The Continuity of Care Document
Although it was originally designed to exchange information on individual patients, the CCD will become a powerful instrument for medical research and public health. Its enforcement of structured data and language provides the first normalized summary that can be generated from any of the 600 plus certified EHRs. As such, CCDs can be interpreted without previous knowledge of the source system, similar to a Web page on the Internet. The CCD uses extensible markup language to represent medical data in a consistent, tagged format. These tags, attached to every data element, identify key context descriptors such as the language being used to encode data (Figure 1). According to federal regulation, the electronic summary must include data on patient demographics, problems, medications, allergies, laboratory results, and procedures. Although these sections represent only a fraction of all medical data, standardization makes them available to systems beyond the originating EHR.
(Enlarge Image)
Figure 1.
A simplified example of tagged diagnosis data within the Continuity of Care Document with explanations on right.
Using the CCD, agencies can create tools to communicate population data that are immediately compatible with any certified EHR. Although the potential applications are many, the following three examples illustrate the promise of CCD applications to improve existing systems for population analysis.
Extending Public Health to Chronic Diseases
More than 20 million people are estimated to have diabetes in the United States, but only 20% of these patients receive the appropriate preventative services as recommended by the American Diabetes Association. Many persons with diabetes are consequently at high risk of complications, comorbidities, and death as a result of the disease. In 2005, the New York City Board of Health approved a novel approach to collect data about the disease and craft better public health responses. The board requires mandatory reporting of patient-identified glycosylated hemoglobin values from laboratories to the local department of public health. This collection format is similar to reporting systems for communicable disease, limited to patient and provider contact information and the lab date and result for glycosylated hemoglobin. As a result of this program, the city now produces quarterly reports for approximately 1600 providers and mails 400 letters each week to patients with high glucose levels.
Although local pilots demonstrate the public health benefit of this and other initiatives, key data from these efforts are missing. Relevant information on medications, preventative practices, and quality performance are not transmitted. An absence of total patient counts and detailed diagnoses prevents accurate incidence calculations. These limit the comprehensiveness of physician reports and patient mailings. As certified EHRs are adopted in a public health region, submitting normalized data via the CCD would enlarge the analytical power of such initiatives while minimizing the reporting burden. Moreover, it would allow expansion of such analytics to other chronic conditions, such as heart failure or hypertension, without increasing the need for new interfaces. Public health authorities exercising their statutory power to collect such information presents a significant opportunity to improve the care for chronic diseases.
Clinical Detail in Death Certificates
In 1999, the US Food and Drug Administration approved rofecoxib, commonly known by its brand name of Vioxx. Within several years, the drug had become a blockbuster for its ability to treat chronic pain without the adverse effects of gastrointestinal ulcers and bleeding. Soon after its launch, however, the safety of rofecoxib was questioned because of postapproval data on the incidence of myocardial infarction. By September 2004, Merck had voluntarily withdrawn the drug from the market because of mounting evidence of cardiovascular harm caused by rofecoxib.
One retrospective study of Kaiser Permanente members examined whether rofecoxib was associated with increased coronary events. By scanning 2.3 million person years, researchers observed an adjusted odds ratio of 3.58 for serious coronary heart disease with high-dose rofecoxib compared with similar drugs. With such a serious increase in risk, the drug clearly caused thousands of deaths before its withdrawal 5 years after approval. One vocal expert and researcher on the study from the US Food and Drug Administration said the agency was "incapable of protecting America against another Vioxx."
Transparency through data can be a valuable safeguard in researching suspected causes of death. One way to increase the power of mortality studies would be to increase data reported at the time of death. Pairing clinical detail on medications and laboratory results, as well as other known conditions, would help agencies and epidemiologists better explore vital statistics data. These data could again be aggregated by CCDs for health facilities by using certified EHR technology. With such tools, death certificate data could be scanned to test hypotheses of medication risk, similar to Kaiser's analysis of rofecoxib. Although that may not be feasible for several years, setting expectations in advance will prepare for an eventual transition to active adverse event detection.
Advancing Biosurveillance
Currently the Centers for Disease Control and Prevention runs a nationwide reporting system named the Influenza-like Illness Surveillance Network (ILINet). ILINet collects information weekly from approximately 1800 outpatient care sites on patients who have a fever and cough or sore throat in the absence of other known causes. Providers submit weekly patient counts stratified by age through fax or the Internet to the Centers for Disease Control and Prevention. Regional information is made publicly available and monitored for potential outbreaks. The estimated time burden for participants is less than 30 minutes per week, but collection of simple patient counts limits the depth of analysis. Novel approaches that electronically extract more comprehensive patient data can improve the depth, speed, and sample size of influenza analytics. One such system pioneered by the Veterans Administration uses patient diagnosis detail from all patient visits documented in its EHR. Using this electronic detail has detected significant shifts in condition type and patient demographics that could not be revealed through ILINet.
CCD-normalized extracts present an opportunity to rapidly scale electronic surveillance. Rather than having registries like ILINet manually collect data, simple applications could be distributed to parse and calculate statistics from certified EHR technology. The flexibility of this approach relies not on transmitting personal health information, but on locally deploying programs to analyze data within a practice's existing infrastructure. Then, richer de-identified summaries of influenza-like encounters could be transmitted weekly, or daily, with minimal effort on the behalf of providers. Although regional pilots exist using older standards for surveillance, the CCD's common structure and language reduces implementation cost. Future public health surveillance systems will unlock new data on health disparity, detailed symptomatology, and therapeutic regimen that will strengthen national efforts to manage infectious disease and environmental risk.