This study was conducted in 5 counties (Caroline, Dorchester, Kent, Queen Anne, and Talbot) on the Maryland Eastern Shore, an area where LD is endemic (Table 1). The study population includes patients living in the 4 counties enrolled in Delmarva Health Plan (DHP, a managed healthcare organization) and non-DHP patients receiving health care from office-based physicians in Kent County from 1997 to 2000. Eligible patients were identified through records of encounters for LD, tick bites, insect bites, and serologic testing for LD antibodies. During 1997 and 1998, identified patients were contacted for informed consent. Patients who indicated that they did not wish to participate were excluded from our database. A cost and risk questionnaire (Appendix 1) was sent to patients who gave informed consent. The response rate of the survey was ≈22%. Interviewers then reviewed patients' charts and consulted relevant sources (e.g., hospital, physician office, laboratory) to obtain the following information: patient demographics; insurance coverage; diagnosis; symptoms; dates of onset and diagnosis; dates of tick bite exposure; dates and costs of primary provider and consultant visits; dates and costs of hospitalizations and emergency department visits; dates, results, and costs of laboratory tests; and dates and costs of antimicrobial drug treatment. All abstracted information was kept confidential. After 1999, an anonymous abstraction of medical records was approved by the institutional review board (IRB) and implemented, allowing inclusion of more patients for all 4 study years, with the exclusion of the records of those who had previously declined participation. All protocols of this study were approved by IRBs from CDC, the state of Maryland, and the University of Maryland. Those patients identified as having received an LD vaccination were not included in this study.

For the purpose of surveillance, a case of LD is defined as physician-diagnosed EM >5 cm or at least 1 late rheumatologic, neurologic, or cardiac manifestation with laboratory evidence of B. burgdorferi infection (16). These criteria were developed as an epidemiologic case definition intended for surveillance purposes only. Although such a standard may aid comparison across clinical studies and facilitate development of research, exposure history and clinical features are critical. For example, treating patients with seasonal (summer) musculoskeletal flulike symptoms in areas where LD is endemic may be clinically appropriate (12). Because the data for this study were collected directly from healthcare organizations and physicians, we used a clinical definition of LD. This definition was based on physicians' determination in the medical record, according to patients' clinical findings, tick exposure, and other relevant details (e.g., laboratory results).

In our study, LD patients were identified by using a final diagnosis code in their medical records. LD patients were then divided into 5 diagnosis groups: clinically defined early-stage LD, clinically defined late-stage LD, suspected LD, tick bite, and other related complaints. Most clinically defined early-stage LD patients had EM; some also had musculoskeletal flulike symptoms such as malaise, fatigue, headache, fever, and chills (12). In this study, clinically defined late-stage LD patients included those with later manifestations (neurologic involvement, cardiac involvement, and arthritis) and patients with chronic LD. The diagnosis groups of suspected LD, tick bite, and other related complaints involved all patients without a clear final diagnosis of LD. Suspected LD referred to patients who had some symptoms that could be indicative of LD without further evidence and thus no definitive diagnosis of LD. Patients with tick bites without symptoms were placed in the tick bite group. The diagnosis group of other related complaints included all other diagnoses that were different from the above 4 diagnosis groups, such as unknown insect bites and screening among asymptomatic persons.

We calculated the following total costs of LD: 1) direct medical costs of LD diagnosis and treatment, 2) indirect medical costs, 3) nonmedical costs, and 4) productivity losses. Intangible costs (e.g., costs incurred because of pain and suffering) were not incorporated. Consumer price index (CPI) for medical care was used to adjust all medical payments into year 2000 dollars (17). For nonmedical costs and productivity losses, we adjusted costs by using the general CPI. We took a societal perspective, which incorporates all costs and all benefits no matter who pays costs or who receives benefits.

Charges were used to estimate the direct medical cost. To determine the direct medical costs associated with LD, we used charge data from both DHP and office-based healthcare providers in Kent County. Direct medical costs of LD included costs (charges) of physician visits, consultation, serology, procedure, therapy, hospitalization/emergency room (ER), and other related costs (Appendix 2).

Indirect medical costs, nonmedical costs, and productivity losses were all acquired from a patient questionnaire used in 1997 and 1998. The questionnaire was sent to LD patients with informed consent forms. Collection of these data was restricted to those 2 years. In this study, indirect medical costs refer to extra prescription and nonprescription drug costs that patients paid out of pocket.

The patient's questionnaire also collected information on nonmedical payments made for home or health aides and miscellaneous services, such as travel (transportation) and babysitting. Each patient's transportation costs to a physician's office were estimated by using the US federal government reimbursement rate, multiplying the reported total travel miles per patient by $0.365/mile. Total travel mileage per patient was calculated by counting the number of physician visits and multiplying total visits by the distance of a round trip to the physician's office.

We used patient-reported time lost from work to estimate productivity losses due to LD on the basis of the human capital method and valued the time lost by using age- and sex-weighted productivity valuation tables (18). Because of the potential complexity of accurately answering the question, we did not ask patients to estimate the time they lost from household production. We did, however, ask patients if they paid anybody to do household tasks because their LD-related infirmities prevented them from doing those tasks. For patients <15 years of age, we assumed that their parents (usually the mother) had to take time off from their work to take care of them. Therefore, their mothers' values of lost days of work were included.

We used the following formula to estimate the average per capita cost of LD, i.e., the mean cost (direct medical costs, indirect medical costs, nonmedical costs, and productivity losses) aggregated across all diagnosis groups of patients:

Expected mean cost of a LD outcome = Σ direct medical costs, indirect medical costs, nonmedical costs, and productivity losses (Mean cost of outcome clinically defined early-stage LD, clinically defined late-stage LD, suspected LD, tick bite, and other related complaints × Probability of outcome clinically defined early-stage LD, clinically defined late-stage LD, suspected LD, tick bite, and other related complaints).

Because the distribution of cost data is often not normal, we also calculated the medians of these costs and used both mean and median to estimate the most likely per capita cost of LD on the Maryland Eastern Shore. The median cost of an LD outcome was calculated by using the following formula:

Expected median cost of a LD outcome = Σ direct medical costs, indirect medical costs, nonmedical costs, and productivity losses (Median cost of outcome clinically defined early-stage LD, clinically defined late-stage LD, suspected LD, tick bite, and other related complaints × Probability of outcome clinically defined early-stage LD, clinically defined late-stage LD, suspected LD, tick bite, and other related complaints)

Differences between annual mean direct medical costs were analyzed by using 1-way analysis of variance followed by a Bonferroni test. Differences were considered significant for p values <0.05. Additionally, we used a multivariate linear regression model to estimate the relative impact of a number of factors on the direct medical costs of LD. The ordinary linear regression (OLS) method was applied by using SAS 8.2 (SAS Institute, Cary, NC, USA) and Stata SE (StataCorp LP, College Station, TX, USA). The dependent variable was total direct medical cost per LD patient. We transformed total direct medical costs by using natural logarithms because the data were highly skewed. Independent variables of the equation included cohort year, LD diagnosis groups, diagnostic and treatment procedures, and patient characteristics (e.g., sex, age). All independent variables, except age, were binomial (yes = 1, no = 0). Baseline costs (i.e., the intercept term in the regression equation) referred to those costs accrued by a woman who had tick bite only (without EM symptoms) diagnosed in 1997 during an office visit. Such a patient had no hospital or ER stay, no serologic tests, no consultation from other physicians, no antimicrobial drug therapy, and no other procedures outside a physician office and hospital/ER. Additional direct medical costs were added or subtracted to the baseline costs for each independent variable of interest if significant (Appendix 3). We tested heteroscedasticity in Stata and corrected mild heteroscedasticity by using "robust" and "hc3" procedures. We also tested both linearity and multicollinearity in SAS and Stata.

Download the PDF athttp://wwwnc.cdc.gov/eid/pdfs/12/4/05-0602-Techapp.pdf.

For each patient, direct medical costs of Lyme disease (LD) and relevant complaints were calculated as follows:Y_i = Σ C_it

Where Y is the total direct medical costs for patient i and C_it is the direct medical cost of type t for patient i. Different types (t) of direct medical costs involved in this study included physician visits, consultation, serology, procedure, therapy, hospitalization/emergency room, and other relevant costs.

Cost of physician visits: refers to total visit charges per patient with primary provider or physician managing LD. It includes charges for procedures only performed in the office at the time of the visit. All other procedures were covered in the cost of procedure category.

Cost of consultation: refers to total charges per patient accrued at the time of consultation including only those procedures (e.g., laboratory procedures or electrocardiography) done in the office at that time. All other procedures ordered by the consult but done off site (e.g., magnet resonance imaging or computerized tomography) were covered in the cost of procedure category.

Cost of serology: refers to total charges of LD serology tests (e.g., enzyme-linked immunosorbent assay or Western blotting test) for each patient.

Cost of procedure: refers to total charges per patient of all other procedures not covered in office visits, consultation, or hospital and emergency room.

Cost of therapy: refers to total charges per patient of therapy including antibiotics and additional costs associated with intravenous therapy (e.g., other medications such as saline solutions, heparin, local anesthetics; registered nurse home visits; X-ray; etc.)

Cost of hospitalization/emergency room: refers to total hospital and/or emergency room charges per patient. It includes physician fees and ambulance fees.

Other relevant costs: refer to other appropriate charges for each patient such as charges for additional laboratory tests.

We used a multivariate linear regression model to estimate the relative impact of a number of factors on the direct medical costs of Lyme disease (LD).Y_i = X_i β + ε_i

where Y_i, the dependent variable, is the total direct medical costs for patient i and X_i is a vector of covariates. ε_i is a mean-zero random error. Baseline costs came from the intercept term, which refers to a female patient who had tick bite only (without erythema migrans symptoms) diagnosed in 1997 during an office visit. Such a patient had no hospital or emergency room stay, no serologic tests, no consultation from other physicians, no antimicrobial drug therapy, and no other procedures outside physician office and hospital/emergency room. Additional direct medical costs were added or subtracted to the baseline costs for each independent variable of interest if statistically significant. Therefore, total direct medical costs per LD patient = baseline costs + cost difference of different patient diagnosis group (early stage LD, late stage LD, suspected LD, other LD relevant complain vs. tick bite ) + costs of serologic test + costs of other procedure outside physician office and hospital/emergency room + costs of hospital and/or emergency room stay + costs of consultation from other physicians + costs of therapy + miscellaneous costs + cost difference with or without erythema migrans + cost difference of gender + cost difference of each additional year of age + cost difference of cohort year, if statistically significant.