Contact Information:
Jeffrey Lidicker, MA
Datastat Services
PO Box 523
Pipersville, PA 18947-0523
215-813-8361

INTRODUCTION
Lyme disease is one of a handful of tick-borne diseases available in North America. Originally discovered in Old Lyme Connecticut, this disease now has the distinction of being the most common of all tick borne diseases.1 In fact, among all vector-borne (includes, mosquitoes, leaches, etc.) diseases , Lyme disease is rapidly emerging in incidence.1 The Federal Center for Disease Control (CDC) recognizes that their counts, being subject to strict guidelines, may underestimate the true number of cases by ten fold.2 Of the 23,763 cases confirmed by the CDC in the United States for the year 2002 alone, it is estimated that 90% of them occur in the North Eastern most states.3 As expenditures for health care in the United States continue to soar, the import of disease incidence rates only gets more serious. In order to better understand the societal and economic burden of the most common vector borne disease, accurate predictions for the number of cases per year are warranted. The intent of this study is to predict up to 2012 the number of CDC reported Lyme disease cases per year in the US.

EXECUTIVE SUMMARY
The estimated number of CDC reported Lyme disease cases in the US in the year 2012 will be between 25,000 and 40,000 cases per year (see Table 10 and Figure 3) with a 95% level of confidence. In the minimum scenario, this represents a small increase from 2002 of only about 1,000 cases per year, but an increase of 15,900 per year in the maximum scenario (67% or 2/3 increase in ten years). However, a total of 32,000 cases per year are expected, producing 8,000 cases more in less than ten years (35% or 1/3 increase). This represents a steady increase in the number of cases per year and a subsequent additional burden on the health care resources. Considering that the CDC has suspected that their counts may under represent the actual number of Lyme disease cases by as much as ten fold, this prediction could actually mean as much as an additional 160,000 cases per year in ten years. The final Time Series estimation model performed well and is of type ARIMA(1, 1, 0) and represented by:

[ (1 + 0.74131*B)(1 ‚ B)Zt = 1,065.9 + at ]

Methods:
The data for the study was garnished from the Federal CDC website were annual counts of cases of Lyme disease in the United States from 1982 through 2002.4,5,6 The CDC only started recording these counts in 1982. Use of the standard Time Series ARIMA analysis is used to identify the model. Parameter estimation and forecasts are done using Maximum Likelihood Estimation. The software used for all statistical analysis is SAS7.

RESULTS

The Data:
The original data has a clear upward trend and appears in Table 1 with the graph in Figure 1. In an effort to eliminate the trend, the first differences were calculated and appear in Figure 2. The differenced data appears to be stationary.

Table 1: Data on Lyme Disease Cases in the US from CDC

Year - Cases
1982 - 400
1983 - 600
1984 - 1,700
1985 - 2,900
1986 - 1,600
1987 - 2,700
1988 - 5,000
1989 - 8,800
1990 - 7,950
1991 - 9,470
1992 - 9,908
1993 - 8,257
1994 - 13,043
1995 - 11,700
1996 - 16,455
1997 - 12,801
1998 - 16,801
1999 - 16,273
2000 - 17,730
2001 - 17,029
2002 - 23,763

Figure 1: Lyme Disease Cases Data from CDC


Figure 2: Lyme Disease Cases Data Differenced from CDC

* The differenced data appears to be stationary except for the last data point, which may reflect different reporting guidelines and less strict definitions for Lyme disease implemented by the CDC in that year.

Model Identification:
In order to determine the underlying model, an ACF and PACF analysis was performed on both the original data and the differenced data. The results for the original data appear in Table 2 and Table 3 while the results for the differenced data appear in Table 4 and Table 5.

From the results of the analysis, it is clear the differencing is necessary as the ACF for the original data decays very slowly but then goes back up in magnitude again in a cyclic nature. The ACF for the differenced data decays slowly but does not come back. Since, the PACF for the differenced data cuts off after lag 1 (and the ACF decays slowly), the model is determined to be a classic example of an ARIMA(1,1,0).

1 | 2 | 3 --->NEXT PAGE [PDF download]
[Abstract][Press Release]
Lyme Disease Cases in the US Projected Through 2012:
Copyright 2005, DataStat Services, WILDER Network, Inc.

Basic Life Support:
Lifesaving CPR alternative.


North Carolina Lyme
Disease Foundation


WELCOME! WILDER
HOTLINE! Public Access. Join our international forum! Enter...


Get the scoop here:
** Featured Editors **


Bowen Microbiological
Lab tests for tick-borne
diseases -worldwide More...

Lyme disease bacteria:
A Motile Menace

Support WILDER Network

All of our services are provided for FREE! Our directors are supporters and volunteers! Your contribution of ANY AMOUNT is appreciated! Kindly click the DONATE button on any page of our web site or send your contribution to: WILDER Network - BOX 116, Forestville, CA 95436. Your support is much needed and greatly appreciated! WILDER Network, Inc. is a non-profit organization.

• < Newsletter, WILD News !
• < Visit the GUESTBOOK !
• < See **Dr Ted's books !
• < Visit the little gift SHOP !
• < Support WILDER Network !