Abstract

Few studies have assessed the accuracy and completeness of Lyme disease surveillance systems. Lyme disease cases were identified through review of medical records for residents of the Marshfield Epidemiologic Study Area (MESA), a population-based cohort of nearly 80,000 in north-central Wisconsin for which comprehensive medical care data are available through the Marshfield Clinic. These cases were compared with cases reported to the Wisconsin Division of Public Health to estimate the completeness of reporting. Annual incidence rates were calculated for MESA using the cases identified from chart review. Division of Public Health data were used to calculate statewide reported incidence rates, as well as rates for an eight-county region that surrounds and includes the MESA region. Of Lyme disease cases meeting the national case definition identified in MESA, 34% were reported to the state. The average incidence of Lyme disease (1992–1998) was 19.1 per 100,000 per year in MESA, 17.0 in the surrounding eight-county region, and 9.0 statewide. Trends in reported incidence across time, gender, and age in the surrounding eight-county region were generally comparable with those observed in MESA. These findings suggest that the passive surveillance system monitored trends in Lyme disease incidence reasonably well despite underreporting of cases.

Lyme disease, caused by infection with the spirochete Borrelia burgdorferi, is the most commonly reported human vector-borne disease in the United States. Early manifestations of the disease include nonspecific symptoms (e.g., fever, malaise, fatigue, headache, myalgia, and arthralgia) and a characteristic rash, erythema migrans. The chronic stages of the disease involve arthritic, cardiac, and neurologic complications. Diagnosis of early Lyme disease is usually based on history and physical examination. Tests such as indirect fluorescent antibody, enzyme immunoassay, and Western immunoblot are useful for the diagnosis of late Lyme disease.

In 1991, the Centers for Disease Control and Prevention established a Lyme disease case definition for use in state-based passive surveillance. A total of 88,967 cases of Lyme disease were reported to the Centers for Disease Control and Prevention between 1992 and 1998, including 3,237 from Wisconsin (1). The reported mean annual incidence rate in Wisconsin was highest among the north-central states and seventh highest nationally (1). Few studies have assessed the accuracy and completeness of these state passive surveillance systems (2–4). In Wisconsin, Lyme disease is often regarded by clinicians as a routine condition. It is frequently managed in urgent care centers or other high-volume outpatient settings, and thus substantial under-reporting of cases may occur.

Population-based studies of Lyme disease incidence in endemic areas have been limited (2, 5, 6). The Marshfield Epidemiologic Study Area (MESA) is a large population-based cohort covering 24 ZIP codes in north-central Wisconsin. Comprehensive inpatient and outpatient medical care data are available through the Marshfield Clinic for the nearly 80,000 residents of MESA. Retrospective review of medical records data for members of this cohort allows estimation of population-based incidence rates for disease and comparison with corresponding rates reported from other sources.

MATERIALS AND METHODS

The Marshfield Clinic is a 600-physician multispecialty group practice that provides medical care in 32 communities in north-central Wisconsin. With over 1.6 million patient encounters and more than 350,000 unique patients seen annually, Marshfield Clinic is the largest group practice in the state of Wisconsin and one of the largest in the country. Marshfield Clinic is the major provider of health care in its 29-county primary service area and the only provider of care in some parts of the region, including the immediate service area surrounding the main campus in the city of Marshfield. A regional data network links patient care and administrative records from the regional clinics with the main clinic's computer system, providing access to data for epidemiologic studies.

The Marshfield Epidemiologic Study Area was established in 1991 to facilitate population-based research using the integrated health care network of the Marshfield Clinic (7). The MESA region is a subset of the geographic area served by the clinic and includes 14 contiguous ZIP codes surrounding the city of Marshfield (MESA Central) and 10 ZIP codes surrounding three of the larger regional care centers (MESA North) (figure 1). These 24 ZIP codes were selected for inclusion in MESA on the basis of very high levels of population and health event coverage, as determined initially by market share data and later validated through random population surveys.

FIGURE 1.

Marshfield Epidemiologic Study Area (MESA) Central, MESA North, and the surrounding eight-county (Wood, Clark, Marathon, Iron, Lincoln, Price, Rusk, Taylor) region in Wisconsin, 1992–1998.

FIGURE 1.

Marshfield Epidemiologic Study Area (MESA) Central, MESA North, and the surrounding eight-county (Wood, Clark, Marathon, Iron, Lincoln, Price, Rusk, Taylor) region in Wisconsin, 1992–1998.

Data pertaining to MESA residents are extracted on a daily basis from the administrative and clinical files of the Marshfield Clinic computer systems to track when persons enter and leave the selected population as a result of birth, death, or migration. Electric company disconnections, the US Postal Service change of address system, and periodic targeted population surveys are also used to monitor in- and outmigration.

Nearly all residents of MESA receive health care from the Marshfield Clinic (7). Previous validation efforts have indicated that the population coverage is 94–96 percent or higher (7) (Marshfield Medical Research Foundation, unpublished data). Across the 24 MESA ZIP codes, the median proportion of the population seeking care within the past 2 years was 86 percent.

In addition to high population coverage, the majority of health care services obtained by the residents of MESA are provided through the Marshfield Clinic or affiliated hospitals, and these events are captured by clinic data systems. In 2000, random sample surveys of the MESA population demonstrated that approximately 90 percent of outpatient visits are captured by the clinic data systems. Recent comparison of inpatient records with the state of Wisconsin hospital discharge database indicates that 93 percent of hospital discharges in MESA are captured. Comparison with state of Wisconsin mortality data indicates that the clinic captures 99 percent of deaths among residents of MESA. The high degree of population coverage and ascertainment of in-patient and outpatient health events makes MESA particularly well suited for studying the occurrence and clinical spectrum of disease in a general population.

Case definition and identification

Computerized International Classification of Diseases, Ninth Revision, codes from the Marshfield Clinic database were used to identify potential cases of Lyme disease that occurred in 1992–1998 in the MESA population. The specific codes identified were codes 088.81, 695.90, and 066.90. The Marshfield Clinic laboratory database was also searched for positive Lyme serologic tests (indirect fluorescent antibody or enzyme immunoassay, with or without Western immunoblot) to identify potential cases. Persons with a history of Lyme disease diagnosis prior to 1992 were excluded from study.

To validate these diagnostic codes and laboratory results, we abstracted data from patients' medical records. Potential cases were grouped into three categories (probable cases, possible cases, and noncases) based on the chart abstraction findings. Probable Lyme disease cases were defined as those patients who met the national case definition (1). The criteria for a probable case included physician-diagnosed erythema migrans of ≥5 cm in diameter or at least one late, noncutaneous manifestation with laboratory confirmation of infection. Late clinical manifestations included objective evidence of recurrent arthritis, cranial neuritis, radiculopathy, meningitis, encephalomyelitis, or cardiac manifestations.

Possible cases included patients with erythema migrans but no documented size information and patients with positive serology and recurrent joint pain or neurologic symptoms that did not meet the criteria for a probable case. A physician (E. A. B.) reviewed all final diagnoses, and patients with valid alternative diagnoses in their medical records were classified as noncases. Examples of valid alternative diagnoses included psoriatic arthritis, Osgood-Schlatter disease, cellulitis, tendonitis, figuratum erythema, palindromic rheumatism, allergic contact dermatitis, intradural schwannomas, carpal tunnel syndrome, and gout.

Probable and possible Lyme disease cases were further defined as either early or late Lyme disease. Patients with documented erythema migrans or a positive B. burgdorferi culture from skin biopsy were classified as early cases, while late Lyme disease cases had evidence of extracutaneous manifestations in addition to a positive diagnostic laboratory test in their medical records.

Estimation of reporting

Lyme disease has been a reportable disease in Wisconsin since 1980. Written case report forms, including information about patient demographics (name, race, gender, and date of birth), dates of onset and diagnosis, and clinical and laboratory information pertinent to the national Lyme disease case criteria, are completed by health care providers and mailed to local county health departments. These reports are then forwarded to the Wisconsin Division of Public Health where each form is reviewed, and those patients that have been determined to meet the national case criteria based on the information provided by the reporting form are included in the state database. Information about the patient's county of residence is included in this database, but no other geographic identifiers such as address or ZIP code were included in this database during the study period.

A list of the Lyme disease patients identified from Marshfield Clinic records was sent to the Wisconsin Division of Public Health, which linked this list of patients by name, birth date, and diagnosis date (within 4 weeks) to computerized records of reported Lyme disease cases that occurred during 1992–1998. This combined list of cases was used to calculate the percentage of Lyme disease cases reported to the state, and chi-squared tests were used to identify any differences in reporting across age, gender, and stage of disease at diagnosis. The Division of Public Health also provided statewide and county-specific reported Lyme disease incidence data for the years 1992–1998 to use in comparisons with the rates described below.

Incidence rate calculation

Annual incidence rates (1992–1998) were calculated for probable cases of Lyme disease in MESA. The denominators for these rates were based on linear interpolation of 1990 and 2000 Census population counts for the MESA ZIP code region. Average annual incidence rates stratified by gender, age (from birth to 20 years, 21–44 years, 45–64 years, and ≥65 years), and stage of Lyme disease at diagnosis (early vs. late) were also calculated. Used in the calculation of these stratified rates were 1990 Census estimates.

Incidence rates were then calculated for the state of Wisconsin on the basis of Division of Public Health surveillance data. However, the entire MESA ZIP code region covers only a small fraction of the state of Wisconsin and primarily lies within a geographic area that has moderate to high reported Lyme disease incidence (8) (figure 2). Therefore, to provide a more appropriate comparison with MESA rates, annual incidence rates based on Division of Public Health surveillance data were calculated for an eight-county region that surrounds and includes the MESA ZIP codes. The eight counties included in this group were Clark, Iron, Lincoln, Marathon, Price, Rusk, Taylor, and Wood. Annual intercensual estimates from the US Census Bureau were used in the calculation of annual incidence rates for the eight-county region and the state. Data from the 1990 US Census were used to calculate rates stratified by age, gender, and stage of disease at diagnosis for these regions.

FIGURE 2.

Mean annual Lyme disease incidence in Wisconsin based on case reports meeting the Centers for Disease Control and Prevention case definition received by the Wisconsin Division of Public Health, 1992–1998.

FIGURE 2.

Mean annual Lyme disease incidence in Wisconsin based on case reports meeting the Centers for Disease Control and Prevention case definition received by the Wisconsin Division of Public Health, 1992–1998.

Ninety-five percent confidence intervals based on the Poisson distribution were calculated for all rates. The equality of rates across two populations was evaluated using a Z statistic, testing in a two-sided fashion whether the difference in rates was significantly different from zero (9).

Human subjects

This study was approved by the Marshfield Clinic Institutional Review Board.

RESULTS

A total of 543 potential Lyme disease cases were identified during the initial review of electronic diagnosis files and laboratory reports from the Marshfield Clinic system. After chart review, 201 illnesses were classified as probable or possible Lyme disease cases. Twelve persons were then excluded from further analyses because valid alternative diagnoses were identified during chart review. Among the 189 cases analyzed, the mean age was 37 years and nearly 53 percent were male. Sixty-eight percent of cases were classified as early Lyme disease, while 32 percent were classified as late (extracutaneous) Lyme disease.

Seven percent of the probable cases and 30 percent of the possible cases were identified for chart review on the basis of positive laboratory results alone and did not receive an International Classification of Diseases, Ninth Revision, code corresponding to Lyme disease. Six (86 percent) of these seven probable cases and 22 (85 percent) of these 26 possible cases were treated with antibiotics recommended for Lyme disease. Eighty-seven (85 percent) of the 102 probable cases had erythema migrans.

Cases reported to the state health department included 35 (34.3 percent) of the 102 probable Lyme disease cases and 17 (19.5 percent) of the 87 possible Lyme disease cases identified from patients' records. Twenty-three (6.7 percent) of the 342 noncases were also reported to the Wisconsin Division of Public Health as Lyme disease. As expected, the reporting rate for probable cases was significantly higher than the reporting rate for possible cases (p = 0.02). No significant differences in reporting were observed across time (data not shown), gender, age, or stage of disease at diagnosis (early vs. late) for probable cases (table 1). Among possible cases, late Lyme disease was three times more likely to be reported than early Lyme disease. Reporting of possible cases was most likely among the youngest patients and decreased substantially among patients older than age 20 years.

TABLE 1.

Lyme disease reporting rates by case status, gender, age, and stage of disease at diagnosis, Wisconsin, 1992–1998

 Probable cases (n = 102)
 
Possible cases (n = 87)
 
Total cases (n = 189)
 
 Reported/total (no.) p value Reported/total (no.) p value Reported/total (no.) p value 
Gender          
 Male 22/55 40 0.19 8/45 18 0.67 30/100 30 0.42 
 Female 13/47 28  9/42 21  22/89 25  
Age (years)          
 0–20 12/36 33 0.84 11/28 39 0.01 23/64 36 0.33 
 21–44 8/27 30  4/25 16  12/52 23  
 45–64 10/24 42  0/18  10/42 24  
 ≥65 5/15 33  2/16 13  7/31 23  
Stage          
 Early 30/87 34 0.93 4/42 10 0.02 34/129 26 0.60 
 Late 5/15 33  13/45 29  18/60 30  
 Probable cases (n = 102)
 
Possible cases (n = 87)
 
Total cases (n = 189)
 
 Reported/total (no.) p value Reported/total (no.) p value Reported/total (no.) p value 
Gender          
 Male 22/55 40 0.19 8/45 18 0.67 30/100 30 0.42 
 Female 13/47 28  9/42 21  22/89 25  
Age (years)          
 0–20 12/36 33 0.84 11/28 39 0.01 23/64 36 0.33 
 21–44 8/27 30  4/25 16  12/52 23  
 45–64 10/24 42  0/18  10/42 24  
 ≥65 5/15 33  2/16 13  7/31 23  
Stage          
 Early 30/87 34 0.93 4/42 10 0.02 34/129 26 0.60 
 Late 5/15 33  13/45 29  18/60 30  

Reported Lyme disease incidence remained relatively constant in Wisconsin from 1992 to 1997, but there was a modest increase in 1998 (table 2). With the exception of 1994, all annual incidence rates observed among MESA residents were higher than the statewide reported rates. For the years 1992–1998, the average MESA incidence rate of probable Lyme disease (19.1 per 100,000 per year) was more than double the state-reported rate during the same period (9.0 per 100,000 per year).

TABLE 2.

Annual Lyme disease incidence rates, 1992–1998

Year of diagnosis Passive state surveillance
 
MESA*
 
p value 
State of Wisconsin
 
Eight-county region
 
Cases (no.) Rate 95% CI* Cases (no.) Rate 95% CI Cases (no.) Rate§ 95% CI 
1992 525 10.5 9.6, 11.4 39 12.6 8.7, 16.6 14 18.4 8.8, 28.1 0.22 
1993 401 7.9 7.2, 8.7 38 12.2 8.3, 16.1 17 22.3 11.7, 33.0 0.03 
1994 409 8.0 7.2, 8.8 36 11.5 7.7, 15.2 6.6 0.8, 12.3 0.23 
1995 369 7.2 6.5, 7.9 38 12.1 8.2, 15.9 13 17.0 7.8, 26.2 0.28 
1996 396 7.7 6.9, 8.4 41 13.0 9.0, 16.9 11 14.4 5.9, 22.8 0.76 
1997 480 9.2 8.4, 10.1 75 23.6 18.3, 29.0 19 24.7 13.6, 35.9 0.86 
1998 657 12.6 11.6, 13.5 108 33.9 27.5, 40.3 23 29.9 17.7, 42.1 0.58 
Total 3,237 9.0 8.7, 9.3 375 17.0 15.3, 18.8 102 19.1 15.4, 22.8 0.31 
Year of diagnosis Passive state surveillance
 
MESA*
 
p value 
State of Wisconsin
 
Eight-county region
 
Cases (no.) Rate 95% CI* Cases (no.) Rate 95% CI Cases (no.) Rate§ 95% CI 
1992 525 10.5 9.6, 11.4 39 12.6 8.7, 16.6 14 18.4 8.8, 28.1 0.22 
1993 401 7.9 7.2, 8.7 38 12.2 8.3, 16.1 17 22.3 11.7, 33.0 0.03 
1994 409 8.0 7.2, 8.8 36 11.5 7.7, 15.2 6.6 0.8, 12.3 0.23 
1995 369 7.2 6.5, 7.9 38 12.1 8.2, 15.9 13 17.0 7.8, 26.2 0.28 
1996 396 7.7 6.9, 8.4 41 13.0 9.0, 16.9 11 14.4 5.9, 22.8 0.76 
1997 480 9.2 8.4, 10.1 75 23.6 18.3, 29.0 19 24.7 13.6, 35.9 0.86 
1998 657 12.6 11.6, 13.5 108 33.9 27.5, 40.3 23 29.9 17.7, 42.1 0.58 
Total 3,237 9.0 8.7, 9.3 375 17.0 15.3, 18.8 102 19.1 15.4, 22.8 0.31 
*

MESA, Marshfield Epidemiologic Study Area (24-ZIP code region located within the eight-county region); CI, confidence interval.

Comparison of eight-county annual rates with MESA annual rates.

Clark, Iron, Lincoln, Marathon, Price, Rusk, Taylor, and Wood counties.

Probable Lyme disease cases per 100,000 per year.

§

Reported Lyme disease cases per 100,000 per year.

The annual Lyme disease incidence rates in MESA paralleled the incidence rates reported in the surrounding eight-county region from 1992 to 1998 (table 2). The mean annual incidence rate of probable Lyme disease in MESA (19.1 per 100,000 per year) was somewhat higher than the rate calculated from Division of Public Health surveillance in the surrounding eight-county region for the same period (17.0 per 100,000 per year). Incidence rates in MESA and the eight-county region increased steeply from 1996 to 1998.

No significant differences across gender or age were observed between the mean annual rates in MESA and the eight-county region (table 3). The rates in MESA and the eight-county region were higher than the statewide reported rates in every age and gender group. Overall, Lyme disease rates were highest in males and persons 45–64 years of age. The lowest Lyme disease rates were observed in persons 21–44 years of age. Within MESA, the incidence of late Lyme disease (2.8 per 100,000 per year) was substantially lower than the incidence of early Lyme disease (16.4 per 100,000 per year). No gender differences in the stage of Lyme disease at diagnosis were observed (data not shown).

TABLE 3.

Mean annual Lyme disease incidence rates (1992–1998) stratified by age and gender

 Passive state surveillance
 
MESA*
 
p value 
 State of Wisconsin
 
Eight-country region
 
 Rate 95% CI* Rate 95% CI Rate§ 95% CI 
Gender        
 Females 8.3 7.8, 8.7 14.4 12.2, 16.7 17.6 12.6, 22.6 0.23 
 Males 10.7 10.2, 11.2 21.0 18.2, 23.7 21.0 15.4, 26.5 0.99 
Age (years)        
 0–20 9.3 8.7, 9.3 17.3 14.2, 20.5 22.0 14.8, 29.2 0.21 
 21–44 7.1 6.7, 7.6 12.8 10.2, 15.3 14.1 8.8, 19.5 0.64 
 45–64 12.2 11.3, 13.1 23.7 18.9, 28.5 26.3 15.8, 36.8 0.65 
 ≥65 10.8 9.8, 11.7 22.2 17.0, 27.4 18.0 8.9, 27.1 0.46 
 Passive state surveillance
 
MESA*
 
p value 
 State of Wisconsin
 
Eight-country region
 
 Rate 95% CI* Rate 95% CI Rate§ 95% CI 
Gender        
 Females 8.3 7.8, 8.7 14.4 12.2, 16.7 17.6 12.6, 22.6 0.23 
 Males 10.7 10.2, 11.2 21.0 18.2, 23.7 21.0 15.4, 26.5 0.99 
Age (years)        
 0–20 9.3 8.7, 9.3 17.3 14.2, 20.5 22.0 14.8, 29.2 0.21 
 21–44 7.1 6.7, 7.6 12.8 10.2, 15.3 14.1 8.8, 19.5 0.64 
 45–64 12.2 11.3, 13.1 23.7 18.9, 28.5 26.3 15.8, 36.8 0.65 
 ≥65 10.8 9.8, 11.7 22.2 17.0, 27.4 18.0 8.9, 27.1 0.46 
*

MESA, Marshfield Epidemiologic Study Area (24-ZIP code region located within the eight-county region); CI, confidence interval.

Comparison of eight-county rates with MESA rates.

Clark, Iron, Lincoln, Marathon, Price, Rusk, Taylor, and Wood counties.

§

Probable Lyme disease cases per 100,000 per year.

Reported Lyme disease cases per 100,000 per year.

The characteristics of probable cases from MESA that met the national case definition were compared with the characteristics of all patients that were actually reported to the Division of Public Health, including probable cases, possible cases, and noncases (table 4). The two groups did not differ in their distributions by gender, age, or year of diagnosis. However, early Lyme disease represented a greater proportion of probable cases and a smaller proportion of all patients reported to the Division of Public Health.

TABLE 4.

Comparison of characteristics between Marshfield Epidemiologic Study Area probable Lyme disease cases and all patients reported to the state health department of Wisconsin, 1992–1998

 Cases meeting the surveillance criteria* (n = 102)
 
All patients reported to the Division of Public Health (n = 75)
 
 No. No. 
Gender     
 Female 47 46.1 34 45.3 
 Male 55 53.9 41 54.7 
Age (years)     
 0–20 36 35.3 27 36.0 
 21–44 27 26.5 23 30.7 
 45–64 24 23.5 16 21.3 
 ≥65 15 14.7 12.0 
Stage of disease at diagnosis     
 Early 87 85.3 34 65.4 
 Late 15 14.7 18 34.6 
Year of diagnosis     
 1992 14 13.7 9.3 
 1993 17 16.7 15 20.0 
 1994 4.9 10.7 
 1995 13 12.8 8.0 
 1996 11 10.8 10 13.3 
 1997 19 18.6 10 13.3 
 1998 23 22.5 19 25.3 
 Cases meeting the surveillance criteria* (n = 102)
 
All patients reported to the Division of Public Health (n = 75)
 
 No. No. 
Gender     
 Female 47 46.1 34 45.3 
 Male 55 53.9 41 54.7 
Age (years)     
 0–20 36 35.3 27 36.0 
 21–44 27 26.5 23 30.7 
 45–64 24 23.5 16 21.3 
 ≥65 15 14.7 12.0 
Stage of disease at diagnosis     
 Early 87 85.3 34 65.4 
 Late 15 14.7 18 34.6 
Year of diagnosis     
 1992 14 13.7 9.3 
 1993 17 16.7 15 20.0 
 1994 4.9 10.7 
 1995 13 12.8 8.0 
 1996 11 10.8 10 13.3 
 1997 19 18.6 10 13.3 
 1998 23 22.5 19 25.3 
*

Classified as probable Lyme disease after medical record review.

Includes probable cases, possible cases, and noncases from the Marshfield Epidemiologic Study Area reported to the Wisconsin Division of Public Health as Lyme disease by the Marshfield Clinic.

Stage of disease at diagnosis not determined for noncases (n = 23).

DISCUSSION

Few studies that are not subject to reporting bias in the past decade have assessed the incidence of Lyme disease in well-defined populations. The incidence observed in MESA (19.1 per 100,000) is higher than the incidence rates reported in population-based studies of Lyme disease conducted in Olmsted County, Minnesota, during the 1980s (2.3 per 100,000) (2) and in northern California during the early 1990s (5.5 per 100,000) (10). However, Lyme disease incidence in MESA is much lower than the incidence rates reported in two studies conducted in high-risk island populations in New York and Massachusetts in the early 1980s (5, 6).

Lyme disease has been a reportable disease in all states since the early 1990s; however, underreporting of Lyme disease appears to be less common in Wisconsin compared with other states. In this study, approximately one third of probable Lyme disease cases were captured by the state surveillance system. Orloski et al. (4) note a 10–36 percent reporting rate in their evaluation of Lyme disease reporting by physicians in New Jersey, and Meek et al. (11) report an estimated Lyme disease reporting rate of 16 percent in their survey of Connecticut physicians. Only four of 17 confirmed cases were reported to the state health department in a study of Lyme disease in Olmsted County, Minnesota (2), and Coyle et al. (3) report estimated Lyme disease reporting rates of 10 percent in 1992 and 8 percent in 1993 in their survey of Maryland physicians. Studies of surveillance for other reportable diseases have shown that sensitivity varies from 6 percent to 90 percent (12).

Direct comparison of probable cases from MESA with the cases that had actually been reported to the Division of Public Health suggests that, despite a degree of both underreporting and overreporting, the surveillance data provide a reasonable surrogate for characterizing the age, gender, and temporal distribution of Lyme disease cases detected in a general population. Reported patients, however, were less likely to have early stage Lyme disease. This is not surprising because early stage Lyme disease is often discovered and managed in high-volume care settings, which may reduce the likelihood of reporting.

The Lyme disease rates in MESA and the surrounding eight-county region were considerably closer in magnitude than anticipated given the observed underreporting of Lyme disease cases. Several factors might have contributed to this finding. First, a large number of patients initially identified by an International Classification of Diseases, Ninth Revision, code corresponding to Lyme disease were found not to meet the case criteria upon chart review. Forty (9 percent) of the persons classified as possible cases or noncases in MESA had been reported to the state and were included in state rate calculations. Inclusion of these possible cases and noncases in Division of Public Health incidence rates likely increased the similarity between the MESA and eight-county rates.

Second, it is expected that a relatively small number of cases of Lyme disease among residents of the MESA region were diagnosed by providers outside the Marshfield Clinic system. If these cases were reported to the Division of Public Health, they would have contributed to the eight-county rate but not to the MESA rate. The most recent MESA validation efforts indicate that a very high proportion of outpatient visits among residents of the MESA region (90 percent) are captured by the Marshfield Clinic data systems. Thus, the Division of Public Health might potentially have knowledge of approximately 10 or more additional Lyme disease cases from the MESA region. Lyme disease rates for MESA based on Division of Public Health data were not directly estimable because of a lack of ZIP code information in the Division of Public Health surveillance database.

Finally, the eight-county region encompasses a larger geographic area and population base than does MESA (322,297 residents vs. 77,280 residents, respectively; 2000 US Census), and the larger eight-county region includes health care providers who are not affiliated with the Marshfield Clinic. Although differences in the underlying risk of Lyme disease between residents of MESA and other residents of the eight-county region are expected to be small, it is plausible that the care-seeking behaviors of patients and the diagnostic and reporting practices of physicians do differ across providers and health care systems. It was not possible to determine the reporting rates of such physicians outside the Marshfield Clinic system.

In both MESA and the surrounding eight-county region, the Lyme disease incidence more than doubled from 1996 to 1998, but statewide rates increased by only 64 percent. This finding is not surprising because MESA is located in a region of Wisconsin with high endemic rates of Lyme disease, while the state rates include cases and denominators from nonendemic as well as endemic areas. This finding could also have been influenced by increased Lyme disease reporting by medical providers in north-central Wisconsin from 1996 to 1998. However, because this increase is reflected both in the passive surveillance data and by medical record abstraction of our population-based cohort, it is unlikely that reporting variability was a major contributing factor.

Findings from two earlier studies of tick populations in Wisconsin suggest that the regional increase in Lyme disease incidence may be related to an increase in the population density and geographic range of Ixodes scapularis (formerly Ixodes dammini), the tick vector of B. burgdorferi (13, 14). This increase was especially noted in the north-central region of the state, while eastern Wisconsin remained relatively uninhabited by ticks during the entire 1981–1994 surveillance period. It may be that the overall state rates do not reflect the dramatic 1996–1998 increase in incidence, because the Lyme disease incidence rates calculated for the entire state were based on data from regions with relatively low and stable tick populations as well as high-risk areas. This finding emphasizes the need to examine incidence trends in smaller geographic areas representing high-risk populations.

In conclusion, these findings suggest that the Wisconsin Division of Public Health surveillance system is monitoring trends in Lyme disease incidence reasonably well across time, gender, and age categories and that underreporting of Lyme disease cases in Wisconsin is less common than reported in other published studies. This study highlights the need for continued assessment and validation of public health surveillance systems with data from population-based cohorts such as the Marshfield Epidemiologic Study Area.

Reprint requests to Dr. Edward Belongia, Epidemiology Research Center, Marshfield Medical Research and Education Foundation, 1000 N. Oak Avenue, Marshfield, WI 54449 (email: belongie@mmrf.mfldclin.edu).

Research support was provided by cooperative agreement U50/CCU514667 from the Centers for Disease Control and Prevention via the Department of Health and Social Services, Wisconsin Division of Public Health.

The authors thank Juanita Herr, Sonia Weigel, Debbie Hilgemann, Katherine Nieman, and Carolee Schueller for their data collection efforts; Lorelle Ziegelbauer for programming assistance; and Carol Beyer for assistance in manuscript preparation.

REFERENCES

1.
Orloski KA, Hayes EB, Campbell GL, et al. Surveillance for Lyme disease—United States, 1992–1998.
Mor Mortal Wkly Rep CDC Surveill Sum
 
2000
;
49
:
1
–11.
2.
Matteson EL, Beckett VL, O'Fallon WM, et al. Epidemiology of Lyme disease in Olmsted County, MN, 1975–1990.
J Rheumatol
 
1992
;
19
:
1743
–5.
3.
Coyle BS, Strickland GT, Liang YY, et al. The public health impact of Lyme disease in Maryland.
J Infect Dis
 
1996
;
173
:
1260
–2.
4.
Orloski KA, Campbell GL, Genese CA, et al. Emergence of Lyme disease in Hunterdon County, New Jersey, 1993: a case-control study of risk factors and evaluation of reporting patterns.
Am J Epidemiol
 
1998
;
47
:
391
–7.
5.
Hanrahan JP, Benach JL, Coleman JL, et al. Incidence and cumulative frequency of endemic Lyme disease in a community.
J Infect Dis
 
1984
;
150
:
489
–96.
6.
Steere AC, Taylor E, Wilson ML, et al. Longitudinal assessment of the clinical and epidemiological features of Lyme disease in a defined population.
J Infect Dis
 
1986
;
154
:
295
–300.
7.
DeStefano F, Eaker ED, Broste SK, et al. Epidemiologic research in an integrated regional medical care system: the Marshfield Epidemiologic Study Area.
J Clin Epidemiol
 
1996
;
49
:
643
–52.
8.
Kitron U, Kazmierczak JJ. Spatial analysis of the distribution of Lyme disease in Wisconsin.
Am J Epidemiol
 
1997
;
145
:
558
–66.
9.
Estève J, Benhamou E, Raymond L. Statistical methods in cancer research. Vol IV. Descriptive epidemiology. Lyon, France: International Agency for Research on Cancer, 1994.
10.
Ley C, Davila IH, Mayer NM, et al. Lyme disease in northwestern coastal California.
West J Med
 
1994
;
160
:
534
–9.
11.
Meek JI, Roberts CL, Smith EV Jr, et al. Underreporting of Lyme disease by Connecticut physicians, 1992.
J Public Health Manag Pract
 
1996
;
2
:
61
–5.
12.
Thacker SB, Berkelman RL. Public health surveillance in the United States.
Epidemiol Rev
 
1988
;
10
:
164
–90.
13.
French JB, Schell WL, Kazmierczak JJ, et al. Changes in population density and distribution of Ixodes dammini (Acari: Ixodidae) in Wisconsin during the 1980s.
J Med Entomol
 
1992
;
29
:
723
–8.
14.
Riehle M, Paskewitz SM. Ixodes scapularis (Acari: Ixodidae): status and changes in prevalence and distribution in Wisconsin between 1981 and 1994 measured by deer surveillance.
J Med Entomol
 
1996
;
33
:
933
–8.