Prospects for Care Coordination Measurement Using Electronic Data Sources

Challenges of Measuring Care Coordination Using Electronic Data and Recommendations to Address Those Challenges

Fulfilling the promise of detecting electronically how well care is coordinated requires an understanding of current challenges that our health system needs to address. Panelists identified a number of challenges in using electronic data for care coordination measurement, which we summarize into six key challenge areas:

  1. Underutilization of health IT system capabilities and clinical workflow barriers.
  2. Lack of data standardization and limited health IT system interoperability.
  3. Unknown clinical data quality in electronic data sources.
  4. Limitations in linking data.
  5. Technical hurdles to accessing data.
  6. Business models that facilitate competition rather than cooperation.

Panelists also discussed ways to address many of these challenges. At the end of each of the six challenge sections, we summarize recommendations that stem from these discussions. These recommendations are based primarily on suggestions from our panel of experts, but also include our own evaluation of promising approaches based on insight from discussion with panelists. Some recommendations are meant specifically for Federal agencies, such as AHRQ, while others are applicable to a wide range of stakeholders within this field, including researchers, measure developers, health IT systems vendors, health care delivery organizations, or systems administrators.

Return to Contents


Key Challenge Area 1: Underutilization of Health IT System Capabilities and Clinical Workflow Barriers

Panelists noted that clinicians generally are not using EHRs to their full capacity. Although features are present in many systems that could make more data available for quality measurement, panelists felt that these are often underutilized.

Challenge 1a: Limited Availability of Structured Data

Panelists identified the predominance of unstructured data in clinical documentation as a key challenge in using information from EHRs for quality measurement. Structured data are contained within specific data fields that specify the type and format of recorded information. For example, a field for "weight" might specify that the information be recorded in kilograms. Unstructured data, in contrast, are generally recorded as free text, with no limitations in the format and often without clear specification of the type of information recorded in a particular location. Progress notes are a common example of unstructured data within EHRs.

Although some research has explored extracting information from free-text sources using a process called natural language processing, panelists agreed that quality measurement using information recorded in unstructured free-text format is not feasible at this time, nor likely to be in the near future. One panelist mentioned that his research team's attempts to extract information from free-text notes using natural language processing failed. They have turned their efforts instead towards developing standards for recording key information in a structured way.

While emphasizing the limits of unstructured data for quality measurement, panelists recognized that medical practice continues to rely heavily on unstructured data and likely will do so for some time. They attributed this to workflow practices that have not yet evolved from the traditional reliance on paper-based documentation. One panelist commented that frequently EHRs are used as elaborate text editors rather than in ways that exploit their database qualities, which is perhaps unsurprising given that clinical documentation has traditionally focused on text rather than structured data fields. Panelists also attributed continued reliance on unstructured documentation to the complexity and nuances of medical practice, which in turn requires a flexible workflow. Panelists did not discuss explicitly whether current EHR systems have abilities to capture these dynamic workflow processes in a structured form, but the absence of such examples likely means that most systems do not have such functionality at this point.

Several panelists emphasized that simply building more structured data fields into EHR systems is unlikely to increase the availability of structured data. They noted that even when structured data fields are available, they are not always used by clinicians, because entering data in a structured format requires a different workflow from entering free text, as has traditionally been done. They highlighted the need to provide incentives to motivate use of structured data fields as part of clinical workflow. Others mentioned the need to educate physicians about the utility of recording information in structured format. Panelists observed that when the advantages of recording structured data are explained—or, better yet, demonstrated—clinicians are more willing to spend the time entering data into structured fields for the sake of having better information at the point of care. In addition to quality measurement, panelists mentioned facilitating information sharing and decision support as advantages of recording structured data. The information sharing advantage also sets up an opportunity for improving care coordination, as well as measuring the presence of best practices in transmission and receipt of necessary information.

Challenge 1b: Health IT Systems Design Reflects Current Workflow

Another challenge identified by panelists was poor documentation of many processes important to care coordination. Panelists emphasized that given the many demands on clinicians' time, only information that is perceived to be critical to patient care delivery or motivated by reimbursement policy is typically recorded. No examples were provided about EHRs that support capturing documentation of coordination activities, although one panelist noted on-going efforts develop such capability within a system used by community health teams. The lack of such capability in many current systems further limits the availability of data on coordination activities. Panelists' discussions suggest that the attention to date seems to be more focused on recording clinical activities during a patient encounter, as opposed to coordination activities needed and performed by teams of clinicians and supporting staff over time and across settings.

Panelists noted that, historically, design of EHRs has been driven by requests from clinicians—primarily physicians—and that EHRs are designed to be customizable to match local workflow patterns. They noted that this is shifting somewhat with recent efforts to develop requirements for EHR certification related to Meaningful Use. However, panelists emphasized that the challenge remains in creating demand among clinicians and health IT system purchasers for features, such as structured data fields and population management functionality, that would enable easier or richer quality measurement. One panelist characterized the problem as getting health IT users to "ask for what they need."

Challenge 1c: Barriers Related to Care Plans

Much of panel's discussion of workflow barriers occurred in the context of care plans. Care plans are a particular concept of interest to care coordination measurement not only because they have been proposed as a potential means of coordinating care (e.g., sharing information among providers in a structured way), but also because a comprehensive care plan has the potential to serve as documentation for many other aspects of care coordination that would be of interest for measurement, such as assessing needs and goals, supporting self-management goals, and establishing accountability or negotiating responsibility. Thus, data that would enable assessment of both the presence of a care plan and evaluation of care plan content are of particular interest for care coordination measurement.

While care plans are not inherent in health IT, panelists' comments often focused on the concept of care plans as it relates to care coordination. Panelists emphasized the lack of consensus in the clinical community about what constitutes a care plan. Panelists with a clinical background agreed that, in their experience, care plans are not typically used in ambulatory settings and in the inpatient setting are generally developed and used only by nurses. Several panelists noted that inpatient nurse care plans are not usually used by physicians and are not transferred out of the hospital when patients are discharged. One panelist commented that because care planning is seen in present clinical culture as a nursing task, a reframing of the concept may be necessary to get buy-in from some physicians about its importance. Others noted that evidence linking the use of care plans with improved patient outcomes would help support their use.

The panel discussed the continued ambiguity surrounding care plans, and how that ambiguity impacts the potential to measure aspects of care coordination related to care planning. Two panelists noted that although some of the data elements included on a list of care plan elements recommended by the National Quality Forum (NQF)5 are usually contained in EHRs in structured format (e.g., medication list, problem list, follow-up appointments, and presence of an advanced directive), they would not typically be grouped in a single location within an EHR.

Using a hypothetical measure (go to the call agenda, Appendix B, first measure listed) as a starting place for discussion, participants debated whether the existence and documentation of various elements of a care plan within an EHR would be sufficient to indicate care coordination, or whether such elements must be grouped in a single location within the record or some other cohesive document. One panelist suggested having health IT systems pull together disparate elements to generate a care plan. In contrast, another panelist suggested that the very need to pull elements of a care plan together from scattered locations throughout a record, or from multiple records (i.e., from specialists, primary care practices (PCPs), or hospitals) rather than finding elements in a single location within a single system, would be a potential indicator of poor coordination.

There was some agreement that a more cohesive plan—one that contains deliberately collected information in a single location—was more indicative of coordination. But panelists noted that few EHRs record information in a way that reflects a discrete care plan, even though some elements of care plans might be located within the record. Panelists also noted that many elements discussed as part of a care plan are not recorded in a standard way, or in a standard location, among different EHRs. Even with a clear specification of data elements required for a care plan, one panelist with experience working for a health IT vendor noted difficulties in integrating care plans into EHRs. Typical EHR design would require creating a care plan for each problem in patients with multiple problems, a result that undermines the utility and intent of care plans for use in care coordination. Integrating patient input into care plans, an important aspect of care coordination, was noted as an additional challenge.

One panelist emphasized that a clear concept of care plans is needed within Health Level 7 (HL7) or another health IT standard in order to provide direction to vendors to create the capability of recording care plans within EHRs. This panelist noted that HL7 has been working on this for some time in a particular work group, but as yet those efforts have not been reported.

The currently recommended Meaningful Use Stage II measure related to care plans specifies only that care plan fields (undefined beyond "treatment goals and patient instructions") be recorded, but does not require that those fields be grouped in a single location or document within the EHR. Currently, there does not appear to be clinical demand for a single care plan location or document within EHRs, suggesting that further changes in clinical practice and work flow would be needed before this would become a common feature of EHRs.

Challenge 1d: Underrepresentation of nonphysician viewpoints in health IT system design

One panelist commented that nursing and social work viewpoints are under-represented in development of EHR content, which has important implications for the kinds of coordination-related information represented in EHRs and highlighted within continuity of care messages (see section on health IT system interoperability). This panelist felt that this was particularly problematic for care coordination measurement, given that many elements of coordination, such as consideration of patient preferences and goals, social and environmental factors, and patient or family needs for support, have traditionally fallen within the scope of nursing, social work, home health, and care management practice. Another informant noted that, to date, both policy and financial support is lacking to encourage health IT-enabled collaboration between social services, case managers, and community-based support organizations with their health care professional counterparts, which further limits opportunities for interdisciplinary care coordination and availability of data on such collaborations when they occur.

Challenge 1e: Clinical workflow and technology integration issues

Overall, discussions with panelists highlighted the importance of considering how EHR users interact with and use their systems when designing any measures that rely on data from such systems. They emphasized that the technological capabilities of systems are less important than their day-to-day usage. Although technological advances and improvements in data standardization may help facilitate quality measurement using health IT systems data, ultimately some evolution in clinical practice patterns, workflow, documentation habits, and demand for EHR features is likely needed before the full richness of information contained in health IT systems can be tapped for quality measurement. One panelist with both clinical and vendor background emphasized that designing EHR technology must balance meeting workflow needs of clinician users, which often requires system customization, with the need for standardizing data and workflows in order to enable additional use of clinical information, such as quality measurement, decision support, and health information exchange.

Recommendations to Address Key Challenge Area 1: Underutilization of Health IT System Capabilities and Workflow Barriers

Panelists suggested several ways to overcome the challenge of clinicians' underutilization of health IT system capabilities and, in particular, continued reliance on free text for documentation. Incorporating structured data fields into EHR systems for key clinical concepts is an important first step in addressing this challenge, but panelists agreed that building structured data fields alone is insufficient to change documentation practices. Changing clinician EHR usage practices is likely to be a slow and difficult process because it requires changes to workflow. Panelists noted that success depends on having a strong business or clinical case (e.g., clear utility of structured data at the point of care, incentives for recording structured data) for making the workflow change, and/or for the consistent commitment of health care practice leadership.

Several panelists discussed ways to motivate clinicians to use structured data fields. All agreed on the need to demonstrate the utility of structured data. Their recommendations included:

  • Align structured data fields with decision support tools. When clinicians find decision support useful, they will see value in taking the time to input the structured data needed for the decision support algorithm.
  • Create protocols for non-physician clinical or support staff to enter some pieces of information in structured fields. Strategies that ease the burden of work on physicians generally increase success in achieving changes in workflow practices.
  • Explain—and, better yet, demonstrate—how structured data improves care coordination and ultimately care delivery. For example, through better patient monitoring and follow-up or by facilitating information exchange that provides additional information at the point of care.
  • To strengthen the above recommendation, provide evidence about how care coordination improves patient outcomes or reduces costs. Evidence will help drive reimbursement of coordination activities, particularly in accountable care organizations, medical homes, and other alternative health care delivery models. It will also demonstrate to clinicians what aspects of coordination are most important to patient health, satisfaction, and quality of life.
  • Make data from structured fields readily available for quality improvement evaluations. This strategy requires buy-in from clinicians that quality improvement is a priority.
  • When clear coding standards are lacking, to the extent possible, align measure specifications with existing guidelines or elements of other quality measures. This will increase the utility of specific data elements, creating a stronger business case for building the data field into an EHR system and actually populating the field during clinical workflow.

Panelists also highlighted that care coordination often depends on team work and dynamic workflows by many health care professionals and that vendors and delivery systems need to design health IT functionality to capture coordination activities more explicitly, both to support team practice and to measure the extent of these activities.

Return to Contents


Key Challenge Area 2: Lack of Data Standardization and Limited Health IT System Interoperability

Another key challenge identified by panelists is lack of standardization in how data are recorded. Standardization refers to the vocabulary or code set used to record the content of information. For example, some might record weight in kilograms, while others record weight in pounds. Even with structured data that allows querying for a particular data field such as weight (as described previously), standardization in data collection is a prerequisite for comparable information across systems. In addition to impeding reliable quality measurement, lack of standardization limits the ability to share information across systems (interoperability), which in turn limits opportunities for care coordination and coordination measurement.

Challenge 2a: Lack of Standardization

To enable quality measurement, information about particular concepts included in the measure definition (data elements) must be standardized. This ensures that a measure calculated from one site means the same thing as the measure calculated from a different site. Two levels of standardization are important to consider for quality measurement using electronic data: (1) presence of a standard code set, and (2) widespread use of the standard.

A recurring theme during discussions with panelists was inconsistency in how clinical information is coded within health IT systems. While standard code sets exist for some particular kinds of information, panelists commented repeatedly that systems vary widely in whether or how those code sets are used, and many systems use code sets developed in-house. The result is lack of an industry standard for how many kinds of clinical information is coded.

Panelists noted several kinds of data of particular interest for measuring coordination (as well as for quality measurement of other clinical concepts) where standard code sets exist, but are not used consistently within the health IT industry today:

  • Laboratory results—Several panelists commented on inconsistencies in how laboratory test and results information are coded in health IT systems. They noted that, although a standard code set exists for lab results—the Logical Observation Identifiers Names and Codes (LOINC) standard—many labs, particularly those within hospitals, continue to use alternative code sets. Many of these alternative code sets were developed in-house and are used only in a single hospital or health system. One panelist noted that this is particularly true of health care organizations that were early adopters of health IT systems, who often developed local coding schemes in the absence of well-established standards. The EHR certification requirements, established by ONC in July 2010, now require that when certified EHRs receive lab results that are coded in LOINC format, they must use LOINC codes when transmitting that information as part of other certification requirements, such as providing clinical summaries or electronic health information to patients.6 To the extent that laboratories report results using LOINC, this new requirement should increase the availability of standardized lab data. But currently there are no requirements that labs report results using LOINC. Panelists agreed that, among systems currently in use, coding of lab results lacks uniformity.
  • Medication/drug information—Several coding systems are available for recording information about medications, each with varying degrees of granularity. One panelist with experience using medication data for research explained that the inpatient setting typically codes medication information using the RxNorm code set (maintained by the National Library of Medicine), while the ambulatory setting typically uses the Food and Drug Administration's National Drug Code (NDC) Directory. However, ambulatory EHRs are expected to shift to using RxNorm in response to the Meaningful Use final rule and EHR certification standards, released in July 2010, which specify that only RxNorm may be used for certifying interoperability of EHR systems. The panelist also noted that, adding further complexity to medication information, NDC codes as used in ambulatory EHRs are not equivalent to those used in pharmacies due to different needs for drug specificity. For example, when prescribing a medication, a primary care provider would choose one of possibly many different generic versions of the drug. When submitted to the pharmacy, this order could be filled using any of the various generic choices, each with its own NDC code. Thus, the NDC code prescribed would not necessarily match the NDC code dispensed, even though the medication received was the same as that prescribed by the physician. Any attempts to link data from ambulatory systems and pharmacies will need to account for this discrepancy.
  • Diagnoses and clinical observations—Variation exists in the systems used to encode clinical concepts, such as diagnoses, within health IT systems. This variation stems partially from system customization that takes place during implementation. Although panelists could not offer information on the frequency with which different code sets are used, two standards in particular were mentioned: ICD-9-CM (and the forthcoming ICD-10-CMix) and SNOMED-CT. ICD-9-CM is one standard used for coding claims data, but outside of reimbursement, it is also used for coding information on diagnoses in some health IT systems. SNOMED-CT is a system that allows coding of a wide variety of clinical concepts beyond diagnoses, which has been recognized as an advantage for use in point-of-care systems such as EHRs. The ONC Health IT Standards Committee has endorsed recommendations that EHR systems record clinical observations using SNOMED-CT by 2015. Currently, certified ambulatory EHR systems must record problem lists using either ICD-9-CM (or ICD-10-CM after 2013) or SNOMED-CT. ONC is continuing to develop final certification requirements.

Inconsistent coding of data requires mapping codes between systems. For example, one panelist mentioned that major lab companies often pay to have their coding systems mapped onto code sets used by major EHR vendors to enable electronic delivery of lab results. However, the cost of this mapping is eventually passed on to EHR users as part of the cost of synchronizing their system with labs. Another panelist noted that most hospitals participating in a health information exchange, particularly those that were early adopters of health IT, must map their internal lab coding system onto the standard used by the HIE. This kind of mapping requires significant time and resources, and adds an additional barrier to information exchange (see next section on interoperability). The panelist noted that depending on its extent and the level of resources committed by the local site, data mapping can take a year or more to complete. Highly accurate mapping is essential whenever exchanged information will be used at the point of care. Another panelist, who designed a database that uses information from a wide range of practices and thus a variety of EHR systems, also noted the need to map EHR data onto the coding systems used by the database. This adds an additional resource burden to sites that want to contribute data to the database. Performing such mapping often requires clinical judgment, particularly when coding systems vary significantly in their structure. This element of judgment can have important implications for interpretation of any measures based on such data.

In contrast, when variation in use of standards is restricted to a limited set of established code sets, as is the case for medication information, the resource burden associated with data mapping is not as significant. For example, cross-walks mapping many drug code sets to RxNorm are already available and the need for mapping between systems should decrease as Meaningful Use and EHR certification requirements increase standardization across health IT systems. Thus, although use of different code sets for drug information adds complexity to use of any medication information linked across systems, panelists did not suggest that this would hamper measurement using medication information.

Other concepts of interest for care coordination lack any established standard for how the information should be recorded or coded. Some examples noted by panelists include:

  • Patient needs and goals.
  • Quality of life.
  • Referrals.
  • Care plans.
  • Self-management plans, goals or supports.
  • Mental health information, such as thoughts of suicide.
  • Tobacco or alcohol use.
  • Environmental and social factors impacting health.

Although sometimes recorded as text within structured data fields, this information is most often included in EHRs as free text within notes, or not available in any documented form. As noted previously, panelists agreed that quality measurement using unstructured data is unlikely to be feasible in the near-term.

Challenge 2b: Limited Health IT System Interoperability

Whether due to incomplete use of existing standards or absence of standards, lack of standardization impacts the ability to share information across systems, termed interoperability. This lack of interoperability increases the resources required to carry out care coordination, or in some cases limits coordination altogether, and, by extension, limits the ability to measure coordination.

Panelists agreed that limited interoperability remains a major hurdle in the exchange of information across health care entities, and thus in the development of health information exchanges, patient registries, and integration of outside information into EHRs. One panelist commented that, to date, interoperability has been more of a promise than a reality. Panelists noted that a major barrier to interoperability is variation in how health IT standards are implemented.

Health Level 7 (HL7) is an international organization that focuses on developing standards for interoperability of health IT systems. While widely applied, we heard repeatedly from panelists that HL7 standards are really guidelines, designed to be highly flexible and customized to local workflows and clinical needs. While this adaptability benefits end-users who are able to customize products to match their practice patterns and workflows (thereby easing health IT adoption), it also leads to widespread variation in how standards are implemented. This variation limits interoperability. Panelists noted that this variation would likewise hamper efforts to use data elements from within EHRs because the way in which that information is coded and structured will vary for each site, even when based on the same standard. One panelist termed this the "Baskin Robbins" problem, because there are too many flavors of the same basic guideline. Another panelists commented that the ability to customize implementation (indeed, the requirement to do so in the absence of a clearly defined standard) slows the progress of standardization.

A component of the HL7 standard frequently noted as promising for care coordination measurement is the Continuity of Care Document (CCD), a standard for transferring information between health IT systems during care transitions, such as hospital discharge or transitions between outpatient practices. The CCD identifies the types of information being transmitted, such as problems or diagnoses, medications, family and social history, procedures, and a plan of care. Ideally, information contained within these and other sections of the CCD should be recognized as such by any receiving system capable of reading a CCD and then be integrated into the receiving system in the appropriate location. A related standard for information transmission, the Continuity of Care Record (CCR) developed by ASTM (formerly American Society for Testing and Materials), contains similar data fields and is used in a similar way.

Although CCD and CCR standards at their current state of implementation are promising for care coordination measurement (e.g., confirming information transfer at transitions or transfer of specific information between health care entities), panelists noted that they have several limitations of. They pointed to lack of consensus within the health IT industry about whether to use the CCD or CCR as the standard for transmitting information during transitions of care. Though very similar in content, the CCD and CCR vary in technical details, and many health IT systems are capable of sending and receiving information in only one format or the other, further limiting interoperability among health IT systems. Panelists noted that efforts are underway to harmonize the CCD and CCR standards through HL7 Clinical Data Architecture consolidation guides, but these efforts are still evolving. One panelist commented that most EHRs could not incorporate a CCD message without significant customization.

In addition to interoperability limitations, CCD and CCR standards specify the type of information transmitted between systems, but do not address how information within each section (e.g., medications) is recorded. As noted above in the discussion on data standardization, variation in how information is coded and structured is one of the core problems hampering interoperability. Furthermore, both CCDs and CCRs allow inclusion of free text, which is not readily usable for quality measurement.

Finally, one clinician noted that interoperability can also be limited within health systems due to use of distinct health IT platforms for different parts of the system. For example, separate databases and software interfaces may be used for the laboratory, radiology department, and social workers. Such duplicity further complicates interoperability within and between health systems.


Recommendations to Address Key Challenge Area 2: Lack of Standardization and Limited Health IT System Interoperability

Panelists were optimistic that standardization will improve significantly in the coming years, particularly in response to the Meaningful Use initiative. They expected that improvements in standardization of data would also improve system interoperability. Panelists agreed that leveraging the Meaningful Use initiative is the most promising way to further enhance standardization in the near term. However, they recommended a number of additional ways to address this challenge:

  • Continue Federal support to develop standards, both in areas where standards are undeveloped, and by motivating adoption of existing standards through incentive programs. The ongoing development of Stage II and Stage III Meaningful Use measures provides insight into areas where more standardization is likely to emerge within the industry.
  • Align other measurement and payment incentive initiatives (e.g., from accountable care organizations, medical homes, or other sources from the Centers for Medicare & Medicaid Services (CMS) with key standards gaps, such as coding of lab results and medication information.
  • Use financial incentives or other means to encourage laboratories to report results using LOINC codes to align with the EHR certification requirement that systems transmit results in LOINC when they are received in this format. Such encouragement would help increase penetration of the LOINC standard throughout the health care system.
  • Develop well-defined measure concepts that will give vendors, EHR users, and HIE administrators clearly defined data elements to build into systems. The NQF Quality Data Model (QDM) will be an important tool for this objective (go to discussion of the QDM in Key Challenge Area 3 and Appendix D).

ix. The conversion from ICD-9-CM to ICD-10-CM in the U.S. within the next few years is widely expected to present a challenge to health systems.

Return to Contents
Proceed to Next Section

Page last reviewed July 2018
Page originally created September 2012
Internet Citation: Challenges of Measuring Care Coordination Using Electronic Data and Recommendations to Address Those Challenges. Content last reviewed July 2018. Agency for Healthcare Research and Quality, Rockville, MD.
Back To Top