The complementary nature of query-based and directed health information exchange in primary care practice

Abstract

Objective

Many policymakers and advocates assume that directed and query-based health information exchange (HIE) work together to meet organizations’ interoperability needs, but this is not grounded in a substantial evidence base. This study sought to clarify the relationship between the usage of these 2 approaches to HIE.

Materials and Methods

System user log files from a regional HIE organization and electronic health record system were combined to model the usage of HIE associated with a patient visit at 3 federally qualified health centers in New York. Regression models tested the hypothesis that directed HIE usage was associated with query-based usage and adjusted for factors reflective of the FITT (Fit between Individuals, Task & Technology) framework. Follow-up interviews with 8 key informants helped interpret findings.

Results

Usage of query-based HIE occurred in 3.1% of encounters and directed HIE in 23.5%. Query-based usage was 0.6 percentage points higher when directed HIE provided imaging information, and 4.8 percentage points higher when directed HIE provided clinical documents. The probability of query-based HIE was lower for specialist visits, higher for postdischarge visits, and higher for encounters with nurse practitioners. Informants used query-based HIE after directed HIE to obtain additional information, support transitions of care, or in cases of abnormal results.

Discussion

The complementary nature of directed and query-based HIE indicates that both HIE functionalities should be incorporated into EHR Certification Criteria.

Conclusions

Quantitative and qualitative findings suggest that directed and query-based HIE exist in a complementary manner in ambulatory care settings.

INTRODUCTION

The process of health information exchange (HIE)—the transmission of patient information across providers—occurs in 2 distinct forms: directed and query-based.^1–6 In directed exchange, structured documents such as test results, imaging reports, and clinical care documents are sent from one provider to another, mimicking the faxing of paper records or email. Directed exchange is also known as “push” HIE because the sender initiates the act of sharing patient information. Often the process is automated, so that key events, like the posting of test results, a hospital admission, or an emergency department visit, trigger the “push” of information as a notification. Alternatively, query-based forms are referred to as “pull” HIE, because the provider who wishes to obtain patient information initiates the exchange. Providers query or search for desired information from a shared information system such as a community-wide longitudinal health record. These shared information systems may be accessible within an electronic health record (EHR) environment or through a stand-alone web-based portal.

Both query-based and directed HIE serve as sources for patient information in healthcare delivery⁶ and both meet federal health policy requirements for accessible and usable patient information.^7,8 Predominately, policymakers and HIE proponents present directed and query-based HIE as different pathways that work together to realize information sharing goals.^2,3,7,9,10

However, the working assumption that directed and query-based exchange are used together to meet information needs is not grounded in a substantial evidence base. In fact, the results of previous studies based on qualitative analyses and survey research are ambiguous or contradictory. For example, a single survey suggests that providers perceive directed exchange as the more favorable approach,¹ but providers report in interviews that directed exchange can be overwhelming and too much information to manage.¹¹ Qualitative research examining health information organizations indicated that leaders of those organizations may view directed and query-based exchange as competing alternatives.¹² However, qualitative research conducted among end users of HIE found that providers reported using query-based exchange to seek in-depth information in response to information received through directed exchange.¹¹ Most importantly, current health information technology policy does not treat both types of HIE as complementary. While the bulk of evidence of the effectiveness of HIE comes from studies of query-based exchange systems,^5,13 only directed exchange capabilities were required as part of the Centers for Medicare and Medicaid Services EHR Certification Criteria.

In the absence of empirical evidence, we sought to clarify the relationship between directed and query-based HIE usage. Using HIE and EHR system log files from multiple ambulatory care practice sites, we investigated whether health care professionals used one type of HIE more than another, whether directed exchange was associated with usage of query-based exchange, and for which patient and visit characteristics were providers more likely to use one HIE type over the other.

MATERIALS AND METHODS

Data from HIE and EHR system log files were combined to model the usage of HIE associated with a patient visit at 3 federally qualified health centers (FQHCs).

Sample and setting

The study sample included all adult encounters (N = 241 868) at 3 FQHCs serving urban and rural counties in Western New York State from 2014 to 2016.

Each FQHC uses the query-based and directed HIE services offered by the Rochester Regional Health Information Organization (RHIO). The Rochester RHIO facilitates community-wide HIE in Western New York. It is a nonprofit organization established in 2006 with a combination of private and state funding that has been actively exchanging patient information for more than 7 years. The Rochester RHIO provides query-based HIE services via a secure, stand-alone web-based portal. Once logged into the portal, authorized users (employees of a RHIO member organization and approved by the RHIO) have access to a consolidated view of patients’ demographic information, prior diagnoses, medication history, radiology reports and images, laboratory results, and discharge summaries from all of the RHIO’s participating providers. Users can query only 1 patient at a time and only the records of patients who have actively provided consent can be accessed (consent rates are >94%). In terms of directed exchange, the RHIO sends electronic laboratory/diagnostic test results, imaging results, and discharge summaries automatically to individual physicians and practices using DIRECT Secure Messaging Protocols. If physicians and practices opt for this service, new test results and information are “pushed” into their EHR systems on a daily basis in the form of interoperable, structured documents. Within the EHR, users are able to review and incorporate data from these documents into the patient record as desired.¹⁴

Data

A combination of 3 datasets furnished a complete description of HIE usage for a given patient encounter. First, patient encounters to FQHC sites were identified from a reporting platform (BridgeIT) operated by the Health Center Network of NY (HCNNY). HCNNY uses BridgeIT to extract and aggregate data from the FQHCs’ eClinical Works EHRs for quality reporting and improvement purposes. The encounter data included patient identifiers, date and time, providers seen, and information on patient characteristics. Second, BridgeIT also accesses the system logs, which document user activity within the EHR. These EHR system log files recorded laboratory results, imaging reports, and other clinical documents electronically delivered to the FQHCs. Importantly, the system log files allowed us to distinguish between electronically delivered information that was a result of a FQHC provider’s order and information that was sent to the FQHC from an external provider (ie, information shared through directed HIE) and to determine if the electronically delivered information was actually reviewed and included in the patient’s EHR (ie, was the exchanged information actually viewed and used). Third, system log files from the Rochester RHIO’s web portal identified usage of query-based HIE and also included patient identifiers, date and time, and user practice location. The web portal was stand-alone from the EHR without single sign-on. The Rochester RHIO and HCNNY provided synthetic patient identifiers for matching records across the different data sources.

HIE usage

HIE usage was measured at the level of the encounter and was defined as HIE activity occurring any time between the date of the patient encounter and the date of their most recent prior encounter. Usage of query-based exchange was defined as any FQHC user’s access of the RHIO’s portal during this window. Portal users had to be located at the same FQHC site as the patient’s visit. Because providers frequently delegate responsibility for accessing HIE information to other staff (ie, proxy usage),¹⁴ we did not limit portal users to the providers seen during visits. Furthermore, only access of clinical data contributed to query-based exchange usage (ie, usage solely for administration or consent management was excluded from the analysis). No evidence of usage of the RHIO’s portal was considered as no query-based exchange usage. Given the portal’s user interface, we could not categorize the specific types of clinical information viewed (ie, common views contained summaries of multiple types of information). Directed exchange usage was defined by the review and inclusion of any patient information delivered to a FQHC’s EHR during the same time window. No evidence of delivered information or incorporation of information into the patient record was considered as no usage. Directed usage was subdivided by use of laboratory, imaging, or other clinical documents. The clinical documents category contained a wide variety of discharge summaries, progress notes, emergency department reports, history and physicals, nursing notes, and consult notes. Because we did not have access to the ordering dates for any of the information generated by non-FQHC providers, we selected the previous encounter date as an anchor for our time period for measuring HIE usage. Previous research indicates that HIE usage is not limited to the day of the encounter, and information is often accessed in advance of a patient’s visit.¹⁵

Covariates

We drew on the FITT (Fit between Individuals, Task & Technology) framework¹⁶ to identify relevant covariates. The FITT framework posits that individual technology usage behavior is the product of the interaction of the attributes of end user, the technology, and work processes, tasks, and objectives. Each of these attributes influence fit, the degree to which a technology supports or assists performing tasks, which in turn affects usage and impact.¹⁷ With a focus on the interrelated nature among the user, the technology, and the task environment, FITT falls within the sociotechnical perspective of information system theories.¹⁸ Work activities (ie, tasks that might increase user reliance on technology during a patient visit) included time of the visit (day of week and whether during morning or afternoon), patient demographics (age, gender), diagnoses associated with the visit, comorbidity, and if the encounter was scheduled or a same day appointment. We applied the Agency for Health Care Research and Quality’s Clinical Classification Software codes to the primary diagnoses associated with each encounter and identified high-prevalence chronic conditions.¹⁹ The provider type seen during the visit (physician, nurse practitioner, or other) reflected the individual characteristic domain. We calculated the FQHC site’s average directed message volume each month and the average number of query sessions each month to describe the organizations’ technology context. These measures were transformed into Z-scores.

Analysis

We described the sample using frequencies, percentages, and means by type of HIE used. For our primary analysis, we test the hypothesis that directed HIE usage was associated with query-based HIE using the following logistic regression model:

$${\mathit{Query}}_j\mathrm{ }={\alpha }_j+\mathrm{ }{\beta }_{1j}L\mathit{ab}+\mathrm{ }{\beta }_{2j}I\mathit{mage}+\mathrm{ }{\beta }_{3j}D\mathit{ocument}+{\mathit{\beta }}_{4\mathit{j}}\mathit{T}\mathit{ask}+\mathrm{ }{\mathit{\beta }}_{5\mathit{j}}\mathit{T}\mathit{ech}+\mathrm{ }{\mathit{\beta }}_{6\mathit{j}}\mathit{I}\mathit{nd}+\mathit{C}+\varepsilon$$

${\mathit{Query}}_j$ is a binary measure of the usage of query-based HIE compared with the reference category of no usage. Lab, Image, and Document are the respective indicators for the type of information shared by directed HIE. A positive coefficient for any would indicate a complementary relationship between directed and query-based HIE, whereas a negative coefficient would be evidence of a substitutionary relationship. $\mathit{Task},\mathrm{ }\mathit{Tech}$ and $\mathit{Ind}$ are the individual patient, provider and contextual factors that describe the task, technology, and individual characteristics that constitute the constructs from the FITT model. $\mathit{C}$ represents the FQHC site fixed effects to control for all between-site time-invariant differences and $\varepsilon$ is random error. Because we are interested in differences between provider types, we cannot include a provider fixed effect. Individual and fully adjusted models included cluster-robust (at the FQHC level) standard errors.²⁰ To facilitate interpretation, we expressed the logistic regression coefficients as marginal effects.

As a secondary analysis, we explored the relationship between the factors suggested by the FITT model and the overall usage of HIE using the following multinomial logit model:

$$\mathit{log}\left(\frac{{\pi }_j}{{\pi }_K}\right)={\alpha }_j+\mathrm{ }{\mathit{\beta }}_{1\mathit{j}}\mathit{T}\mathit{ask}+\mathrm{ }{\mathit{\beta }}_{2\mathit{j}}\mathit{T}\mathit{ech}+{\mathit{\beta }}_{3\mathit{j}}\mathit{I}\mathit{nd}+\mathit{C}+\varepsilon \mathrm{ }K=1,\mathrm{ }2,\mathrm{ }3,\mathrm{ }4$$

${\pi }_j$ is the probability of each type of HIE usage: directed only, query-based only, or both. ${\pi }_K\mathrm{ }$is the reference category of no usage. The categorical outcomes provided flexibility for analyzing the relationship between each HIE approach under different situations. The models included FQHC fixed-effects and cluster-robust standard errors and followed the same model fitting strategy described above. Again, we expressed the regression coefficients as marginal effects. Given the large number of comparisons in multinomial models, we presented a subset of notable findings as a series of forest plots and provided the full results as an Supplementary Appendix.

Supplemental interviews

Log files are useful for objectively recording behavior, but they have inherent limitations around understanding users’ information needs and goals.²¹ To interpret findings, we conducted key informant interviews with a convenience sample of 8 HIE and EHR end users from the study sites. With the cooperation of our data partners and FQHC contacts, we identified clinical (4 nurses and a physician) and nonclinical users (administration and information managers) familiar with workflows and EHR and HIE usage to be interviewed. All identified end users agreed to participate. Interviews followed a semistructured format. We analyzed interview transcripts for relevant themes using a template analysis approach, where a list of codes (with definitions) is developed a priori and applied to the data.²² We developed a coding template based on interview notes and an initial reading of 2 transcripts (see Supplementary Appendix). Codes were refined through joint discussion. Two readers independently applied the template to all transcripts (kappa for prompted query-based usage = 0.79). We looked for frequent and co-occurring codes to provide additional insights related to our primary question of how directed and query-based HIE relate in ambulatory care practice.

RESULTS

Overall usage

A higher proportion of encounters had directed HIE usage (23.5%) than query-based HIE usage (3.1%). The proportion of encounters with both directed and query-based HIE usage was 1.4% (Table 1). As a product of the large sample size, nearly all comparisons between measures reflective of the FITT model and type of HIE usage were statistically significant. Nevertheless, several broad trends were notable. All types of HIE usage (directed only, query only, or both) were more common for encounters among patients with higher comorbidity scores, those with high-prevalence chronic disease diagnoses (with the exception of depression) and during office visits or after emergency department or hospital discharge visits. HIE usage also tended to be higher during morning appointments. Usage of directed HIE was more common among encounters with physicians, whereas nurse practitioners had higher odds of query-based usage and use of both types of HIE.

Table 1.

Characteristics of primary care encounters associated with preceding query-based and directed health information exchange usage

	Total	No exchange	Directed	Query-based	Both	P a
	N = 241 868	n = 174 118	n = 56 872	n = 7499	n = 3379
Task
Patient characteristics
Female	63.6	62.8	66.3	59.8	69.4	<.0001
Age, y	45.7 ± 16.3	44.7 ± 163.3	48.1 ± 16.0	48.0 ± 16.1	49.5 ± 15.2	<.0001
Comorbidity score	3.5 ± 0.88	3.5 ± 0.89	3.7 ± 0.84	3.5 ± 0.86	3.7 ± 0.84	<.0001
Primary diagnosis at visit
Hypertension	8.1	7.8	9.0	9.0	7.4	<.0001
Congestive heart failure	0.1	0.1	0.2	0.2	0.5	<.0001
Coronary artery disease	0.2	0.2	0.2	0.4	0.5	<.0001
Cardiac arrhythmias	0.3	0.2	0.3	0.4	0.3	<.0001
Hyperlipidemia	1.0	0.9	1.3	1.1	1.0	<.0001
Stroke	0.1	0.1	0.2	0.1	0.2	<.0001
Arthritis	0.5	0.5	0.7	0.5	0.7	<.0001
Asthma	1.1	1.1	1.1	1.2	1.0	<.0001
Cancer	0.2	0.1	0.3	0.4	0.4	<.0001
COPD	1.0	0.9	1.0	1.4	1.2	<.0001
Depression	2.5	2.6	2.1	2.5	2.5	<.0001
Diabetes	6.0	5.9	6.3	5.2	6.1	<.0001
Hepatitis	0.7	0.7	0.7	0.9	0.7	<.0001
HIV	0.3	0.3	0.3	1.0	0.5	<.0001
Schizophrenia	0.1	0.1	0.2	0.2	0.1	<.0001
Substance use	2.0	2.2	1.8	1.2	1.1	<.0001
External injury	0.2	0.2	0.3	0.3	0.3	<.0001
Visit type
Office	79.9	78.1	84.3	85.8	86.6	<.0001
Specialist	6.8	7.4	5.8	2.0	2.0	<.0001
Postdischarge/emergency	0.6	0.4	1.1	1.3	3.9
Substance abuse treatment	1.5	1.7	1.2	0.5	0.4
Other/unclassified	11.1	12.4	7.6	10.4	7.2
Same-day appointment	24.0	25.7	19.4	20.9	18.6	<.0001
Morning appointment	46.7	45.8	49.3	47.5	50.9	<.0001
Individual
Provider type
Physician	45.3	44.3	48.8	41.4	45.6	<.0001
Nurse practitioner	23.9	24.2	21.8	32.0	22.7
Other	30.8	31.5	29.4	26.6	31.7
Technology
Monthly pull usage	180.5 ± 77.7	179.7 ± 77.5	184.2 ± 77.5	178.8 ± 78.1	166.7 ± 85.8	<.0001
Monthly directed usage	1084.1 ± 555.1	1071.2 ± 550.7	1136.9 ± 560.4	1024.1 ± 564.0	993.7 ± 605.5	<.0001

Values are % or mean ± SD.

COPD: chronic obstructive pulmonary disease; HIV: human immunodeficiency virus.

^aComparison of type of exchange usage.

Association between directed and query-based HIE

Without controlling for other factors, use of directed HIE with each of the 3 types of information was associated with increased odds of query-based HIE usage during an encounter (Table 2). Directed HIE use for laboratory results increased the odds of using query-based HIE by 43%, which was equivalent to a 1.7-percentage-point increase (or 4.2% of encounters without to 5.9% for encounters with directed usage). Use of directed HIE for imaging information was associated with a 2.1-percentage-point increase (4.2%-6.3% of encounters), a 52% relative increase, in query-based HIE usage (odds ratio, 1.52; 95% confidence interval, 1.01-2.31). The largest effect was observed for HIE use for clinical documents, which increased the odds of query-based HIE usage by more than 4 times (odds ratio, 4.28; 95% confidence interval, 3.97-4.63) or 12 percentage points. Compared with general office visits, the odds of query-based HIE usage were lower for specialist and substance abuse visits. Encounters classified as a post-discharge or post emergency department visit had a significantly higher probability (9.3 percentage points higher) of query-based HIE usage than general office visits. The odds of query-based HIE usage were higher among nurse practitioners compared with physicians and the probability of query usage was higher when sites had overall higher query-based HIE activity.

Table 2.

Associations between usage of directed health information exchange for laboratory, imaging, or clinical documents and query-based health information exchange in 3 New York federally qualified health centers

	Any query usage		Any query usage
	Odds ratio (95% CI)	Marginal effect (%)	Adjusted odds ratio (95% CI)	Marginal effect (%)
Directed usage laboratory	1.43 (1.00-2.03)a	1.7	1.28 (0.99-1.65)	1.0
Directed usage imaging	1.52 (1.01-2.31)a	2.1	1.16 (1.08-1.24)b	0.6
Directed usage documents	4.28 (3.97-4.63)b	12.0	3.07 (2.54-3.71)b	4.8
Female	0.98 (0.63-1.51)	0.0	1.06 (0.74-1.52)	0.3
Age	1.01 (1.01-1.01)b	0.0	1.01 (1.01-1.01)b	0.0
Comorbidity score	1.02 (0.82-1.27)	0.0	0.96 (0.78-1.17)	–0.2
Visit type
Office	Reference		Reference
Specialist	0.28 (0.23-0.34)b	–3.4	0.32 (0.27-0.39)b	–3.2
Postdischarge/emergency	3.25 (2.20-4.80)b	9.3	2.54 (1.51-4.29)b	6.4
Substance abuse treatment	0.31 (0.30-0.32)b	–3.2	0.37 (0.33-0.40)b	–2.9
Other/unclassified	0.74 (0.57-0.96)a	–1.2	0.83 (0.60-1.15)	–0.8
Same-day appointment	0.80 (0.66-0.99)a	–0.9	0.86 (0.71-1.04)	–0.6
Morning appointment	1.08 (1.02-1.14)c	0.3	1.00 (0.95-1.05)	0.0
Provider type
Physician	Reference		Reference
Nurse practitioner	1.42 (1.30-1.55)b	1.7	1.45 (1.30-1.61)b	1.7
Other	0.90 (0.84-0.96)c	–0.4	0.94 (0.89-1.00)	–0.2
Monthly pull usage	1.95 (1.10-3.44)a	2.9	2.00 (0.87-4.59)	2.9
Monthly directed usage	2.56 (1.81-3.61)b	4.0	0.90 (0.17-4.69)	–0.5

CI: confidence interval.

^aP< .05. ^bP< .01. ^cP< .001.

In the fully adjusted model (Table 2), usage of directed HIE continued to be associated with increased usage of query-based HIE: 0.6 percentage points higher for imaging information and 4.8 percentage points higher for clinical documents. The probability of query-based HIE usage remained lower for specialist visits, higher for post-discharge visits, and higher for encounters with nurse practitioners.

Factors associated with usage of directed and query-based HIE

In the fully adjusted model, the association between HIE usage and several factors representing different task, individual, and technology characteristics differed by HIE type (Figure 1). In the task domain, the largest effects on HIE usage were associated with encounters classified as a postdischarge or post emergency department visit: the frequency of directed HIE use increased 11 percentage points, query-based usage increased 4.2 percentage points, and usage of both increased 4 percentage points. In contrast, any type of HIE usage was lower for specialist visits or substance abuse treatment visits compared with office visits. Also, within the task domain, indicators of increasing patient complexity were associated with increased HIE usage. For example, encounters among patients with a diagnosis of cancer or congestive heart failure had higher usage rates of both directed and query-based HIE. Similarly, encounters among patients with a stroke diagnosis had higher directed usage, query-based HIE was higher in encounters among patients with an human immunodeficiency virus diagnosis, and a 1-unit increase in comorbidity score increased the usage of directed HIE by nearly 4 percentage points. In terms of individual characteristics, encounters with nurse practitioners had lower directed exchange use (–2.8 percentage points), but usage of query-based HIE that was 1.5 percentage points higher. The FQHC site’s average directed message volume each month and the average number of query sessions, the measures of the technology domain, were not consistently associated with usage. Usage of query-based HIE was the exception, which was 3 percentage points higher in months when the FQHC had higher overall rates of query-based usage (see Supplementary Appendix for full results).

Figure 1.

Associations among selected individual, task, and technology factors and type of health information exchange usage. HIV: human immunodeficiency virus.

Qualitative findings

All 8 key informants identified multiple situations in which receipt of information via directed HIE had prompted their usage of the query-based system. Query-based HIE was used in response to all 3 of the directed types of HIE. A commonly reported reason was the need for additional information, since clinical documents received via directed exchange do not always include all relevant information. A nurse case manager noted how the need for additional information resulted in query-based usage:

Sometimes the [discharge summaries] don't have everything…[the discharge summary] will reference a consultation report that they didn't send and I'll go out and get that. Sometimes there will be labs pending and by the time I get [the discharge summary] the labs are complete and I'll have to go out on the RHIO to get it.

Similarly, a nurse in charge of managing diagnostic imaging results noted how a colonoscopy report received via directed HIE would often be missing the associated pathology report (considered a laboratory report), which would prompt her to use the query-based exchange. Other end users described similar situations of seeking additional information to account for incomplete laboratory results or unavailable imaging reports, or to see related progress notes.

A transition of care could also serve as a prompt for query-based usage. Discharge summaries could result in query-based usage to obtain diagnostic results, medication lists, and patient contact information (if their own information was out of date) in preparation for a postdischarge appointment. Additionally, even the receipt of diagnostic results could inform nurses and case managers of hospitalization or emergency department visits, of which the FQHCs were previously unaware. A nurse reported:

“Oh, this patient had some labs done not ordered by us. I wonder what's going on?” So we'll look and find they were seen in the emergency room two days ago and we don't have a report from the hospital yet. So that'll prompt us to go and get more information.

A third reason repeatedly noted for accessing the query-based system was the receipt of abnormal diagnostic results. A physician and a nurse case manager both reported that abnormal laboratory or diagnostic imaging results would prompt searches for prior encounters, procedures, and results. As a nurse recounted:

There was an abnormal lab and [the physician] wanted me to figure out why. Where did it come from? What's going on? Were they hospitalized? What happened?

DISCUSSION

Quantitative and qualitative findings suggest that directed and query-based HIE exist in a complementary manner in ambulatory care settings. Directed HIE usage was the much more commonly used approach. However, directed HIE usage was associated with an increase in the usage of query-based HIE. Although query-based HIE can supply the additional information not always available from directed HIE, current U.S. policy does not require EHR features and functionality that enable the effective usage of both approaches to information exchange.

In our adjusted models, directed HIE for imaging information and for clinical documents were associated with usage of query-based HIE. The qualitative interviews provide some insights to the possible mechanisms underlying these associations. For example, clinical documents shared via directed exchange do not always include all the information necessary for end users or the documents could signal a transition in care. Both instances could prompt additional information seeking. Likewise, imaging results could also prompt queries for additional information or related results. Prior qualitative studies of HIE users tend to support the experiences of the key informants in this study, noted previously. For example, absence of available laboratory reports can prompt query-based exchange,¹¹ and incomplete information in one system can drive users to obtain needed information in other ways.^23,24 Understanding how different approaches work together in such examples is relevant to the HIE use cases of avoidance of repeat or potentially unnecessary imaging,^25–28 and support for transitions of care.²⁹

In contrast, after controlling for other factors, usage of directed exchange for laboratory information was not associated with query-based usage. Access of additional information from query-based HIE may not be very common, given that the large proportion of routine laboratory tests are not markedly abnormal^30,31 and that EHRs are better able to integrate and display, in a longitudinal context, the structured and quantitative laboratory results.¹¹ Nevertheless, qualitative interviews suggested use of query-based HIE might occur due to directed HIE laboratory results.

Evidence of complementary usage in primary care settings is striking given that much of the negative perceptions around HIE have been in the query-based HIE context³² and that providers may have more positive perceptions of directed exchange.¹ In addition, directed exchange has been portrayed as the “easier” of the 2 approaches,⁷ which may be the case as directed HIE’s better integration into EHRs alleviated many of the workflow barriers associated with query-based portal usage.¹⁴ Despite these perceptions and workflows, continued and even complementary usage strongly suggests that query-based exchange provides additional value to end users. Given this complementary usage, it is incumbent on health care organizations utilizing HIE, and Health Information Organization providing HIE services, to provide the functionality to effectively utilize both approaches.

The evidence of the complementary nature of directed and query-based HIE highlights a potential U.S. health information technology policy shortcoming. If the prevailing view is that both work together,^2,3,7,9,10 if both are beneficial to patient care,¹³ and if use of one approach prompts usage of the other, then both directed and query-based HIE functionalities should be incorporated into EHR Certification Criteria. However, that has not been the case. The Meaningful Use Program (now the Promoting Interoperability Programs),³³ which has been the most significant driver and definer of interoperable health information technologies for nearly a decade,³⁴ only includes directed exchange requirements. Without specific criteria that support query-based exchange, such as required single sign-on or health information organization exchange participation, it is not surprising that query-based exchange usage is much less common. Given the financial importance of EHR certification, it is not surprising that nonrequired features, like query-based HIE, received much less attention and investment from vendors.^35,36 Better integration of EHRs and query-based exchange platforms exist, but that was accomplished in the absence of broad policy support.

Additionally, these findings provide more detailed insights into the usage of HIE in ambulatory care settings. As has been seen in prior studies, query-based HIE was used in the minority of patient encounters.³⁷ Again, the absence of a single sign-on and a separate portal outside the EHR environment likely inhibits some usage.³⁸ However, the finding that directed HIE was used in one-quarter of all encounters is new and comparable with estimates of how often patient information is unavailable to providers, such as in Smith et al³⁹ and Tham et al,⁴⁰. Also consistent with the existing literature^15,38,41 was the repeatedly indicated relationship between patient complexity and both directed and query-based HIE usage, whether measured by a comorbidity index, presence of specific chronic conditions, or previous utilization. Notably, the large marginal effects associated with an encounter that occurred postdischarge or after an emergency department visit indicate the use of HIE to support transitions of care.

Limitations

This study has several limitations inherent in log file-based analyses. We cannot comment on the need, application, or successful retrieval of specific data elements. We do know which specific data elements or information gaps may have motivated individual usage of either directed or query-based HIE. Likewise, we do not know if usage of either directed or query-based HIE was successful in meeting the end user’s information needs or what (if any) data elements obtained through HIE were actually applied to the delivery of care. However, from our qualitative interviews, we know that end users sought additional information on diagnostic results, encounter history, procedure history, and medications, in addition to patient contact information. These findings emphasize that future work on the motivation behind information seeking through HIE and the success of searches will need to be attentive to the multiple options available to providers. Additionally, while the study sample included multiple FQHCs, generalizability may be limited because each had the same EHR vendor and participated in the same health information organization. Importantly, our study focused on the individual end user. Individual usage is different than organizational adoption and it is possible that while individual end users apply both directed and query-based HIE in a complementary manner, organizations may treat the 2 as substitutable technology in adoption decisions. Last, we cannot say how much usage of each type would occur in the absence of the other.

CONCLUSION

Ambulatory care providers utilize directed and query-based HIE in a complementary manner. Health information organizations, EHR vendors, and policymakers should support utilization of both approaches to information exchange.

FUNDING

This work was supported by the Agency for Healthcare Research and Quality grant number 1R01HS024556-01A1 (PI: JRV). This research was supported in part by Lilly Endowment, Inc, through its support for the Indiana University Pervasive Technology Institute. This material is based on work supported by the National Science Foundation under Grant No. CNS-0521433.

AUTHOR CONTRIBUTIONS

JRV, MAU, LPC, and JSS conceived the research study. JRV and MAU led the data analysis. JRV, MAU, LPC, and JSS drafted the manuscript and reviewed for critical content.

ETHICS APPROVAL

This study was approved by the Indiana University Institutional Review Board.