Push and Pull: Physician Usage of and Satisfaction with Health Information Exchange

Abstract

Federal policy toward health information exchange (HIE), the electronic transfer of patient data across organizations, has evolved to support two forms—push, or point-to-point data transmission, and pull, or multisource data aggregation. HIE usage is associated with user satisfaction, but existing quantitative research is limited to settings where only pull HIE is available. To address this gap, we surveyed 99 physicians regarding usage of and satisfaction with push HIE and pull HIE available in their communities as well as effects of HIE on practice and overall HIE satisfaction. In five of nine measures, respondents reported being very satisfied with push HIE more often than pull HIE (p < 0.05). Physicians were at least four times as likely to report being very satisfied with HIE overall if they were pediatricians, were very satisfied with push HIE, or noted that HIE improved their access to complete information. Findings have implications for HIE implementation and policy.

Introduction

Health information exchange (HIE), the electronic transfer of patient data across organizations, has the potential to improve healthcare quality, reduce healthcare costs, and strengthen public health efforts [1]. Through provisions of the Health Information Technology for Economic and Clinical Health (HITECH) Act of 2009, physicians and hospitals must now demonstrate meaningful use of electronic health records (EHRs) capable of HIE to be eligible for Medicare reimbursement [2]. In support of this goal, 56 states and territories are currently developing HIE capacity through the HITECH-funded State HIE Cooperative Agreement Program. Although there is considerable effort directed toward HIE today, optimal approaches are unknown.

Broadly speaking, HIE can be accomplished through at least two forms of data transmission—push and pull—that address different patient data needs [3]. Push HIE emulates existing paper-based processes that support point-to-point transactions, such as delivery of results from a laboratory to a physician and referral of a patient from one physician to another. This is called “push” because data is electronically deposited in a recipient’s system after a sender initiates transfer. In contrast, pull HIE aggregates data from multiple healthcare institutions across a community to provide comprehensive information for clinical care, emergency response, biosurveillance, and quality activities. This is called “pull” because a user who seeks data must actively access a system and query relevant data. Push HIE typically involves transmission of messages and documents to physicians’ EHRs or clinical inboxes, such as in the Indiana Health Information Exchange’s DOCS4DOCS system [3], while pull HIE usually involves the query of patient data in a centralized or federated repository, such as in the MidSouth eHealth Alliance’s secure web portal [4]. Whereas existing privacy laws for paper-based clinical data apply to push HIE, regulations for pull HIE vary across states and communities and require patients to provide or withdraw consent for participation [3].

Over time federal policy has promoted both push HIE and pull HIE, including most recently a set of technologies called the Direct Project for enabling push HIE [3, 5]. Because push HIE and pull HIE serve complementary purposes, the Nationwide Health Information Network (NwHIN) will likely require elements of both forms of HIE [3]. Existing studies have evaluated usage of [4, 6, 7] and satisfaction with [8] pull HIE, but little quantitative research has been conducted about usage of and satisfaction with HIE overall when both push and pull mechanisms are available to users. User satisfaction has been associated with increased usage of pull HIE [8], and understanding determinants of user satisfaction with HIE overall is critical to ensure regular usage across settings, especially when multiple HIE mechanisms are available. The objectives of this study were to describe physician usage of and satisfaction with push HIE and pull HIE and to identify factors related to physicians’ overall satisfaction with HIE when both push and pull mechanisms are available.

Background

Since 2004 New York State has funded EHR, HIE, and other health information technology (HIT) efforts through the Healthcare Efficiency and Affordability Law for New Yorkers (HEAL NY) [9]. Phase 1 of HEAL NY awarded funds to 26 grantee organizations including health systems, municipal health departments, independent practice associations, and regional health information organizations (RHIOs). Two of the RHIOs that received HEAL 1 awards, HEALTHeLINK serving greater Buffalo and Rochester RHIO serving greater Rochester, implemented the same HIE platform from Axolotl Corporation (San Jose, CA) to provide push HIE and pull HIE to their respective communities. The two RHIOs began implementing push HIE in 2007–2008 and pull HIE in 2008–2009. At the time of this study, users in all practices served by the two RHIOs had access to push HIE and pull HIE.

Push HIE delivered laboratory and radiology results to the certified EHRs and “lite EHRs” of physicians who ordered tests and/or were designated by ordering physicians to receive test results. “Lite EHRs” provided result viewing, electronic prescribing, and limited clinical documentation through a secure website. Pull HIE enabled users to search for patients and view all laboratory and radiology results ordered by physicians from across greater Buffalo or greater Rochester through a secure web portal. Pull HIE also allowed users to access patient demographics and transcribed reports (e.g. progress notes). Hospitals, laboratories, and radiology centers provided data for access via pull HIE. For clinicians and staff to access a patient’s data via pull HIE, New York State policy required a patient to provide affirmative consent. Of patients approached to participate in pull HIE through HEALTHeLINK and Rochester RHIO during the study period, 93%–97% provided affirmative consent. Physicians could also access patient data via pull HIE without affirmative consent for emergency purposes per a “break the glass” policy, and patients had the ability to opt out of such access when deciding not to provide affirmative consent for pull HIE.

We recognize that push HIE and pull HIE manifest in configurations different from those studied. Specifically, the clinical content, sender, recipient, data transmission trigger, user interface, patient population, and organizational structure may vary. Despite these potential differences, there is sufficient similarity between the push HIE studied and the Direct Project as well as between the pull HIE studied and the NwHIN Exchange [3] for findings of this investigation to be useful in other settings.

Methods

We performed a cross-sectional survey of 584 physicians whose practices were provided HIE services by HEALTHeLINK (n=311) or Rochester RHIO (n=273) as of July 2010. The Weill Cornell Medical College Institutional Review Board approved this study.

Survey design

We identified domains of questions to include in the survey based on a model of user acceptance of new information technologies [10]. We then conducted a literature review of surveys addressing adoption and use of HIT to identify previously validated questions that could be adapted for the current survey [11–13]. When previously validated questions were not available, we drafted questions as needed. Physicians and researchers at Weill Cornell Medical College and HEAL NY grantee organizations who had expertise in HIT reviewed the survey and provided feedback. All survey questions were closed-ended and multiple choice. Domains and types of questions included: physician and practice characteristics, usage of push HIE and pull HIE, satisfaction with push HIE and pull HIE, effects of HIE overall on physician practice, and satisfaction with HIE overall. Survey questions specifically asked about “electronically viewing laboratory and radiology results” through HIE.

Study population and data collection

Cornell University Survey Research Institute (SRI) administered the survey and received contact information for practice administrators from HEALTHeLINK and Rochester RHIO. Beginning in July 2010, SRI sent practice administrators initial emails and physical letters inviting physicians to complete the online survey. Practice administrators, if they were not physicians, were asked to forward a hyperlink to physicians and/or print the survey for physicians to complete and return by fax. Between August and October, SRI sent practice administrators five reminder emails about the survey. Cornell SRI sent a $15 check to physicians who completed the survey and a $10 check to non-physician practice administrators who facilitated survey completion. Data collection ended in December 2010.

Data analysis

We calculated descriptive statistics to characterize respondents’ demographics, usage of HIE, satisfaction with HIE, perceived effects of HIE on practice, and overall satisfaction with HIE. Responses from HEALTHeLINK and Rochester RHIO were pooled after testing for heterogeneity. To facilitate analysis, we dichotomized responses to (a) five-point Likert scale questions about usage as always or most of the time versus sometimes, rarely, and never and (b) four-point Likert scale questions about satisfaction as very satisfied versus somewhat satisfied, somewhat dissatisfied, and very dissatisfied. We based these decisions on median responses being skewed toward most of the time and sometimes for usage as well as very satisfied and somewhat satisfied for satisfaction. We also dichotomized responses to three-point Likert scale questions about perceived effects of HIE as improved versus no change or worsened in order to focus on improvement.

To compare physicians’ usage of and satisfaction with push HIE and pull HIE as well as perceived effects of HIE, we used Pearson’s χ² tests or Fisher’s exact tests where appropriate. Using Pearson’s χ² tests, or Fisher’s exact test where appropriate, as well as univariable logistic regression models, we examined bivariate relationships between physicians’ overall satisfaction with HIE (e.g. very satisfied versus all other responses combined) and demographics, HIE mechanisms used (only push HIE, only pull HIE, or both push HIE and pull HIE), satisfaction with push HIE overall, satisfaction with pull HIE overall, and perceived effects of HIE. Finally, we constructed a multivariable logistic regression model using variables found to be significant in bivariate testing. We then used backward elimination, removing the variable with the least significance and refitting the model until all variables remaining were significant. The threshold for significance was p < 0.05 for all comparisons. We managed data using Microsoft Excel 2010 and conducted analysis using Stata 10.1.

Results

112 physicians (19%)—64 from HEALTHeLINK and 48 from Rochester RHIO—completed the survey. After excluding those who did not rate the primary study outcome of overall satisfaction with HIE, 99 respondents remained (Table 1). Compared to respondents, more non-respondents had practiced medicine for 30 years or more and were specialists. More than 75% of respondents identified themselves as board certified in the primary care disciplines of family practice, general pediatrics, and general internal medicine [14]. The majority of physicians practiced in an office not affiliated with a hospital or medical center (n=64). More than half of physician office practices were single specialty groups/partnerships (n=54), and the bulk of main practice sites had between two and nineteen practicing providers (e.g. MD, DO, NP, PA, CNM).

Table 1.

Characteristics of physician respondents meeting inclusion criteria and non-respondents. Categories may not sum to column totals due to missing data. For non-respondents, race or ethnic group was not available (NA).

	Respondents (n=99)	Non-Respondents (n=472)
Gender
Male	60 (60%)	256 (54%)
Female	38 (38%)	204 (43%)
Race or ethnic group
White	89 (90%)	NA
Non-white	9 (9%)	NA
Years practicing medicine
< 10	19 (19%)	63 (13%)
10–19	38 (38%)	202 (43%)
20–29	25 (25%)	56 (12%)
> 29	14 (14%)	120 (25%)
Physician specialty
Family practice	46 (46%)	73 (15%)
General pediatrics	16 (16%)	48 (10%)
General internal medicine	13 (13%)	32 (7%)
Other specialties	24 (24%)	176 (37%)
Primary setting of main practice
Hospital or medical center (MC)	14 (14%)	60 (12%)
Physician office affiliated with hospital or MC	12 (12%)	47 (10%)
Physician office not affiliated with hospital or MC	64 (65%)	195 (41%)
Community health center	7 (7%)	22 (5%)
Type of physician office
Solo practice	23 (23%)	29 (6%)
Multi-specialty group/partnership	6 (6%)	41 (9%)
Single specialty group/partnership	54 (54%)	198 (42%)
Other	2 (2%)	89 (19%)
Number of providers at main practice
1	14 (14%)	45 (10%)
2–5	24 (24%)	109 (23%)
6–10	20 (20%)	93 (20%)
11–19	27 (27%)	34 (7%)
> 19	13 (13%)	52 (11%)

Usage of push and pull HIE

Of the 99 physicians included in the sample, 78 (80%) used push HIE and 52 (53%) used pull HIE; 41 (41%) used only push, 15 (15%) used only pull, 37 (37%) used push and pull, and 6 (6%) did not specify. A significantly greater proportion of physicians reported using push HIE always or most of the time (68%, n=53) compared to pull HIE (19%, n=10) (p=0.001).

Satisfaction with push and pull HIE

As shown in Figure 1, greater proportions of physicians were very satisfied with specific aspects of push HIE than pull HIE. Significantly greater proportions of physicians were very satisfied with the accuracy (p=0.001), usefulness (p=0.001), timeliness (p=0.002), and completeness of information (p=0.002) as well as system speed (p=0.003) of push HIE compared to pull HIE. For push HIE, more than half of respondents were very satisfied with the accuracy (66%, n=52), usefulness (59%, n=47), and timeliness (51%, n=40) of information. Physicians least frequently reported being very satisfied with system speed (32%, n=25) and ease of use (30%, n=24) of push HIE. For pull HIE, physicians reported most frequently being very satisfied with presentation of information (32%, n=16) and least frequently being very satisfied with system speed (n=5, 10%). Additionally, 16% (n=8) of pull HIE users reported being very satisfied with obtaining patient consent for pull HIE. A greater proportion of physicians reported being very satisfied overall with push HIE (30%, n=24) compared to pull HIE (17%, n=9), but the difference was not statistically significant (p=0.148).

Figure 1.

Proportions of physicians (push users: n=78; pull users: n=52) very satisfied with aspects of push and pull HIE in order of push proportions.

Perceived effects of HIE on physician practice

Physicians largely reported improvements to their practice as a result of HIE. As depicted in Figure 2, physicians most frequently reported improvement in access to timely information (73%) and access to complete information (60%). Areas of least improvement were privacy and security of patient health information (34%) and ordering potentially redundant tests (32%).

Figure 2.

Proportions of physicians reporting that HIE improved aspects of practice (n=99).

Perceived effects of HIE on physician practice by physicians who used push HIE and pull HIE

Compared to those physicians who used only one form of HIE (n=56), physicians who used both push HIE and pull HIE (n=37) more frequently reported improvements to their practice as a result of HIE (Figure 3). Significantly greater proportions of physicians who used push HIE and pull HIE noted improved access to accurate information (p=0.014), quality and safety of care (p=0.003), and privacy and security of personal health information (p=0.005) as compared to physicians who used only push HIE or pull HIE. Of physicians who used both push HIE and pull HIE, 73% (n=27) were primary care physicians, and almost two thirds (n=24, 65%) practiced in an office not affiliated with a hospital or medical center.

Figure 3.

Proportions of physicians who used push HIE and pull HIE (n=37) versus those who used push HIE or pull HIE (n=56) reporting that HIE improved aspects of practice.

Bivariate analysis

Of the 99 total respondents, 38 (38%) expressed they were very satisfied with HIE overall. Bivariate relationships between physicians being very satisfied with HIE overall and the following predictors were significant (p < 0.05): being a pediatrician, use of only push HIE, being very satisfied with push HIE, identification of improved access to timely information, identification of improved access to complete information, identification of improved access to accurate information, identification of improved administrative efficiency, identification of improved communication with other providers, identification of improved quality and safety of care, identification of improved privacy and security of personal health information, and identification of improved/reduced ordering potentially redundant tests. Of note, use of both push HIE and pull HIE was not associated with being very satisfied with HIE (p=0.608).

Multivariate analysis

Through multivariate analysis using backward elimination, we identified three predictors of physicians being very satisfied overall with HIE (p < 0.05): being a pediatrician (OR 4.90; 95% CI 1.02–23.62), being very satisfied with push HIE (OR 7.99; 95% CI 2.15–29.78), and identification of improved access to complete information through HIE (OR 15.69; 95% CI 2.72–90.64).

Discussion

To our knowledge, this is the first study to quantify physicians’ usage of and satisfaction with push HIE and pull HIE when both forms of health information exchange are available to users. Physicians reported using push HIE more often than pull HIE and being very satisfied more frequently with aspects of push HIE than pull HIE. In six of eight measures of physician practice, more than 50% of physicians noted improvement as a result of HIE. Physicians who used both push HIE and pull HIE more frequently noted improvement as a result of HIE compared to physicians who used only one form of HIE. Physicians were at least four times as likely to report being very satisfied with HIE overall if they were pediatricians, were very satisfied with push HIE, or noted that HIE improved their access to complete information.

This investigation quantifies many observations described in previous qualitative studies of HIE [15, 16]. Rudin and colleagues noted that HIE users perceived improvements to quality of care due to HIE providing more timely, complete, and accurate information, as well as improvements to administrative efficiency through reduction of phone calls to other physician offices [16], all of which our findings support. This indicates that HIE provides physicians with better information overall. Studies in two other states have identified a lack of complete information as a barrier to HIE usage [15, 16], and in our study 60% of physicians noted improved access to complete information as a result of HIE. Physicians appear to value complete information available through HIE, and complete information may be critical to success of HIE efforts. Only 34% of survey respondents perceived an improvement in privacy and security of patient health information, reflecting the privacy and security concerns for HIE expressed in the literature [17]. Reduction of redundant laboratory and radiology testing has been cited as a potential [18] and perceived effect [15] of HIE, but respondents in our survey were least likely to identify this as an area that they perceived as having improved. Additional research is needed to understand the impact of HIE on laboratory and radiology utilization.

In our study, physicians expressed a general preference for push HIE versus pull HIE as evidenced by reports of greater usage of and satisfaction with push HIE compared to pull HIE. This suggests that push HIE provided physicians with the laboratory and radiology information they desired, which was by and large for tests they ordered as opposed to tests ordered by physicians with whom they might not have a relationship. Furthermore, workflow integration is critical to HIE usage [7, 16], and push HIE accessible through existing EHRs rather than a stand-alone pull HIE system might have been more amenable to physicians with limited time to use systems or train to use new systems. HIE facilitated by Direct Project technologies meets the requirements of this use case as well, and physicians may similarly use and be satisfied with Direct Project-based transactions. In this study, physicians who used both push HIE and pull HIE noted improvements to practice as a result of HIE more frequently than those physicians who used only one form of HIE. This supports the view that push HIE and pull HIE serve different but complementary patient information needs [3]. As the number of physicians using EHRs and the volume of HIE transactions increases, physicians in primary care and other specialties may express a need for data not provided through push HIE, which pull HIE may provide. Pull HIE may also provide value to other stakeholders, including emergency department physicians, medical and surgical specialists, and public health officials, all of whom were not surveyed in the current study.

In this investigation, we identified three independent predictors of physicians being very satisfied with HIE overall. First, physicians were more likely to be very satisfied if they were pediatricians. In another study, researchers noted that “[a] pediatrician who used the HIE infrequently did not believe many pediatric care visits had problems with missing clinical information because consulting physicians usually forwarded their medical notes back to this clinician via fax” [16]. In the communities we studied, pediatricians might have begun relying on communications and results accessible via HIE instead of fax to streamline clinical information processing. Second, physicians were more likely to be very satisfied if they identified improved access to complete information through HIE. Physicians want complete information through HIE [15, 16]. Expanding the number of institutions and physicians participating in HIE may improve the completeness of information and thus physicians’ overall satisfaction with HIE. Finally, physicians who were very satisfied with push HIE were very satisfied overall with HIE.

This study has limitations. Although our survey had a lower response rate than anticipated, the sample size of 99 respondents exceeds the sample size of similar HIE evaluations conducted using interviews [16] and focus groups [15] by almost five-fold. More than 75% of respondents were board certified in primary care disciplines [14], and thus the findings from this study may not reflect attitudes of specialist physicians. This study focused on physicians using HIE to view laboratory and radiology results only, whereas other investigations have considered access of additional types of data, including narrative notes and medications, in gauging HIE user perspectives [15, 16, 19]. The push HIE and pull HIE studied may provide benefits to physicians beyond the result viewing we investigated. For example, pull HIE may be better suited than push HIE for sharing certain types of data, such as narrative notes, that are less likely to have been pushed by another physician, thereby increasing the completeness of information for a patient. Finally, as in many clinical informatics investigations [20], perceptions are likely to be shaped by specific features of the particular systems in use, including the quality of implementation and maturity of functionality. Future studies should examine different implementations of the systems studied as well as other vendor systems and consent models.

This study provides perspectives on physicians’ usage of and satisfaction with two forms of health information exchange. Physicians reported using push HIE more frequently than pull HIE and were more often very satisfied with push HIE compared to pull HIE for electronically viewing laboratory and radiology results. Findings suggest that current policy for expanding push HIE may lead to increased usage of and satisfaction with HIE overall for physicians. However, as evidenced by physicians who used both push HIE and pull HIE noting improvement as a result of HIE more frequently than those who used only one form of HIE, availability and usage of both push HIE and pull HIE may be necessary to improve healthcare processes overall. Additional research is required to understand the value of different HIE forms as EHR adoption increases and HIE infrastructure matures.