Higher energy prices are associated with ... - Wiley Online Library

Article Safety

Higher energy prices are associated with diminished resources, performance and safety in Australian ambulance systems Lawrence H. Brown School of Public Health, Tropical Medicine and Rehabilitation Sciences, James Cook University, Queensland

Abstract Objective: To evaluate the impact of changing energy prices on Australian

Taha Chaiechi

ambulance systems.

Discipline of Economics, School of Business, James Cook University, Queensland

Methods: Generalised estimating equations were used to analyse

Petra G. Buettner School of Public Health, Tropical Medicine and Rehabilitation Sciences, James Cook University, Queensland

contemporaneous and lagged relationships between changes in energy prices and ambulance system performance measures in all Australian State/Territory

Deon V. Canyon School of Public Health, Tropical Medicine and Rehabilitation Sciences, James Cook University, Queensland; John A Burns School of Medicine, University of Hawaii at Manoa, United States

J. Mac Crawford

ambulance systems for the years 20002010. Measures included: expenditures per response; labour-to-total expenditure ratio; full-time equivalent employees (FTE) per 10,000 responses; average salary; median

Ohio State University College of Public Health, United States

and 90th percentile response time; and injury compensation claims. Energy price

Jenni Judd School of Public Health, Tropical Medicine and Rehabilitation Sciences, James Cook University, Queensland

data included State average diesel price, State average electricity price, and world crude oil price. Results: Changes in diesel prices were inversely associated with changes in

T

salaries, and positively associated with

he potential adverse impacts of energy scarcity and rising energy prices on health services are the subject of much opinion, but little scientific evaluation. Three decades ago, noting that health facilities are dependent on energy, Bailey 1 raised concerns about energy scarcity and energy costs in the context of the United States’ (US) dependence on imported oil. More recently, several authors have warned that energy is a critical input for all health services,2-8 and that there are several pathways through which energy scarcity and energy costs could affect health services. These include: the cost and availability of medical supplies and equipment; the cost and availability of healthrelated transport; the cost of lighting, heating and air conditioning health facilities; impacts on food security leading to increased demand Submitted: March 2012

on health services; and economic impacts that disrupt funding for health services.3,8 These concerns are shared by the general public,9 and there is some empirical basis for them. In a time series analysis of US petroleum and health care prices between 1973 and 2008, Hess et al.5 found a 1% increase in oil price inflation was associated with a 0.03% increase in medical care prices after an eight-month lag, although the effect was much stronger in the 1970s than in recent years. Ambulance services are an important component of the health system, providing emergency medical care and/or medical transport to more than three million patients in Australia and New Zealand (NZ) each year.10 Recent reports in the lay media have anecdotally described the adverse effects of rising fuel prices on the operating budgets of

Revision requested: July 2012

Accepted: August 2012

changes in ambulance response times; changes in oil prices were also inversely associated with changes in salaries, as well with staffing levels and expenditures per ambulance response. Changes in electricity prices were positively associated with changes in expenditures per response and changes in salaries; they were also positively associated with changes in injury compensation claims per 100 FTE. Conclusion: Changes in energy prices are associated with changes in Australian ambulance systems’ resource, performance and safety characteristics in ways that could affect both patients and personnel. Further research is needed to explore the mechanisms of, and strategies for mitigating, these impacts. The impacts of energy prices on other aspects of the health system should also be investigated. Key words: ambulance services, energy prices, health economics, Australia

Correspondence to: Lawrence H. Brown, Anton Breinl Centre for Public Health and Tropical Medicine, James Cook University, Townsville, QLD 4811; e-mail: [email protected]

2013 vol. 37 no. 1

Aust NZ J Public Health. 2013; 37:83-9

AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH © 2013 The Authors. ANZJPH © 2013 Public Health Association of Australia

doi: 10.1111/1753-6405.12015

83

Brown et al.

Article

ambulance services in the US, the United Kingdom and Canada,11-15 but the extent of those impacts and how they manifest have not been systematically studied in any setting. Increasing energy prices would be expected to lead to increased total operational costs, but ambulance services are complex systems involving vehicles, equipment, buildings, communication systems and personnel.16 A need to divert financial resources to meet increasing energy prices could conceivably have an impact on any or all of these system components, potentially in ways that ultimately affect system performance. This study aimed to empirically evaluate, for the first time, the impact of changing energy prices on Australian ambulance systems. We used a panel data approach to test the null hypothesis that changes in ambulance service resource, performance and safety characteristics are not associated with changes in energy prices.

Energy price data

Methodology This retrospective study was approved by the Human Research Ethics Committee at James Cook University, Australia (HREC #3982) with the understanding that the results would be aggregated and reported in a manner such that individual system performance could not be determined.

Setting In Australia, responsibility for ambulance services rests with State and Territory governments. In New South Wales (NSW), South Australia (SA) and Tasmania (TAS) ambulance services are provided by the Department of Health. In Queensland (QLD) and the Australian Capital Territory (ACT) they are provided by the departments responsible for emergency services. St John Ambulance, an independent not-for-profit organisation, is contracted to provide services in the Northern Territory (NT) and Western Australia (WA). In Victoria (VIC), ambulance services are organised within the Department of Health, but prior to 2008 they were provided by two independent statutory authorities: the Metropolitan Ambulance Service (MAS) serving the greater Melbourne area, and Rural Ambulance Victoria (RAV) serving the outlying rural areas. In the 2009/10 financial year, Australian ambulance systems performed slightly more than 3.5 million emergency and nonemergency responses.10

Ambulance system data Data on the resource, performance and safety characteristics of the ambulance systems in Australia’s six states were obtained from publicly available annual reports published online at each agency’s website, or its parent agency’s (e.g. Department of Health) website, for 2000/01 through 2009/10. While Ambulance Victoria data for the entire state of VIC were available from 2008/09 forward, only annual reports for MAS were available for prior years. Because of differences in the structure and geography of these agencies, we treated them as individual systems in this analysis. Limited resource and performance data for the ambulance agencies serving Australia’s two Territories (NT and ACT) were extracted from Australia’s

84

Council of Ambulance Authorities (CAA) annual reports, also available online. In some cases, prior year data were reported in the first available report, allowing collection (or calculation) of data for one additional year. The extracted annual data included: number of ambulance responses; total expenditures; labour-related expenditures; number of full time equivalent (FTE) employees; number of work-related injury compensation claims; median ambulance response time; and 90th percentile ambulance response time. When response time data were not included in the individual state ambulance agency annual reports, these indicators were instead extracted from the CAA annual reports. From these data, eight system-related measures were identified or calculated: four resource indicators, two performance indicators, and two safety indicators (Table 1).

The Australian Institute of Petroleum (AIP) reports annual average terminal gate prices (TGP) for diesel fuel and petrol in Australia’s capital cities, with the exception of Canberra, ACT. TGP reflects the wholesale price of diesel or petrol with goods and services tax added. TGP data for diesel fuel for financial years 2004/05 through 2009/10 were obtained from the AIP website,17 and the annual average TGP for diesel fuel in each capital city (as cents per litre) was used as a measure of vehicle fuel prices for each State or Territory. The annual average TGP for Sydney, NSW was used as a proxy measure for the cost of diesel in the ACT. State average annual electricity prices (as dollars per megawatt hour) for financial years 2000/01 through 2009/10 were obtained from the Australian Energy Market Operator (AEMO).18 AEMO data for TAS were limited to the 2004/05 through 2009/10 financial years, and the AEMO data does not include electricity price data for WA, NT or ACT. Electricity prices for NSW were used as a proxy of electricity prices in ACT, but because of their remoteness, distinctive geography and unique economies, a proxy measure of electricity prices for NT and WA was not attempted. Since TGP data were available for only six financial years, average annual world crude oil prices were used as an additional indirect measure of vehicle fuel prices with a longer historical record. Oil prices were obtained from the US Energy Information Administration (EIA) for financial years 2000/01 through 2009/10.19 EIA data are reported in US dollars, so annual average exchange rates reported by the Australian Tax Office were used to convert these prices into Australian dollar prices (as dollars per barrel).20

Model selection and specification This panel data analysis followed the approach outlined by Markus,21 Hsiao,22 and Twisk.23 generalised estimating equation (GEE) modelling was used to evaluate the relationships between changes in the energy price measures and changes in the resource, performance and safety measures while accounting for the repeated-measures nature of the dependent variables. GEE models marginal expectations (or changes) in the dependent variables and, unlike conditional fixed effects models, does not require full


2013 vol. 37 no. 1

Safety

Australian ambulance systems affected by energy price

specification of joint distribution of observations. In GEE, the relationships between the variables in the model at different time points are analysed simultaneously;23 conceptually it can be thought of as regression modelling with correction for the dependency of the serial observations.24 GEE generates robust and model-based standard errors, and produces consistent normal solutions even if the underlying correlation structure is incorrectly specified.23,25 These attributes make GEE well suited for unbalanced panel data analyses. The financial data were adjusted to 2009/10 Australian dollars – standardising the values to most recent year of included data – using consumer price index (CPI) multipliers published by the Australia Bureau of Statistics (ABS).26 Changes in energy prices and changes in the dependent variables were calculated as the ‘first difference’, i.e. the difference between the value for any given year and the value for the prior year. To avoid spurious regression, GEE requires that data are ‘stationary’, i.e. the mean and variance do not change over time or position. We used the Augmented Dickey Fuller test, allowing for a random walk and a drift, and the Phillips-Perron test to check for unit roots (or non-stationarity) in the data. The analysis then evaluated both the contemporaneous (within the same financial year) and one-year lagged relationships between energy price changes and changes in each individual resource, performance and safety indicator within each State ambulance system. Both contemporaneous and lagged relationships must be considered to achieve unbiased statistical parameter estimations. System administrative structure (dichotomised as ‘health department based’ or ‘non health department based’) was included as a possible covariate, and time was modelled as a linear trend. The analysis used an auto-regressive GEE model with calculation of robust standard errors, assuming a Gaussian distribution and an identity link, as: ΔYit = β0 + β1Admini + β2ΔDiesel$it + β3ΔDiesel$it-1 + β4ΔElect$it + β5ΔElect$it-1 +β6ΔYit-1 + β7t + εit or ΔYit = β0 + β1Admini + β2ΔOil$it + β3ΔOil$it-1 + β4ΔElect$it + β5ΔElect$it-1 +β6ΔYit-1 + β7t + εit where ΔYit is the change in the resource, performance or safety indicator for the ambulance system in State or Territory i for year t; Admini is the administrative structure of each ambulance system; ΔDiesel$it is the change in the average terminal gate price for diesel in each State or Territory for year t; ΔDiesel$it-1 is the prior year change in average terminal gate price for diesel in each State or Territory; ΔOil$t is the change in average world crude oil price for year t; ΔOil$t-1 is the prior year change in average world crude oil price; ΔElect$it is the change in average electricity price in each State or Territory for year t; ΔElect$it-1 is the prior year change in average electricity price in each State or Territory; ΔYit-1 is the prior year change in the value of the dependent variable for each ambulance system; and ε is the error term. As recommended by Twisk, the autoregressive GEE modelling was conducted using an independent correlation structure.23 Finally, because the inclusion of electricity price in the models eliminated NT and WA from the analyses, both the contemporaneous

2013 vol. 37 no. 1

and lagged relationships between diesel price and oil price changes and changes in the resource, performance and safety indicators for those two ambulance systems were modelled separately, omitting electricity price from the models described above. Model fit was evaluated by calculating the explained variance (R2) in each model as “1 – (variance of the model / variance of the dependent variable)”. For all analyses, an alpha value of 0.05 was used to establish statistical significance.

Results During the years included in this study, Australian ambulance systems performed more than 20 million ambulance responses, with annual median ambulance response times ranging between 7.2 and 11.0 minutes at an average (± standard deviation) CPI-adjusted expense of $616 ± $131 per response. The aggregate measures of all the performance indicators are shown in Table 1. None of the resource, performance or safety indicators were available for all ambulance agencies for all years, except for 90th percentile ambulance response time. Table 1 also shows the number of systemyears (out of a maximum possible of 80) for which data were available. Safety-related measures were the most under-reported indicator, available for only four (out of eight) systems for a total of 20 system-years. Figure 1 shows the evolution of CPI-adjusted energy prices and resource indicators (panel A), performance indicators (panel B) and safety indicators (panel C) over time. The presence of a unit root could be rejected for all of the first-difference data except for injury compensation claims per 10,000 responses (data not shown); thus that variable was excluded from the final analysis.

Modelling diesel and electricity prices Table 2 shows the results of the GEE models evaluating the associations between changes in diesel and electricity price and changes in the resource, performance and safety indicators in those systems for which electricity data were available. CPI-adjusted expenditures per response increased over time, as did the ratio of labour-related expenditures to total expenditures. There was a significant contemporaneous, positive association between changes in electricity price and changes in expenditures per ambulance response (coefficient [standard error] = 3.169 [1.595]); changes in electricity price also manifested in changes in the ratio of labour-related expenditures to total expenditures (0.0009 [0.0004]) and changes in average salary (0.662 [0.201]). The relationship between changes in energy price and changes in labour-related expenditures was also seen for lagged changes in both electricity price (0.0016[0.0005]) and diesel price (0.0009 [0.0004]). There was a lagged positive association between changing diesel price and changes in median ambulance response time (0.022 [0.006]), but the explained variance for that model was low (5.6%). There was, however, both a contemporaneous (0.049 [0.019]) and a lagged (0.031 [0.014]) positive association between changes in diesel price and changes in 90th percentile ambulance response time, and the explained variance for that m odel exceeded 95%.


85

Brown et al.

Article

Results for systems without electricity data The only significant associations between changes in diesel price and the outcome measures in the systems without electricity data were both a contemporaneous (-0.001 [0.0002]) and lagged (-0.0006 [0.0001]) negative association between changes in diesel price and changes in the ratio of labour expenditures to total expenditures, and a significant positive association between changes in diesel price and

Panel A L:Tratio Exp/Resp ($) FTE/Resp AvgSal ($ 000) Oil$ ($ / barrel)

2009‐10

2008‐09

2007‐08

2006‐07

2005‐06

2004‐05

2003‐04

2002‐03

2001‐02

Diesel$ (¢ / L) 2000‐01

Elect$ ($ / MWh)

Financial Year

Panel B RTMed (min) RT90ile (min) Diesel$ (¢ / L) Elect$ ($ / MWh)

2009‐10

2008‐09

2007‐08

2006‐07

2005‐06

2004‐05

2003‐04

2002‐03

Oil$ ($ / barrel) 2001‐02

160 140 120 100 80 60 40 20 0

Financial Year 160 140 120 100 80 60 40 20 0

Panel C Comp/Resp Comp/FTE Diesel$ (¢ / L) 2009‐10

2008‐09

2007‐08

2006‐07

2005‐06

2004‐05

2003‐04

2002‐03

Oil$ ($ / barrel) 2001‐02

Table 2 also shows the results of the GEE models evaluating changes in oil and electricity price over a longer historical record. In this analysis, workplace injury claims per 100 FTE diminished over time, but health-based systems had higher workplace injury claims when compared with non-health-based systems. The associations between changes in energy price and resource indicators differed when modelling oil price over ten years instead of diesel price over six years: there was still a positive association between lagged changes in electricity price and changes in the ratio of labour-related expenditures to total expenditures (0.0014 [0.0006]), but changes in electricity price were not associated with changes in expenditures per response. Lagged changes in oil price, however, were negatively associated with changes in expenditures per response (-3.992 [1.949]) and FTE per 10,000 responses (-0.085 [0.042]). There was a lagged negative association between changes in oil price and changes in average salary (-0.633 [0.255]), but there was a contemporaneous positive association between changes in electricity price and changes in average salary (0.380 [0.151]). When modelling changes in oil and electricity price, there was a significant positive association between changes in electricity price and changes in workplace injury claims per 100 FTE that was seen for both contemporaneous (0.305 [0.080]) and lagged (0.644 [0.237]) changes in electricity price.

160 140 120 100 80 60 40 20 0

2000‐01

Modelling oil and electricity prices

Figure 1: Evolution of energy prices and Australian ambulance system resource A), performance Figure 1: Evolution of energy prices and(Panel Australian ambulance system resource (Panel A), (Panel B) (Panel and safety indicators 2000-2010. performance B) and safety indicators (Panel (Panel C),C), 2000 - 2010.

2000‐01

There were insufficient observations to model the associations between changes in diesel and electricity price and changes in workplace injury claims.

Elect$ ($ / MWh)

Financial Year

Diesel – diesel price; Oil$ – oil price; Elect$ – electricity price; Exp/Resp – expenditures per Diesel – diesel price; Oil$ – oil price; Elect$ – electricity price; Exp/Resp – response; L:TRatio – labour-to-total expenditure ratio; FTE/Resp – full time equivalent employees per expenditures per response; L:TRatio – labour-to-total expenditure 10,000 responses; AvgSal – average salary; RTMed – median response time; ratio; Comp/FTE – injury FTE/Respclaims – full per time100 equivalent employees per 10,000 responses; compensation full time equivalent employees; $ 000 – thousandAvgSal dollars; L – litre; – average salary; RTMed – median response time; Comp/FTE – injury MWH – megawatt hour; min – minutes; all financial data in 2009-10 Australian dollars. Note: The compensation claims per 100 full time equivalent employees; $ 000 – scale of the Y axis varies by indicator, as shown in the legend. thousand dollars; L – litre; MWH – megawatt hour; min – minutes; all financial data in 2009-10 Australian dollars. Note: The scale of the Y axis varies by indicator, as shown in the legend.

Table 1: Ambulance system resource, performance and safety indicators extracted or calculated from the annual reports, total system-years of observations, and aggregate results for the entire study period. Indicator

Calculation

N System-Years Mean (max possible = 80) (± Standard Deviation)

Expenditures per Response ($)

Total Expenditures / Total Responses

61

542.04 ± 125.99

Labour to Total Expenditure Ratio

Labour Expenditure / Total Expenditures

64

0.65 ± 0.05

FTE per 10,000 Responses

FTE / Total Responses X 10,000

60

40.95 ± 5.32

Average Salary ($’000)

Labour Expenditures / FTE

60

85.57 ± 16.53

Median Response Time (min)

As Reported

70

9.26 ± 0.90

90th Percentile Response Time (min)

As Reported

80

17.56 ± 2.92

Compensation Claims per 10,000 Responses

Compensation Claims / Total Responses X 10,000

20

6.80 ± 2.93

Compensation Claims per 100 FTE

Compensation Claims / FTE X 100

20

17.75 ± 5.98

Resource Indicators

Performance Indicators

Safety Indicators

$’000 – thousand dollars; min – minutes; FTE – full time equivalent employees; all financial data in 2009-10 Australian dollars.

86


2013 vol. 37 no. 1

Safety

Australian ambulance systems affected by energy price

Table 2: Associations between changes in oil and electricity costs and changes in Australian ambulance system resource, performance and safety indicators. (Model)

Dependent Variables – coefficient [standard error]

Independent Variable

ΔExp/Resp

ΔL:TRatio

ΔFTE/Resp

ΔAvgSal

ΔRTMed

ΔRT90ile

76.354 [11.483] 66.129 [42.158] -1.634 [2.699] -0.968 [1.800] 3.169 [1.595] 2.720 [2.010] -0.354 [0.486] 75.9%

0.036 [0.008] 0.002 [0.016] 0.0004 [0.0009] 0.0009 [0.0004] 0.0009 [0.0004] 0.0016 [0.0005] 0.106 [0.235] 80.8%

1.154 [1.329] 0.174 [0.944] 0.120 [0.064] 0.044 [0.070] -0.029 [0.049] -0.072 [0.040] -0.173 [0.311] 91.5%

5.066 [6.409] 3.844 [3.802] -0.494 [0.325] -0.723 [0.204] 0.662 [0.201] 0.310 [0.182] -0.377 [0.339] 86.1%

0.107 [0.114] -0.215 [0.289] 0.013 [0.013] 0.022 [0.006] -0.002 [0.007] -0.003 [0.010] 0.086 [0.318] 5.6%

0.288 [0.511] -0.286 [0.361] 0.049 [0.019] 0.031 [0.014] -0.006 [0.014] -0.009 [0.005] 0.115 [0.155] 95.5%

-3.951 [10.583] 6.958 [23.398] -3.084 [1.741] -3.992 [1.949] 0.849 [0.937] 0.704 [1.785] 0.100 [0.139] 65.5%

-0.0005 [0.001] 0.004 [0.008] -0.0014 [0.0008] -0.0008 [0.0004] -0.0001 [0.0003] 0.0014 [0.0006] 0.076 [0.052] 69.2%

-0.149 [0.337] -.0187 [0.307] -0.043 [0.043] -0.085 [0.042] -0.020 [0.044] -0.014 [0.043] -0.048 [0.069] 84.9%

0.769 [1.204] -0.701 [2.560] -0.264 [0.195] -0.680 [0.223] 0.380 [0.151] -0.050 [0.183] 0.031 [0.296] 83.3%

-0.011 [0.019] -0.098 [0.101] -0.005 [0.006] 0.010 [0.006] 0.001 [0.005] 0.003 [0.006] 0.220 [0.137] 30.7%

-0.028 [0.024] -0.024 [0.185] 0.007 [0.009] 0.009 [0.011] -0.006 [0.006] 0.008 [0.008] 0.155 [0.113] 95.5%

ΔComp/FTE

(ΔDiesel$ & ΔElect$) Year Health-Based ΔDiesel$ 1L ΔDiesel$ ΔElect$ 1L ΔElect$ 1L ΔDependent Variable Explained Variance (ΔOil$ & ΔElect$) Year Health-Based ΔOil$ 1L ΔOil$ ΔElect$ 1L ΔElect$ 1L ΔDependent Variable Explained Variance

insufficient observations

-2.165 [0.948] 2.571 [1.066] -0.050 [0.058] -0.024 [0.060] 0.305 [0.080] 0.644 [0.237] -0.329 [0.395] 93.4%

p