Objective: In this study, it was aimed to compare the performance results of the methods modeled by using generalized additive models (GAM) and distributed lag non-linear models (DLNM) methods from real data of three different outcome variables of three separate diseases related to air pollution. Material and Methods: The data were retrospectively obtained from three hospitals under the General Secretariat of Gaziantep province public hospitals for a total of 1,916 days between 01 January 2009 and 31 March 2014. Response variables were number of the emergency unit admission, hospitalization and mortality due to asthma, chronic obstructive pulmonary disease (COPD) and pneumonia. The response variables were estimated by GAM and DLNM methods by building four different models and the performances of the models were compared. Results: When the estimation performances of GAM and DLNM methods are compared for each of the dependent variables in the prediction of hospitalizations due to asthma, GAM model IV [Akaike Information Criteria (AIC) (4,280.63)] values were found to perform the best. It was observed that DLNM method performed better than GAM in models established for the prediction of almost all other dependent variables. For when compare the odds ratio (OR) plot estimated on particulate matter (PM10); it was seen that GAM method made predictions with lower standard error compared to DLNM methods. Conclusion: When the models created with each dependent variable were compared; it was generally observed that superior performance was obtained from the DLNM method. However, the lowest standard error in the OR charts were observed in the models using the GAM method.
Keywords: Generalized additive models; distributed lag non-linear models; environmental epidemiology; epidemiologic modeling
Amaç: Bu çalışmada, hava kirliliğiyle ilişkili olduğu düşünülen, 3 farklı hastalığın 3 farklı sonuç değişkeni, gerçek veriler üzerinden genelleştirilmiş eklemeli modeller (GAM) ve dağıtılmış gecikmeli doğrusal olmayan modeller (DLNM) yöntemleri kullanılarak, modellenmesi ve model performanslarının karşılaştırılması amaçlanmıştır. Gereç ve Yöntemler: Gaziantep ili Kamu Hastaneleri Genel Sekreterliğine bağlı 3 hastaneden 1 Ocak 2009-31 Mart 2014 tarihleri arasında toplam 1.916 gün boyunca geriye dönük olarak izlenmesiyle elde edilen veriler kullanılarak oluşturuldu. Cevap değişkenleri; astım, kronik obstrüktif akciğer hastalığı (KOAH) ve pnömoni nedeniyle 'acil servislerine başvurular', 'hastanede yatış' ve 'mortalite' sayısı şeklindedir. Tahminlerde GAM ve DLNM yöntemleri kullanılmış, aynı yöntemle kurulan 4 farklı modelden en iyi performansa sahip model, ilgili yöntem için karşılaştırma modeli olarak kullanılmıştır. Bulgular: Bağımlı değişkenlerin her biri için GAM ve DLNM metotlarının performansları karşılaştırıldığında, astım tanısı ile hastanede yatan sayısının tahmin edildiği IV. modelde GAM [Akaike Information Criteria (AIC) (4.280,63)] yöntemi en iyi performansı göstermiştir. Geriye kalan bağımlı değişkenlerin neredeyse tümünün tahmini için oluşturulan modellerde ise DLNM yönteminin GAM'den daha iyi performans gösterdiği gözlenmiştir. Partikül maddesi (PM10) üzerinden tahmin edilen göreceli olasılıklar oranı (OR) değerleri için GAM ve DLNM yöntemlerine göre daha düşük standart hataya sahip olduğu görülmüştür. Sonuç: İncelenen bağımlı değişkenlerle oluşturulan modeller kıyaslandığında, genelde DLNM yönteminin, GAM yönteminden daha iyi performans gösterdiği gözlenmiştir. Fakat PM10 üzerinden tahmin edilen OR grafiklerinde GAM yönteminin, DLNM yöntemlerinde daha düşük standart hata ve dolayısıyla daha dar güven aralığına sahip tahminler yaptığı görülmüştür.
Anahtar Kelimeler: Genelleştirilmiş eklemeli modeller; dağıtılmış gecikmeli doğrusal olmayan modeller; çevre epidemiyolojisi; epidemiyolojik modelleme
- Sundell J. On the history of indoor air quality and health. Indoor Air. 2004;14 Suppl 7:51-8. [Crossref] [PubMed]
- Hales S, Blakely T, Woodward A. Air pollution and mortality in New Zealand: cohort study. J Epidemiol Community Health. 2012;66(5):468-73. [Crossref] [PubMed] [PMC]
- Artun GK, Polat N, Yay OD, Üzmez ÖÖ, Arı A, Tuygun GT, et al. An integrative approach for determination of air pollution and its health effects in a coal fired power plant area by passive sampling. Atmos Environ. 2017;150:331-45. [Crossref]
- Guarnieri M, Balmes JR. Outdoor air pollution and asthma. Lancet. 2014;3;383(9928):1581-92. PMID: ; PMCID: [Crossref] [PubMed] [PMC]
- Patz JA, Engelberg D, Last J. The effects of changing weather on public health. Annu Rev Public Health. 2000;21:271-307. [Crossref] [PubMed]
- McCullagh P. Generalized linear models. Eur J Oper Res. 1984;16(3):285-92. [Crossref]
- Zuur A, Ieno EN, Smith GM. Additive and generalised additive modelling. Analyzing Ecological Data. 2nd ed. New York: Springer; 2007. p.120-33.
- Wood S, Augustin NH. GAMs with integrated model selection using penalized regression splines and applications to environmental modelling. Ecol Model. 2002;157(2-3):157-77. [Crossref]
- Hastie TJ, Tibshirani RJ. Smoothing. Generalized Additive Models. 1nd ed. New York: Routledge; 1990. p.38-65.
- Austin PC. A comparison of regression trees, logistic regression, generalized additive models, and multivariate adaptive regression splines for predicting AMI mortality. Stat Med. 200710;26(15):2937-57. [Crossref] [PubMed]
- Wood S. Generalized Additive Models: An Introduction with R. 2nd ed. Florida: CRC Press; 2017. p.11-138. [Crossref]
- Guisan A, Edwards TC, Hastie T. Generalized linear and generalized additive models in studies of species distributions: setting the scene. Ecol Model. 2002;157(2):89-100. [Crossref]
- Yee TW, Mitchell ND. Generalized additive models in plant ecology. J Veg Sci. 1991;2(5):587-602. [Crossref]
- Wood SN. Stable and efficient multiple smoothing parameter estimation for generalized additive models. J Am Stat Assoc. 2004;99(467):673-86. [Link]
- Hastie T, Tibshirani R. Generalized additive models: some applications. J Am Stat Assoc. 1987;82(398):371-86. [Link]
- Bayram H, Bogan M, Kul S, Oktay MM, Akpinar-Elci M, Al B, et al. Effects of desert dust storms and meteorological variables on emergency room visits and hospitalization due to COPD in South East Turkey. Am J Respir Crit Care Med. 2015;191:A6167. [Link]
- Xia Y, Tong H. Cumulative effects of air pollution on public health. StatMed. 2006;25(20):3548-59. [Crossref] [PubMed]
- Dobson AJ. An Introduction to Generalized Linear Models, 2nd ed: Florida: CRC Press; 2010. p.11-90.
- Gasparrini A. Distributed lag linear and non-linear models in R: the package dlnm. J Stat Softw. 2011;43(8):1-20. [Crossref] [PubMed] [PMC]
- Gasparrini A, Armstrong B, Kenward MG. Distributed lag non-linear models. Stat Med. 2010;20;29(21):2224-34. [Crossref] [PubMed] [PMC]
- Rodriguez G. Smoothing and non-parametric regression. Springer; 2001:1-12. [Link]
- Ma W, Sun X, Song Y, Tao F, Feng W, He Y, et al. Applied mixed generalized additive model to assess the effect of temperature on the incidence of bacillary dysentery and its forecast. PLoS One. 2013;29;8(4):e62122. [Crossref] [PubMed] [PMC]
- Yang L, Qin G, Zhao N, Wang C, Song G. Using a generalized additive model with autoregressive terms to study the effects of daily temperature on mortality. BMC Med Res Methodol. 2012;12(1):165. [Crossref] [PubMed] [PMC]
- Wang Q, Gao C, Wang H, Lang L, Yue T, Lin H, et al. Ischemic stroke hospital admission associated with ambient temperature in Jinan, China. PLoS One. 2013;19;8(11):e80381. [Crossref] [PubMed] [PMC]
- Peng RD, Dominici F, Louis TA. Model choice in time series studies of air pollution and mortality. J R Stat Soc A Stat. 2006;169(2):179-203. [Crossref]
- Masselot P, Chebana F, Bélanger D, St-Hilaire A, Abdous B, Gosselin P, et al. Aggregating the response in time series regression models, applied to weather-related cardiovascular mortality. Sci Total Environ. 2018;628:217-25. [Crossref] [PubMed]
- Gasparrini A. Modeling exposure-lag-response associations with distributed lag non-linear models. Stat Med. 2014;28;33(5):881-99. Erratum in: Stat Med. 2014;28;33(5):900. [Crossref] [PubMed] [PMC]
.: Process List