Peridomestic and community-wide landscape risk factors for Lyme disease across a range of community contexts in Pennsylvania

Katherine A. Moon, Jonathan Pollak, Melissa N. Poulsen, Annemarie G. Hirsch, Joseph DeWalle, Christopher David Heaney, John Aucott, Brian S Schwartz

Research output: Contribution to journalArticle

Abstract

Land use and forest fragmentation are thought to be major drivers of Lyme disease incidence and its geographic distribution. We examined the association between landscape composition and configuration and Lyme disease in a population-based case control study in the Geisinger health system in Pennsylvania. Lyme disease cases (n = 9657) were identified using a combination of diagnosis codes, laboratory codes, and antibiotic orders from electronic health records (EHRs). Controls (5:1) were randomly selected and frequency matched on year, age, and sex. We measured six landscape variables based on prior literature, derived from the National Land Cover Database and MODIS satellite imagery: greenness (normalized difference vegetation index), percent forest, percent herbaceous, forest edge density, percent forest-herbaceous edge, and mean forest patch size. We assigned landscape variables within two spatial contexts (community and ½-mile [805 m] Euclidian residential buffer). In models stratified by community type, landscape variables were modeled as tertiles and flexible splines and associations were adjusted for demographic and clinical covariates. In general, we observed positive associations between landscape metrics and Lyme disease, except for percent herbaceous, where associations differed by community type. For example, compared to the lowest tertile, individuals with highest tertile of greenness in residential buffers had higher odds of Lyme disease (odds ratio: 95% confidence interval [CI]) in townships (1.73: 1.55, 1.93), boroughs (1.70: 1.40, 2.07), and cities (3.71: 1.74, 7.92). Similarly, corresponding odds ratios (95% CI) for forest edge density were 1.34 (1.22, 1.47), 1.56 (1.33, 1.82), and 1.90 (1.13, 3.18). Associations were generally higher in residential buffers, compared to community, and in cities, compared to boroughs or townships. Our results reinforce the importance of peridomestic landscape in Lyme disease risk, particularly measures that reflect human interaction with tick habitat. Linkage of EHR data to public data on residential and community context may lead to new health system-based approaches for improving Lyme disease diagnosis, treatment, and prevention.

Original languageEnglish (US)
Article number108649
JournalEnvironmental research
Volume178
DOIs
StatePublished - Nov 1 2019

Fingerprint

Lyme disease
Lyme Disease
risk factor
Health
Buffers
Electronic Health Records
forest edge
confidence interval
Satellite Imagery
disease treatment
Odds Ratio
Confidence Intervals
disease incidence
Satellite imagery
Clinical Laboratory Techniques
tick
patch size
Ticks
Land use
NDVI

Keywords

  • Forest fragmentation
  • Land cover
  • Lyme disease
  • Tick-borne disease

ASJC Scopus subject areas

  • Biochemistry
  • Environmental Science(all)

Cite this

Peridomestic and community-wide landscape risk factors for Lyme disease across a range of community contexts in Pennsylvania. / Moon, Katherine A.; Pollak, Jonathan; Poulsen, Melissa N.; Hirsch, Annemarie G.; DeWalle, Joseph; Heaney, Christopher David; Aucott, John; Schwartz, Brian S.

In: Environmental research, Vol. 178, 108649, 01.11.2019.

Research output: Contribution to journalArticle

@article{02e79f2f55d04d3c98161bfd557c72c5,
title = "Peridomestic and community-wide landscape risk factors for Lyme disease across a range of community contexts in Pennsylvania",
abstract = "Land use and forest fragmentation are thought to be major drivers of Lyme disease incidence and its geographic distribution. We examined the association between landscape composition and configuration and Lyme disease in a population-based case control study in the Geisinger health system in Pennsylvania. Lyme disease cases (n = 9657) were identified using a combination of diagnosis codes, laboratory codes, and antibiotic orders from electronic health records (EHRs). Controls (5:1) were randomly selected and frequency matched on year, age, and sex. We measured six landscape variables based on prior literature, derived from the National Land Cover Database and MODIS satellite imagery: greenness (normalized difference vegetation index), percent forest, percent herbaceous, forest edge density, percent forest-herbaceous edge, and mean forest patch size. We assigned landscape variables within two spatial contexts (community and ½-mile [805 m] Euclidian residential buffer). In models stratified by community type, landscape variables were modeled as tertiles and flexible splines and associations were adjusted for demographic and clinical covariates. In general, we observed positive associations between landscape metrics and Lyme disease, except for percent herbaceous, where associations differed by community type. For example, compared to the lowest tertile, individuals with highest tertile of greenness in residential buffers had higher odds of Lyme disease (odds ratio: 95{\%} confidence interval [CI]) in townships (1.73: 1.55, 1.93), boroughs (1.70: 1.40, 2.07), and cities (3.71: 1.74, 7.92). Similarly, corresponding odds ratios (95{\%} CI) for forest edge density were 1.34 (1.22, 1.47), 1.56 (1.33, 1.82), and 1.90 (1.13, 3.18). Associations were generally higher in residential buffers, compared to community, and in cities, compared to boroughs or townships. Our results reinforce the importance of peridomestic landscape in Lyme disease risk, particularly measures that reflect human interaction with tick habitat. Linkage of EHR data to public data on residential and community context may lead to new health system-based approaches for improving Lyme disease diagnosis, treatment, and prevention.",
keywords = "Forest fragmentation, Land cover, Lyme disease, Tick-borne disease",
author = "Moon, {Katherine A.} and Jonathan Pollak and Poulsen, {Melissa N.} and Hirsch, {Annemarie G.} and Joseph DeWalle and Heaney, {Christopher David} and John Aucott and Schwartz, {Brian S}",
year = "2019",
month = "11",
day = "1",
doi = "10.1016/j.envres.2019.108649",
language = "English (US)",
volume = "178",
journal = "Environmental Research",
issn = "0013-9351",
publisher = "Academic Press Inc.",

}

TY - JOUR

T1 - Peridomestic and community-wide landscape risk factors for Lyme disease across a range of community contexts in Pennsylvania

AU - Moon, Katherine A.

AU - Pollak, Jonathan

AU - Poulsen, Melissa N.

AU - Hirsch, Annemarie G.

AU - DeWalle, Joseph

AU - Heaney, Christopher David

AU - Aucott, John

AU - Schwartz, Brian S

PY - 2019/11/1

Y1 - 2019/11/1

N2 - Land use and forest fragmentation are thought to be major drivers of Lyme disease incidence and its geographic distribution. We examined the association between landscape composition and configuration and Lyme disease in a population-based case control study in the Geisinger health system in Pennsylvania. Lyme disease cases (n = 9657) were identified using a combination of diagnosis codes, laboratory codes, and antibiotic orders from electronic health records (EHRs). Controls (5:1) were randomly selected and frequency matched on year, age, and sex. We measured six landscape variables based on prior literature, derived from the National Land Cover Database and MODIS satellite imagery: greenness (normalized difference vegetation index), percent forest, percent herbaceous, forest edge density, percent forest-herbaceous edge, and mean forest patch size. We assigned landscape variables within two spatial contexts (community and ½-mile [805 m] Euclidian residential buffer). In models stratified by community type, landscape variables were modeled as tertiles and flexible splines and associations were adjusted for demographic and clinical covariates. In general, we observed positive associations between landscape metrics and Lyme disease, except for percent herbaceous, where associations differed by community type. For example, compared to the lowest tertile, individuals with highest tertile of greenness in residential buffers had higher odds of Lyme disease (odds ratio: 95% confidence interval [CI]) in townships (1.73: 1.55, 1.93), boroughs (1.70: 1.40, 2.07), and cities (3.71: 1.74, 7.92). Similarly, corresponding odds ratios (95% CI) for forest edge density were 1.34 (1.22, 1.47), 1.56 (1.33, 1.82), and 1.90 (1.13, 3.18). Associations were generally higher in residential buffers, compared to community, and in cities, compared to boroughs or townships. Our results reinforce the importance of peridomestic landscape in Lyme disease risk, particularly measures that reflect human interaction with tick habitat. Linkage of EHR data to public data on residential and community context may lead to new health system-based approaches for improving Lyme disease diagnosis, treatment, and prevention.

AB - Land use and forest fragmentation are thought to be major drivers of Lyme disease incidence and its geographic distribution. We examined the association between landscape composition and configuration and Lyme disease in a population-based case control study in the Geisinger health system in Pennsylvania. Lyme disease cases (n = 9657) were identified using a combination of diagnosis codes, laboratory codes, and antibiotic orders from electronic health records (EHRs). Controls (5:1) were randomly selected and frequency matched on year, age, and sex. We measured six landscape variables based on prior literature, derived from the National Land Cover Database and MODIS satellite imagery: greenness (normalized difference vegetation index), percent forest, percent herbaceous, forest edge density, percent forest-herbaceous edge, and mean forest patch size. We assigned landscape variables within two spatial contexts (community and ½-mile [805 m] Euclidian residential buffer). In models stratified by community type, landscape variables were modeled as tertiles and flexible splines and associations were adjusted for demographic and clinical covariates. In general, we observed positive associations between landscape metrics and Lyme disease, except for percent herbaceous, where associations differed by community type. For example, compared to the lowest tertile, individuals with highest tertile of greenness in residential buffers had higher odds of Lyme disease (odds ratio: 95% confidence interval [CI]) in townships (1.73: 1.55, 1.93), boroughs (1.70: 1.40, 2.07), and cities (3.71: 1.74, 7.92). Similarly, corresponding odds ratios (95% CI) for forest edge density were 1.34 (1.22, 1.47), 1.56 (1.33, 1.82), and 1.90 (1.13, 3.18). Associations were generally higher in residential buffers, compared to community, and in cities, compared to boroughs or townships. Our results reinforce the importance of peridomestic landscape in Lyme disease risk, particularly measures that reflect human interaction with tick habitat. Linkage of EHR data to public data on residential and community context may lead to new health system-based approaches for improving Lyme disease diagnosis, treatment, and prevention.

KW - Forest fragmentation

KW - Land cover

KW - Lyme disease

KW - Tick-borne disease

UR - http://www.scopus.com/inward/record.url?scp=85071119620&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85071119620&partnerID=8YFLogxK

U2 - 10.1016/j.envres.2019.108649

DO - 10.1016/j.envres.2019.108649

M3 - Article

C2 - 31465993

AN - SCOPUS:85071119620

VL - 178

JO - Environmental Research

JF - Environmental Research

SN - 0013-9351

M1 - 108649

ER -