Your browser doesn't support JavaScript. Please upgrade to a modern browser or enable JavaScript in your existing browser.
Skip Navigation U.S. Department of Health and Human Services www.hhs.gov
Agency for Healthcare Research Quality www.ahrq.gov
Archive print banner

This information is for reference purposes only. It was current when produced and may now be outdated. Archive material is no longer maintained, and some links may not work. Persons with disabilities having difficulty accessing this information should contact us at: https://info.ahrq.gov. Let us know the nature of the problem, the Web address of what you want, and your contact information.

Please go to www.ahrq.gov for current information.



Quality Interagency Coordination (QuIC) Task Force
Return to QuIC Home
About QuIC
Steering Group
Workgroups
Press Releases
Related Links
Site Map
Written Statement

spacer

National Summit on Medical Errors and Patient Safety Research

Panel 4: Reporting Issues and Learning Approaches

Testimony of N. Stephen Ober, M.D., M.B.A President and Chief Executive Officer, Synergy Health Care, Inc., of the Informatics Service Group of QUINTILES Transnational


The first National Summit on Medical Errors and Patient Safety Research was held on September 11, 2000, in Washington, DC. Sponsored by the Quality Interagency Coordination Task Force (QuIC), the Summitís goal was to review the information needs of individuals involved in reducing medical errors and improving patient safety. More importantly, the summit set a coordinated and usable research agenda for the future to answer these identified needs.

Individuals were selected by the Agency for Healthcare Research and Quality (AHRQ) to testify at the summit as members of the witness panels. Each submitted written statements for the record before the event, documenting key issues that they confront with regard to patient safety as well as questions to be researched. Other applicants were invited to submit written statements.

Disclaimer and Copyright Statements


Introduction

Good afternoon. My name is Stephen Ober. I am a physician and President and CEO of Synergy Health Care, a health research and data analytics company headquartered in Waltham, Massachusetts. Synergy is a subsidiary of Quintiles Transnational Corporation, the largest contract pharmaceutical organization (CPO) in the world and a leader in healthcare informatics services. I am very pleased to participate in today's critical discussion of the need to reduce the occurrence of errors in the practice of medicine and to improve the safety of our patients.

As illustrated in the Institute of Medicine report, To Err Is Human, current medical error reporting systems focus on the hospital inpatient experience, where data is commonly available in medical records and where errors can have rapid, catastrophic consequences. While often effective in identifying the "root cause" of individual incidents, these passive and reactive systems—by their very design—do not provide an active, prospective approach to universally monitor for medical errors and assure patient safety. And, despite the fact that "more care and increasingly complex care is provided in ambulatory settings," reporting systems that rely on inpatient data provide little insight into either the incidence or the impact of medical errors in the ambulatory care setting.

My testimony today will propose a prototype safety surveillance system that utilizes real-time, de-identified medical and pharmaceutical transaction information to identify the incidence and impact of pharmaceutical and other medical errors in the ambulatory care setting. Such a system would scan millions of patient medical and pharmacy encounters per day, and—with the application of appropriate analytics—provide users with actionable measures toward which to focus their interventions.

Today, information systems that utilize electronic healthcare transaction data are being deployed by providers, payors, manufacturers, and government agencies to perform population-based research, post-marketing surveillance of pharmaceutical products, and ad-hoc marketing studies. When properly designed, these systems are real-time and always up-to-date. These systems can collect and aggregate de-identified data from a variety of sources, and these data can be linked at the patient level to create the longitudinal patient history that is critical to monitoring patients in a non-fragmented manner across the continuum of care. Let me very briefly illustrate our efforts. Today, Synergy has assembled the largest real-time, integrated patient-level database in the world—a database that contains more than 1.5 billion de-identified hospital, medical and pharmaceutical claims, which can be linked at the individual level to provide robust healthcare data on 100+ million Americans. Refreshing the database with over three million new claims per day, we currently perform real-time analyses of new and existing drugs, including prescribing trends, concomitant therapies, diagnoses and co-morbidities, dosage by indication, and the like. These analyses are provided to customers using state-of the-art query tools over the Internet. Figures 1 and 2 depict samples of a typical user interface.

The success of these applications proves that the data and technology to address the occurrence of medical errors in real-time already exist, and the prospect of developing new analytic tools for application to the public health is an exciting one.

My testimony will briefly review these data and technologies, suggest certain patient safety measures to monitor, and most importantly, propose areas of research that should be explored so these systems can be appropriately applied to detect medical errors and monitor patient safety.

Data and Technology

As you know, the Health Insurance Portability and Accountability Act (HIPAA) encourages the submission of electronic data to payors to improve administrative efficiency. Today over 65% of all health care insurance claims are submitted electronically. This represents 3 billion of the nearly 5 billion transactions submitted by providers in the United States today 1. In appropriately de-identified form, it is this data that is driving the development of 'applied' informatics in the private sector.

There are currently three types of electronic data submitted by providers that may be used by informatics organizations: retail pharmacy claims, medical claims and hospital claims. Figure 3 shows the structure of data flows from providers to payors and the point at which these data convene to be aggregated and stored in a consolidated data warehouse. These stored data can then be analyzed by specific safety measures and results can be reported in a timely fashion.

While the data transfer technology utilized today consists mostly of direct dial and other point-to-point connections, the Internet is playing a greater role in data transfer. Advances in communication and data transfer technology will assure that electronic claims are transferred more efficiently and accurately in years to come. The science of data warehousing has advanced with remarkable speed. Today it is now possible in the commercial environment to store 2 terabytes (2000 gigabytes—the capacity of 200 personal computers) in a space the size of a telephone booth. A patient safety surveillance system using real time electronic claims data is feasible today due to these revolutionary technical advances.

Table 1 briefly describes each data type, the data formats, the source the current volume estimates in the U.S., and a sample of key data elements.

Select for a text version of Table 1.

Table 1.

Pharmacy Data Data Medical Data Hospital Data

Format:

NCPDP

Format:

HCFA 1500

Format:

UB 92

Source:

Retail pharmacies

Source:

Physician offices

Source:

Hospitals

Volume:

(# electronic claims 2000)
1,814,000,000

Volume:

(# electronic claims 2000)
789,000,000

Volume:

(# electronic claims 2000)
399,000,000

Key Data Elements:

  • Encrypted patient ID
  • Patient age
  • Patient gender
  • Prescription date
  • Prescription name
  • Quantity dispensed
  • Days supply
  • Refill flag
  • Prescribing physician
  • Payor
  • Pharmacy
  • Amount charged ($)
  • Amount paid ($)

Key Data Elements:

  • Encrypted patient ID
  • Patient age
  • Patient gender
  • Service date
  • Diagnosis (ICD9 code)
  • Procedure (CPT4 code)
  • Physician
  • Payor
  • Location of care
  • Amount charged ($)
  • Amount paid ($)

Key Data Elements:

  • Encrypted patient ID
  • Patient age
  • Patient gender
  • Admission date
  • Discharge date
  • Diagnosis (ICD9 code)
  • Procedure (CPT4 code)
  • Diagnosis related group (DRG code)
  • Physician
  • Payor
  • Location of care
  • Amount charged ($)
  • Amount paid ($)

As envisioned by HIPAA and noted above, these data are in standard formats. This greatly decreases the data integration time and scrubbing required to prepare these data for proper analysis. While the data elements available lack detailed clinical information that may be contained in a patient's medical record (such as blood pressure, physical signs, symptoms, and laboratory results), claim-based data elements do contain valuable diagnostic and treatment information which can be utilized to develop sensitive measures to monitor patient safety.

A unique, encrypted patient identifier that spans all three data types is critical for analyses that require medical and pharmaceutical information. While challenging to develop, technology now exists to create such an identifier that follows an individual over time and is independent of payor, pharmacy, medical provider or location of care. Standard data elements that are present in all three data types—such as a patient's gender—can be concatenated, irreversibly encrypted and successfully used to follow a patient from cradle to grave.

Safety Measures

As I mentioned before, Synergy applies data analysis tools to patient-centric healthcare claims data. For instance, our Rx Market Monitor is a real-time, web-based, interactive application that allows users to monitor prescribing patterns of physicians across the U.S. at the regional and MSA levels. Measures such as patient age and gender, physician specialty, prescription type (new vs. refill), drug indications and patient co-morbidities are included. This analytic currently contains some 20 therapeutic classes/condition-specific modules with specific information on almost 500 individual drugs. Similarly, Rx Dosage Insight allows users to monitor prescribing patterns, and includes measures such as drug compliance, titration and prescription dynamics (average daily dose, average days supply, etc.) and can include patient age/gender and physician specialty.

New measures developed for an electronic patient safety surveillance system focused on ambulatory care settings will necessarily be limited by the type and frequency of data available. However, by using the data previously described and exploiting the real time, patient level attributes of these data, with adequate funding for development of new analytics, several types of measures could be deployed:

  1. Drug-drug interactions—It is common knowledge that certain drugs interact adversely with other drugs. This is especially true when patients are on multiple medications (polypharmacy). While pharmacies often screen for these potential problems at the point of prescription dispensing, the system is far from comprehensive. For example, medications prescribed by several different physicians and filled at different pharmacies may not be scrutinized. Because the electronic patient safety surveillance system proposed here follows patients across pharmacies and providers, there is a much greater chance of identifying potential serious drug-drug interactions than exists today.
  2. Drug-disease interactions—Certain drugs should be avoided in patients suffering from some medical conditions. For example, a common class of antihypertensive agents (Beta Blockers) can severely impact the breathing of patients with chronic obstructive pulmonary disease (COPD). Again, due to the often fragmented system of care in the U.S. patients with COPD could mistakenly receive Beta Blockers. Because the electronic patient safety surveillance system can link medical and pharmaceutical data at the patient level, it is possible to determine the disease states of patients who have received adverse medications. This linking provides a much greater chance of identifying patient populations at risk for potential serious drug-disease interactions than exists today.
  3. Drug-patient interactions—Many medications are targeted for patients in certain age groups. In addition, many drugs are contraindicated for patients of specific age and/or genders. Children and senior citizens tend to be the most adversely affected by medications not specifically developed for them. While some drug-patient interactions are well known and happen rarely, others depend on the frequency and length of drug use. For example, a common antibiotic can permanently stain children's teeth. Common pain medications should be used judiciously in the elderly as they can cause kidney damage and severely affect a senior citizen's mental status. By knowing basic patient demographics such as age and gender, emerging or previously unknown drug-patient interactions can be discovered, and known drug-patient interactions can be easily monitored.
  4. Untoward medical events due to drugs—A common technique in the science of traditional quality assurance is to look for the presence of certain medical events following the administration of medication. For example, a patient with an emergency room encounter several days after starting a new antihypertensive medication could indicate a drug side effect. A diabetic patient hospitalized after an insulin adjustment could indicate an inappropriate insulin dosage change. While not absolutely specific in isolating true medical errors, these indirect measures can be most useful in focusing attention on cases that may require further investigation. These types of measures are also very useful in provider profiling—the science of analyzing patterns of care to identify aberrant trends and high rates of untoward medical events among providers or provider groups.
  5. Medical complications—Medical data alone is useful in identifying complications of invasive procedures and other medical interventions. While many complications are known risks and accepted risks of medical therapy, many are the result of medical errors. Useful measures in this area include monitoring reoperations, rehospitalizations, emergency room visits after hospital discharge and repeat medical procedures. Post surgical infections—a very common and devastating complication—can be monitored in this manner. While monitoring complications has been used for decades in quality assurance departments across the country, there is no standard system in place to monitor these events on a national basis. It is also very difficult for most health care institutions to gain insight to activities in the outpatient setting, where patients suffering from many of these complications may first present.

  6. Outcome measures—Outcomes research has blossomed in the past decade. Many good outcome measures have been developed and the field has broadened to include measures of patient functional status, satisfaction, quality of life and economics. Because of its longitudinal, patient centric architecture, electronic claims data can be utilized to develop indicators that measure outcomes of medical and pharmaceutical interventions. Unlike searching for complications associated with specific medical events, outcome measures are more broad based and focus on whether a medical intervention really worked or not. For example, patients switched from one medication to another in certain therapeutic classes could be the result of an adverse outcome. Patients presenting multiple times to a physician prior to initiating therapy could signify a delay in diagnosis. Like complication monitoring, these measures do not posses a high degree of specificity and must be viewed as a tool to identify case for further exploration.

These six areas of patient safety and medical error surveillance monitoring have their strengths and weaknesses. The strengths include:

  • The ability to monitor potential problems in real time and take appropriate actions early or even before a medical error occurs.
  • The ability to monitor larger volumes of patient experiences than is available by any system currently in existence today.
  • The ability to follow patients across the continuum of care irrelevant of payor, provider, or pharmacy.

The weaknesses of such an electronic safety surveillance system must be addressed by further research. Although successful in commercial applications, there is little experience using these data sets to monitor patient safety and medical errors.

Suggested Research Agenda

While both the technology and the data for an electronic patient safety surveillance system are currently available, several research questions require further exploration. These questions fall into three major categories. These include validity, measure development and reporting.

First, the validity, reliability, and sensitivity of these systems must be tested for use in safety surveillance. While successful for other purposes, there has been little research to date on these important statistical and epidemiological issues—especially the validation of electronic data against paper medical records. Several studies could be conducted to answer these questions. It is obvious that these data sets were originally created for reimbursement purposes—not to accurately monitor medical interventions and outcomes. Over the past decade however these types of data have been used on a small scale for these purposes. The National Committee for Quality Assurance (NCQA) allows these data sets to be used to report results for it Health Employer Data and Information Set (HEDIS) measures. Managed care organizations routinely use claims data to develop and monitor the impact of their disease management and quality improvement initiatives. Despite these successes, there is very little in the literature concerning the validity and accuracy of claims data. This problem is especially important when using aggregated claims data across payors. While one payor may fully understand the accuracy of its own claims information, these observations may not be generalizible to other payors, providers and regions of the country. Since reimbursement policies drive claims submission, it is entirely likely that the accuracy of claims data for use in patient safety surveillance may vary with these policies. Studies that correlate claims information with medical records need to be performed across payors, provider types and regions of the country. A firm understanding of the validity and reliability of claims data is mandatory before accurate patient safety measures can be developed.

Second, specific indicators and outcome measures must be developed. As discussed previously, while measures such as drug-drug interactions, drug-disease interactions, adverse outcomes and complications of certain medical interventions can be monitored, a comprehensive research effort is required to carefully select those indicators that can be most accurately and consistently measured by these systems. Measurement development is a complex field. The Joint Commission on the Accreditation of Healthcare Organizations (JCAHO), the National Committee for Quality Assurance (NCQA) and the Agency for Healthcare Research and Quality (AHRQ) have all done an outstanding job of developing quality indicators over the years. These efforts have resulted in techniques and measures that have become the very foundation on which modern quality assurance is built. However, most all these indicators were designed for different purposes and use different data sets than we have today. While many of these traditional measures may be relevant for patient safety and medical error monitoring, it is unknown how or if they can be applied to real time electronic claims on a national scale. Studies must be performed to select appropriate measures that are most applicable to the size and speed of electronic data sets currently available.

Finally, assuming the availability of specific electronic patient safety and medical error measures, it is unclear who should have access to results. Research must be performed to identify the appropriate stakeholders who can accurately interpret outcomes and implement appropriate courses of action. In many ways this is the most important aspect of developing a patient safety surveillance system. Since the proposed system will have varying levels of sensitivity, those parties utilizing results of the system must be cognizant of the strengths and weaknesses of such a system. If the system is truly national in scope, it could have very universal appeal. Regulators, providers, employers, and even consumers may demand access to results. Those familiar with and appropriately trained in the art and science of quality measurement and patient safety will most likely use these results for the goal in which they were intended—to identify and eliminate medical errors and improve patient safety. Studies must be performed to determine the legitimate interests of the relevant stakeholders and to define the appropriate limits for their access to such results. In addition, policies need to be established that address important issues of actions taken upon providers and others based upon the data. Further, the question of protecting individual privacy versus the potential to identify not only individual physicians but also patients and perhaps thereby prevent medical errors must be discussed. As stated above, a national real time patient safety surveillance system will have strengths and weaknesses. Those accessing the results of such a system must realize this. A thorough evaluation of these issues is critical.

Conclusion

Electronic transaction data systems can be the fuel for a revolutionary, cost-effective patient safety surveillance system that can track the delivery of care, and the occurrence of medical errors, across many care settings, from hospitals and physicians' offices to outpatient clinics and pharmacies. Such a data-based surveillance system could offer stakeholders many immediate advantages—reducing medical errors, improving patient safety and, most importantly, providing major advances toward improving the quality of health care for all Americans.

Notes:

1. Health Data Directory, 2000 Edition, Faulker and Gray, 1999.

Current as of September 2000


Return to Testimony Directory
National Summit on Medical Errors and Patient Safety
QuIC Home Page
Department of Health and Human Services


The information on this page is archived and provided for reference purposes only.

 

AHRQ Advancing Excellence in Health Care