Supplementary Material Changes in Relative Fit of Human Heat Stress Indices to Cardiovascular

Supplementary material— Changes in relative fit of human heat stress indices to cardiovascular, respiratory and renal hospitalizations across five Australian urban populations

James Goldie1, 2, 3 (orcid.org/0000-0002-5024-6207)

Lisa Alexander1, 2 (orcid.org/0000-0002-5635-2457)

Sophie Lewis2, 3 (orcid.org/0000-0001-6416-0634)

Steven C Sherwood1, 2 (orcid.org/0000-0001-7420-8216)

Hilary Bambrick4 (orcid.org/0000-0001-5361-950X)

1. Climate Change Research Centre, UNSW Australia, Sydney, New South Wales, Australia.

2. ARC Centre of Excellence for Climate System Science, UNSW Australia, Sydney, New South Wales, Australia.

3. Fenner School of Environment & Society, Australian National University, Acton, Australian Capital Territory, Australia.

4. School of Public Health and Social Work, Queensland University of Technology, Brisbane, Queensland, Australia.

Corresponding author: James Goldie

Fenner School of Environment & Society

The Australian National University

Building 141, Linnaeus way

Canberra, ACT, 2601

+61 421 747 208

1. HadISD station table and processing

We used the following stations from HadISD v1.0.3.2014f (Dunn et al. 2012):

Adelaide: 946725-99999 ADELAIDE INTL ARPT
Brisbane: 945780-99999 BRISBANE AERO
Cairns: 942870-99999 CAIRNS AERO
Perth: 946100-99999 PERTH AIRPORT
Sydney: 947670-99999 SYDNEY AIRPORT AMO

HadISD observations are subdaily but not regular—that is, the number of observations per day can vary across time series and across stations. However, the stations we selected for this analysis varied in two predictable respects.

Firstly, the precision of temperature data shifted between integer (whole degrees) and subinteger (one tenth of a degree) at regular hours of the day. We identified a three-hourly series starting at 0200 UTC that was available at subinteger precision across the desired time range.

However, for summer periods only from the year 2000 onward, this data was instead available from 0100 UTC onward. After conferring with the Robert Dunn, the maintainer of the dataset, we believe that this is the result of a change in daylight savings policy in the dataset that HadISD uses: the observers switched from keeping constant UTC time to constant local time. As a result, we decided to use the three-hourly series from 0100 during these periods but shift them forward 1 hour before using them for further calculation.

2. City boundaries and population bases

Hospital admissions data in Sydney (SYD), Adelaide (ADE) and Perth (PER) was delivered based on their respective Statistical Divisions (SDs) in the 2006 ASGC (Australian Bureau of Statistics 2015). These SD boundaries were used to clip SLA boundaries from the 2001 and 2006 ASGCs, as well as the 2011 ASGS.

Hospital admissions for Brisbane (BRI) and Cairns (CAI) were delivered along with the rest of Queensland, so admissions for these cities needed to be isolated. Brisbane's boundaries were extracted from the 2006 Statistical Division, like the southern cities, but Cairns was selected from a group of SLAs, as no other suitable geostatistical unit was found.

The census population figures of 2001 and 2006 Statistical Divisions were used where available; in all other cases the populations of constituent SLAs were summed and used.

Constituent Cairns SLAs included:

350052066: Cairns (C) - City
350052065: Cairns (C) - Central Suburbs
350052076: Cairns (C) - Western Suburbs
350052068: Cairns (C) - Mt Whitfield
350052074: Cairns (C) - Trinity

Table S1: populations reported for each city in the 2001, 2006 and 2011 censuses. Brisbane and Cairns populations are the sum of constituent Statistical Local Areas (SLAs).

Adelaide / Brisbane / Cairns / Perth / Sydney
Year / Unit / Pop / Unit / Pop / Unit / Pop / Unit / Pop / Unit / Pop
2001 / SD / 1066 / SD / 1609 / SLAs / 53 / SD / 1325 / SD / 3948
2006 / SD / 1106 / SD / 1763 / SLAs / 54 / SD / 1445 / SD / 4119
2011 / SLAs / 1239 / GCCSA / 2130 / SLAs / 57 / GCCSA / 1752 / SLAs / 4748

(thousands)

We retrieved the populations of each age group at each Census to ensure that differing age distributions would not have a material effect on the analysis (Figure S1) (Australian Bureau of Statistics 2017). In 2011, we could not retrieve population figures by age group for Statistical Local Areas, so Greater Capital City Statistical Areas (GCCSAs) were used in Adelaide and Sydney for this purpose only.

3. Diagnosis groups

Table S2: International Classification of Diseases 9th edition (ICD-9) and 10th edition (ICD-10) codes for the selected conditions in each group.

Disease group / Description / ICD-9 Codes / ICD-10 Codes / Used in
Cardiovascular / Ischaemic heart diseases / 410–414 / I20–I25 / C1, C4
Heart failure / 428 / I50 / C1, C4
Respiratory / Pneumonia and lower respiratory infections / 480–486 / J12–J18, J20–J22 / C2, C4
Chronic lower respiratory conditions / 491, 492, 494, 496 / J40–J44 / C2, C4
Renal / Renal failure / 584–585 / N17–N19 / C3, C4

4. Additional results

Table S3: Akaike Information Criterion (AIC) ranks for heat stress models of all selected admissions (C4). Lower ranks (shaded green) indicate better fit than higher ranks (shaded yellow).

Index / Type / CAI / BRI / SYD / ADE / PER
n = 3639 / n = 71 688 / n = 147 910 / n = 69 893 / n = 47 762
Max dewpoint (°C) / H / 3 * / 3 * / 9 * / 1 * / 2 *
Mean dewpoint (°C) / H / 2 * / 6 * / 13 / 3 * / 7 *
Min dewpoint (°C) / H / 1 * / 13 * / 15 / 10 / 16
Max AT (°C) / HH / 8 / 1 * / 1 * / 6 * / 1 *
Mean AT (°C) / HH / 11 / 5 * / 4 * / 5 * / 8
Min AT (°C) / HH / 13 / 11 * / 12 * / 8 / 12
Max sWBGT (°C) / HH / 7 / 2 * / 3 * / 4 * / 3 *
Mean sWBGT (°C) / HH / 5 / 4 * / 6 * / 7 * / 10
Min sWBGT (°C) / HH / 6 / 9 * / 7 * / 9 / 13
Max temperature (°C) / T / 14 / 8 * / 2 * / 13 / 5 *
Mean temperature (°C) / T / 16 / 7 * / 5 * / 14 / 15
Min temperature (°C) / T / 15 / 12 * / 11 * / 15 / 14
3DAT (-2 days) (°C) / HW / 10 / 10 * / 8 * / 16 / 11
3DMT (-2 days) (°C) / HW / 4 * / 15 / 10 * / 12 / 4 *
EHF (-2 days) (°C2) / HW / 9 / 14 * / 16 / 2 * / 6 *
Parent model (no index) / P / 12 / 16 / 14 / 11 / 9

* (Bolded, italicized) Heat stress model shows a statistically significant difference in fit (p < 0.05) compared to its parent model using a Likelihood Ratio Test.

Table S4: Effect sizes, expressed in the increase (positive, shaded purple) or decrease (negative, shaded green) in admissions per unit increase in the index, of heat stress indices in models of all selected admissions (C4).

Index / Type / CAI / BRI / SYD / ADE / PER
n = 3639 / n = 71 688 / n = 147 910 / n = 69 893 / n = 47 762
Max dewpoint (°C) / H / +2.3% * / +0.7% * / +0.3% * / +0.3% * / +0.4% *
Mean dewpoint (°C) / H / +2.5% * / +0.6% * / +0.2% / +0.3% * / +0.2%
Min dewpoint (°C) / H / +2.3% * / +0.3% * / +0.1% / +0.1% / +0.0%
Max AT (°C) / HH / +1.0% / +0.8% * / +0.3% * / +0.1% * / +0.3% *
Mean AT (°C) / HH / +1.0% / +0.8% * / +0.4% * / +0.2% * / +0.1%
Min AT (°C) / HH / +0.6% / +0.5% * / +0.2% * / +0.1% / +0.1%
Max sWBGT (°C) / HH / +1.6% / +1.0% * / +0.5% * / +0.3% * / +0.5% *
Mean sWBGT (°C) / HH / +2.0% / +0.9% * / +0.5% * / +0.3% * / +0.2%
Min sWBGT (°C) / HH / +1.7% / +0.6% * / +0.4% * / +0.3% / +0.1%
Max temperature (°C) / T / -0.8% / +0.7% * / +0.4% * / +0.1% / +0.3% *
Mean temperature (°C) / T / -0.4% * / +0.9% * / +0.4% * / +0.1% / +0.0%
Min temperature (°C) / T / +0.5% / +0.7% * / +0.4% * / +0.0% / -0.0%
3DAT (-2 days) (°C) / HW / -2.1% / +1.0% * / +0.5% * / +0.0% / +0.1%
3DMT (-2 days) (°C) / HW / -1.9% * / +0.3% / +0.4% * / -0.1% / +0.4% *
EHF (-2 days) (°C2) / HW / -1.9% / +0.6% * / -0.1% / -0.1% * / -0.1% *

* (Bolded, italicized) Heat stress model shows a statistically significant difference in fit (p < 0.05) compared to its parent model using a Likelihood Ratio Test.

References

Australian Bureau of Statistics. 2015. QuickStats. Available: http://www.abs.gov.au/websitedbs/censushome.nsf/home/quickstats?opendocument&navpos=220 [accessed 13 April 2016].

Australian Bureau of Statistics. 2017. TableBuilder. Available: http://www.abs.gov.au/websitedbs/censushome.nsf/home/tablebuilder [accessed 12 May 2017].

Dunn RJH, Willett KM, Thorne PW, Woolley EV, Durre I, Dai A, et al. 2012. HadISD: a quality-controlled global synoptic report database for selected variables at long-term stations from 1973–2011. Clim. Past 8:1649–1679; doi:10.5194/cp-8-1649-2012.