People versus Machines?

For years, microbiology laboratories have performed many of the same tasks – microbial detection, identification, resistance testing, and a host of similar applications involving test tubes, flasks, and above all, time. Microbiology is a resource-intensive discipline, which can cause problems when labs simply don’t have the necessary equipment, staff or skills. One solution that improves year on year is laboratory automation – letting machines assist with, or even independently handle, some of the tasks that are tricky or take additional time to do by hand. But how can pathology labs access this kind of technology? How can they make best use of it? And how can they ensure a balance between automated contributions and the “human factor?”

As a healthcare scientist, I have four decades of experience in both the public and independent sectors. For the last 18 years, I’ve been head of the microbiology service at the Royal Free Hospital (RFH) in London – a responsibility that includes the high-level isolation unit pathology laboratory in which we process samples from patients with viral hemorrhagic fever. I’ve also held a variety of other pathology-related roles, from laboratory information management system (LIMS) implementation to project management. Recently, pathology at the RFH has entered into a joint venture in which I am the scientific lead: we’ve merged microbiology services for two separate London National Health Service (NHS) Trusts in concert with an independent provider. The result – Health Services Laboratories – is our way of concentrating expert services on one site, with the goal of improving patient services and more efficiently using NHS resources.

Thanks to the service reconfiguration, we now have a lot of intriguing new opportunities. My colleagues and I have been working together to develop a clinically-led model for the future of infection services within Health Services Laboratories. Our laboratories are new and state-of-the-art, providing us with an environment in which best practice can be achieved. But most of all, we’re excited about the future development of molecular and automated sciences – advancements that will let us focus our attentions on the areas of greatest need (for instance, antibiotic resistance or syndromic medicine and surveillance) to improve outcomes for our patients.

Historical obstacles

Until recently, microbiology hasn’t had the benefit of significant automation like the kind we’ve seen in clinical biochemistry or hematology. Instead, it has relied on well-educated, highly trained staff to first process patient samples and then read hundreds of cultures and antibiotic sensitivity plates! Traditional microbiology testing generally includes microscopy, antigen detection, serology and culture – and, where possible, we’ve tried to introduce automation with the use of multipoint inoculators, microplate readers, and antibiotic disc readers. In the case of bacterial and fungal infections, clinical specimens are inoculated onto a range of selective and differential media, then incubated at various temperatures and atmospheres. In most cases, biochemical testing is then performed to identify species, and plating or broth dilution is used to determine antimicrobial susceptibilities. The only rapid, reliable method available is Gram staining – which, even now, results in a 17 percent decrease in mortality when the results are reported in under an hour (1).

Unfortunately, not all of these tasks have historically lent themselves to automation. As a result, staffing is a major and costly issue within pathology, because we need to ensure that sufficient levels of qualified and support staff are available around the clock for timely patient care. In certain areas of the United Kingdom – like London, where I work – that can be a challenge, because it’s difficult to employ large numbers of staff due to the high cost of living; in other areas, there might be different problems, such as few trained staff available or a lack of resources to out-compete larger laboratories. In theory, staff numbers can be reduced if we increase our use of automated platforms, but here, too, there’s a caveat; automation is only as good as the quality of staff operating the service, as it’s their job to ensure that the systems are verified, validated, quality-controlled, and that the results make sense in clinical context.

Traditional microbiology cultures were thought to represent a “gold standard” of diagnosis, but with today’s technological advancement, that’s no longer really the case. In fact, their sensitivity and specificity are poor for many reasons (the distribution of organisms in patient samples, poor sampling and transport techniques that could kill microorganisms, misidentification, and even organisms like Chlamydia species that can’t be cultured in a routine setting) when compared to modern nucleic acid amplification techniques. Phenotypic testing presents a challenge in terms of identification to species level, whereas 16S rRNA gene sequence comparisons allow universal phylogenetic tree comparisons. Even antimicrobial sensitivity testing results can vary based on the media used, atmospheric conditions, and the person reading the end result.

One significant drawback trumps all of the others, though: the time to result, or turnaround time. In some circumstances, traditional testing may take days or even weeks to return results – and in the case of organisms like mycobacteria or fungi, a delayed or incorrect diagnosis can result in inappropriate antimicrobial prescription and adverse effects for the patient, which could be life-threatening! In severe disease, even an hour’s delay in treatment may significantly affect the outcome. We need testing methods that are not only accurate, but fast – because ensuring that patients receive the right drug at the right time is good both for individual outcomes and for antimicrobial stewardship.

Tackling preconceptions

It seems clear that the benefits of automation are in demand, which explains why, since the mid-2000s, companies have worked hard to introduce automation packages to suit most diagnostic environments – sample spreaders, antibiotic sensitivity and agar plate readers, and an ever-increasing array of nucleic acid amplification analyzers and sequencers.

Automation doesn’t mean just pushing thousands of agar plates around a system that incubates them and takes images. It’s much more than that – a combination of methods and science. So how can it help overcome the issues with traditional testing? The answer to that question depends on the type of automation, what’s incorporated into the system, and to what degree the methods are applied directly to patient samples. Some well-known applications for detecting pathogens directly from the patient use simple lateral flow devices or high-throughput molecular systems to detect chlamydia, gonorrhea, methicillin-resistant Staphylococcus aureus, fecal pathogens, or dermatophytes from patients’ nails. All of those tests yield results within a few hours and can be automated for even faster readouts. But that’s not the only area in which automation can help. To offer just one more example, it undoubtedly improves the diagnosis of sepsis. The use of blood culture monitoring systems in combination with MALDI-TOF mass spectrometry, rapid fluorescence in situ hybridization, and automated minimum inhibitory concentration (MIC) sensitivity testing, enables a meaningful result in eight to 12 hours (compared with the 48 to 72 hours of conventional testing). The outcome? Rapid treatment and reduced morbidity and mortality.

Central to automation is information technology (IT). From hospital order communications to LIMS, it’s the key element that bonds the individual components of automation together. In concert with expert rules, IT can help users enhance data flow – and, in some circumstances, even complete tasks with no intervention from medical staff. For instance, computers can apply expert rules so that isolates from specific, fully sensitive sample types can be authorized directly, while isolates that fail the expert rule are directed to the medical team for further investigation. With the advent of the smartphone, patient data can be sent directly to the clinician via the LIMS, saving time and money in antibiotic stewardship. But human beings aren’t the only things for which IT reduces the need; another benefit is that it’s usually a paper-light, or even paper-free, system.

In our microbiology service, we currently use a variety of automation options. We perform automated blood cultures, antibiotic sensitivity testing, microbial identification using MALDI-TOF, molecular pathogen detection (for MRSA, CRE, Clostridium difficile, chlamydia, gonorrhea, tuberculosis, other mycobacteria, enteric pathogens, and dermatophytes) and 16s 18siITC sequencing. Wherever possible, all systems are interfaced using either our own LIMS or the integration engine associated with the machines. At the moment, we’re anticipating an upcoming move to a new, state-of-the-art center for infection, at which point we will be stepping up our use of automation, making strategic use of a KIESTRA lab automation system (BD) to deal with a complex workload from a range of clients.

Automating appropriately

Despite automation’s many advantages, potential users should remain mindful that all that glitters is not necessarily gold. It’s important to carefully consider the patient population and the microbial epidemiology when bringing the component parts together, ensuring that each test’s positive and negative predictive values meet your laboratory’s needs. Even more important is avoiding the “black box” mentality – the belief that a department can use automation to reduce skilled personnel numbers or replace them with staff who have only basic scientific training. This is fraught with ignorance – after all, just because the system produces a result doesn’t mean that the result is correct. For instance, if we examine nucleic acid amplification testing and threshold cycle (Ct) values, where this is the intersection between an amplification curve and a threshold line, we get a measure of the concentration of target in the polymerase chain reaction. But many factors besides that concentration impact the absolute Ct value. As human beings, we can untangle those complicating factors to understand the test output. But in automated testing, does the software interpret the Ct values correctly? We need a skilled laboratory professional to verify that – because if it doesn’t, and there’s no operator looking at the curves, then as with any automated tool, we may end up with a result that is meaningless, or worse, incorrect in a way that places a patient at risk.

Automation isn’t one-size-fits-all, either. The size and scope of the laboratory can influence the type of system that can be used. The footprint for high-throughput analysis, for instance, is not small, especially in circumstances where more than one analyzer is needed. Does your laboratory’s infrastructure have room to accommodate all the necessary workflows of best practice for molecular diagnostics while minimizing environmental contamination? Not every service can make the jump to automation in the same way, or with the same tools.

Investing in people

It seems counterintuitive to say that people are the most important aspect of automation, but it’s also true. The key to an efficient laboratory is getting the skill mix right. That means not just using staff at a basic grade, but investing in the workforce. Hiring the right people is the obvious first step, but what’s next? Providing continuing professional development keeps healthcare scientists up to date so that they can develop and troubleshoot equipment and assays. Employees with growth opportunities stay motivated, and motivated staff take less time off and make fewer errors. In an ideal situation, an automated laboratory should be well-staffed, run 24/7, accommodate the vagaries of molecular workflow, and incorporate automated microbial culture so that samples flow through the system without the need to batch additional tests like MALDI-TOF or antimicrobial resistance detection. But a system like this comes at a price, and with the NHS and so many other healthcare systems financially stretched, that’s more of an obstacle than ever. How can we overcome it? Healthcare scientists and clinicians must work together to demonstrate that automation offers better patient outcomes, more efficient bed management, fewer nosocomial infections, and a reduction in the use of expensive antimicrobials.

Importance of industry

People are the main drivers of the future of automation, but others – industry, innovation, and point-of-care testing opportunities – also play a role. MALDI-TOF mass spectrometry is a great example of the way industry plays into advancing automation. The technique has revolutionized microbial identification by increasing speed and accuracy, forcing manufactures of phenotypic methods to rethink their approaches to keep up. Convenience is also important; inevitably, patients and clinicians require diagnostics as close to the patient as possible, so the use of smart devices and the ability to monitor patient parameters in real-time are major incentives for development. My hope in that regard is that we’ll soon improve our ability to distinguish colonization from true infection. At present, we culture sputum for so-called pathogens – but we know that sputum is a poor sample to use for diagnosis, because it’s possible that the organism we detect is only colonizing the respiratory tract. I think the solution to this problem, and to many of automation’s weaknesses, lies in industry. Diagnostic companies should be developing automated systems that can detect pathogens with resistance profiles directly from patient samples by applying a combination of molecular detection and mass spectrometric quantification in real-time. In fact, I suspect that this may well be the future of the diagnostic microbiology laboratory!

As our financial and technological situation changes, the landscape of pathology is changing along with it. Wherever possible, it’s important for local departments to work together, sharing equipment and resources so that we can all move toward fully staffed, 24/7, best-practice service together. Automation is an increasingly significant part of that. In my opinion, it has already improved patient outcomes, at both relatively simple levels of automation like using software to reduce transcription errors, and more complex aspects like using mass spectrometry and NAAT to reduce the time needed to detect a pathogen from days to hours. But the biggest benefits come from applying science wisely. There’s no point investing in automation if the science isn’t focused on the most vulnerable, immunocompromised patients. Why? That’s the population that generally has the worst outcomes and costs service providers the most money through prolonged stays and expensive treatments.

With the emergence of totally treatment-resistant bacteria, and the rapid evolution of new diseases and new drugs, it’s vital that we use the best science available to us to reduce the spread of these organisms. With services thoughtfully designed to incorporate the careful use of automation, I think we’ll find ourselves much closer to the solutions we seek.

Simon Rattenbury is Head of Laboratory Service in HSL Microbiology at The Royal Free London NHS Foundation Trust and scientific lead London for Health Services Laboratories, UK.