## Hierarchical Bivariate Time Series Models: A Combined Analysis of the Effects of Particulate Matter on Morbidity and Mortality

### BibTeX

@MISC{F_hierarchicalbivariate,

author = {Dominici F and Zanobetti A and Zeger S. L and Schwartz J and Samet J. M},

title = {Hierarchical Bivariate Time Series Models: A Combined Analysis of the Effects of Particulate Matter on Morbidity and Mortality},

year = {}

}

### OpenURL

### Abstract

In this paper we develop a hierarchical bivariate time series model to characterize the relationship between particulate matter less than 10 microns in aerodynamic diameter (P M10) and both mortality and hospital admissions for cardiovascular diseases. The model is applied to time series data on mortality and morbidity for 10 metropolitan areas in the United States from 1986 to 1993. We postulate that these time series should be related through a shared relationship with P M10. At the first stage of the hierarchy, we fit two seemingly unrelated Poisson regression models to produce city-specific estimates of the log relative rates of mortality and morbidity associated with exposure to P M10 within each location. The sample covariance matrix of the estimated log relative rates is obtained using a novel generalized estimating equation approach that takes into account the correlation between the mortality and morbidity time series. At the second stage, we combine information across locations to estimate overall log relative rates of mortality and morbidity and variation of the rates across cities. Using the combined information across the 10 locations we find that a 10 µg/m3 increase in average P M10 at the current day and previous day is associated with a 0.26 % increase in mortality

### Citations

1696 | Generalized Additive Models
- Hastie, Tibshirani
- 1990
(Show Context)
Citation Context ...i et al., 2003a). Poisson time series regression models (Liang and Zeger, 1986; Zeger and Liang, 1992; Fahrmeir and Tutz, 2001; McNeney and Petkau, 1994; Albert, 1999) or generalized additive models (=-=Hastie and Tibshirani, 1990-=-) have been widely used to analyze univariate time series data of air pollution and health in selected locations (Dockery and Pope, 1994; Schwartz, 1995; American Thoracic Society, 1996a,b; Korrick et... |

831 |
Longitudinal Data Analysis using Generalized Linear Models.” Biometrika 73(1
- Liang, Zeger
- 1986
(Show Context)
Citation Context ...P M10 concentrations were positively associated with mortality and morbidity outcomes on average across locations (Samet et al., 2000b; Dominici et al., 2003a). Poisson time series regression models (=-=Liang and Zeger, 1986-=-; Zeger and Liang, 1992; Fahrmeir and Tutz, 2001; McNeney and Petkau, 1994; Albert, 1999) or generalized additive models (Hastie and Tibshirani, 1990) have been widely used to analyze univariate time ... |

372 |
Approximate inference in generalized linear mixed models
- Breslow, Clayton
- 1993
(Show Context)
Citation Context ... et al., 2000a; Hwang and Chan, 2001) in which site-specific data on air pollution and health are assembled under a common framework. Hierarchical models (DuMouchel and Harris, 1983; DuMouchel, 1990; =-=Breslow and Clayton, 1993-=-; Carlin and Louis, 1996) are a suitable approach for analyzing univariate time series data from multiple locations (Dominici et al., 2000, 2002; Hwang and Chan, 2001). In comparison to analyses of da... |

347 |
Multivariate Statistical Modelling Based on Generalized Linear Models
- Fahrmeir, Tutz
- 1994
(Show Context)
Citation Context ...d with mortality and morbidity outcomes on average across locations (Samet et al., 2000b; Dominici et al., 2003a). Poisson time series regression models (Liang and Zeger, 1986; Zeger and Liang, 1992; =-=Fahrmeir and Tutz, 2001-=-; McNeney and Petkau, 1994; Albert, 1999) or generalized additive models (Hastie and Tibshirani, 1990) have been widely used to analyze univariate time series data of air pollution and health in selec... |

346 |
Bayes and Empirical Bayes Methods for Data Analysis
- Carlin, Louis
- 1996
(Show Context)
Citation Context ...han, 2001) in which site-specific data on air pollution and health are assembled under a common framework. Hierarchical models (DuMouchel and Harris, 1983; DuMouchel, 1990; Breslow and Clayton, 1993; =-=Carlin and Louis, 1996-=-) are a suitable approach for analyzing univariate time series data from multiple locations (Dominici et al., 2000, 2002; Hwang and Chan, 2001). In comparison to analyses of data from a single site, p... |

131 |
Air pollution and daily mortality: a review and meta analysis.” Environmental Research 64
- Schwartz
- 1994
(Show Context)
Citation Context ...he mid-twentieth century by a series of well-documented air pollution “disasters” in the US and Europe. By the early 1990’s, time series studies with data from single locations (Dockery et al., 1993; =-=Schwartz, 1994-=-; American Thoracic Society, 1996a; Pope, 2000), showed that air pollution, even at much lower concentrations than existed during the earlier disasters, was associated with increased rates of mortalit... |

74 |
Acute respiratory effects of particulate air-pollution. Annual Review of Public Health 15,Metals associated with both the water-soluble and insoluble fractions of an ambient air pollution particle catalyze an oxidative stress
- Dockery, Pope
- 1994
(Show Context)
Citation Context ...etkau, 1994; Albert, 1999) or generalized additive models (Hastie and Tibshirani, 1990) have been widely used to analyze univariate time series data of air pollution and health in selected locations (=-=Dockery and Pope, 1994-=-; Schwartz, 1995; American Thoracic Society, 1996a,b; Korrick et al., 1998). Critics of single-site studies questioned the choice of particular cities and asked if models had been selected that gave e... |

60 |
Combining evidence on air pollution and daily mortality from the twenty largest US cities: a hierarchical modeling strategy (with discussion). J R Stat Soc A 2000;163:263–302
- Dominici, JM, et al.
(Show Context)
Citation Context ... NMMAPS mortality analyses estimated associations between all-cause and cause-specific mortality and particulate matter less than 10 microns in aerodynamic diameter (P M10) for 90 cities in the U.S. (=-=Samet et al., 2000-=-b; Dominici et al., 2002). The NMMAPS morbidity analyses estimated associations between hospitalization in the elderly and P M10 for 14 cities in the U.S. (Samet et al., 2000b; Schwartz, 2000; Zanobet... |

55 |
Fine particulate air pollution and mortality
- Samet, Dominici, et al.
- 2000
(Show Context)
Citation Context ... NMMAPS mortality analyses estimated associations between all-cause and cause-specific mortality and particulate matter less than 10 microns in aerodynamic diameter (P M10) for 90 cities in the U.S. (=-=Samet et al., 2000-=-b; Dominici et al., 2002). The NMMAPS morbidity analyses estimated associations between hospitalization in the elderly and P M10 for 14 cities in the U.S. (Samet et al., 2000b; Schwartz, 2000; Zanobet... |

49 |
Short term effects of ambient sulphur dioxide and particulate matter on mortality in 12 European cities: results from time series data from the APHEA project
- KATSOUYANNI, TOULOUMI, et al.
- 1997
(Show Context)
Citation Context ...asked if models had been selected that gave estimates of effect that were biased upwards (Lipfert and Wyzga, 1993; Li and Roth, 1995). These criticism have been addressed by using multi-site studies (=-=Katsouyanni et al., 1997-=-; Samet et al., 2000a; Hwang and Chan, 2001) in which site-specific data on air pollution and health are assembled under a common framework. Hierarchical models (DuMouchel and Harris, 1983; DuMouchel,... |

47 |
Nonconjugate Bayesian estimation of covariance matrices and its use in hierarchical models
- DANIELS, KASS
- 1999
(Show Context)
Citation Context ... implementation of the simulation-based techniques (Gilks et al., 1996), this distribution is not flexible enough to elicit non-informative priors on the variances and on the correlation coefficient (=-=Daniels and Kass, 1999-=-; Daniels, 1999). Instead of the conjugate prior for the entire covariance matrix, we assume that the two variance components σ 2 M and σ2 H have a priori a half-normal distribution on (0, ∞) with mea... |

45 | Assessing confounding, effect modification, and thresholds in the association between ambient particles and daily deaths. Environ Health Perspect
- Schwartz
(Show Context)
Citation Context ....S. (Samet et al., 2000b; Dominici et al., 2002). The NMMAPS morbidity analyses estimated associations between hospitalization in the elderly and P M10 for 14 cities in the U.S. (Samet et al., 2000b; =-=Schwartz, 2000-=-; Zanobetti et al., 2000a). Methodological approaches and substantive results of the separate mortality and morbidity analyses have been reported (Samet et al., 2000a; Daniels et al., 2000; Zeger et a... |

41 | Air pollution and mortality: estimating regional and national dose-response relationships. J Am Stat Assoc 2002;97:100–11
- Dominici, Daniels, et al.
(Show Context)
Citation Context ...lyses estimated associations between all-cause and cause-specific mortality and particulate matter less than 10 microns in aerodynamic diameter (P M10) for 90 cities in the U.S. (Samet et al., 2000b; =-=Dominici et al., 2002-=-). The NMMAPS morbidity analyses estimated associations between hospitalization in the elderly and P M10 for 14 cities in the U.S. (Samet et al., 2000b; Schwartz, 2000; Zanobetti et al., 2000a). Metho... |

41 |
Airborne particles are a risk factor for hospital admissions for heart and lung disease. Environ Health Perspect
- Zanobetti, Schwartz, et al.
(Show Context)
Citation Context ...., 2000b; Dominici et al., 2002). The NMMAPS morbidity analyses estimated associations between hospitalization in the elderly and P M10 for 14 cities in the U.S. (Samet et al., 2000b; Schwartz, 2000; =-=Zanobetti et al., 2000-=-a). Methodological approaches and substantive results of the separate mortality and morbidity analyses have been reported (Samet et al., 2000a; Daniels et al., 2000; Zeger et al., 1999; Schwartz and Z... |

35 |
An overview of methods for the analysis of longitudinal data
- Zeger, Liang
- 1992
(Show Context)
Citation Context ...re positively associated with mortality and morbidity outcomes on average across locations (Samet et al., 2000b; Dominici et al., 2003a). Poisson time series regression models (Liang and Zeger, 1986; =-=Zeger and Liang, 1992-=-; Fahrmeir and Tutz, 2001; McNeney and Petkau, 1994; Albert, 1999) or generalized additive models (Hastie and Tibshirani, 1990) have been widely used to analyze univariate time series data of air poll... |

26 |
Air pollution and mortality
- Kelsall, Samet, et al.
- 1997
(Show Context)
Citation Context ...rage lag 0 and lag 1 P M10 daily time series and potential confounding factors for the mortality and hospitalization data such as long-term trends, seasonality, and weather (Samet et al., 1995, 1997; =-=Kelsall et al., 1997-=-; Dominici et al., 2000; Samet et al., 2000a). Note that X M t might be the same as X H t in some situations. In this application, we specify model (1) as over-dispersed Poisson with a linear term for... |

25 | Bayesian Metaanalysis - DUMOUCHEL - 1990 |

24 |
Inference for multivariate normal hierarchical models
- Everson, Morris
- 2000
(Show Context)
Citation Context ...c , e.g. again uniform prior on all the entries of the matrix B. Additional details on these prior distributions, including the definitions of the prior densities and software implementations are in (=-=Everson and Morris, 2000-=-). The posterior distributions of the overall log relative rates of mortality and hospital admissions (αM, αH) (first row), and of the between-city standard deviations (σM, σH) (second row) under our ... |

24 | Confounding and effect modification in the short-term effects of ambient particles on total mortality: results from 29 European cities within the APHEA2 project. Epidemiology 2001;12:521–31 - Katsouyanni |

23 |
Bayes methods for combining the results of cancer studies in humans and other species
- DUMOUCHEL, HARRIS
- 1983
(Show Context)
Citation Context ...site studies (Katsouyanni et al., 1997; Samet et al., 2000a; Hwang and Chan, 2001) in which site-specific data on air pollution and health are assembled under a common framework. Hierarchical models (=-=DuMouchel and Harris, 1983-=-; DuMouchel, 1990; Breslow and Clayton, 1993; Carlin and Louis, 1996) are a suitable approach for analyzing univariate time series data from multiple locations (Dominici et al., 2000, 2002; Hwang and ... |

15 | Health effects of outdoor air pollution, Part 2 - Society, Bascom - 1996 |

15 | A prior for the variance in hierarchical models
- DANIELS
- 1999
(Show Context)
Citation Context ...imulation-based techniques (Gilks et al., 1996), this distribution is not flexible enough to elicit non-informative priors on the variances and on the correlation coefficient (Daniels and Kass, 1999; =-=Daniels, 1999-=-). Instead of the conjugate prior for the entire covariance matrix, we assume that the two variance components σ 2 M and σ2 H have a priori a half-normal distribution on (0, ∞) with mean zero and a la... |

15 |
Air pollution and mortality: issues and uncertainty
- Lipfert, Wyzga
- 1993
(Show Context)
Citation Context ...1996a,b; Korrick et al., 1998). Critics of single-site studies questioned the choice of particular cities and asked if models had been selected that gave estimates of effect that were biased upwards (=-=Lipfert and Wyzga, 1993-=-; Li and Roth, 1995). These criticism have been addressed by using multi-site studies (Katsouyanni et al., 1997; Samet et al., 2000a; Hwang and Chan, 2001) in which site-specific data on air pollution... |

13 |
Epidemiology of Particle effects
- Pope, Dockery
- 1999
(Show Context)
Citation Context ...h lower concentrations than existed during the earlier disasters, was associated with increased rates of mortality and morbidity in cities in the United States, Europe, and other developed countries (=-=Pope and Dockery, 1999-=-). One key limitation of these studies was use of data from a single, or at most a few locations of uncertain representativeness of broader geographic regions. The National Morbidity, Mortality, and A... |

12 |
Daily mortality analysis by using different regression models
- Li, Roth
- 1995
(Show Context)
Citation Context ...1998). Critics of single-site studies questioned the choice of particular cities and asked if models had been selected that gave estimates of effect that were biased upwards (Lipfert and Wyzga, 1993; =-=Li and Roth, 1995-=-). These criticism have been addressed by using multi-site studies (Katsouyanni et al., 1997; Samet et al., 2000a; Hwang and Chan, 2001) in which site-specific data on air pollution and health are ass... |

11 |
Invited Commentary: Particulate Matter-Mortality Exposure-Response Relations and Thresholds
- Pope
- 2000
(Show Context)
Citation Context ...ented air pollution “disasters” in the US and Europe. By the early 1990’s, time series studies with data from single locations (Dockery et al., 1993; Schwartz, 1994; American Thoracic Society, 1996a; =-=Pope, 2000-=-), showed that air pollution, even at much lower concentrations than existed during the earlier disasters, was associated with increased rates of mortality and morbidity in cities in the United States... |

10 |
Longitudinal data analysis (repeated measures) in clinical trials
- ALBERT
- 1999
(Show Context)
Citation Context ...across locations (Samet et al., 2000b; Dominici et al., 2003a). Poisson time series regression models (Liang and Zeger, 1986; Zeger and Liang, 1992; Fahrmeir and Tutz, 2001; McNeney and Petkau, 1994; =-=Albert, 1999-=-) or generalized additive models (Hastie and Tibshirani, 1990) have been widely used to analyze univariate time series data of air pollution and health in selected locations (Dockery and Pope, 1994; S... |

7 |
Models for longitudinal data: a generalised estimating equation approach
- ZEGER, LIANG, et al.
- 1988
(Show Context)
Citation Context ... Secondly, to take into account the joint correlation function for the bivariate mortality and morbidity time series, we estimate the sample covariance between ˆ β c M and ˆ β c H equations approach (=-=Zeger et al., 1988-=-). by using a novel generalized estimating At the second stage, we assume that the vector of true log relative rates of mortality and morbidity (β c M , βc H ) has a bivariate normal distribution with... |

5 |
Estimating PM10-mortality dose-response curves and threshold levels: an analysis of daily time series for the 20 largest US cities
- DANIELS, DOMINICI, et al.
- 2000
(Show Context)
Citation Context ...et et al., 2000b; Schwartz, 2000; Zanobetti et al., 2000a). Methodological approaches and substantive results of the separate mortality and morbidity analyses have been reported (Samet et al., 2000a; =-=Daniels et al., 2000-=-; Zeger et al., 1999; Schwartz and Zanobetti, 2000; Schwartz, 2000; Zanobetti et al., 2000b). The analyses showed that P M10 concentrations were positively associated with mortality and morbidity outc... |

5 | Air pollution, weather and mortality - Samet, Zeger, et al. - 1997 |

4 |
A Special Report to the
- DOMINICI, MCDERMOTT, et al.
- 2003
(Show Context)
Citation Context ...wartz, 2000; Zanobetti et al., 2000b). The analyses showed that P M10 concentrations were positively associated with mortality and morbidity outcomes on average across locations (Samet et al., 2000b; =-=Dominici et al., 2003-=-a). Poisson time series regression models (Liang and Zeger, 1986; Zeger and Liang, 1992; Fahrmeir and Tutz, 2001; McNeney and Petkau, 1994; Albert, 1999) or generalized additive models (Hastie and Tib... |

4 |
The Association of Mortality and Particulate Air Pollution. Health Effects Institute
- Samet, Zeger, et al.
- 1995
(Show Context)
Citation Context ...ign matrices including average lag 0 and lag 1 P M10 daily time series and potential confounding factors for the mortality and hospitalization data such as long-term trends, seasonality, and weather (=-=Samet et al., 1995-=-, 1997; Kelsall et al., 1997; Dominici et al., 2000; Samet et al., 2000a). Note that X M t might be the same as X H t in some situations. In this application, we specify model (1) as over-dispersed Po... |

4 |
Is there a threshold in the relation between particulate air pollution and daily deaths: a meta-smoothing approach
- SCHWARTZ, ZANOBETTI
- 2000
(Show Context)
Citation Context ...i et al., 2000a). Methodological approaches and substantive results of the separate mortality and morbidity analyses have been reported (Samet et al., 2000a; Daniels et al., 2000; Zeger et al., 1999; =-=Schwartz and Zanobetti, 2000-=-; Schwartz, 2000; Zanobetti et al., 2000b). The analyses showed that P M10 concentrations were positively associated with mortality and morbidity outcomes on average across locations (Samet et al., 20... |

4 |
A Special Report to the Health Effects Institute on the Revised Analyses of the NMMAPS II Data: Morbidity and Mortality among Elderly Residents of Cities with Daily PM Measurements
- SCHWARTZ, ZANOBETTI, et al.
- 2003
(Show Context)
Citation Context ...nd seasonality, and 3 df to adjust for temperature and barometric pressure. These choices are made on the basis of our previously published results and on recent re-analyses and sensitivity analyses (=-=Schwartz et al., 2003-=-; Dominici et al., 2003a). In the results section we explore the sensitivity of the overall log relative rates to the df in the natural cubic splines of time, temperature and barometric pressure. We e... |

4 |
Harvesting-resistant estimates of pollution effects on mortality. Epidemiology
- SL, Dominici, et al.
- 1999
(Show Context)
Citation Context ...artz, 2000; Zanobetti et al., 2000a). Methodological approaches and substantive results of the separate mortality and morbidity analyses have been reported (Samet et al., 2000a; Daniels et al., 2000; =-=Zeger et al., 1999-=-; Schwartz and Zanobetti, 2000; Schwartz, 2000; Zanobetti et al., 2000b). The analyses showed that P M10 concentrations were positively associated with mortality and morbidity outcomes on average acro... |

3 |
Overdispersed Poisson regression models for studies of air pollution and human health. The Canadian
- MCNENEY, PETKAU
- 1994
(Show Context)
Citation Context ...idity outcomes on average across locations (Samet et al., 2000b; Dominici et al., 2003a). Poisson time series regression models (Liang and Zeger, 1986; Zeger and Liang, 1992; Fahrmeir and Tutz, 2001; =-=McNeney and Petkau, 1994-=-; Albert, 1999) or generalized additive models (Hastie and Tibshirani, 1990) have been widely used to analyze univariate time series data of air pollution and health in selected locations (Dockery and... |

2 |
Improved semiparametric regression models of air pollution and mortality
- DOMINICI, MCDERMOTT, et al.
- 2003
(Show Context)
Citation Context ...wartz, 2000; Zanobetti et al., 2000b). The analyses showed that P M10 concentrations were positively associated with mortality and morbidity outcomes on average across locations (Samet et al., 2000b; =-=Dominici et al., 2003-=-a). Poisson time series regression models (Liang and Zeger, 1986; Zeger and Liang, 1992; Fahrmeir and Tutz, 2001; McNeney and Petkau, 1994; Albert, 1999) or generalized additive models (Hastie and Tib... |

2 |
Bayesian inference for generalized additive models based on Markov random field
- FAHRMEIR, LANG
- 2001
(Show Context)
Citation Context ...sed to extend our modeling approach. For example, we could have used a Bayes approach via MCMC sampling for inference in generalized additive models with city-specific random effects as suggested by (=-=Fahrmeir and Lang, 2001-=-). This approach would have avoid our reliance on the assumption of normality at the first stage of the hierarchical model. At the other end, to properly estimate the sample covariance between the log... |

2 |
Air pollution effects on daily clinic visits for lower respiratory illness
- HWANG, CHAN
- 2002
(Show Context)
Citation Context ...imates of effect that were biased upwards (Lipfert and Wyzga, 1993; Li and Roth, 1995). These criticism have been addressed by using multi-site studies (Katsouyanni et al., 1997; Samet et al., 2000a; =-=Hwang and Chan, 2001-=-) in which site-specific data on air pollution and health are assembled under a common framework. Hierarchical models (DuMouchel and Harris, 1983; DuMouchel, 1990; Breslow and Clayton, 1993; Carlin an... |

2 |
Generalized additive distributed lag models
- ZANOBETTI, WAND, et al.
- 2000
(Show Context)
Citation Context ...., 2000b; Dominici et al., 2002). The NMMAPS morbidity analyses estimated associations between hospitalization in the elderly and P M10 for 14 cities in the U.S. (Samet et al., 2000b; Schwartz, 2000; =-=Zanobetti et al., 2000-=-a). Methodological approaches and substantive results of the separate mortality and morbidity analyses have been reported (Samet et al., 2000a; Daniels et al., 2000; Zeger et al., 1999; Schwartz and Z... |

1 | Effects of ozone and other pollutants on the pulmonary function of adult hikers. Environmental Health Perspectives 106 - Neas, Dockery, et al. - 1998 |

1 | Air pollution and daily mortality in - unknown authors - 1995 |