Anaemia constitutes a major global public health problem. Globally, it currently affects two billion people (32.9% of the world’s population) [1]. However it is disproportionately concentrated in developing nations where it is four times more prevalent than in developed world [2]. Anaemia also results in one million deaths annually, with three-quarters of these occurring in Africa and South East Asia [3]. Anaemia is associated with major health consequences including decreased well-being, fatigue, lethargy, impaired physical activity and work performance [4]. In children it is associated with increased mortality and decreased cognitive and physical development [5, 6]. In pregnant women it is associated with low-birth weights, increased risk of maternal and perinatal mortality [7].

Despite the fact that the burden of anaemia is highest in developing nations, standard medical testing facilities are often limited and inaccessible to most patients [8]. Thus most cases of anaemia will go undiagnosed and untreated [9]. The economic realities in developing nations makes it cost-prohibitive to setup and run standard laboratory facilities at all levels of the health care system [8, 10, 11]. The use of point-of-care (POC) haemoglobin meters is potentially economically advantageous and time-saving and may also be of tremendous help in critical care units where rapid and frequent testing are required [11, 12]. These devices are much cheaper than automated analysers and do not require sophisticated expertise training to be able to operate them [13]. As such they are usually operated by a non-laboratory staff such as a physician who is directly involved in patient care [12]. They are battery-powered and will be highly beneficial in settings with inconsistent electric power supply. They are either bench-top or hand-held thus do not require sophisticated infrastructure [14]. They require smaller blood samples and thus are valuable in conditions that require multiple blood draws such as in critical care units and neonatal/paediatric units [12]. This reduces the risk of iatrogenic anaemia. They require capillary blood samples and will be appropriate in situations where it is technically difficult and distressing to perform phlebotomy [15].

Despite the potential benefits of using POC haemoglobin meters in clinical practice, there is still a lot of concern surrounding their accuracy [16, 17]. Do they underestimate haemoglobin values and thus expose patients to needless and sometimes harmful treatment for anaemia such as blood transfusions or do they overestimate haemoglobin values and as such prevent patients with anaemia from receiving appropriate treatment for anaemia? All these concerns have slowed their adoption in clinical practice. We aimed to evaluate the technical and clinical accuracy of two POC haemoglobin meters and their agreement with a reference analyser in diagnosing anaemia in the African setting.

A total of 228 participants (104 [45.6%] being men) with 8 (6.45% of women) being pregnant were included. Ages ranged from 4 to 85 years with a mean of 42.4 ± 17.3 years. The general characteristics of the study population are shown on Table 2.

In our study we noticed a positive and significant correlation between each haemoglobin meter and the reference analyser (r = 0.90 for MissionHb® and r = 0.89 for Urit12®). However, both devices did not meet clinical accuracy requirements using total error allowable. Both meters significantly overestimated haemoglobin values and both had a weak agreement with the reference analyser in diagnosing the presence and severity of anaemia among our participants.

The positive correlation implies there was a strong direct linear relation between the measurements of each haemoglobin meter with those of the reference analyser. These results were comparable to that obtained for a bench-top haemoglobin meter-the HemoCue®(HemoCue, Angelholm, Sweden) by Lamhaut et al., in France [17]. HemoCue® is the most popular and widely studied haemoglobin meter. Although other testing systems are available, scientific literature and on-field evaluation of non-HemoCue systems are extremely limited or even lacking for some 22]. Lamhaut et al., included 44 participants undergoing potentially haemorrhagic surgery and they had a correlation of 0.85 between the HemoCue® and a reference analyser. The HemoCue® is technically more sophisticated than the devices and this similarity in results might mean that their simplicity does not compromise their technical accuracy [17]. Our results were lower than that obtained by Nkrumah et al., in their study of the HemoCue® in Ghana [22]. They had a correlation coefficient of 0.995. This difference can be explained by the fact that the HemoCue® unlike the two devices has greater calibration functionality.

We also observed good agreement between the two haemoglobin meters and the reference analyser in the measurement of haemoglobin as shown on the Bland-Altman graphs. These results were better than those obtained by Schapkaitz et al., in their study of the HemoCue® in South Africa [12]. Using a sample of 100 participants of all age groups, they had only 93% of the difference plots within the 95% limits of agreement (− 3.9, 5.5) and their limits of agreement were wider than those observed in our study. This difference can be explained by the fact that they used a smaller sample size than we did.

We also observed that the overestimated mean haemoglobin values, when compared to the reference analyser were statistically significant in both devices. However, only the MissionHb® meter had a mean haemoglobin significantly higher than the clinically allowable range. This implies that we could potentially fail to identify patients with borderline anaemia in some instances and provide adequate treatment with the use of these devices. Nevertheless, it is worth noting that both meters were overall excellent at identifying the presence of anaemia given their significantly high areas under the receiver operating curves. The ability of a haemoglobin meter to detect the presence or not of anaemia is very important in clinical settings, but its ability to determine the severity of the anaemia is even more important as the management of anaemia is entirely depending on its severity. Both meters had weak agreement with the reference analyser in determining the severity of the anaemia among our participants when we adopt a more logical interpretation of Cohen’s kappa statistic [20]. The respective levels of overestimation of haemoglobin by both meters might be an underlying reason for this finding. A weak level of agreement in diagnosing the severity of anaemia between both meters and the reference analyser, implies the severity of anaemia in several patients could be incorrectly assessed when using these meters and consequently lead to undue exposure of some patients to blood transfusions when not absolutely required. As such extreme clinical judgment would be imperative with the use of these devices.

Our study may be limited by a lack of a large representation of patients in the different ranges of anaemia (mild, moderate and severe). However, our study focused on hand-held haemoglobin meters thus adding to the extremely limited or even lacking literature of non-HemoCue systems. In addition, the results of our study may not be readily translated to all populations world-wide where the causes of anaemia may be very different from those of our study population (usually iron deficiency anemia). On the other hand, our study seems to be the first to assess the clinical implications of error in haemoglobin measurement using point-of-care haemoglobin meters.

In conclusion, MissionHb® and Urit12® had good technical accuracy when compared to the reference analyser. However, the MissionHb® meter failed to meet clinical accuracy requirements, and both meters overestimated haemoglobin values. Even though both meters were excellent at detecting the presence of anaemia, they both had weak agreement with the reference analyser in determining the severity of the anaemia. Our findings suggest that although the use of these devices in resource-limited settings might offer enormous advantages compared to a reference analyser, extreme caution and clinical judgment are imperative to complement their use.