5. METODOLOGÍA
5.3. Objetivo específico 3: Determinar la influencia de los elementos de interés provenientes del
5.3.1. Agua superficial y subterránea
Health evidence was assessed along with the environmental proxy. The health evidence focused only on the CSG area (i.e., hospitalisation rates in the CSG area) and the environmental proxy made use of the selected environmental indicator for CSG development (i.e., monthly well counts) to determine the level of CSG well development activity. Details of health evidence, as well as the environmental proxy and the corresponding analyses, are provided below.
2.2.1 Health evidence
As with Chapter 6, population data were matched to the hospital admissions data provided by Queensland Health (Chapter 4, Section 2.2.2) so that accurate hospitalisation rates could be calculated.
Estimated resident population (ERP) data for the CSG area were obtained from the Australian Bureau of
Statistics (ABS) for every calendar year (1995-2011). This was divided by 12 to obtain an estimate of monthly population counts. Monthly hospital admissions data were provided by Queensland Health for the same area, for the duration of the study period to calculate monthly hospitalisation rates to align with monthly well data. Details on obtaining population data were outlined in Chapter 3, Section 3.
Hospitalisation rates per 1,000 persons were calculated for each month of each calendar year of the study period. The following rates were calculated:
all-age, all-cause rates;
all-age, cause-specific rates;
age-specific, all-cause rates; and
age-specific, cause-specific rates.
2.2.2 The environmental proxy: CSG well development activity
CSG development is essentially a closed process once well development is completed (the gas flows in pipelines). Within the whole of the UNGD process, the well development stage is reported as presenting the highest environmental health-related impact (Jagals, 2013). Due to limited environmental data, monthly CSG well count (i.e., growth in CSG well numbers) was used as the indicator for resource development in the CSG area. Gas well count, rather than gas production, served as a proxy for CSG well development activity (i.e., the activities and associated construction related to the well development stage rather than to the processing stage). Details on obtaining CSG well data were described in Chapter 4, Section 2.2.1.
A variable was created to partition the monthly time periods and numbers of wells into categories relevant to the timeline of CSG well development (shown in Figure 7.1).
Figure 7.1. Monthly well count data for the coal seam gas (CSG) area, 1995-2011 (reference category: ‘very low’ period of January 1995-March 1999).
As in Figure 7.1, gas well development (GWD) activity data were divided and categorised as follows: ‘very low’ = January 1995-March 1999; ‘low’ = April 1999-June 2003; ‘medium’ = July 2003-September 2007; and ‘intense’ = October 2007-December 2011. Each well category had a total of 51 months, with the early (‘very low’) period serving as the reference category.
2.3 Analysis
Regression analyses were conducted on monthly hospital admissions data (either all-age, all-cause;
all-age, cause-specific; age-specific, all-cause; or age-specific, cause-specific) to determine if hospital admission rates in the CSG area increased significantly over time in GWD periods with increasing CSG development activity compared to the reference period. This was done for conditions previously
identified as increasing over time relative to either the CM or RA areas (in Chapter 6). Again, the analyses presented here are preliminary and are intended to guide future research.
The analyses were completed using SAS 9.4 (SAS Institute, 2013). The Vuong test (SAS Institute;
Vuong, 1989) was used to check each regression to determine which model provided the best fit for the data (Poisson, negative binomial regression, or zero-inflated negative binomial regression). Additionally, the deviance, dispersion, Akaike Information Criterion, and Bayes Information Criterion were checked to assess goodness of fit (Ismail & Jemain, 2007; Lord & Park, 2010; Zhang & Liu, 2013). As with the previous analyses, counts were modelled with a log link function, and models were offset by the log of the population. For these analyses, the primary outcome of interest was to determine whether admission rates increased over time in any of the latter periods of GWD activity compared to the referent.
This analytical method allowed for the calculation of rate ratios (RR; 95% CI) to determine whether there was an increase in hospitalisation rates for a given health outcome over time in the CSG study area relative to GWD activity. As above, the ‘very low’ category was the reference, so rate ratios were calculated comparing ‘very low’ GWD activity with ‘low’ GWD activity, ‘very low’ with ‘medium’
GWD activity, and ‘very low’ with ‘intense’ GWD activity.
Due to the magnitude of the analyses, a strategy was used to guide presentation and interpretation of results in the most comprehensible way. (Note: all results are presented in Appendix L.) This strategy determined which results to focus on in this section of the thesis and was as follows:
Statistically significant increases over time that were observed in hospitalisation rates in the CSG area relative to both the CM and the RA areas for all-ages, or within specific age groups (as discussed in Chapter 6), were of primary interest.
o Within this previously identified group of conditions, it was necessary to determine whether there were increases over time in hospitalisation rates for the identified health condition as a function of each GWD activity category (‘low’, ‘medium’, ‘intense’) compared with the reference category of ‘very low’. Similar to a typical dose-response relationship, a positive relationship between GWD activity and hospitalisation rates was considered to be indicative of a stronger association between GWD activity and hospitalisation rates.
o Increases in hospitalisation rates across some of the GWD categories (but not all) compared with the reference category of ‘very low’ were also considered. The rationale for this was that there could potentially be lag times between exposures associated with GWD activity and the development of health consequences serious enough to result in hospitalisation. In this way, health consequences (measured using hospitalisation rates within ICD chapters) that occurred in the CSG study area as a function of increasing GWD activity were identified.
Statistically significant increases over time that were observed in hospitalisation rates in the CSG area relative to either the CM or RA areas for all-ages, or within specific age groups (Chapter 6), were also of relevance, but to a lesser extent compared to the conditions where hospitalisation rates increased over time in the CSG area compared to both the CM and RA areas.
o The same groupings, as described above, were then used for assessing the strength of
association for the health conditions where the hospitalisation rates increased in the CSG area relative to either the CM or RA areas. Again, a positive dose-response type of relationship between GWD activity and hospitalisation rates provided the strongest association between GWD activity and hospitalisation rates.
As scoped in Chapter 1, it was not the intent of this ERHI assessment to weigh costs and benefits;
instead, the focus was on potential adverse health outcomes over time as a function of GWD activity. This approach was used for the previously identified health conditions across all-age, all-cause rates; all-age, cause-specific rates; age-specific, all-cause rates; and age-specific, cause-specific rates.
3 Results
The results are presented for the impact on all-ages, followed by the impact on children/adolescents and adults, separately, as measured through hospitalisation rates.