Herein, we suggest a conceptual approach for selecting and implementing the type of diagnostic tools for implementation of SNAPS (see Figure 1
). The examples that follow in the subsequent section used a five-step process that followed a closed-loop approach similar to CLISAR and other circular economic models [25
]. The first step is to understand the specific problem as well as the social and economic context where decision-support technology may be needed. The next step is to identify readily available resources and then design diagnostic tools for creating a technology portfolio (sensors, analytics software, portable hardware, etc.). The third step involves the selection of the most appropriate tools to create SNAPS based on technical capabilities as well as interactive feedback from stakeholders. In step four, scientists and end-users test technology prototypes in field conditions by using established participatory methodologies. Finally, the results from the proof-of-concept testing are used to evaluate and refine the technology. This process is repeated until a solution meets user expectations and desired performance characteristics. The concept is based on principles of circular systems and convergent thinking [25
], where technology refinement may occur by using reductionist or parallel approaches. Below, we present two examples of how this conceptual model is applied in real-world settings. The first example is in advanced stage field-testing (refinement and technology improvement, with some elements in the second circular phase), while the second example is in the early phase of development (tool selection and technology transfer).
2.1. Early Assessment of Tuberculosis in Vulnerable Populations
In 2017, 1.6 million people died from tuberculosis (TB) globally, and there were 10 million new TB cases that occurred in the same year [27
]. TB has surpassed HIV as the leading infectious disease killer worldwide since 2014 [28
]. Furthermore, multidrug-resistant and extensively drug-resistant TB (MDR/XDR-TB) are current global public health threats. The 2017 Moscow Ministerial Declaration on ending TB, involving 120 countries and over 800 partners, identified “to advance research and development of new tools to diagnose, treat and prevent TB” as one of four action items [29
]. This meeting was followed in 2018 by a United Nations (UN) General Assembly first-ever high-level meeting to accelerate efforts to end TB [30
The care of TB patients starts with accessible and affordable diagnosis. The majority of TB patients live in poor conditions and in geographically remote areas. Culture-based techniques are the gold standard for diagnosis, but this is relatively expensive and results take six-to-eight weeks [31
]. For decades, TB diagnosis has relied on direct sputum smear microscopy (SSM) in many countries [31
]. SSM is fast, inexpensive, facile, and specific for detecting Mycobacterium tuberculosis
(Mtb) in high incidence areas [31
]. SSM does not require a highly specialized apparatus and is therefore very suitable for low-resource settings [31
]. However, the accuracy of SSM is only 25–65%, which is considerably lower than the standard culture technique, and its limit of detection is about 10,000 colony forming units per milliliter (CFU/mL) [34
]. In a recent study involving hundreds of specimens tested with culture, SSM, and the Xpert MTB/RIF system, the SSM method exhibited an average accuracy of 54% for respiratory samples and 50% for non-respiratory samples [36
]. Furthermore, the overall performance of SSM depends on different variables including the type of lesion, the type and number of specimens, the specific Mycobacterial
species, the staining technique, and the competence of the microscopist [35
]. In a 2014 survey, 22 high-burden countries conducted 78 million sputum smears valued at 137 million USD in 43,000 microscopy centers; about 61% of the analyses were conducted in the BRICS countries (Brazil, Russian Federation, India, China and South Africa) [37
]. About 79% of the smears performed in the BRICS countries were used for initial diagnosis. On average, the unit cost for a smear was 1.77 USD, including materials, labor, and overhead expenses [35
]. Several studies had shown that the accuracy of SSM improved when specimens were subjected to liquefaction, followed by the concentration of the Mycobacteria
through overnight sedimentation or centrifugation [34
]. However, the enhanced SSM performance provided by these pretreatment steps may not be sufficient to offset their increased cost, the complexity of their process, and potential biohazards.
Recent advances in bacteria preconcentration and the diagnosis of TB and multi-drug resistant tuberculosis (MDR-TB) include sophisticated techniques such as Xpert MTB/RIF, TB beads, liquid culture, centrifugation, filtration, and line probe assays [43
]. However, these techniques are not necessarily accessible or affordable for those who need them the most [48
]. Considering the high accuracy (~97%) and specificity (~99%) of the Xpert system relative to the culture standard [36
], the World Health Organization issued a recommendation in 2010 to use Xpert MTB/RIF for the diagnosis of all persons with signs and symptoms of TB. However, the Xpert MTB/RIF assay entails a price of US$
10 per cartridge. Thus, if this method was to be implemented for all people with presumed TB, the cost would exceed 80% of the total TB spending in low-income countries such as India, Bangladesh, Indonesia and Pakistan [49
]. In 2014 and 2015, there were 33 and nine SSMs for every Xpert MTB/RIF test procured, respectively [50
]. While high-end diagnostic methods are more accurate and/or specific than SSM, these techniques remain cost-prohibiting and inaccessible for people living in low-income countries where Mtb has a high prevalence.
An essential aspect of TB is the substantial financial burden placed on patients and their families due to treatment and associated costs. For example, TB patients are often required to take absence leave from work, which, is unpaid in some cases, leading to a higher risk of financial struggle in the household [51
]. Tanimura et al. reported the distribution of financial burden for the TB patient as 20% due to direct medical costs, 20% due to direct non-medical costs, and 60% due to income loss [52
]. On average, the total cost was equivalent to 58% of reported annual individual income and 39% of reported household income [52
In this context, accurate, rapid, and cost-effective diagnostic tests are paramount for reducing TB infection and its unacceptably high mortality rates, especially for an easily treatable disease [53
]. The ambitious goal of the global “End TB Strategy” to diminish TB incidence by 90% and reduce TB mortality by 95% by the year 2035 is unlikely achievable without highly accurate yet low-cost tools to address epidemics in settings of poverty [54
]. New tools must include improved point-of-care diagnostic tests that are delivered to low-income communities and at the first point-of-contact by patients in the healthcare system. Ideally, TB tests should be performed with the use of non-invasive sampling procedures, and results should be promptly delivered to the patients, allowing for a quick turnaround time for treatment in a single clinical encounter and hence avoiding the loss of patient follow up [54
Thus, our strategy was to develop low-cost biosensing assay for rapid TB detection by employing modern advances in nanoparticle science and glyco-chemistry, thus resulting in an accuracy matching the performance of Xpert MTB/RIF [55
] and standard culture. The nanoparticle-based colorimetric biosensing assay (NCBA) is based on the concept of the magnetically activated cell enrichment (MACE) technique using glycan-coated magnetic nanoparticles (GMNP). In this technique, the Mtb cells are isolated and enriched by applying a magnetic field to activate nanoparticle-bound Mtb cells without using any expensive antibodies or energy-consuming centrifuge instruments, thus eliminating the need for time-consuming growth of Mtb. The NCBA test involves the utilization of iron oxide nanoparticles with superparamagnetic properties. The incorporation of magnetic nanoparticles (MNPs) allows for significant improvements over other pre-concentration techniques due to their high surface-area-to-volume ratio and physicochemical properties. The MNP solution is colloidal in nature, providing stability, low sedimentation rates, and minimal precipitation due to gravitation forces. The MNPs are coated with glycan to facilitate their attachment to the bacterial cell wall through carbohydrate-binding protein sites, providing selectivity to the biosensing mechanism. There are three stages of specificity involved in this method: First, glycan–cell interaction is specific to the bacteria cell membrane through carbohydrate–protein binding. Second, the Ziehl–Neelsen staining used in the NCBA test is specific to acid-fast bacilli Mycobacteria
. Third: the Mycobacteria
present in sputum due to respiratory hemoptysis (i.e., intense coughing) is likely TB-causing bacteria.
The NCBA has been used to test sputum samples in Nepal (500 samples), Peru (1108 samples), and Mexico (24 samples) [55
]. In the case of Nepal, all sputum samples were tested for TB by using three different methods: SSM, Xpert MTB/RIF, and the NCBA. In this study, SSM detected only 40% of the true-positive specimens, while Xpert and the NCBA successfully detected 100% of the true-positive samples. Neither one of the methods yielded false-positive results. Table 1
presents the results from the SSM (left panel) and the NCBA tests (right panel), using Xpert MTB/RIF as the standard for defining the number of true-positive and true-negative TB cases. Table 2
presents the performance characteristics for both SSM and the NCBA, including sensitivity, specificity positive predictive value (PPV), negative predictive value (NPV), and accuracy. As shown in Table 2
, at a 95% confidence interval, SSM had a relatively low sensitivity of only 40% (29−52%), while the NCBA exhibited high sensitivity comparable to the Xpert system (95−100%). The accuracy of SSM was 90% (87–93%), while the accuracy of the NCBA was 100% (99–100%). Given the sample size and nature of the collected samples, the calculated prevalence for this cohort of patients was 16% (80 out of 500).
When samples were positive, the Xpert MTB/RIF system reported the bacterial load set by the manufacturer as very low, low, medium, and high. These four categories were used to estimate the equivalent load in SSM and the NCBA by matching the corresponding samples with Xpert results. Table 3
shows a comparison of the detection limit and dynamic range of the detection of the two techniques with respect to the Xpert system. As seen in the table, the NCBA yielded the same results as Xpert MTB/RIF at all levels of bacterial load. Conversely, SSM was unable to detect positive samples at the very low level and detected only 14% of true-positives at the low level, 48% at the medium level, and 79% at the high level. TB positive samples are normally distributed around the medium level, at which SSM exhibited a poor detection rate of less than 50%.
The NCBA method significantly outperformed SSM with a lower detection limit for acid fast bacilli (AFB) of 102
CFU/mL and a fast analysis time of 10–20 min. This diagnostic tool is facile (Figure 2
), easily scalable, and inexpensive (0.10 USD/test). According to the Ministry of Health of Nepal, a low-cost TB diagnostic test with 70% accuracy could potentially save 300,000 lives just in Nepal over the next five years [58
]. The NCBA technique shows promising potential for improving the TB control program in Nepal and other high-prevalence low-income countries. The deployment of the NCBA in remote rural areas would help increase case finding and case notification, thus supporting public health programs for fighting drug-resistant TB. There are nearly 600 microscopy centers distributed throughout Nepal in which the immediate implementation of the NCBA is possible. Similarly, this technique is applicable in many of the high TB-burden countries. In 2013, Desikan hypothesized that a universally accessible and rapid detection method with a sensitivity of 85% and specificity of 97% could save about 392,000 lives every year worldwide [33
]. Thus, the developed NCBA technology may enable the “End TB Strategy” and lead towards a TB-free world.
2.2. Alerting Mercury Exposure in Artisanal Gold Mining Communities
In South America, Africa, and Asia, millions of individuals are exposed to dangerous levels of mercury concentrations as a result of artisanal small-scale gold mining (ASGM) [59
]. ASGM is a rudimentary gold mining approach that is performed by individuals or groups with little or no mechanization, often in informal (illegal) operational settings with toxic chemicals [60
]. ASGM is composed of three main steps: crushing the ore into fines, mixing the fines with liquid mercury, and separating the mercury from gold by evaporating the mercury [61
]. Often in unregulated occupational conditions, workers perform mercury evaporation by using open pits, which not only have severe adverse health effects for the workers that inhale the mercury vapor but also release the toxic vapor into the environment. ASGM recently exceeded combustion of coal as the leading anthropogenic source for mercury emissions globally [62
]. The risk of exposure to mercury can lead to detrimental effects on the nervous, immune, reproductive, and digestive systems, induce infertility, reduce mental function, and induce kidney failure [63
The global responsibility for reducing mercury emissions was recognized by the Minamata Convention in Switzerland in 2013. At the convention, over 140 countries signed a treaty committing to protect human health from mercury exposure [62
]. The signatory countries pledged to “ban new mercury mines, phase-out existing mines, ensure the phase out and phase down of mercury use in a number of products and processes, develop control measures for emissions, and regulate the informal sector of ASGM” [62
]. In order to mitigate mercury exposure and regulate mining operations, it is prudent for marginalized communities to monitor the presence of mercury in their water through low-cost, rapid, and facile devices.
Several analytical methods have been developed for mercury determination in water. Standard laboratory techniques include cold vapor atomic absorption spectroscopy (CV-AAS) [68
], cold vapor-atomic fluorescence spectrometry (CV-AFS) [70
] and inductively coupled plasma mass spectrometry (ICP-MS) [72
]. These spectroscopic techniques are highly sensitive and accurate but are often impractical for environmental applications due to the high cost of analysis. In addition, these standard methods require extensive user training, and the results often require days or even weeks to produce results, making them less suitable for rural communities [74
]. Some field capable units are commercially available, namely based on direct mercury analysis (DMA) and handheld nanosensors/biosensors [77
]. DMA is based on the principle of thermal decomposition (vaporization), followed by amalgamation and subsequent atomic absorption spectroscopy. While extremely accurate, DMA is cost prohibitive for low-income communities because commercial prices of US-manufactured equipment range between 13k and $
30k USD. Perhaps inexpensive nanosensors/biosensors that are coupled with low-cost electrochemical techniques on portable devices are likely to be more suitable as tools for the on-site analysis of mercury, especially where ASGM is in practice.
While there are many types of transduction methods for the low-cost determination of mercury, electrochemical methods are sensitive, quantitative, and may be the mechanism of choice for cost-effective rapid detection in the field [79
]. The most common electrochemical method for ionic mercury detection is that of the anodic linear stripping voltammetry (ASV) techniques [74
]. ASV is a two-step method of deposition/accumulation during the reduction of mercury ions and stripping during the oxidation of mercury ions along the surface of the electrode. As the mass transfer limit is reached in the reaction, the oxidative current forms a well-defined peak that can be used to calculate the concentration of mercury in the sample [81
]. The efficiency of any electrochemical stripping test can be determined by calculating the percent change in oxidative current relative to baseline.
Carbon-based nanomaterials are a popular choice for improving the electrochemical detection of mercury, as this type of material exhibits a high surface area, strong mechanical strength, excellent thermal conductivity, and high conductivity [82
]. Some of the carbon nanomaterials in recent literature include glassy carbon [85
], carbon nanotubes [87
], graphene [88
], and reduced graphene oxide [89
]. While each of these nanocarbon materials is efficient for mercury detection via stripping voltammetry, some of the materials are complicated to fabricate and exhibit poor water solubility [90
]. Among carbon nanomaterials, graphene and reduced graphene oxide (rGO) have the highest water solubility and one of the lowest fabrication costs. For these reasons, there is a growing trend to develop disposable, low-cost, graphene-based electrodes for field applications.
Examples of low-cost graphene electrodes include screen-printed electrodes and conductive paper and plastic [74
]. In 2014, Lin et al. (2014) [92
] discovered a low-cost, one-step, conductive material when reducing graphene on a commercial polymer with a carbon dioxide infrared laser. Since then, multiple researchers have shown that laser scribing could be used to design electrodes to sense biomolecules by using infrared and ultra-violet light lasers [93
]. While graphene is indeed a useful material in sensing, one of its problems is the tendency of graphene and graphene oxide to bind to a variety of materials in aqueous phase [97
]. For this reason, sensor labs typically metallize graphene electrodes with a noble metal that has a specific interaction with mercury ions. These metals can be deposited by using simple electrodeposition methods or advanced techniques such as pulsed sono-electrodeposition [98
]. Recently, Abdelbasir et al. 2018 [99
] showed that copper nanoparticles recovered from waste cables can be used to detect ionic mercury by using linear sweep stripping voltammetry (LSSV).
Low-cost, portable, mobile phone-based acquisition systems have been developed for mercury analysis in the field [100
]. While this is significant for deploying sensors in low-income regions, the inexpensive-portable sensor-systems lack data analytics capability to transform the data into meaningful information that could be useful for the user. For example, the maximum concentration level for inorganic mercury in drinking water is 6 ppb [101
]. However, bodyweight, ingestion rate, length of exposure, form and pathway of the contaminant, health of the individual, and concentration of mercury influences the degree of mercury toxicity [102
]. Thus, a SNAPS tool may assist communities in acquiring data and extracting actionable information for decision support.
Our group is currently working on developing the SNAPS platform for estimating the toxicity risk associated with the ingestion of mercury-contaminated water. This SNAPS platform is composed of a disposable graphene–nanocopper sensor that is coupled with a low-cost handheld potentiostat and a smartphone. The working mechanism of the platform starts with the detection of mercury present in the sample by using the graphene–nanocopper sensor. Next, selective electrochemical interactions between mercury and the electrode generate an electrical signal. The electrical signal is acquired and processed by the potentiostat to produce a current output. Then, computer software records the current output and transforms it into concentration data via calibration curves. Finally, a smartphone app is used by the user to enter the data for the following parameters: mercury concentration in water (from the sensor), bodyweight of the user, water ingestion rate, and length of exposure. Based on these parameters, the app runs an algorithm that includes a hazard quotient formula to generate an estimation of the risk of toxicity for the user [105
We recently conducted a proof-of-concept demonstration of this SNAPS platform in a rural area that has been dramatically impacted by ASGM known as La Toma in Cauca, Colombia. Even though this SNAPS platform is in an early stage of development, it represents an example of how rural communities in developing countries may use sensors as a service to access data on mobile devices and extract actionable information to help make informed decisions. Figure 3
shows the progression of the proof-of-concept demonstration of the technology.
Mercury enters natural aquatic systems primarily due to the burning of mercury amalgam during the extraction of gold from raw ore.