These last few weeks I have been struggling to get my head around what, on the surface, looks to be a simple problem, namely printing quality photographs of my granddaughter. The image that appears on my monitor has perfect colour, excellent luminance and the balance of highlights to shadows is just what was intended. On quality paper the image lacks brilliance, is too dark and generally lacks what it needs to be a print worthy of hanging on the wall. Of course what looks perfect on my MAC at home also looks drab on a windows system at work. This same inability to transfer truthful colours across different platforms has also bedevilled Alison and my attempts to develop an honours board for the web-site.
The issue is that the only true image is in fact the virtual image in the computer memory. Each device, whether it be a monitor or a printer, has a profile that is essentially a lookup table that relates defined colours (red, green and blue components each have a value between 0 and 255) to what the device can display. The lookup table essentially contains corrections to compensate for deficiencies in what the monitor can display. For colours that are outside the profile limit of the monitor, the colour is interpolated. Thus the monitor and printer each have different profiles so colours will appear differently depending on where you output them. The virtual image is the same. Of course printing is made even more complicated by the fact that inkjet printers don't print in RGB (red, green, blue), but rather CMYK (cyan, magenta, yellow and black) and sometimes multiple blacks! All very complicated for what on the surface appears to be a simple problem. Colour management in digital photography is really all about error management.
This error management across platforms got me thinking about errors in general. Recently we have been introduced to disposable cardboard spacers. These have been welcomed with open arms in some quarters and in others there has been resistance "because the validation work has not been done". What are the errors associated with administering inhalers by any means - with or without spacers; with cardboard, large volume or small volume spacers?
Consider the steps in administering Salbutamol with a spacer:
- Prep the MDI
- Prep the spacer - eg if a new volumatic is being used, has it been washed and left to dry before use?
- Load the spacer
- Have the patient inhale
- Have the patient hold their breath at full inspiration
- Have the patient exhale slowly.
- Repeat inhalation to specified maximum dose.
- At a specified time test for changes
Errors can occur at every step of this chain. Errors at points 1-3 are obvious. The rate and depth of the patient inhalation will affect how much and where the salbutamol is deposited. How much settling of the drug occurs depends in part on how good the breathhold is. Whether the patient coughs following administration or is able to return gently to FRC can affect delivery. There is considerable variability in the dose administered between labs and in the time to retesting. This is hardly a robust procedure as many of these factors are outside our control.
A related question is "Why are we administering the salbutamol in the first place?" More often than not we are seeking evidence of reversibility of airways obstruction. Once over a threshold, the answer is either a yes or no. We are not seeking to determine either the highest possible FEV1 or the sensitivity to salbutamol. The actual dose administered is of secondary importance. This is probably fortunate as, at best, only about 35% of the dose get to the lungs from a large volume spacer anyway. Does it really matter whether the spacer delivers 20, 30 or 40%? Surely the more important thing is consistency of delivery. In the scale of things how much does the spacer itself contribute to the overall errors associated with the process?
We have of late been exposed to considerable discussion about reference values for our tests. It is indeed important to use a set of equations that accurately describe our local population. However, those equations and their associated errors describe a population NOT an individual. We can describe and define biases for height and age and come up with all manner of statistics to better define our reference range. The truth remains though that at least 5% of the individuals are going to be outside the reference range as defined by the population. We cannot define racial correction adequately - we cannot control the variables than define the relationship between standing height and chest volume. In Australia and New Zealand we have populations that are of very diverse backgrounds, many individuals have mixed parentage, some having both parents of the same ethnicity and being brought up in Australia or in their homeland. The errors, that we cannot control, associated with matching the individual to the population are such that focussing on a percentage predicted is potentially risky business when making a diagnosis.
As scientists I think we fail to pay enough attention to errors and the cumulative effect they have on the outcome of a process. I well remember spending longer calculating measurement errors after physics practical classes at University than I did writing up the rest of the experiments. An understanding of sources of errors in what we do is critically important and I often wonder whether we should be including measurement error in our reports. This would not help with errors in BD administration or in matching the individual to the population but at least it would raise sources of error in our consciousness.
Sadly this doesn't help me with my colour management. I will just have to keep experimenting.
'til next month,