Effect of Surveillance Method on Reported Characteristics of Lyme Disease, Connecticut, 1996–2007

The epidemiology of Lyme disease varies by surveillance method.

L yme disease, a multisystem disease caused by the spirochete bacterium Borrelia burgdorferi, is the most commonly reported vector-borne disease in Connecticut and in the United States (1). During 1996-2007, Connecticut contributed 33,457 (15%) cases to the national surveillance case count, with a mean of 83.4 cases per 100,000 population reported annually, and consistently led the nation in reported annual incidence rate during the study period (2). Diagnosis of Lyme disease is based on clinical fi ndings, serologic evidence of infection, and history of exposure to Ixodes scapularis ticks. Early stages of illness are most readily diagnosed by identifi cation of erythema migrans. Later-stage illness can involve the musculoskeletal, neurologic, or cardiovascular systems.
Positive serologic results are necessary for identifying and classifying patients with later manifestations.
During 1996-2007, the Connecticut Department of Public Health (CDPH) received Lyme disease reports through 4 surveillance methods: passive physician, active physician, enhanced laboratory, and mandatory laboratory. Physician-based surveillance (passive and active) was conducted during the entire study period and relied on health care providers to report new diagnoses of Lyme disease. Active surveillance comprised a voluntary network of health care providers who reported cases 1× per month. Enhanced laboratory surveillance, conducted during 1996-1997, required participating Connecticut laboratories to send supplemental case report forms with each positive B. burgdorferi result to the ordering physician. In January 1998, to study the effectiveness of a newly released Lyme disease vaccine, mandatory laboratory surveillance was implemented that required all laboratories to report positive and equivocal results to CDPH. Follow-up, conducted by CDPH staff, involved sending a letter and supplemental report form to the ordering physician. To assist the physician, demographic and patient-identifying information from the laboratory report was incorporated into the form. Mandatory laboratory surveillance ended after 2002 when the Lyme disease vaccine was removed from the market. In 2007, mandatory reporting of positive Lyme disease results was reinstated for laboratories with electronic reporting capability. Two large commercial laboratories provided electronic reports. Follow-up was reestablished by using the previous method, i.e., CDPH staff sent a letter and supplemental report form to the ordering physician.
Public health surveillance methods for infectious diseases change over time, depending on program priorities and resources, advancements in diagnostic testing, modifi cations to surveillance case defi nitions, and changing reporting modalities (e.g., electronic laboratory reporting). Lyme disease surveillance data provide a measure of the relative geographic distribution of this disease and its effect on public health in Connecticut and have been used to assess the effectiveness of control and prevention activities (3)(4)(5). These data also form part of the risk communication messages provided to the general public, advocacy groups, media, political leaders, health care providers, and public health professionals. We examined how surveillance method affected the classifi cation of reported clinical and demographic characteristics of case-patients and the incidence of Lyme disease in Connecticut, during 1996-2007.

Surveillance Case Defi nition
Lyme disease reports were categorized by using the national surveillance case defi nition issued in 1996 (6). A case was defi ned as 1) physician report of erythema migrans of >5 cm in diameter or 2) at least 1 objective late manifestation (i.e., musculoskeletal, neurologic, or cardiovascular) with laboratory confi rmation of infection with B. burgdorferi by enzyme immunoassay, immunofl uorescent assay, or Western immunoblot. CDPH classifi ed reports that did not meet the case defi nition as not a case. Because clinical information is required for case classifi cation, when supplemental follow-up reports were not returned, they were considered lost to followup. The distinct report forms used for each surveillance method contained the following data elements: case-patient demographic characteristics (sex, age, race, ethnicity), clinical fi ndings (erythema migrans or late manifestations), seasonality, and case status. Seasons were defi ned as winter (December-February), spring (March-May), summer (June-August), and fall (September-November).

Data Collection
The statewide Lyme disease surveillance system was maintained by an average of 1.5 full-time employees. Reports were entered into the National Electronic Telecommunications System for Surveillance (NETSS), a public health surveillance information system that used Epi Info 6.0 (7). In 2007, NETSS and Microsoft Excel (Microsoft, Redmond, WA, USA) were used. Supplementary variables necessary for follow-up and maintenance of reports were added to the standard NETSS variables by CDPH staff and included health care provider name and contact information, license number, and origin of report.
A potential case could have been reported through >1 surveillance method. For consistency in classifi cation, the origin of case reports was entered in the following hierarchy: active surveillance, passive surveillance, enhanced laboratory surveillance, and mandatory laboratory surveillance. Data were cleaned and duplications were removed at the end of each year.

Data Analyzed
The following statistics were calculated across surveillance methods: annual mean number of reports and cases, annual incidence rates, proportion of reports by case status, demographic characteristics of case-patients, seasonality of cases, and clinical and laboratory fi ndings. Incidence per 100,000 population was determined by using decennial census data covering the year of data collection (1990 or 2000). Statistical tests were performed by using Epi Info 6.0. We used χ 2 test with the Yates continuity correction. A p value <0.05 was considered signifi cant. The positive predictive value (PPV) of reports of potential cases was calculated for each type of surveillance method by determining the ratio of cases to reports.

Overall Analysis
During 1996-2007, CDPH staff processed 87,174 Lyme disease reports, of which 7,278 (8.3%) were duplicate entries and were removed from the database. A total of 79,896 individual reports were analyzed. Of these, 43,767 (54.8%) were reported through mandatory laboratory surveillance, 19,350 (24.2%) through passive physician surveillance, 13,040 (16.3%) through active physician surveillance, and 3,739 (4.7%) through enhanced laboratory surveillance. Overall, 33,457 (41.9%) reports were classifi ed as cases, and 26,318 (32.9%) as not cases; 20,121 (25.2%) were lost to follow-up (Table 1). Except for calculation of PPV, reports classifi ed as lost to followup were excluded from further analyses.
During 1996, Connecticut had 5,473 reports of Lyme disease. The number of reports increased with the successive implementation of enhanced and mandatory laboratory surveillance reporting, peaking at 12,947 in 2002 ( Figure 1). In 1998, the fi rst year of mandatory laboratory surveillance, the overall number of Lyme disease reports  (Figure 1), and the incidence nearly doubled to 89.8 cases per 100,000 population. An average of 16.0% more cases were reported through physician-based surveillance during years with mandatory laboratory reporting.
The PPV varied across surveillance methods and was highest for physician-based surveillance methods (Table  1). Less than 25% of reports received through mandatory laboratory surveillance were classifi ed as cases. Cases reported through this method accounted for nearly one third (31.9%) of all cases during the study period.

Demographic Characteristics
The median age of case-patients was 38 years (range 34-43 years). Case-patients <20 years of age were more likely to be reported through physician-based surveillance (p<0.001); laboratory-based surveillance was more likely to report case-patients >40 years of age (p<0.001) ( Figure 2). Overall, whites accounted for 82.0% of cases, similar to the state's racial distribution, and the distribution did not differ signifi cantly by surveillance method. Ethnicity data were available for approximately one third (32.6%) of case-patients; only 1.2% were reported as Hispanic. Laboratory-based surveillance reported an average of 32 Hispanic case-patients annually, compared with 20 reported through physician-based surveillance ( Table 2). On average, 9.6% more male than female casepatients were reported by each surveillance method.

Seasonality
In 72.9% of cases, illness onset occurred during the summer (76.3% physician-based vs. 66.8% laboratorybased cases) ( Table 5). Erythema migrans occurred in 84.2% of cases with onset during the summer. Erythema migrans was signifi cantly more likely to be reported during the summer through physician-based surveillance than through laboratory-based surveillance (71.3% vs. 28.7%; p<0.001). Late manifestations were 2× more likely to be reported through laboratory-based surveillance during the summer months (17.5% vs. 8.2%; p<0.001).

Discussion
In Connecticut, data obtained through 4 surveillance methods during 1996-2007 demonstrated that the epidemiology of Lyme disease is subject to variation by surveillance method. The number of reports, proportion of reports classifi ed as cases, incidence, and demographic and clinical characteristics of case-patients differed between physician-based and laboratory-based surveillance. Although some of the annual fl uctuation in reports and cases might be attributable to an actual increase or decrease in disease, the substantial changes seen indicate that the principal factor most likely resulted from changes in surveillance method over time. As these surveillance artifacts show, changes in surveillance methods can cause changes in trends. Therefore, the nature of the surveillance method and the effect of changes in the method are necessary to consider when interpreting the data.
Lyme disease surveillance methods ultimately rely on physicians to report the necessary clinical information to classify cases. Because health care providers in outpatient settings often underreport commonly seen illnesses (8,9), in Connecticut, follow-up for mandatory laboratory surveillance might help serve as a reminder system for physicians to report cases. This fact could explain the 16.0% increase in the average annual number of cases  reported through physician-based surveillance during years when laboratory surveillance was mandatory.
Our data showed that physician-based surveillance, combined with laboratory-based surveillance, resulted in more comprehensive clinical and demographic information and higher incidence of illness than each method alone. Of all reported cases, nearly one third (31.9%) originated through laboratory-based surveillance. However, use of laboratory-based surveillance is ineffi cient: only 24.3% were classifi ed as cases. Case-patients reported through laboratory-based surveillance also differed signifi cantly in age groups, reporting older case-patients; clinical information, reporting more late stage illness; and seasonal data, reporting more cases during the fall and winter months. Our combined surveillance methods contributed to, and broadened, the overall epidemiologic description of Lyme disease in Connecticut.
This study has several limitations. First, laboratorybased surveillance was diffi cult to evaluate independently of physician-based surveillance. Because health care professionals could potentially report by using each of the surveillance methods, we used a hierarchy to help reduce bias toward 1 surveillance method over another. Second, because active surveillance providers volunteered to participate, these physicians were more likely to be those most interested in Lyme disease surveillance. Therefore, physicians who volunteered to participate in active surveillance might have been more likely to report cases in a strictly passive surveillance system.
To satisfy the sometimes confl icting goals of surveillance methods and allocation of public health resources, collection of case data needs to be streamlined. Two potential alternatives may be the following: modeling by using sampling schemes or greater use of electronic information systems, which is planned in Connecticut. Electronic laboratory reporting, automation of follow-up requests, and Web-based provider reporting will conserve resources, and provide incident data information demanded of public health agencies.
When determining which methods of Lyme disease surveillance to use, the purpose of that surveillance and available resources need to be considered. In Lyme disease-endemic states where the epidemiologic purpose might primarily be to monitor geographic, clinical, and demographic trends, intensive statewide surveillance is not essential. Rather, surveillance needs to be conducted   consistently over time. Intensive surveillance efforts may even be counterproductive when not sustainable because of limited resources or when resources are diverted from other public health activities. Replacing traditional casereporting surveillance methods with less labor-intensive data collection methods, such as regular populationbased surveys, may be suitable for following trends and estimating disease (10).