Diagnostic sample transport via pneumatic tube systems: data logger and their algorithms are sensitive to transport effects

Introduction

Various pneumatic tube systems (PTS) are used in hospitals around the world to transport diagnostic samples [1]. PTS are capable to reduce turn-around-time and to improve preanalytical workflows to accelerate result-oriented patient treatment [2], [3], [4], [5]. In addition to conventional PTS (cPTS) utilizing cartridges, innovative PTS (iPTS) directly send unpacked samples into a bulk loader of laboratory automation systems.

Validation of each new PTS is mandatory to ensure that the quality of diagnostic samples is not impaired. Hemolysis is one of the most frequent and severe preanalytical error [6]. It affects numerous measurands [1, 7], [8], [9]. Hemolysis can occur due to a disease (in vivo hemolysis) or due to sample collection and sample handling (in vitro hemolysis). In vitro hemolysis does not reflect the patients’ status and, consequently, bears a risk for patients if altered results are released. In contrast, in vivo hemolysis, e.g. due to infectious diseases, hereditary disorders, drugs, valvular transplant or immune-mediated, reflects the patients’ status and needs to be identified, for example by measuring haptoglobin and hemopexin [10, 11]. Especially, as in-vitro hemolysis can be induced by a number of pre-analytical processes, such as repeated fist closing or prolonged tourniquet application during blood withdrawal, it is important to investigate transportation systems separately [7, 12, 13]. A widely used practice is to collect diagnostic samples from healthy volunteers. Samples are split among transport types considered: the gold standard, often courier and the transport type in question.

It is recommended to evaluate PTS after installation and on a regular basis thereafter [13]. To identify alternative ways for validation the use of data loggers has been proposed [13, 14]. During transport, the data logger collects information on acceleration in three axes. Specific algorithms are used to predict the effects of sample transport on the integrity of the diagnostic sample. For algorithm development, original diagnostic samples are inevitable as for example described by Streichert et al. who used a commercial data logger [13]. Furthermore, an improved data logger prototype was developed at the University Medicine Greifswald, including an improved algorithm [15].

In our study, we evaluated two different data loggers and their respective algorithms and compared the outcome with measurement results from diagnostic samples. The investigated transport types encompassed courier, cPTS and iPTS.

Materials and methods

Results

The loggers rotated evenly between courier, cPTS and iPTS resulting in one recording per transport type for each study day. Recorded average transport times were 4:00 min, 1:49 min and 0:30 min for courier, cPTS and iPTS, respectively. Recordings of g-forces in all three axes were displayed in graphs as shown in Supplementary Figure S1 (MSR145^® logger) and Supplementary Figure S3 (prototype logger). The maximum g-force for each logger (16 g for MSR145^® logger and 200 g for prototype logger) was achieved in all runs with iPTS.

During eight study days, two different data loggers were carried once a day using each transport type. The distribution of the measured stress values during each transport type are displayed using boxplots (n=8).

Each logger calculated its own arbitrary unit to indicate the impact of transport. AUC-values in arbitrary units from the MSR145^® logger for 10 Hz ranged from 0 to 6 for courier transport, from 158 to 213 for cPTS and from 311 to 415 for iPTS. For 50 Hz, the AUC for courier transport ranged from 1 to 29, for cPTS from 757 to 928 and for iPTS from 1,455 to 1,745. St-values for the prototype logger ranged from 296 to 403 kmsu for courier transport, from 301 to 586 kmsu for cPTS and from 1,334 to 1,569 kmsu for iPTS. All calculated AUC- and St-values are given in Table 1. The distribution of AUC- and St-values for each transport type are visualized in Figure 1.

Figure 1:

Distribution of established stress values separated by data logger and way of transport.

During eight study days, two different data loggers were carried once a day using each transport type. The measured stress values during each day and each transport type are displayed in this Table, as well as calculated mean and standard deviation.

Compared to courier, cPTS and iPTS showed higher AUC- and St-values, respectively. Among the PTS-transports, iPTS gave higher AUC- and St-values than cPTS. When having raised the recording frequency from 10 to 50 Hz in the MSR145^® logger, the AUC-values increased for cPTS and iPTS, but were more pronounced for iPTS.

Median AUC-values for courier transport were 0 for 10 Hz and 2 for 50 Hz, for cPTS 181 and 879 arbitrary units and for iPTS 322 and 1,531 arbitrary units, respectively. The median St-value of the data logger prototype in iPTS showed higher values than the ones of the other two types of transport. The median St was 375 kmsu for courier transport, 445 kmsu for cPTS and 1,387 kmsu for iPTS.

A mixed model was used to analyze the association between the transport type and AUC- as well as St-values. The results are presented in Table 2. Both algorithms identified, independent of the recording frequency, transport via iPTS as significantly more stressful than courier transport. Based on both frequencies of the MSR145^® logger, the results further showed that the transport via cPTS is significantly more stressful than courier transport. In contrast, no statistically significant difference between these two transport types could be found in the results of the data logger prototype.

During eight study days, two different data loggers were carried once a day using each transport type. Because of repeated measurements, a mixed model was used to analyze the association between transport type (exposure) and the respective stress value (outcome).

A total of 16 (0.4 %) measurements were invalid due to low sample volume or lipemia. Supplementary Table S1 displays the number of valid measurements as well as median (25 %-; 75 %-quartile) of each measurand depending on the transport type. Friedman test showed that the median results between the three transport types were statistically different for only three of the investigated measurands: free hemoglobin (fHb), lactate-dehydrogenase (LDH) and thyrotropin (TSH).

A more detailed view on the distribution of the measurement values for the hemolysis sensitive measurands fHb, LDH, ASAT and potassium separated by transport type is given in Figure 2. For all other measurands, these measurement distributions are visualized in Supplementary Figures S4 and S5. Regarding the hemolysis-sensitive measurands, only for potassium no significant statistical differences in measurement values between transport types were observed. The measurement values for ASAT and LDH differed significantly when comparing both PTS transport types with courier transport, but not between each other. fHb measurements were significantly elevated in both PTS transport types, with higher results in iPTS.

Figure 2:

Distribution of measurement values for the laboratory measurands aspartate aminotransferase, potassium, lactate-dehydrogenase and free hemoglobin separated by transport type.

Distribution of measurement values for the different laboratory measurands was visualized with boxplots. The number of samples measured after each transport type is shown at the left corner of each Figure. Wilcoxon signed-rank test was used for group comparisons. The corresponding p-values are displayed in the Figure.

Furthermore, for each sample and measurand the difference between courier transport and iPTS or cPTS was calculated, respectively (Supplementary Table S2). We observed that the relative difference in measurement values between transport types was less than 5 % for all but one of the 37 measurands. The only exception is the hemolysis sensitive measurand fHb, which was higher in both PTS types compared to courier transport.

Discussion

This study evaluated the ability of two data loggers and their respective algorithms to monitor the quality of diagnostic sample transport.

The results from hematology, coagulation and clinical chemistry were well comparable between all investigated transport types, with the relative differences not exceeding 5 %, except for fHb. The hemolysis sensitive measurands fHb, LDH and ASAT showed statistically significant differences between transport types. From a clinical point of view, these differences can be regarded as negligible. As shown by the median and 25 %-; 75 %-quartile for courier, cPTS and iPTS (Supplementary Table S1), the measurement results were clearly comparable between transport types. For example, the median concentration of ASAT was 0.29 μkat/L for courier transport and 0.30 μkat/L for both PTS types. Likewise, median potassium concentration did not differ by transport type (3.80 mmol/L for all three transport types) and medians for courier, cPTS and iPTS transports for fHb (16.0, 17.5 and 19 μmol/L) and LDH (2.69, 2.74 and 2.81 μkat/L) were very similar from a clinical perspective. One measurand, TSH, which is not known to be particularly affected by hemolysis, also showed a statistically significant difference depending on the transport type, but this can also be regarded as clinically irrelevant as shown by comparing the median concentration of each transport type (courier: 1.61 mU/L, cPTS: 1.56 mU/L, iPTS: 1.55 mU/L).

Overall, the impact of any of the two PTS transport types on measurement results was low. At the same time, both investigated data loggers showed differences between the cPTS and iPTS: AUC and St were considerably higher in iPTS. For the AUC recorded at 10 Hz, the values for both cPTS and iPTS remained below 500 arbitrary units, as suggested by Streichert et al. as a threshold to indicate a beginning impact on diagnostic samples [13]. A previous evaluation of the iPTS used in this study employing patient samples and the MSR145^® logger also confirmed AUC values below the suggested 500 arbitrary units when using the 10 Hz AUC-results [4]. At a frequency of 50 Hz, the AUC in PTS increased considerably, reaching values up to 1,500 arbitrary units in the present study, with iPTS showing higher AUC than cPTS (Figure 1). This demonstrates the impact of the recording frequency on the algorithms.

In comparison, the St was recorded at a considerable higher frequency of 1,600 Hz. Since algorithms and units differed, a comparison could only be descriptive: both algorithms showed differences between courier and PTS transport, which is reasonable since courier transport over short distances is regarded as an established standard with very low acceleration forces.

The investigated data loggers seem to detect the stress on the diagnostic sample earlier than the analytical quality and consequently patient safety is affected. This is highly desirable, since tools in monitoring transport, that are capable to serve as an early warning system to allow enough scope of action for corrective measures, are needed. With an increasing recording frequency from 10 to 50 Hz, the difference between courier and PTS transport became more pronounced when AUC was used (Figure 1, Table 2). To our knowledge, the AUC was used in three studies with 10 Hz [4, 13, 16]. Further three studies used the AUC concept with different frequency or without reporting the recording frequency [14, 18, 19]. One study reported a possible impact of the reported frequency [19]. Considering our findings on the impact of recording frequencies, further investigations should be conducted to evaluate at which recording frequency a steady state is reached and results are reliable.

Conclusions

Both, AUC and St, showed differences between the cPTS and iPTS, with St emphasizing these differences more clearly. The statistical, but not clinical, differences in the results for fHb, LDH and ASAT may be regarded as an early sign of preanalytical impairment of diagnostic blood samples. Results from data loggers – calculated by different algorithms – record this important interval of beginning mechanical stress with a high resolution, demonstrating that they are sensitive to transport effects. Thus, the investigated data loggers and their respective algorithms AUC and St have the potential to identify mechanical stress on a diagnostic sample before analytical and consequently clinical impairment occurs. They may be used for continuous monitoring of PTS transport and early detection of preanalytical impairment of blood samples. Further studies should investigate the continuous increase of stress on diagnostic samples using both types of loggers and algorithms until the analytical, and thus clinical, impact can be measured. These kinds of studies could help to identify reliable thresholds for AUC and St.

Currently, evaluation and monitoring of diagnostic sample transport should not only rely on data loggers, but should include both, diagnostic samples and data loggers, with appropriate algorithms.