The Relationship between the Number of Community Acquired Pneumonia Patients and the Weather among the Patients Who Visit ER: A Poisson Regression with Variable Selection Via Elastic net

Choi, Hanzo; Park, Sanghyun; Kwak, Myoung Kwan; Pyo, Changhae; Park, Keunhong; Kim, Hahnbom; Shin, Seoungyul

J Korean Soc Emerg Med > Volume 27(1); 2016 > Article

Original Article

Journal of The Korean Society of Emergency Medicine 2016;27(1): 22-29.

응급실 내원 환자 중 지역사회획득 폐렴 발생 환자수와 기상 요소와의 연관성: 엘라스틱 넷 변수 선택 기법을 이용한 포아송 회귀분석

최한조, 박상현, 곽명관, 표창해, 박근홍, 김한범, 신승열

서울의료원 응급의학과

The Relationship between the Number of Community Acquired Pneumonia Patients and the Weather among the Patients Who Visit ER: A Poisson Regression with Variable Selection Via Elastic net

Hanzo Choi, Sanghyun Park, Myoung Kwan Kwak, Changhae Pyo, Keunhong Park, Hahnbom Kim, Seoungyul Shin

Department of Emergency Medicine, Seoul Medical Center, 156 Sinnae-ro, Jungnang-gu, Seoul, Republic of Korea

Correspondence

Sanghyun Park ,Tel: 02) 2276-7424, Fax: 02) 2276-7418, Email: emergency70@hanmail.net,

Received: July 20, 2015; Revised: July 20, 2015 Accepted: October 19, 2015. Published online: February 29, 2016.

ABSTRACT

Purpose:
This study shows the relationship between meteorological factors and the number of community acquired pneumonia (CAP) patients in the emergency room and lag effect of meteorological factors affecting CAP.
Method:
A retrospective study was conducted. Patients diagnosed with CAP in the emergency room between January 2012 and December 2014 were enrolled. The patients were over 18 years old and lived in Seoul, Korea. Meteorological factors (highest daily temperature, lowest temperature, mean temperature, diurnal temperature, rainfall, relative humidity, amount of sunshine, and powdery dust under 10 μg/m³ (PM10)) between December 2011 and December 2014 in Seoul were acquired from the Korea Meteorological Administration. Multiple Poisson regression (Generalized Linear Model) was used with daily patient’s number of CAP as the response variable and meteorological factors as the explanatory variable. Variable selection was performed via Elastic net.
Results:
A total of 568 CAP patients were checked. Highest temperature (before 6 days), rainfall (before 1 day), relative humidity (before 20, 15, 13, 6, 2, and 1 days), and PM10 (before 27, 24, 17, and 13 days) showed relationship and lag effect with the incidence of CAP.
Conclusion:
This study showed that meteorological factors (highest temperature, rainfall, relative humidity, and PM10) had relationship and lag effect with the incidence of CAP. We can make a prediction model with health weather index for prevention of CAP and redistribution of medical facilities and resources.

Key words: Pneumonia, Weather, Regression analysis