed continuous data that were highly skewed. We used SAS 9.2 and Microsoft Excel 2007 to perform calculations and manage data.
3.
Results
3.1.
Practice sites
Of practice sites registered to use the system in each community, access occurred in 18% (n = 59) of sites in Community A, 30% (n = 80) in Community B, and 82% (n = 28) in Community C during the first nine months of system availability (Table 2). After 27 months, access had occurred from 60% (n = 200) of sites in Community A and 59% (n = 156) in Community B. In each community, the majority of practice sites from which system access occurred were outpatient. In Community A and Community B, the majority of system sessions occurred from outpatient sites, while in Community C inpatient sites accounted for the majority of sessions.
3.2.
Users
The number, roles, and primary practice site of users who accessed the system varied across the three communities (Table 3). Community A registered the greatest number of users to access the system (n = 3658) followed by Community B (n = 3461) and Community C (n = 118). After 27 months, the number of users accessing the system in Community A and Community B had more than tripled. Over time more than half of users in Community A and Community B were nurses and staff, while after nine months more than two thirds of users in Community C were providers (e.g. physicians and physician extenders). In all three communities, the majority of users practiced primarily in outpatient practice sites. Of note, the proportion of users primarily practicing in inpatient sites was substantially greater in Community C after nine months compared to Community A and Community B over time.
3.3.
Patients
As shown in Table 4, during the first nine months of system availability, the number of patients who affirmatively consented to HIE participation and the proportion of patients whose data were accessed varied widely across the three communities. After 27 months in Community A and Community B, the number of affirmatively consented patients had increased more than six-fold while the proportion of patients whose data were accessed remained relatively similar. In each of the three communities, patients whose data were accessed were older than those whose data were not accessed. Additionally, patients whose data were accessed tended to be older than the overall population (Table 1). Of patients whose data were accessed after nine months, 60% (n = 936) in Community A, 7% (n = 1724) in Community B, and 27% (n = 296) in Community C had detailed access rather than default only access. After 27 months, the proportion of patients with detailed access rather than default only access remained relatively similar in Community A (53%, n = 5936) and had almost doubled in Community B (13%, n = 28,036). During the first nine months of system availability, the majority of patients whose data were accessed had their data accessed for the first time on the same day as consent in Community A (77%, n = 1096) and Community B (96%, n = 23,867). However, in Community C 17% (n = 196) of patients had their data accessed for the first time on the same day as consent, and the median (IQR) number of days from consent to first access was 17 (1–84). After 27 months, first time patient data access occurred on the same day as consent for 55% (n = 6003) of patients in Community A and 86% (n = 183,771) of patients in Community B. Over time the majority of patients in Community A and Community B had their data accessed in an outpatient setting, while after nine months the majority of patients in Community C had their data accessed in an inpatient setting (Table 5). During the first nine months, the number of patients whose data were accessed from two or more practice sites was 33 (0.1%) in Community A, 843 (1.8%) in Community B, and 12 (0.01%) in Community C. After 27 months, the number of patients whose data were accessed from two
815
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
Table 3 – Characteristics of users who accessed patient data. Months 1–9
Users by role, n Nurses and staff, n (%) Nurses, n (%) Staff, n (%) Providers, n (%) Physicians Internal medicine Emergency medicine Family medicine Pediatrics Obstetrics and gynecology Other specialty Oncology/hematology Radiology Physician extenders Other clinicians, n (%) Users by practice site type Outpatient, n (%) Inpatient Emergency Other
Months 1–27
Community A
Community B
Community C
Community A
Community B
155 110 (71) NA NA 39 (25) 32 (21) 5 3 10 4 1 9 0 0 7 (4) 6 (4)
399 240 (60) 35 (9) 205 (51) 104 (26) 76 (19) 15 3 16 1 3 24 4 10 28 (7) 55 (14)
88 22 (25) 20 (23) 2 (2) 60 (68) 54 (61) 20 14 8 7 4 1 0 0 6 (7) 6 (7)
672 487 (72) NA NA 134 (20) 105 (16) 13 11 27 11 6 32 5 0 29 (4) 51 (8)
1318 717 (54) 134 (10) 583 (44) 398 (30) 295 (22) 77 35 53 5 9 70 17 29 103 (8) 202 (15)
147 (95) 2 (1) 4 (3) 2 (1)
334 (84) 21 (5) 19 (5) 25 (6)
43 (49) 29 (33) 16 (18) –
637 (95) 8 (1) 17 (3) 10 (1)
1131 (86) 44 (3) 91 (7) 51 (4)
Available data sources in Community A did not distinguish between nurse and staff users. Physician extenders included nurse practitioners and physician assistants. Other practice sites included public health, long-term care, care and disease management, and home care settings.
or more practice sites increased to 315 (0.1%) in Community A and 41,210 (11.6%) in Community B.
3.4.
Data
As depicted in Fig. 2, data report access increased over time in all three communities. After nine months, the number of data reports accessed ranged from 5111 in Community A and 31,065 in Community B to 7690 in Community C. After 27 months, the number of data reports accessed increased more than six-fold to 32,784 in Community A and more than ten-fold to 382,920 in Community B. In Community B and Community C, users most frequently accessed only the default report, while in Community A users most frequently accessed a detailed report. Of detailed data reports accessed in Community A and Community B, users accessed laboratory reports most frequently followed by radiology, transcribed (e.g. discharge summary, history and physical, consultation), and admission-dischargetransfer (ADT) reports, which were structured documents generated by individual hospitals’ patient tracking systems. As shown in Table 6, in Community B staff accessed data reports more frequently than other user roles, and in Community A, where data sources did not distinguish between nurse and staff users, nurses and staff accessed data reports more frequently than other user roles. However, in Community C, providers accessed detailed data reports most frequently while nurses accessed the majority of default reports.
4.
Discussion
This study is the first to our knowledge to quantify patterns of usage of a commercial query-based health information
exchange system implemented on a large, multi-community scale. Across the three communities, between 59% and 82% of practice sites registered to use the system actually accessed the system. In two communities, outpatient staff primarily accessed patient data, while in one community inpatient providers were the main users. Consistent with previous studies [18,20], patients whose data were accessed tended to be older than those whose data were not accessed, as older patients may have more complex conditions that motivate users to access HIE [17,18,20]. Users primarily accessed summaries of recent results as well as detailed reports of laboratory and radiology studies. Over time, system usage increased in each community but with varying proportions of patients whose data were accessed. While the proportions of patients whose data were accessed in Community C (1%) and Community A (5%) are comparable to patient-level access rates reported in the literature [16], Community B’s exceptionally high 53–60% may be due to consent workflow procedures. The predominant consent workflow in Community B involved users logging into the HIE, searching for a patient, adjusting consent status in a popup window, and then landing on the recent result summary by default. Of patients who provided affirmative consent in Community B during the first nine months of system availability, 96% had their data accessed for the first time on the same day as consent, suggesting that users sought access to HIE data almost immediately upon being granted permission. In contrast, during the same time period, users in Community A updated consent through a separate credentialing application and accessed patient data on the same day as consent for 77% of patients while users in Community C updated consent through an EHR form and accessed data on the same day as consent for 18% of patients. Furthermore, after 27 months,
816
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
Table 4 – Characteristics of patients who affirmatively consented to have their data accessed.
Months 1–9 Community A n (%) Female Male Age at consent 18–24, n (%) 25–44 45–64 65 and older Community B n (%) Female Male Age at consent 18–24, n (%) 25–44 45–64 65 and older Community C n (%) Female Male Age at consent 18–24, n (%) 25–44 45–64 65 and older Months 1–27 Community A n (%) Female Male Age at consent 18–24, n (%) 25–44 45–64 65 and older Community B n (%) Female Male Age at consent 18–24, n (%) 25–44 45–64 65 and older
Unique patients affirmatively consented
Unique patients whose data were accessed
Unique patients whose data were not accessed
31,356 17,245 (55) 14,099 (45)
1571 (5) 963 (61) 608 (39)
29,785 (95) 16,282 (55) 13,491 (45)
3014 (10) 8045 (26) 11,439 (36) 8853 (28)
61 (4) 284 (18) 582 (37) 643 (41)
2953 (10) 7761 (26) 10,857 (36) 8210 (28)
46,948 35,223 (75) 11,723 (25)
24,765 (53) 17,500 (71) 7263 (29)
22,183 (47) 17,723 (80) 4460 (20)
2569 (5) 10,960 (23) 19,653 (42) 13,766 (29)
1672 (7) 5964 (24) 9326 (38) 7803 (32)
897 (4) 4996 (23) 10,327 (47) 5963 (27)
129,692 84,023 (65) 45,665(35)
1107 (1) 677 (61) 430 (39)
128,585 (99) 83,346 (65) 45,235 (35)
17,186 (13) 44,747 (35) 43,313 (33) 24,446 (19)
17 (2) 69 (6) 428 (39) 593 (54)
17,169 (13) 44,678 (35) 42,885 (33) 23,853 (19)
219,647 141,725 (65) 77,890 (35)
11,263 (5) 6697 (59) 4566 (41)
208,384 (95) 135,028 (65) 73,324 (35)
16,688 (8) 55,278 (25) 89,285 (41) 58,389 (27)
424 (4) 2001 (18) 4418 (39) 4419 (39)
16,264 (8) 53,277 (26) 84,867 (41) 53,970 (26)
354,161 238,116 (67) 115,919 (33)
212,586 (60) 142,834 (67) 69,648 (33)
141,575 (40) 95,282 (67) 46,271 (33)
27,201 (8) 88,085 (25) 148,006 (42) 90,866 (26)
14,070 (7) 49,645 (23) 89,735 (42) 59,136 (28)
13,131 (9) 38,440 (27) 58,271 (41) 31,730 (22)
Patients whose data were accessed tended to be older than those whose data were not accessed.
Table 5 – Unique patients for whom data were accessed stratified by practice site type. Months 1–9
Unique patients, N Outpatient, n (%) Inpatient Emergency Other
Months 1–27
Community A
Community B
Community C
Community A
Community B
1571 1515 (96%) 14 (1%) 44 (3%) 0 (0%)
24,765 20,677 (82%) 2235 (9%) 1466 (6%) 918 (4%)
1107 278 (25%) 778 (70%) 57 (5%) –
11,263 10,697 (95%) 64 (1%) 277 (2%) 271 (2%)
212,586 188,418 (83%) 18,860 (8%) 19,020 (8%) 1982 (1%)
Patients can be counted more than once if accessed in more than one type of practice site. As a result, sum of n’s may exceed total N.
817
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
Table 6 – Data reports accessed by users. Months 1–9
Laboratory reports, N Staff Nurses Providers Other clinicians Radiology reports Staff Nurses Providers Other clinicians Transcribed reports Nurses Providers Other clinicians ADT Staff Nurses Providers Other clinicians Recent result summary Staff Nurses Providers Other clinicians Patient information Staff Nurses Providers Other clinicians
Months 1–27
Community A
Community B
Community C
Community A
Community B
2491 1454 (58%) – 888 (36%) 149 (6%) 1152 872 (76%) – 255 (22%) 25 (2%) 221 120 (54%) – 96 (43%) 5 (2%) 68 48 (71%) – 20 (29%) – 1179* 730 (62%) – 387 (33%) 62 (5%) NA NA NA NA NA
3221 1785 (55%) 136 (4%) 918 (29%) 382 (12%) 1521 896 (59%) 41 (3%) 362 (24%) 222 (15%) 261 177 (68%) 16 (6%) 54 (21%) 14 (5%) NA NA NA NA NA 25,655* 23,330 (91%) 81 (<1%) 329 (1%) 1915 (7%) NA NA NA NA NA
300 – 6 (2%) 204 (68%) 90 (30%) 196 – 4 (2%) 190 (97%) 2 (1%) 44 0 (0%) 2 (5%) 40 (91%) 2 (5%) 278 1 (<1%) 58 (21%) 131 (47%) 88 (32%) 1739 0 (0%) 221 (13%) 1030 (59%) 488 (28%) 5133* 130 (3%) 4571 (89%) 345 (7%) 87 (2%)
14,663 11,076 (76%) – 2813 (19%) 774 (5%) 6168 4858 (79%) – 929 (15%) 381 (6%) 2144 1656 (77%) – 372 (17%) 116 (5%) 68 48 (71%) – 20 (29%) – 9741* 7090 (73%) – 1284 (13%) 1367 (14%) NA NA NA NA NA
58,713 30,223 (51%) 5388 (9%) 14,902 (25%) 8199 (14%) 29,846 18,613 (62%) 1954 (7%) 5938 (20%) 3341 (11%) 7210 3053 (42%) 834 (12%) 2426 (34%) 895 (12%) 1354 528 (39%) 164 (12%) 385 (28%) 277 (20%) 281,124* 267,793 (95%) 2627 (1%) 2391 (1%) 8313 (3%) NA NA NA NA NA
Community A data sources did not distinguish between nurses and staff. As denoted by *, in Community A and Community B the recent result summary was the default report whereas in Community C the patient information report was the default report. Transcribed reports included discharge summary, history and physical, consultation, etc. ADT: admission-discharge-transfer.
the proportion of patients with detailed access—beyond the default landing page—in Community B (7.9%) was substantially greater than in Community A (2.7%). In Community B, the practice of logging into the HIE to update consent may have become engrained in clinical culture and created a habit for users to access HIE data in the course of care. Substantial increases in the numbers of reports accessed after nine months versus after 27 months in Community B, especially when compared to Community A, support this hypothesis. Updating patient consent status directly through the querybased HIE portal may be inconvenient and time consuming with respect to other clinical tasks, but the tradeoff of workflow efficiency and regular HIE system usage may warrant consideration in other HIE portal implementations requiring affirmative patient consent. Future investigations should use qualitative methods to understand HIE usage with respect to consent workflow, system usability, and utility of clinical data for different practice types and communities. Across the three communities, clinicians and staff accessed a variety of data reports using the system, but they accessed only the default landing page with greatest frequency. This is consistent with a study of query-based HIE portal usage for medically indigent patients where users most frequently accessed only the default, or “gateway,” screen listing a patient’s most recent encounters [21]. Together these
findings emphasize the importance of data displayed on the default landing page of query-based portals, as the data may provide clinical utility to users without needing to display a detailed report, and consideration of levels of access, such as default only and detailed, when examining usage of querybased HIE portals [21]. Nurses and staff often prepare HIE data for physicians to use by generating printouts [22], and researchers have suggested that HIE designs incorporate user- and role-specific display of data to accommodate differing information needs [21,22]. Additionally, another study noted that regular HIE usage by nurses and staff was associated with increased overall levels of HIE access in practice sites [17]. Findings from our investigation support these observations, as staff in outpatient settings predominantly accessed the HIE system in Community A and Community B, the communities with the highest degree of access, while providers in inpatient settings mostly accessed the HIE in Community C. Although outpatient and inpatient clinical workflows diverge and may explain the difference in system usage between Community C and the two other communities, the majority of users in Community C (49%) practiced primarily in an outpatient setting, suggesting an opportunity to increase overall system usage through increased nurse and staff usage in outpatient settings. Additionally, further study is required to understand inpatient
818
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
Fig. 2 – Data reports accessed over time in Communities A, B, and C. Access of data reports through the system increased over time in the three communities studied. Default data reports accessed are presented on the secondary axis due to scaling issues resulting from their frequent access in Community B. Users accessed 1354 ADT reports in Community B (not shown). “Summary Only” and “Patient Info Only” refer to default only access as described in Section 2.
usage of HIE systems and whether nurse and staff usage in inpatient areas is related to overall levels of HIE access. To increase system usage, researchers and practitioners should engage all members of the care team and explore role- and practice setting-specific presentation of HIE data. From the current investigation and previous studies of query-based exchange, common observations—including older patients having HIE data accessed more frequently than younger patients [18,20], nurse and staff system usage being related to overall increased HIE usage [17,21], and importance of default patient data displays in HIE [17,21]—may
indicate emergence of theory for query-based exchange that can increase usefulness of HIE systems. Novel observations from this study, such as the potential relationship between opt-in consent workflow and increased system usage, may also contribute to nascent query-based exchange theory. Researchers have documented a dearth of theory in biomedical informatics research [23,24], and additional study in query-based HIE may help fill this gap in the literature. Our investigation may be overly conservative in measuring HIE usage. We examined access of individual static data reports, such as laboratory results and transcribed notes, and excluded aggregated dynamic data views, including just-intime insurance eligibility verification and medication history lookup, that were in early stages of development at the time of the study. Because users must first access a default landing page prior to accessing dynamic transaction features, our measures of access captured any use of these features by proxy. Furthermore, we examined usage of one web portal system for query-based HIE, and particularities of the system and its implementation, including but not limited to user interface design, workflow integration, organizational commitment, and training, may have affected user behavior [25]. Finally, we investigated usage of HIE in communities funded by New York State’s unique health information technology program, which may not exist in other settings. However, in contrast to previous approaches, the commercial query-based exchange system implementations studied were facilitated by community organizations without involvement of academic institutions with substantial informatics resources [17,26] and targeted all patients community-wide rather than only the medically indigent [27], which may enable findings to generalize to other settings. Patient privacy concerns [28,29], healthcare system structure [30,31], and patient population characteristics [32] influence HIE infrastructure and implementation decisions, and findings from this study may only be useful to settings where opt-in patient consent, communitywide data sharing, and query-based exchange match social norms. This investigation illustrates usage of a commercial querybased exchange system across multiple communities and care settings. Consistent with other studies, increased HIE usage by staff members appeared to be related to higher levels of system usage by all clinicians. Additionally, users most frequently accessed only the system’s default landing page as opposed to detailed clinical reports. Workflow for updating patient consent may have contributed to higher cultural acceptance of HIE system usage in one community, which may be useful for increasing usage of HIE portal systems in other settings requiring opt-in consent. User role, practice site type, and consent workflow may affect patterns of query-based HIE system usage in the communities studied and elsewhere.
Authors’ contributions TRC conceptualized the manuscript, guided data collection, authored the manuscript, and incorporated co-authors’ feedback into the manuscript. AME performed statistical calculations and provided critical manuscript feedback. SBJ
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
819
references Summary points What was already known • Usage of query-based health information exchange (HIE) portal systems has been associated with healthcare cost reductions. • Studies of query-based HIE portal system usage have addressed only emergency department and safety net settings in two communities. Additionally, the two communities implemented different query-based HIE systems, limiting the ability to compare system usage patterns across communities. What this study added to our knowledge • This study is among the first to quantify patterns of usage of a commercial query-based health information exchange system implemented on a large, multi-community scale. • Consent workflow requiring users to login to a querybased HIE portal system, as opposed to a separate system, may have contributed to higher overall HIE usage in one community, where 60% of patients had their data accessed. • Across the three communities studied, users most frequently accessed only default patient data, emphasizing the importance of data initially presented to users in HIE systems, followed by detailed laboratory reports, indicating a need for more research of HIE system usage in multiple communities and settings.
provided critical manuscript feedback. RK conceptualized the manuscript and provided critical manuscript feedback.
Conflicts of interest The authors declare that they have no conflicts of interest in the research.
Acknowledgments The study was conducted as part of the Health Information Technology Evaluation Collaborative (HITEC), the multidisciplinary academic consortium charged with evaluating the effects of New York State’s investment in health information technology. Authors of the manuscript include members (TRC and AME) and the director (RK) of HITEC. This study was conducted with funding from the New York State Department of Health (contract number C023699). The authors would like to thank the community organizations for participating in the study; Lisa M, Kern, M.D., M.P.H. for conceptual feedback; Kenneth S. Boockvar, M.D., M.S. for data collection assistance; Renny V. Thomas, M.P.H. for data collection assistance; and Jessica S. Ancker, M.P.H., Ph.D. for manuscript assistance.
[1] J.R. Vest, L.D. Gamm, Health information exchange: persistent challenges and new strategies, J. Am. Med. Inform. Assoc.: JAMIA 17 (2010) 288–294. [2] T. Mäenpää, T. Suominen, P. Asikainen, et al., The outcomes of regional healthcare information systems in health care: a review of the research literature, Int. J. Med. Inf. 78 (2009) 757–771. [3] L. Lenert, D. Sundwall, M.E. Lenert, Shifts in the architecture of the Nationwide Health Information Network, J Am Med Inform Assoc: JAMIA 19 (2012) 98–502. [4] C. Williams, F. Mostashari, K. Mertz, et al., From the Office of the National Coordinator: the strategy for advancing the exchange of health information, Health Aff. (Project Hope) 31 (2012) 527–536. [5] G.J. Kuperman, Health-information exchange: why are we doing it, and what are we doing? J. Am. Med. Inform. Assoc.: JAMIA 18 (2011) 678–682. [6] O. Heinze, M. Birkle, L. Köster, et al., Architecture of a consent management suite and integration into IHE-based Regional Health Information Networks, BMC Med. Inform. Decision Making 11 (2011) 58. [7] T. Schabetsberger, F. Wozak, B. Katt, et al., Implementation of a secure and interoperable generic e-Health infrastructure for shared electronic health records based on IHE integration profiles, Stud. Health Technol. Inform. 160 (2010) 889–893. [8] J. Zhang, K. Zhang, Y. Yang, et al., Grid-based implementation of XDS-I as part of image-enabled EHR for regional healthcare in Shanghai, Int. J. Comput. Assisted Radiol. Surg. 6 (2011) 273–284. [9] S. Liu, B. Zhou, G. Xie, et al., Beyond regional health information exchange in China: a practical and industrial-strength approach, AMIA. Annual Symposium Proceedings/AMIA Symposium AMIA Symposium 2011 (2011) 824–833. [10] S. Devore, R.W. Champion, Driving population health through accountable care organizations, Health Aff. (Millwood) 30 (2011) 41–50. [11] A. Tzeel, V. Lawnicki, K. Pemble, The business case for payer support of a community-based health information exchange: a Humana pilot evaluating its effectiveness in cost control for plan, Am. Health Drug Benefits 4 (2011) 207–216. [12] M.E. Frisse, K.B. Johnson, H. Nian, et al., The financial impact of health information exchange on emergency department care, J. Am. Med. Inform. Assoc.: JAMIA 19 (2012) 328–333. [13] J.S. Ancker, L.M. Kern, E. Abramson, et al., The Triangle Model for evaluating the effect of health information technology on healthcare quality and safety, J. Am. Med. Inform. Assoc.: JAMIA 19 (2012) 61–65. [14] C.P. Friedman, “Smallball” evaluation: a prescription for studying community-based information interventions, J. Med. Library Assoc.: JMLA 93 (2005) S43–S48. [15] K.B. Johnson, C. Gadd, Playing smallball: approaches to evaluating pilot health information exchange systems, J. Biomed. Inform. 40 (2007) S21–S26. [16] J.R. Vest, J. Jasperson, What should we measure? Conceptualizing usage in health information exchange, J. Am. Med. Inform. Assoc.: JAMIA 17 (2010) 302–307. [17] K.B. Johnson, K.M. Unertl, Q. Chen, et al., Health information exchange usage in emergency departments and clinics: the who, what, and why, J. Am. Med. Inform. Assoc.: JAMIA 18 (2011) 690–697. [18] J.R. Vest, L.D. Gamm, R.L. Ohsfeldt, et al., Factors associated with health information exchange system usage in a
820
[19]
[20] [21]
[22]
[23]
[24]
[25]
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 2 ( 2 0 1 3 ) 810–820
safety-net ambulatory care clinic setting, J. Med. Syst. 36 (2012) 2455–2461. Virtual Health Record – OptumInsight.com. http://www. optuminsight.com/physicians/solutions/health-information -exchange/optum-virtual-health-record/overview/ (accessed 20.04.13). J.R. Vest, Health information exchange and healthcare utilization, J. Med. Syst. 33 (2009) 223–231. J.R. Vest, Jasperson ‘jon Sean. How are health professionals using health information exchange systems? Measuring usage for evaluation and system improvement, J. Med. Syst. 36 (2012) 3195–3204. K.M. Unertl, K.B. Johnson, N.M. Lorenzi, Health information exchange technology on the front lines of healthcare: workflow factors and patterns of use, J. Am. Med. Inform. Assoc.: JAMIA 19 (2012) 392–400. B. Kaplan, Evaluating informatics applications—some alternative approaches: theory, social interactionism, and call for methodological pluralism, Int. J. Med. Inf. 64 (2001) 39–56. M. Chiasson, M. Reddy, B. Kaplan, et al., Expanding multi-disciplinary approaches to healthcare information technologies: what does information systems offer medical informatics? Int. J. Med. Inf. 76 (Suppl. 1) (2007) S89–S97. C.P. Friedman, J.C. Wyatt, Evaluation Methods in Biomedical Informatics, Springer, 2010.
[26] K.B. Johnson, C.S. Gadd, D. Aronsky, et al., The MidSouth eHealth Alliance: use and impact in the first year, AMIA. Annual Symposium proceedings/AMIA Symposium AMIA Symposium (2008) 333–337. [27] J.R. Vest, H. Zhao, J. Jasperson, et al., Factors motivating and affecting health information exchange usage, J. Am. Med. Inform. Assoc.: JAMIA 18 (2011) 143–149. [28] A. Geissbuhler, S. Spahni, A. Assimacopoulos, et al., Design of a patient-centered, multi-institutional healthcare information network using peer-to-peer communication in a highly distributed architecture, Stud. Health Technol. Inform. 107 (2004) 1048–1052. [29] A. Geissbuhler, Lessons learned implementing a regional health information exchange in Geneva as a pilot for the Swiss national eHealth strategy, Int. J. Med. Inf. (2013), doi:10.1016/j.ijmedinf.2012.11.002. [30] H. Park, S.-I. Lee, Y. Kim, et al., Patients’ perceptions of a health information exchange: a pilot program in South Korea, Int. J. Med. Inf. 82 (2013) 98–107. [31] T. Bowden, E. Coiera, Comparing New Zealand’s “Middle Out” health information technology strategy with other OECD nations, Int. J. Med. Inf. (2012), doi:10.1016/j.ijmedinf.2012.12.002. [32] W.T. Steward, K.A. Koester, S.P. Collins, et al., The essential role of reconfiguration capabilities in the implementation of HIV-related health information exchanges, Int. J. Med. Inf. 81 (2012) e10–e20.