Is P4P Possible Without an EHR?

February 8, 2008
MDNG Psychiatry, December 2007, Volume 9, Issue 8

Pay-for-performance programs, in which providers are reimbursed at a higher rate if they can demonstrate that they deliver high-quality services, are being introduced in most states as part of an...

Pay-for-performance programs, in which providers are reimbursed at a higher rate if they can demonstrate that they deliver high-quality services, are being introduced in most states as part of an aggressive attempt at controlling healthcare costs without adversely affecting the quality of care. Indeed, many of these programs promise better quality as a result of being able to measure key indicators that are thought to drive clinical outcomes, as well as patients’ perception of the level of care.

At issue, however, is whether these programs can be effectively put in place ahead of the installation of complete electronic health records (EHRs). Many in the industry argue that, without an EHR, it will be impossible to establish controlled processes for the measurement of these indicators in a reliable manner, such that the measurements are truly comparable across providers, because the data necessary for these calculations will only be available after the problem of EHR

interoperability has been addressed. Others argue that P4P programs can be established, but they would consist of weak indicators with all the bureaucratic limitations of HEDIS-like calculations, citing similar data-availability issues.

This article will argue that a robust P4P program—with meaningful indicators and an affordable mechanism for auditing and verifying the measurements—can indeed be established in advance of universal availability of EHR technology. In fact, the greatest barriers to such a program are not rooted in the availability of data, but rather in the adoption of regional data-sharing agreements among all the industry players.


The healthcare industry, which accounts for approximately 15% of the US gross national product, is poised for a radical transformation aimed at reducing the rate of growth of its costs and at increasing the quality and reliability of its services. The industry is universally perceived to have high costs, and the traditional view that high costs correlate with high medical quality are being challenged by a crude but effective first pass at the measurement of healthcare outcomes—which suggests otherwise. Driving this transformation is the work of a coalition of large payers (including the employers and the government), insurers, and providers, and its primary vehicle will be an overhauled payment system for healthcare services.

In order to complete the transformation, the industry must be able to measure the “value” and the cost of the healthcare delivered. These measures must be objectively derived from information in the medical record, and must be comparable across providers. The cost of obtaining the information used in the measures would, ideally, not significantly exceed the costs already incurred in documentation of the medical record, or else the measurement process will itself make

healthcare less affordable.

The EHR does not help much

We will start our argument by debunking the idea that the EHR will make the task of creating a robust P4P program much easier. Everyone involved in the measurement of healthcare agrees that coordinating the efforts necessary to create such a program is a difficult task. But how would an EHR simplify this?

Claim #1: EHRs make available all the data that is needed to calculate the P4P measures.

This is patently untrue. Most of the data elements necessary to calculate clinical P4P measures are already available in other electronic systems that are much more common than EHRs. Lab values are available from lab systems and lab vendors. Demographics, procedures, and diagnoses are available from billing and registration systems. Pharmacy data is available within the dispensing industry and from the payers (insurance companies and the government).

Visit-utilization information is available from scheduling and billing systems. While the EHR contains additional clinical information, much of it is in text form (progress notes, for example) and is not readily of use in refining a P4P program.

Claim #2: EHRs will make the data available from a single source, rather than from disparate systems, therefore making the task easier because all the data for a single individual is tied together under a single ID, whereas separate systems might not share the same identifier for a unique patient. At first glance, this seems to make sense—after all, no one wants to manage the difficult task of matching a lab value to a patient record without an identifier. However, let us remember how the data in the EHR was brought together under a single identifier to begin with: by tackling the difficult question of providing a unique ID to each patient and requiring your lab vendor, for example, to include that identifier in the HL7 transaction that brings back the result. Therefore, an EHR is not necessary in order to ensure a common identifier across data systems of the same provider organization. By contrast, external data that does not carry the unique identifier of a provider organization is equally difficult to integrate, whether one is working with an EHR system or with separate systems. In other words, the industry must solve this problem whether or not we fi rst install EHRs everywhere. Master Patient Index functionality is available in most commercial registration systems, and can also be utilized at the level of Regional Health Information Organizations (RHIOs) in order to overcome this problem.

Claim #3: EHRs contain additional clinical information that allows the calculation of measures that are more accurate than those calculated using only “administrative” data. For example, EHRs contain “problem lists” populated by the physicians, which may be a more accurate way of identifying a diabetic than by using the ICD9 diagnosis code in an encounter, which may actually be indicating a “rule-out” of the disease. Additionally, a provider can indicate valid reasons that disqualify a patient from a particular protocol that is used in a measure (for example, allergy to a medication or a “not-medically-necessary” situation such as that of a 95-year-old diabetic with terminal cancer of whom the provider probably won’t require bi-annual glyco-hemoglobin testing).

This is the most valid claim. It appeals to our distrust of early HEDIS-type measures that inaccurately identified diabetics, or the possible individual contraindications to receiving aspirin post myocardial infarction. However, the first concern has been demonstrated to be false by the local successes involving RHIOs or other community registries that collectively pool data from multiple providers and correctly maintain accurate diabetic registries without relying on a problem list. Instead, these systems identify possible diabetics by looking at multiple possible indicators (ICD9, lab values, insulin-type medications) with greater accuracy and then making those lists available to the providers for verification—a process we at Palo Alto Medical Foundation have discovered to be much more accurate than relying on the provider to add the diagnosis to the problem list of an EHR, which shows greater variability across providers who are comfortable or facile with the complexity of a full EHR. The second concern, contraindications or allergies, is real, but hardly a reason to wait years for EHRs to be available universally rather than adding this capability to regional integrators of data.

If not EHRs, then what?

Having asserted that the EHR does not add much that is not already available, we must propose an alternative mechanism for the support of P4P programs. This approach must maximize the use of available information and leapfrog some of the challenges presented by the lack of a universal patient identifier. It must then perform two distinctly separate but complementary tasks. First, it must gather the data from every available electronic source and perform systematic cleanup functions. Second, it must be able to calculate hundreds of clinical, operational, and financial (efficiency) indicators in a transparent and easily auditable manner. We have been using the term “tactical warehouse” to refer to such systems. Many versions of such systems have arisen from local efforts of large integrated systems, medical groups, RHIOs, or other large collaborative efforts. Although some lack flexibility or a robust architecture, others represent the birth of a new class of healthcare IT that promises to address not only P4P needs, but also the portability of essential elements of our medical records.

The Palo Alto Medical Foundation tactical data warehouse, called Solutions, consists of three engines:

The Harvesting Engine gathers data from any available source system. It contains data cleansing software that was designed based on the premise that most important measures in healthcare come from relatively few fields (see the “Pareto rule”). It uses extraction technology (not the more complicated live interface) to access data regardless of system brand. It draws data from systems including EHRs, lab systems, registration, billing, general ledger and payroll, etc. And it houses a simple Master Patient Index tool that facilitates the creation of a tree of patient identifi ers that ties a unique patient across source systems. Its output is a clean, simplified EHR-like set of tables containing the fields that are required for most P4P calculations. Lab tests and other elements have been merged such that, for example, all fasting glucose tests are identified by a single code.

The Calculation Engine applies the definition of each measure to create “metric tables,” which are refreshed weekly, monthly, or quarterly as required by each metric. In the process, this engine generates useful “value-added” tables such as disease rosters and patient activity tables that create a rich, standardized

environment ripe for the easy generation of ad hoc queries.

The Presentation Engine catalogs the measures and makes them available to the users singly (for use in products such as dashboards and scorecards), in data series that represent the value of a metric over a period (used in run charts), or in series representing value ranges (for histograms). We maintain more than 150 metrics and their 1,100 variants in a catalog that includes clinical (eg, percent of diabetics with HgA1C <7, by provider), operational (eg, panel size or see-your-own ratios, by provider), and financial (eg, direct cost per practice RVU, by department measures, along with their “roll-ups”) to departmental, site, and organization levels.

The industry would do well to embrace local and regional efforts toward the creation of such systems by pressuring all players (providers, insurers, pharmacies, lab vendors, etc) to enter into HIPAA-compliant data-sharing agreements that would feed these clearinghouse and data aggregation services. Agencies such as AHRQ, NQF, and others, backed by HRSA and CMS, should initiate aggressive certification services that would enable successful tactical warehouse efforts to offer these calculation services, along with data portability services, with a stamp that guarantees that the calculations conform to registered measure defi nition standards.

If we follow that path, we will enable robust P4P programs which, if aggressively funded to motivate change, just might prevent an explosion of costs that are looming once the “Baby Boom” generation hits the Medicare rolls. And by bypassing the EHR, we will get there five years sooner.

Tomas Moran, MS, is Senior Director of Quality and Planning for the Palo Alto Medical Foundation for Health Care, Research and Education, a not-for-profit pioneer in both multispecialty group practice of medicine and outpatient medicine. The foundation is an affiliate of Sutter Health. To learn more about the organization, visit their website at

Related Content:

Publications | Psychiatry