Table 1

Daily time-series studies with air pollution exposures and myocardial infarction (MI) outcomes: description of studies

First author and year of publicationPopulation/data sourceLocation and time periodNumber of events included (mean per day for time-series studies)Air pollution exposure variable(s)Potential confounders includedMI ascertainmentLags considered (days, except where noted)
Daily time-series studies
Fatal and non-fatal events
    Cendon 200611Hospital admissions data (112 hospitals: infirmaries and ICUs); age >64 onlySao Paulo, Brazil 1998–919272* (26.4)PM10 (24 h average)Season and trend, temperature (non-linear, 2-day moving average), humidity, day of weekEvents with ICD-10 codes suggesting MI in the Public Health Data Analysis System Division0–7 inclusive
    Lanki 200612AMI registers and hospital discharge registers5 European cities (Augsburg, Barcelona, Helsinki, Rome, Stockholm) 1992–2000 (3–7 year period per city)26 854 (between 0.9 and 8.4 per city)PM10, O3 (8 h average, summer only), NO2, CO, modelled particle number conc. (proxy for PM <0.1 μg/m3)Season and trend, apparent temperature (non-linear, same day and average of lag days 1–3), barometric pressure, weekday indicator, holiday indicatorRecords with ICD9 code 410 in hospital registers (two cities); or records meeting MONICA definition of MI in AMI registers (three cities)580–3 inclusive
    Koken 200313Hospital admissions data (11 hospitals, covering ages 65+ years)Denver county, USA 1993–7 (July and August only)1576* (5.1)PM10, O3, NO2, SO2, CO (all 24 h average)Daily maximum temperature (lag days 0–4), dew point temperature, day of week, calendar year, population sizePrimary discharge diagnosis (ICD9 = 410.XX)0–4 inclusive
    Mann 20025Records from a health maintenance organisationSouthern California, USA 1988–9519 690 (6.7*)PM10 (24 h average), O3 (8 h average), NO2, (24 h average) CO (8 h average)Season and trend, temperature (non-linear, same day), relative humidity, calendar year, day of week, annual population sizeRecords with ICD9 code 4100–5 days inclusive
    Ye 200114Hospital emergency transports records (four hospitals, ages 65+ years)Tokyo, Japan 1980–95 (July and August only)3200* (3.28)PM10, O3, CO, NO2, SO2, (all daily average)Annual trends, daily maximum temperature (lag days 0–4), population sizeAs diagnosed by emergency doctor, based on presenting symptoms0 (adjusted for 1–4 inclusive)
    Linn 200015Hospital admissions dataLos Angeles, USA 1992–5Not reportedPM10, O3, CO, NO2 (all 24 h average)Season and trend, day of week, holidays, mean temperature (same day), barometric pressure, indicators for hot days, cold days, rainy daysRecords with an all-patient-refined diagnosis-related group code of 111, 115, or 121Different lags considered, exact strategy unclear
    Poloniecki 199716Hospital episode statisticsLondon, UK 1987–9468 300* (26.7)O3 (8 h average); NO2, SO2, CO, black smoke (all 24 h average)Season and trend, temperature (lag day 1), humidity, day of week, public holidays, influenza epidemic indicatorRecords with ICD9 code 4101
Fatal events only
    Murakami 200617Vital statistics of Japan data (34 districts)34 districts, Japan 1990–414 430 (7.9*)Suspended particulate matter (hourly measurements)Time of day, temperature (non-linear, same day), regionRecords with ICD9 code 410Exposure windows from 1 to 48 h
    Sharovsky 200418Death registry dataSao Paulo, Brazil 1996–812 007 (16.4)PM10, CO, SO2 (daily average)Season and trend, mean temperature (non-linear, up to lag day 7), relative humidity, atmospheric pressure, day of week, holidays, influenza levelsDeath certificates with MI (ICD10 = I21) listed as primary cause0, and moving average of up to previous 7 days
    Rossi 199919Vital statistics department mortality dataMilan, Italy 1985–91600* (0.9)Total suspended particlesSeason and trend, temperature (non-linear, lag days unclear), relative humidity, day of week, holidays, epidemics, pollutionDeaths with ICD9 codes of 410Different lags considered, exact strategy unclear
Case–crossover studies
Fatal and non-fatal events
    Barnett 200620Hospital admissions data from seven citiesAustralia (five cities) and New Zealand (two cities) 1998–200128 818*PM2.5 (24 h average), PM10 (24 h average), O3 (8 h average), CO (8 h average), NO2 (24 h average)Temperature (lag days 0–1), change in temperature from previous day, humidity, hot and cold days, pressure, day of week, holiday, rainfallRecords with ICD9 code 410 or ICD10 code I21-22Average of 0–1
    Zanobetti 200621Hospital admissions data from the US Medicare programme (ages 65+ years)Boston metropolitan area, USA 1995–915 578PM2.5, PM non-traffic (modelled), O3, CO, NO2, black carbonApparent temperature (non-linear, lag day 1); also matched for same day temperature), day of weekRecords with ICD9 code 4100, 1, and mean of 0 and 1
    Peters 200522Coronary event registry (cases surviving first 24 h only)Augsburg, Germany 1999–2001851PM2.5, total number concentration (proxy for ultrafine particles), O3, SO2, CO, NO2 (all 24 h average; 1 h average also considered for PM)Temperature (non-linear, same day), day of weekPatients meeting MONICA definition of MI580–5 (also 0–6 h for hourly analysis)
    Ruidavets 200523AMI registryToulouse, France 1997–9399O3 (highest 8 h average of the day), SO2 (24 h average), NO2 (24 h average)Day of week (matched), min and max temperature (same day), humidity, influenza levelsClinical, ECG and enzyme data available to support diagnosis0–3 days inclusive
    Sullivan 200524Community database linking emergency service and hospital outcome dataWashington State, USA 1988–945793Increase in short-term average PM2.5 (derived from fine PM), defined as 10 μg/m3 increase in 1, 2, 4, 24 h averaged PM2.5). Similar for SO2 and COTemperature (non-linear, same day), relative humidityDischarge diagnosis of AMI confirmed by enzyme and ECG changes0–2 days inclusive
    Zanobetti 200525Hospital admissions data from the US Medicare programme (ages 65+ years)21 Cities, USA 1986–99302 453PM10 (daily average)Day of week (matched), apparent temperature (non-linear, lag days 0–1)Medicare claims where primary diagnosis had ICD9 code 4100–2 days inclusive
    Peters 200426KORA MI registryAugsburg, Germany 1999–2001691Exposure to traffic as measured by retrospective diary for the 4 days preceding eventNone specifiedRecords meeting MONICA definition of MI580–6 days inclusive
    D’Ippoliti 200327Regional hospital admissions dataRome, Italy 1995–76531Total suspended particles, CO, SO2, NO2 (all 24 h average)Day of week (matched), temperature (non-linear, lag day 1), humidity, air pressureRecords with ICD9 code of 4100–4, and mean of 0–2 days
    Peters 200128Coronary care unit admissions recordsGreater Boston, USA 1995–6772PM2.5, PM10, ozone, SO2, NO2, CO, black carbonSeason, day of week, minimum daily temperature (non-linear, same day), relative humidityPatients had all of: ⩾1 CK above upper limit of normal, positive MB isoenzymes, symptoms0–5 inclusive (also 0–5 h for hourly analysis)

  • *Derived from reported mean daily rate, and length of period under study.

  • AMI, acute myocardial infarction; ICI, intensive care unit; PM, particulate matter.