Automation and robotics are entering many aspects of our lives and changing the way things are done. This trend extends into the laboratory where the decision to add automation is similarly complex. There is the potential for labor savings, increased precision, better accuracy and automated data logging, which must be balanced against capital costs and the need for training to operate and maintain the potentially more complex system. Automated systems can become costly to maintain when they are underutilized and can also require more highly trained personnel to operate, negating reductions in personnel required. This article explores these and other aspects of the decision to automate an analytical procedure in a generic manner.
Most people are familiar with the binary decision of make or buy. In the internal laboratory of a food processor, this decision becomes whether to send the samples out to a contract laboratory or to set up to test them internally. For the contract laboratory, the decision is whether to subcontract the work. To make these decisions, it is important to understand the purpose of an analysis and information that will be generated from the data.
We can push this binary example further. Often these decisions become a question of whether it is time to make a change or make do. Future uncertainty complicates the make-or-buy decision we face in the laboratory, forcing the decision to be make, buy or make do. One must consider both the potential benefits and the changing nature of the need as part of the decision. The availability of capital will also affect the decision.
Focusing specifically on the decision to automate an analysis, there are two cases that need to be considered. There is the trivial case where automation is required, and no other decision will be satisfactory. Similarly, if there are hazards or other constraints that make human intervention impractical or impossible, the decision to automate is trivial. This case is not typical and will not be considered further. The balance of this article will consider the non-trivial case and the parameters that need to be considered with regard to automating a procedure in the laboratory.
The Purpose of Testing
The decision to automate an analytical procedure will always be driven by an understanding of the purpose for doing the test. For the analytical laboratory, this decision is often made remotely by the person making the request, even when the laboratory is an internal resource, like a contract laboratory. Under these conditions, the decision to automate is reduced to a cost-benefit analysis. The purpose of the testing becomes the need to satisfy the customer, and all other considerations are moot. However, it is extremely important to consider whether the analyses should be sent to a laboratory that specializes in this particular procedure or analysis.
For example, the determination of milk solids is an example where a specialized laboratory is so much more efficient that to do the traditional manual methods will almost never be cost-effective. Using an automated infrared method with specialized instrumentation, the protein and fat content of milk can be measured fast and reliably. The cost per test is incredibly low compared with using the more traditional and general nitrogen content and lipid extraction method. Thus, unless there are other overriding considerations, the outside laboratory that is doing thousands of these tests has an unassailable economy of scale that renders doing the tests in-house unreasonable. Automation has clearly prevailed in this case and tends to drive customers to have other testing done by such laboratories.
Table 1 includes questions that target the nature and need for analytical results in cases where the purpose of testing will impact the need for automation. The answers to these questions will provide a list of requirements that must first be met by the selected alternative and then eventually drive the decisions of whether and how to automate.
With these answers, one or more business needs will hopefully be identified, delineated and then broadly characterized to involve safety, quality, process control or environmental factors. The bottom line on the purpose of testing is that data that is not analyzed and converted to information need not be generated, as automation will be a waste of resources that could be spent elsewhere. A simple degree Brix measurement in a juice-batching operation is analyzed if it is used by the operator to release a batch of juice for further processing. It has been used to make a decision even if no further examination of the data is made. In contrast, if data is placed on log sheets but never considered, it has no purpose.
Identifying the Best Procedure
Some analytical procedures are inherently more amenable to automation. The selection of the procedure is, therefore, an important aspect of the decision to automate. Thinking outside the box can often yield an advantage over more traditional approaches. As discussed above, infrared methods provide great benefit for milk analysis. Measuring residual adenosine triphosphate (ATP) for monitoring sanitation is similarly advantageous over traditional microbial swab testing after a plant wash down. However, ATP measurement can be so fast and portable that low levels of automation are needed. Selecting the best procedure to generate data that yields the needed information or renders decision making possible is critical to the automation decision under consideration.
For example, consider a hypothetical lemon juice processor who pays growers for lemons based on acid content. If we assume that we have representative juice samples from each lot of lemon juice, what is the best way to determine the acid content of the samples?
An obvious first choice is the traditional titration with base. There are numerous published and recognized procedures for this purpose that have been automated to various degrees as discussed below. It would be hard to fault someone for electing to take this traditional approach. However, there may be parameters that render other procedures better or preferable.
In Europe, enzyme-based assays are common to measure the level of citric acid and could be used for measuring the acidity of lemon juice if the standard of payment were based on the actual citric acid content rather than the titratable acidity. This procedure is colorimetric and thus quite amenable to automation. Various commercial systems are available if the need justifies the capital expense and the reagent costs.
A high-pressure liquid chromatography (HPLC) procedure is another alternative. This procedure directly measures the concentration of citric acid and is independent of the degree of buffering. In practice, buffering is not normally an issue, but if there is an available HPLC with an autosampler, a chromatographic procedure could become more cost-effective than other automation strategies. The data system associated with a modern HPLC system can readily transfer data to a laboratory information system, reducing the need for human intervention in the analysis process. Sample preparation time would be an important consideration. The pulp in the lemon juice will influence filtration, but numerous successful approaches are available. However, consumable costs may be higher than for a more traditional titration procedure.
Taking this example further, we can consider spectroscopic approaches. Near-infrared requires no reagents and can be made continuous or semi-continuous. It would be very fast, and consumable cost would be negligible. Data can be transferred with little or no human intervention to a laboratory information system. However, setting up a flow-through system with a flow cell might make an auto-sampler type of operation less desirable than other procedures. Still, this might be the approach of choice. However, even under conditions where near-infrared would be the best approach, it is unlikely to even be considered if an organization is focused on automating the traditional titration procedure just because it is traditional.
Table 2 includes some questions to assist in identifying and beginning to compare alternative procedures. These parameters will either identify procedures that meet the requirements defined by the purpose discussed above or will need to be optimized to achieve the greatest benefit.
Degrees and Types of Automation
The decision to automate a procedure is not binary. There are degrees of automation or at least degrees of mechanical and electronic assistance. The potential degrees of automation are as varied as the analytical procedures that are utilized. It is important to consider the full range of options in making a decision to make the best fit to the specific situation. In considering this, I have used the broadest definitions of automation to be inclusive. This concern is probably best illustrated with an example.
Let us return to the lemon juice processor discussed above. Let us further assume that titration is the procedure of choice. With these assumptions, we can focus on the degree of automation that will optimize the analysis. The simplest set-up for acid titrations includes a burette, a balance or scale for weighing samples (assuming a weight basis of reporting), a standardized base (probably purchased), a supply of Erlenmeyer flasks and an indicator. With a little training and practice, a technician can perform a large number of titrations and will be required to both judge when the indicator changes color and develop the knack of swirling the flask during the titration to insure mixing. The burette will need to be filled manually and be read at both the beginning and end of each titration. The calculation is simple and can easily be done with a calculator and reported on the appropriate log form or entered into a computer, depending on the sophistication of the information management system used in the laboratory and the requirements for documentation.
In my experience, a magnetic stirrer is the first bit of assistance observed as laboratories acquire more than the bare minimum of equipment. A magnetic stirrer avoids the swirling wrist action. It also requires adding and removing the magnetic stirrer. I am not sure this really saves time or money, but in my experience, it is usually preferred by technicians. This is discussed further in the next section as an aspect of difficulty and source of tedium.
To remove some of the art from the procedure, a pH meter is substituted for the indicator. In practice, this seems to reduce operator-to-operator variability. However, it adds to the maintenance burden as it is commonly used for acid-base titrations. However, I am not certain it fits even my broad definition of automation. It is mentioned here because it is a necessary element that would eliminate additional work when later automation options are considered. A pH meter is generally preferred to an indicator, so this maintenance can be considered part of the basic assay.
The next step in automation or method improvement is a self-filling burette. Is this a form of automation? Maybe; maybe not. A self-filling burette obviates the need to read the starting value for the titration because it is fixed by the apparatus. This means that only the weight of the sample and the volume of titrant need to be recorded or used for the calculation. With a good installation, this will speed up the analysis process. It can increase the precision and accuracy of the analysis by removing some of the uncertainty associated with the starting point for the titration. The capital cost for a self-filling burette is not large compared to a good regular burette, but it will require a little extra maintenance to maintain the valves and reagent reservoirs. Such systems work best when used on a regular basis. Sitting idle for several days will require that system be drained and the set-up procedure be repeated. This is a characteristic of most automated systems.
Escalating the level of automation further, one can add a data-logging scale or balance. This eliminates the need to enter the sample weight manually. This will reduce the potential for operator error in transcription but is only useful when an information system is in place to take advantage of this time savings. However, this system could be as simple as placing the samples weights in a spreadsheet where the titrant volumes can be entered and the acid calculations completed.
The next level of assistance or automation would be an auto-titrator. This degree of automation should meet everyone’s concept of automation. Typically, an auto-titrator system would include all the features discussed above plus automatic control of the titrant flow. These features add to the overhead but can provide great returns. Again, idle time, set-up and maintenance will be more challenging than just using a simple burette or a self-filling burette.
The ultimate in automation would be a robotic system that actually handles the weighing out of the sample for titration. This further increases the overhead. However, such systems are in use, so there are conditions where this is the solution of choice.
For any protocol, there will be a range of options for reducing the time and effort involved. The laboratory will have resources that will make some options better than others. This will depend on the range of analyses done in the laboratory and will feed heavily in the laboratory decision to automate.
Degree of Difficulty and Tedium
Automation can mitigate difficulty and reduce the tedium associated with generating analytical data. People often joke that such activity is work, and work is not supposed to be fun by definition. Nevertheless, job satisfaction will play a role in performance. Additonally, to the extent that automation can achieve these ends at a reasonable cost, automation should merit consideration.
Regarding difficulty, some analyses require skill and experience to generate useful data. Even a simple titration can be difficult when the color of the sample is close to the color of the indicator. Many microbiological analyses require a practiced hand to make the numerous sterile transfers required. Immunoassays and enzyme-based assays can also be somewhat challenging with the need for special glassware and other supplies. Increasingly, vendors supply kits that eliminate these challenges and simplify data generation. However, one pays for the convenience and the technology.
Removing the tedium from the data-generation process with automation generally comes at a price. The required level of training and expertise generally goes up with the level of automation. The routine operation of the instrument may not require a highly skilled analyst, but there generally must be a person with special training and experience to troubleshoot and maintain the automated analytical systems. Installing a sophisticated automated method in a laboratory setting where only simple, wet chemistry-style testing is done will require significant training and should not be considered lightly. However, as illustrated above, it is possible to reduce the tedium associated with data generation without incurring the full overhead of complete automation. Clearly, balancing automation with laboratory needs will minimize the tedium and facilitate testing.
Benefits and Potential Costs
In making a decision to automate an analytical procedure, there is frequently a desire to leap to the cost-benefit analysis. However, has we have already discussed, other areas need to be considered first. One must know what is needed and what will be the best approach. These decisions can get circular with a cost-benefit analysis, but there is no way to complete such an analysis without a list of approaches that need to be considered.
When looking at the cost-benefit analysis associated with the various degrees of automation, there are a number of benefits and costs to consider:
• There is the potential for personnel savings as machines and computers provide control for the analysis. Unfortunately, the dollar savings per result may be illusory as the analyst using the automated system may be paid more per hour than a technician who might perform a manual assay.
• Increased precision is often possible with automated systems. Computers and digital control will reduce the variance in analytical procedures.
• Accuracy may or may not be increased. Generally, the manual system and the automated system will have very similar detection limits and accuracies because these features are inherent to the ability to test for a particular analyte. If more accuracy is needed, an alternative approach may be required.
• The economic benefits of automated data logging can be hard to quantify. The greatest benefit accrues from avoiding transcription errors and the increased reliability of determinations. This reliability increases confidence in decisions based on the logged data, but prior to the automation, every effort was already made to get the right answers.
• Capital costs and costs per analysis for materials are important to the decision to automate. Automated systems tend to rely on specialized consumables that facilitate sample handling. These consumables are generally single-use and it is not cost-effective to wash them.
Decision Making
Ultimately, few decisions are better than the information upon which they are made. Better information is usually better than luck or intuition. The decision to automate an analytical procedure can only be made in the context of the need and purpose for testing. This establishes the requirements that must be met. There are also a series of parameters that should be either maximized or minimized to achieve optimal performance. These elements are the “musts” and “wants” by which alternative protocols and automation schemes can be compared. The best alternative will include the proper degree of automation and, if information is developed to the fullest extent possible, will even help in selecting the particular model and brand of automation.
Unfortunately, even with the best intentions, information will be incomplete and imperfect. There will still be some subjective aspects to the automation decision, such as the personal relationship with the automation vendor and the specific technical support personnel. Not all vendors perform equally in all parts of the country. Getting testimonials from respected colleagues regarding the performance of various alternatives will be critical to making the best possible decision. Testimonials from vendors can be screened for successful installations and not include organizations where the equipment was misplaced or was a poor fit.
We have examined the various criteria that can be used to compare various approaches to generating analytical data. These approaches will include various degrees of automation and associated capital costs. Each approach will have costs and benefits. This matrix of approaches and criteria will provide a solid basis for making a sound decision.
Eric Wilhelmsen, Ph.D. is a recognized world authority in food authentication, serving for over 25 years in both academic and industrial positions. He can be reached at the Alliance of Technical Professionals.
