Parameters for Validating a Hospital Pneumatic Tube System

Address correspondence to this author at: Washington University School of Medicine, 425 S. Euclid Ave., Campus Box 8118, St. Louis, MO 63110. Fax 314-362-1461; e-mail gronowski@wustl.edu.

An abstract of this study has been submitted to the AACC for presentation at the 2019 annual meeting.

Clinical Chemistry, Volume 65, Issue 5, 1 May 2019, Pages 694–702, https://doi.org/10.1373/clinchem.2018.301408

Cite

Christopher W Farnsworth, Daniel M Webber, James A Krekeler, Melissa M Budelier, Nancy L Bartlett, Ann M Gronowski, Parameters for Validating a Hospital Pneumatic Tube System, Clinical Chemistry, Volume 65, Issue 5, 1 May 2019, Pages 694–702, https://doi.org/10.1373/clinchem.2018.301408

Abstract

Pneumatic tube systems (PTSs) provide rapid transport of patient blood samples, but physical stress of PTS transport can damage blood cells and alter test results. Despite this knowledge, there is limited information on how to validate a hospital PTS.

We compared 2 accelerometers and evaluated multiple PTS routes. Variabilities in PTS forces over the same routes were assessed. Response curves that demonstrate the relationship between the number and magnitude of accelerations on plasma lactate dehydrogenase (LD), hemolysis index, and potassium in PTS-transported blood from volunteers were generated. Extrapolations from these relationships were used to predict PTS routes that may be prone to false laboratory results. Historical data and prospective patient studies were compared with predicted effects.

The maximum recorded g-force was 10g for the smartphone and 22g for the data logger. There was considerable day-to-day variation in the magnitude of accelerations (CV, 4%–39%) within a single route. The linear relationship between LD and accelerations within the PTS revealed 2 PTS routes predicted to increase LD by ≥20%. The predicted increase in LD was similar to that observed in patient results when using that PTS route.

Hospital PTSs can be validated by documenting the relationship between the concentrations of analytes in plasma, such as LD, with PTS forces recorded by 3-axis accelerometers. Implementation of this method for PTS validation is relatively inexpensive, simple, and robust.

Hospital pneumatic tube systems (PTSs) 3 provide rapid and efficient transport of samples ( 1). However, the physical stress induced by a PTS on blood can induce cellular lysis ( 2– 4). As a result, multiple laboratory results are susceptible to error. Several studies have demonstrated that lactate dehydrogenase (LD), potassium (K+), and hemolysis index (HI) are prone to increases owing to PTS transport ( 5– 7), with LD being the most sensitive to acceleration ( 3, 4, 6, 8, 9). Despite the known impact of PTS transport on cellular lysis, there is no consensus on how to best evaluate a PTS for its effects on clinical test results ( 10).

Previous studies have evaluated PTSs by correlating number and magnitude of accelerations, measured by various 3-axis accelerometers with changes in laboratory results ( 2, 4, 11). False increases in LD, K+, and HI have been associated with prolonged time in the PTS, excessive number of accelerations with g-forces >3g ( 4), and excessive cumulative accelerations measured by area under the curve (AUC) of the acceleration distribution ( 2). Although these studies have been successful in demonstrating the impact of acceleration changes within the PTS on patient samples, many questions remain with regard to how to best “validate” a PTS to ensure accurate reporting of clinical tests. For example, yet undefined variables include the type of device that should be used to measure the number and magnitude of shock forces, parameters that best correlate with clinically significant changes in test results, day-to-day variability in PTS forces, and appropriate individuals to use for validation studies.

A recent review of this subject called for all laboratories to perform a validation of their PTS ( 10). Given the lack of guidance on how to best undertake a validation, we set out to define the parameters for evaluating a hospital PTS.

Materials and Methods

MEDICAL CENTER AND PTS

The Barnes-Jewish Hospital/Washington University (BJH/WU) PTS is primarily a Swisslog system, although a few areas include Pevco equipment. The PTS transports at approximately 20 feet/s and has 36 zones (express and pressure/vacuum), about 130 stations, 150 transfer units, 45 traffic control units, 45 nonvariable frequency drive blowers, and 2 multilinear transfer units. All samples and data loggers were transported in Swisslog NexSeal carriers lined with foam padding.

DATA LOGGERS

The number and magnitude of accelerations within the PTS were measured using a PCE-VD3 3-axis acceleration monitor (PCE Instruments) or an iPhone 6s (Apple) utilizing the Google Science Journal app. The published acceleration range for the PCE-VD3 was ±16g with an accuracy of ±0.5g. To our knowledge, there are no published data for the specifications of the accelerometers used on an iPhone 6s. Previous studies have claimed an acceleration range of ±8g on an iPhone 5 ( 4). For comparison studies, the iPhone and data logger were transported wrapped in bubble wrap in the same carrier.

PTS ROUTES

Two types of PTS routes were selected for this study: (a) routes from areas of potential clinical consequence (including the outpatient cancer center and emergency department) and (b) routes that were representative for a building. Most of the stations send carriers to 1 of several transfer units within a building before redirecting them to the laboratory via express routes. Therefore, the forces from 1 tube station within a building generally reflect the whole building. Intraday and interday PTS variability was determined for 2 routes by capturing data logger parameters 5 times each on 1 day and over 5 separate days.

HUMAN PARTICIPANTS AND RETROSPECTIVE DATA

Validation studies.

Five tubes of blood were collected from 15 healthy individuals by a single experienced phlebotomist into lithium heparin tubes with separator gel (Becton Dickinson). One sample from each individual was walked to the laboratory. The other 4 samples were transported via PTS with the data logger. One tube from each individual was then removed, and the remaining samples were walked back to the same PTS station. The remaining samples were then transported through the PTS. This was repeated a total of 4 times through the PTS for each individual. Tubes were centrifuged simultaneously, after all samples arrived in the laboratory.

Patient studies.

Fourteen patients with confirmed chronic leukemia or lymphoma were recruited. Paired samples were collected in lithium heparin tubes with separator gel. Samples were collected by antecubital draw or from indwelling central venous catheters by multiple nurses over 2 days. One sample was walked to the core laboratory. A separate sample was transported via PTS.

Informed consent was obtained from all volunteers. All studies were approved by the Washington University Institutional Review Board.

Retrospective studies.

Historical data were collected by querying the laboratory information system (Cerner).

MEASUREMENT OF LD, HI, AND K+

Samples were centrifuged at 5000 rpm (2500g) for 6 min. LD, K+, and HI were all measured using a Roche Cobas 501 analyzer. HI was determined with spectrophotometric readings at 600/570 nm. All samples from a single patient were performed on the same analyzer. All testing was performed according to the manufacturer's instructions.

CORRELATION OF PTS PARAMETERS WITH LABORATORY RESULTS

The percent change in LD and absolute change in HI and K+ were calculated for each healthy individual by taking the difference from the sample sent through the PTS and the sample walked to the laboratory. Response curves were generated by plotting the percent change in LD against the cumulative number of forces >3g, 5g, 10g, 15g, AUC, and elapsed time in the tube. Linear regression was performed using a least-squares ordinary fit with the linear regression line being constrained to pass through the origin. This was repeated for absolute change in HI and K+. Estimated acceptable forces from the PTS were calculated by extrapolating 20% for LD (defined to be a clinically significant change) and an absolute change of 60 in HI (the threshold used to trigger our laboratory comments on hemolysis) from the generated curves. Predicted change in LD, HI, and K+ was estimated by using the mean parameters for each route and extrapolating from the generated curves using the slope of the best-fit line.

DATA AND STATISTICAL ANALYSIS

Using the iPhone, resulting accelerations in the x, y, and z direction were recorded and exported as a CSV file. Microsoft Excel was used to calculate total g-forces experienced by the device during transit. The g-forces recorded by the PCE-VD3 device were extracted using Acceleration data logger software version 2.7 (Silicon Labs) and exported as a CSV file to Excel. All statistical calculations, AUCs, and number of force calculations were performed using GraphPad Prism version 7. To determine AUC of the acceleration distribution, a baseline of 1.15g was set. The numbers of peaks were quantified by setting the minimum peak height to 3, 5, 10, or 15g. Comparisons between 2 groups were evaluated for statistical significance by paired t-test or unpaired t-test.

Results

SELECTION OF A DATA LOGGER DEVICE

To determine the type of device best suited for PTS evaluation, an iPhone 6s and a PCE-VD3 data logger were compared. Elapsed time, number of accelerations, and magnitude of accelerations were captured by both devices simultaneously transported through multiple PTS routes a total of 11 times. Fig. 1A represents an overlay of accelerations measured by iPhone and PCE-VD3 for a single representative trip through the PTS or walked. Although the iPhone captured more subtle accelerations between 1.15 and 3g (data not shown), larger magnitude accelerations were detected using the PCE-VD3 device. The maximum g-force recorded was 10g for the iPhone and 22g for the PCE-VD3. As a result, the AUC was significantly greater for the PCE-VD3 device ( Fig. 1B). The 2 devices recorded a similar number of accelerations with g-forces >3 ( Fig. 1C). However, the PCE-VD3 was able to detect more accelerations with g-forces >5, as well as significantly greater numbers of accelerations with g-forces >10 and >15 ( Fig. 1, D–F). Given the limitations of the iPhone at higher g-forces, the PCE-VD3 device was used for all further experiments.

Comparison of iPhone and PCE-VD3 data logger for detecting accelerations within the PTS.

Comparison of the number and magnitude of accelerations when transported by courier or PTS as measured by iPhone or PCE-VD3 data logger. Devices were wrapped in bubble wrap and placed in a padded Swisslog carrier. The carrier was sent via multiple routes in the PTS 11 times or walked by courier. Accelerations from a single representative route are shown in (A). AUC, as calculated by the area >1.15g for each of the 11 transports as measured by iPhone or PCE-VD3 data logger (B). Number of accelerations with g-forces >3 (C), g-forces >5 (D), g-forces >10 (E), and g-forces >15 (F) for each of the 11 observations as measured by iPhone or PCE-VD3 data logger. Paired t-test, ***P < 0.001.

MAPPING MULTIPLE PTS ROUTES

To determine areas in our hospital that may be more susceptible to PTS-related hemolysis, we used the PCE-VD3 device to examine various PTS routes. Fig. 2 illustrates 8 different routes. Mean elapsed time varied from 100 s to 350 s for the 8 routes ( Fig. 2A). The routes that experienced accelerations with the greatest g-forces were route 5 (a cardiac floor) (mean number of accelerations >15g = 44; 95% CI, 35–52) and route 8 (an outpatient cancer center) (mean number of accelerations >15g = 29; 95% CI, 22–35) ( Fig. 2, B–F).

Characterization of 8 different PTS routes within the BJH/WU medical center.

Elapsed time for each route (A). AUC for each route (B). Number of accelerations with g-forces >3 (C). Number of accelerations with g-forces >5 (D). Number of accelerations with g-forces >10 (E). Number of accelerations with g-forces >15 (F). Bars represent mean ± SE.

INTRADAY AND INTERDAY VARIABILITY IN PTS PARAMETERS

Intraday and interday variability in measured PTS parameters from 2 different routes is shown in Table 1. Interday CV of cumulative accelerations, AUC, and elapsed time ranged from 4% to 39%. Route 5 exhibited considerable day-to-day variation in all parameters measured. The number of g-forces >15 appeared to be the most variable parameter for both routes 5 and 8. Interestingly, there was an inverse relationship between the number of blood tubes transported in a carrier and the magnitude of g-forces (see Fig. 1 in the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vol65/issue5). In other words, the fewer tubes in the carrier, the greater the forces recorded within the carrier. This effect appears to be because of reduced mobility within the carrier, as wrapping the data logger with bubble wrap also reduced the magnitude of the g-forces experienced by the data logger relative to when shipped without protective wrapping (see Fig. 2 in the online Data Supplement). In contrast, increasing the weight of the carrier by 20% or 50% had limited effects.

Intraday and interday variability in 2 routes of a hospital PTS. a

Intraday n = 5; interday n = 5 over 5 separate days.

Intraday and interday variability in 2 routes of a hospital PTS. a

Intraday n = 5; interday n = 5 over 5 separate days.

PTS PARAMETERS ASSOCIATED WITH CLINICALLY SIGNIFICANT CHANGE IN LABORATORY RESULTS

To determine how much agitation causes a clinically relevant change to LD, we collected blood from 8 healthy male individuals and transported the samples through the PTS up to 4 times on 2 days from 2 separate locations. Initial studies showed that samples from men had greater susceptibility to PTS-induced increases in LD and HI than those from women (see Fig. 3 in the online Data Supplement) despite similar median LD concentrations in couriered samples [male, 180 U/L and interquartile range (IQR), 157–189; female, 175 U/L and IQR, 167–198]. To reduce variability and simulate a worst case scenario, we limited this aspect of the study to healthy male participants. Response curves were generated based on the various parameters collected from the PCE-VD3 ( Fig. 3). Using a 20% change in LD as clinically significant, cutoffs were established to define parameters that should not be exceeded to avoid falsely increased LD ( Fig. 3, arrows). The same method was used to plot absolute change in HI and absolute change in K+ (see Figs. 4 and 5 in the online Data Supplement).

Correlation of various PTS parameters with percent change in LD.

Percent change in LD <[(LD_PTS − LD_walked)/LD_walked]*100> in samples from 8 healthy male volunteers. Elapsed time (A). AUC (B). Number of accelerations with g-forces >3 (C), >5 (D), >10 (E), and >15 (F). Each point is the mean of 3 or 4 patient samples. Diagonal line represents linear regression. Bars represent ±SE. Arrow represents clinically significant change in LD defined as 20%. Gray dashed lines show 20% LD change extrapolation.

PREDICTING THE EFFECT OF VARIOUS PTS ROUTES ON LABORATORY RESULTS

The predicted effect of the PTS on LD, HI, and K+ for each route was extrapolated from the data generated in Figs. 2 and 3 (or see Figs. 4 and 5, respectively, in the online Data Supplement) and is shown in Table 2 (see also Tables 1 and 2 in the online Data Supplement). The PTS routes with the greatest predicted impact on LD were route 5 (average, 27.2%) and route 8 (average, 23.4%). PTS route 5 also had the largest predicted impact on HI (average, 31) and K+ (average, 0.11 mmol/L) (see Table 2 in the online Data Supplement).

Mean (SD) predicted percent change in LD for various routes within the medical center. a

Shown for each parameter is the mean (SD) percent change in LD after PTS transport predicted by that parameter. Values were derived by taking the mean parameter for each route (elapsed time, AUC, g-forces >3, g-forces >5, g-forces >10, or g-forces >15) in Fig. 2 and extrapolating percent change in LD from the response curves in Fig. 3.

Mean (SD) predicted percent change in LD for various routes within the medical center. a

Shown for each parameter is the mean (SD) percent change in LD after PTS transport predicted by that parameter. Values were derived by taking the mean parameter for each route (elapsed time, AUC, g-forces >3, g-forces >5, g-forces >10, or g-forces >15) in Fig. 2 and extrapolating percent change in LD from the response curves in Fig. 3.

VALIDATION OF ESTIMATED PTS EFFECT WITH PATIENT DATA

During our investigations, an astute physician working in our outpatient leukemia and lymphoma clinic in the cancer center perceived an increase in the number of samples with falsely increased LD (i.e., results were not increased when repeated at an outside laboratory). Importantly, this perceived increase occurred after closing a satellite laboratory that was located within the same building as the outpatient cancer center. The previous route from the cancer center to the satellite laboratory was route 7, and the new route from the cancer center to the core laboratory was route 8 ( Fig. 2).

To address the perceived increase in LD, historical data were extracted from the laboratory information system for a 3-month period after the satellite laboratory closure and the same 3-month period 1 year prior. The data revealed a 14% increase in median LD when samples were transported via route 8 in the PTS compared with route 7 ( Fig. 4A). Interestingly, the predicted increase in LD in Table 2, from route 7 to route 8, was estimated to be 14% (23.4% − 9.4% = 14%). The increase in median LD was associated with an increase in the median HI of 200% (absolute change = 8) during this period (see Fig. 6A in the online Data Supplement). Because of the potential impact on patient care, a courier was implemented to walk samples from the cancer center to the core laboratory. LD results from the 2-month period before courier were compared with 2 months after courier. The difference in median LD between walking and PTS was 7.4% ( Fig. 4B). Samples transported by PTS were associated with a greater percentage of samples with LD concentrations >200 U/L relative to those walked by courier ( Fig. 4C). Samples transported by PTS had an increase in HI by 85.7% (absolute change = 6) relative to those transported by courier (see Fig. 6B in the online Data Supplement).

The effect of PTS route change on LD results from a cancer center.

LD results (median and IQRs) from the cancer center for the 3-month period before (route 7, April 2017–June 2017) and 3 months after (route 8, April 2018–June 2018) closing a satellite laboratory (A). LD results for 2 months using route 8 (July 2018–August 2018) compared with 2 months of courier (September 2018–October 2018) (B). Relative frequency of LD in 50 U/L bins for 2 months before and after implementation of courier (C). Paired samples from 14 patients with lymphoma or leukemia. One tube was walked by courier and the other was sent via PTS (route 8). Percent difference in LD relative to courier was calculated (D). LD testing in the core laboratory was performed on a Roche Cobas 501 chemistry analyzer and in the satellite laboratory on a Roche Cobas Modular P chemistry analyzer. A lack of bias between these methods (Roche 501 vs Modular P) was confirmed as part of weekly quality control (data not shown).

To confirm the effect of route 8 on LD, paired samples were collected from patients with lymphoma or leukemia. One sample was walked to the core laboratory and 1 sample was sent via PTS (route 8). The PTS station in the old satellite laboratory had been closed, so samples could not be shipped to that location for comparison. There was an 8% increase in LD concentrations observed in samples transported by route 8 relative to those walked by courier ( Fig. 4D). Route 8 was also associated with a 150% increase in HI (absolute increase of 12) (see Fig. 6C in the online Data Supplement).

Discussion

The PTS is the largest piece of equipment that a hospital owns, yet repairs, maintenance, speed, and other factors involving the PTS often fall outside of the laboratory's purview, and there are no requirements or guidelines for how to evaluate a hospital PTS ( 10). Nonetheless, recent articles have called for laboratorians to “validate” their PTS ( 10). In this study, we describe tools for assessing the clinical utility of a hospital PTS.

Smartphones and data loggers have both been used for PTS evaluation. In multiple studies, large acceleration changes have been associated with hemolysis ( 2, 4, 11). To our knowledge, this is the first study to compare 2 different types of accelerometers for assessing the accelerations within a PTS. Although comparable in their ability to measure the number of accelerations and the number of accelerations with g-forces >3, the iPhone did not record g-forces >10, in contrast to the data logger, which consistently logged measurements >15g. The iPhone also had a lower sampling frequency relative to the data logger (0.067 s vs 0.05 s for the data logger). Importantly, there are no specifications released with iPhone accelerometers or from downloadable applications (i.e., Google Science Journal) that address accuracy or precision. Commercial data loggers are relatively inexpensive (about $150) and are released with detailed specifications on accuracy and precision. Furthermore, the device used here is similar in size, shape, and weight to a tube of blood, closely replicating a single tube sent through a PTS carrier. For these reasons, we recommend that validations of PTSs are performed with 3-axis accelerometers capable of measuring the maximum forces generated by a PTS.

Previous publications have not addressed the importance of testing multiple PTS routes within a hospital. In our hospital, there are approximately 130 PT stations, approximately 150 transfer units, 45 traffic control units, 45 nonvariable frequency drive blowers, and 2 multilinear transfer units, all of which can affect the number and magnitude of accelerations experienced by carriers. We approximate that if performed in duplicate, it would take >80 h to test every route. Furthermore, our data demonstrate significant intraday and interday PTS variability, indicating that performing 2 replicates is insufficient to capture the total variability associated with PTS transport. Therefore, our approach was to select routes with the following characteristics: (a) from sites, like a cancer center, that closely monitor LD and sites such as the emergency department for which hemolyzed samples can cause significant delay in patient treatment and diagnosis; and (b) sites from different buildings that send PTS carriers via different transfer units. By understanding the PTS structure, we were able to select locations within each building that best reflected the entire building. Therefore, we recommend that laboratorians become familiar with their PTS and that they collaborate with appropriate personnel to learn how their PTS functions, the routes that carriers may take within their PTS, and the potential areas within the PTS that may cause preanalytical variability.

Previous studies have demonstrated the usefulness of determining the relationship of changes in LD and the number of accelerations that exceed a certain g-force ( 2, 4, 11). Using blood from healthy volunteers, we built on these previous studies. Our approach allowed prediction of the number of shock forces >3g, >5g, >10g, >15g, AUC, and elapsed time in the PTS associated with a clinically significant increase in LD. The usefulness of this approach is reflected by the calculated increase in LD after closing a satellite laboratory (changing from route 7 to route 8, 14% increase in median LD) matching our estimated difference from route 7 to route 8 (14.0%). The importance of mapping different routes to estimate the impact of the PTS is 2-fold. First, it allows laboratories to quickly address which routes currently exist within a hospital that may cause clinically unacceptable changes to specific analytes. Second, this approach allows laboratories to assess new PTS stations and routes before transporting patient samples. The importance of this cannot be overstated. Unfortunately, we started validating our PTS after closing a satellite laboratory that was servicing our outpatient cancer center (route 7). This change rerouted these samples to our core laboratory via a route associated with longer transport time, greater AUC, more total accelerations, and more accelerations with g-forces >3, >5, >10, and >15 (route 8). The adverse effect on LD concentrations was clear in historical data, after implementing a courier, and in paired patient samples. In hindsight, if we had understood the impact of altering the PTS route, we may have deemed route 8 inappropriate for shipping samples from our cancer center, and might have kept the original laboratory open. Our data also demonstrate that PTS-related hemolysis is unlikely to negatively affect laboratory results from other units or buildings within our hospital. For example, samples transported via PTS from the emergency department (route 3) had an 11.3% (SD, 1.8%) predicted increase in LD and a predicted increase in K+ of 0.04 mmol/L (SD, 0.01 mmol/L). This was supported by retrospective data comparing 2 months before and after moving our core laboratory (data not shown). Route 5 had the largest predicted increase in LD (31% ± 7.2%); however, this is a cardiac unit that seldom orders LD. Furthermore, K+ was predicted to increase only by 0.11 mmol/L (SD, 0.02 mmol/L) after transport through route 5.

The parameter recorded by 3-axis accelerometers that most affects laboratory samples is unknown. It is likely a combination of number of accelerations, magnitude of the force, and duration of transport. Therefore, we used the average of each collected parameter to estimate the percent change from each route. Future studies with the capability of introducing controlled forces to samples are needed to help elucidate the exact parameters that most affect cell lysis. As an understanding of the PTS factors most affecting cell lysis emerges, this approach to the evaluation of PTSs can be further refined. We observed a striking difference between men and women in changes in LD and hemolysis after PTS transport. Although this study had a relatively small sample size, the trend was substantial enough to bifurcate the data. These findings are likely multifactorial and may reflect higher hematocrit in men relative to women. However, previous studies have also demonstrated increased hemolysis of stored red blood cells isolated from men compared with women ( 12) and significant interindividual variability of hemolysis when transporting through a PTS ( 7). Therefore, we suggest that male individuals are used when validating PTS for the impact on hemolysis.

There are several weaknesses associated with this study. In the process of limiting the number of routes tested, we may have missed PTS routes that may pose a risk to sample integrity. Also, there is concern that anemic patients, patients in the emergency department and intensive care unit, and patients with high cell counts are more susceptible to hemolysis ( 10). Therefore, what may seem like a “safe” route in this study may not be safe for all samples. Nonetheless, further studies addressing at-risk patient populations might be useful. Furthermore, we did not assess the impact of centrifugation speed and associated forces with increased LD—a topic with conflicting data in the literature ( 13, 14). Finally, our study was limited to LD, HI, and K+, analytes known to be sensitive indicators of hemolysis. It is possible that other analytes are adversely affected by PTS forces. Further studies must be performed to validate other laboratory tests.

In conclusion, we describe a method for validating a hospital PTS using a 3-axis accelerometer and blood samples from healthy men to estimate increases in LD, HI, and K+ as a result of PTS transport. Implementation of this method for validation is relatively inexpensive, simple, and robust. Importantly, validation of the PTS allows for modifications that limit preanalytic variability and will ensure accurate reporting of patient results.