Appendix 1: Life Cycle Assessment Methods
For this appendix, we primarily draw upon past explanations about life cycle assessment generally,[43–45] and from several previous publications from our broader group.[4,13,17,18] Life cycle assessment is a scientific method to determine the entire "cradle to grave" environmental and financial effects of processes and products.[43,45] The Society for Environmental Toxicology and Chemistry (Pensacola, Florida) defined the components of a life cycle assessment in 1991: (1) raw material acquisition; (2) processing and manufacturing; (3) distribution and transportation; (4) use, reuse, and maintenance; (5) recycling; and (6) waste management. Everything we use and do has an environmental footprint, whether this is for a tangible product or a service such as an admission to hospital. Life cycle assessments have a "system boundary," i.e., a limit to which one examines the environmental effects of a product or process. This system boundary is defined by local Australian and international standards.[14,19] For example, if we are examining a plastic syringe, the system boundary could be defined to include the manufacture of the plastic and ongoing maintenance of installed infrastructure, but not the actual manufacture of such installed infrastructures which are in turn used to make the syringe.
Environmental factors beyond carbon dioxide equivalent emissions, including water consumption; petrochemical use; air, water, and terrestrial pollution; and release of toxic byproducts, can be accounted for in life cycle assessment. We have focused upon carbon dioxide equivalent emissions as they are an important focus due to the increasing health concerns of climate change. In the late 1990s, standardization of how life cycle assessments should be conducted was achieved when the International Organization for Standardization released the ISO-14000 series.
Using the ISO-14040 standards, we defined our study's functional unit as all anesthesia for a total knee replacement in a public hospital in Victoria, Australia. The ISO-14040 standards life cycle assessment system boundary defines inclusions/exclusions. We did not include data for heating/ventilation/air conditioning, or any surgical equipment.
Importantly, once one has details about the components making up a process/procedure, their masses/amounts, and their origins, then one can then undertake a life cycle assessment with the relevant software and application. For example, for a general anesthetic, we obtained quantified data about (1) electricity used for cleaning/sterilizing reusable equipment, the patient air warmer, scavenging, and the anesthetic machine; (2) plastics, steel, cotton, and so forth; (3) pharmaceuticals; and (4) volatile anesthetics and oxygen use. Data related to the source/origin of the electricity, plastics, and so forth were also important. With these input data, we then turned to quantifying the outputs with life cycle inventories. We obtained the power rating for the patient air warmer (0.8 kilowatt-hours/h) from online data for Model 775, Bair Hugger, USA. Anesthetic machine electricity use (0.08 kilowatt-hours/h) was obtained from Chakladar, and anesthetic scavenging (0.4 kilowatt-hours/h) from Barwise.
Life Cycle Inventories
Life cycle assessments make use of life cycle inventories. A life cycle inventory is a catalog of flows to and from nature, with inputs such as energy, water, and raw materials, and outputs (releases) to air, land, and water. There can be a large number of inventory flows numbering in the hundreds to thousands, in such a way that the life cycle inventory of even a simple plastic syringe requires multiple flows of petrochemical resource extraction, manufacture, transport, and use. To examine all of these details de novo every time a life cycle assessment is undertaken would be prohibitively exhaustive and expensive. It is ideal to obtain as much primary/foreground data (e.g., measurement of electricity use for a hospital sterilizer) as possible in order to reduce the uncertainty of the data. Nevertheless, multiple secondary/background sources of information are usually required for life cycle assessments (e.g., details of plastic manufacture).
Large national and international databases are the routine sources for such secondary data, such as EcoInvent and the Australian Life Cycle Inventory, which incorporate geographically specific average industry data. For example, the estimated carbon dioxide emission from burning coal from a defined region is obtained from such environmental databases. Such average industry data can have greater associated uncertainty than directly measured (primary) data.[27,44] Care must then be taken to ensure that the secondary data indicate the local conditions of the life cycle assessment in question (e.g., local coal-fired electricity versus hydroelectric electricity used for the secondary data).
A process diagram/tree (Figure A1.1) is developed from all of the inputs that make up an output. We have included the process diagram for spinal anesthesia as an example. One can see that electricity forms a large part of the total carbon dioxide equivalent emissions as indicated by the wide red lines associated with electricity, with oxygen also being important on the right-hand side of the process diagram. Note that in this diagram, in order to be able to visualize some of the complexity of life cycle assessment methods, we have included a "cutoff" of only items that contribute greater than 1% of the final carbon dioxide equivalent emissions to general anesthesia. In reality, we included all inputs (at least several hundred) that contributed to the final carbon dioxide equivalent emissions.
Statistical Analyses: The Pedigree Matrix and Uncertainty
The life cycle inventory thus has inputs (such as electricity from coal) that are combined to form an output (e.g., a plastic syringe). Every input in every process from secondary databases has a degree of uncertainty associated with it. This uncertainty routinely cannot be derived directly from the available information, so a standard procedure was developed to derive uncertainty factors from a qualitative assessment of the data, known as the Pedigree Matrix. The Pedigree Matrix is a commonly used qualitative scoring system derived from the secondary data's reliability, completeness, temporal and geographical proximity to the process or item being assessed, and further technological factors,[27,44] with a score from 1 (good) to 5 (poor) for each factor. The Pedigree Matrix relies upon expert judgment. For example, if the secondary data for carbon dioxide equivalent emissions per kilowatt-hour of electricity produced was obtained recently from all local coal fired power stations, this would have better reliability, completeness, and temporal and geographical proximity than secondary data from an overseas-derived database that sampled one coal-fired power station a decade ago. As the Pedigree Matrix is based upon expert opinion, it is open to a perception of irregularities. The Pedigree Matrix has been updated to incorporate some of these concerns with greater emphasis upon direct empirical values for each of the factors.[17,46]
There are also uncertainties associated with all life cycle assessment primary inputs that are directly measured. For example, the plastic syringes used by anesthesiologists in our study were transported from the Philippines to Australia. There is little uncertainty associated with the carbon dioxide emissions from such shipping as the distance traveled is known and the variability in fuel consumption of container ships is small. Similarly, the sterilization of the reusable plastic spinal trays in our study had little uncertainty as we had measured the sterilizer's electricity use more than 1,000 times with different load types. If we had measured this sterilizer electricity use but once, the carbon dioxide equivalent emissions from such electricity use would have a greater associated uncertainty. As for secondary data from life cycle inventory databases, the Pedigree Matrix for primary input data is a qualitative scoring system.
To combine the values and frequency distributions of these hundreds of inputs to obtain outputs such as carbon dioxide equivalent emissions, we used Monte Carlo analyses (routine for life cycle assessment). Monte Carlo methods are a broad class of computational algorithms that rely on repeated random sampling to obtain numerical results. Monte Carlo methods are useful when there are large numbers of inputs and where it is impractical to obtain data for each of these inputs de novo.[27,44]
When there is a range of possible values for a result, there are a number of approaches to how to determine the best estimate and the frequency distribution with CIs around this result. Monte Carlo methods take data points from within the frequency distributions for all inputs to develop a final output result, frequency distribution, and the plausible range, including the central tendency of the frequency distribution. The greater the number of "runs" by Monte Carlo analysis, the better the estimate of the most likely value and the associated frequency distribution. A final 95% CI for a process is achieved based on the random sampling anywhere within the 95% CIs for all inputs. A Monte Carlo analysis includes at least 1,000 "runs" of random samples to reduce the chance of unusual results—that is, taking input data from the extremes of the 95% CIs. The 95% CI of the mean/average (or any other result) indicates what the variability of the results could be if the study was performed a large number of times. The 95% CI of the mean/average from Monte Carlo analysis may not be closely aligned with the directly obtained minima/maxima results. The 95% CI may lie within or beyond the minimum/maximum. This is because the 95% CI is reflective of the mean only; it is not immediately relevant to the other directly obtained results such as the minimum/maximum (range).
Modeling and the Final Results
As noted in the Materials and Methods section, we used two life cycle inventories (Ecoinvent and the Australian Life Cycle Inventory) to obtain carbon dioxide equivalent emissions associated with devices and processes. For all processes involving local electricity consumption (kilowatt-hours), we have used the Australian inventory. This is particularly relevant to electricity for patient warming, anesthetic scavenging, cleaning/sterilizing, liquid oxygen compression, and waste management. Importantly, Australian carbon dioxide equivalent emissions per kilowatt-hour are considerably higher than the European average due to coal-fired electricity sources of electricity in Australia. For all devices (e.g., manufacture of plastic endotracheal tubes), we used the Ecoinvent inventory to obtain the associated carbon dioxide equivalent emissions. Because most common products (e.g., plastics, steel, cotton) are traded on the international market, their origin can be varied and multiple, and it can be difficult to trace the precise origins of their makeup. Ecoinvent thus uses a "rest of the world" approach, averaging the associated carbon dioxide equivalent emissions. For example, if we know the carbon dioxide equivalent emissions/kilogram plastic polypropylene manufacture for 30 countries, we use the average carbon dioxide equivalent emissions per kilogram for that process.
Data were modeled in SimaPro-9 LCA (life cycle assessment) software (PRé Consultants). We developed an inventory that quantified materials and energy used, and modeled this using the Ecoinvent (version 3.5) and Australian Life Cycle Inventory databases. We used the International Reference Life Cycle Data System 2016 (European Commission) impact assessment method to translate the inventory into environmental impact scores, along with Monte Carlo software algorithms (SimaPro) to obtain results and 95% CIs. We divided our data on environmental impacts by an average Australian person's total daily environmental effects in order to compare the environmental impacts with peoples' routine activities. To ascertain a global perspective, we modeled our results (carbon dioxide equivalent emissions) with Ecoinvent electricity data with those for identical anesthetics being provided in China, the European Union, and the United States. Note that the aforementioned rest of the world average approach across at least 30 countries means that the carbon dioxide equivalent emissions arising from other items such as plastics manufacture will not vary between countries. Only variations in the carbon intensity of electricity generation will lead to intercountry variability in carbon dioxide equivalent emissions.
It is routine to provide 95% CIs in life cycle assessment around the summated data, but atypical to do so for all further modeled data. For example, figure 4 gives the carbon dioxide equivalent emissions for different countries for general, spinal, and combination anesthesia. There are 12 bars in this figure, so any 95% CI analysis would be prolonged. There are reasons though why such effort would be quite superficial. By definition, the same items/processes are being used in Australia and China/Europe/the United States (e.g., electricity for multiple processes, single-use plastics, pharmaceuticals). Only the carbon dioxide equivalent emissions per kilowatt-hour or kilogram plastic will vary. The uncertainty associated with the carbon dioxide equivalent emissions for each of these common items/processes is thus proportional. For example, if 1 kg of carbon dioxide equivalent emissions is produced by 1 kilowatt-hour of electricity in Australia, but only 0.5 kg of carbon dioxide equivalent emissions in the United States, the 95% CI is approximately (not precisely, but near enough) half that in the United States compared with Australia. If a process is highly uncertain in Australia, then it will be highly uncertain elsewhere, just relatively so (according to the associated carbon dioxide equivalent emissions). The same model is being used to determine the carbon dioxide equivalent emissions and the uncertainty.
The authors thank the nurses, engineers, electricians, and doctors of Williamstown Hospital operating rooms (Western Health), Melbourne, Australia. Catherine O'Shea, R.N., M.Env.Sci. (Western Health), Laura Elliot-Jones, B.Clin.Sci., M.D., and Lisa Dahl, M.B.B.S. (Western Health) assisted in data collection.
In-kind support (no cash funding) was provided solely from Western Health Anaesthesia Department sources (Melbourne, Australia).
Anesthesiology. 2021;135(6):976-991. © 2021 American Society of Anesthesiologists | Lippincott Williams & Wilkins