Clinical, Operative, and Economic Outcomes of the Point-of-Care Blood Gases in the Nephrology Department of a Third-Level Hospital

An observational ambispective study was conducted including patients with venous BGA sent to the laboratory from the nephrology department of our hospital during the last trimester of 2016 (LAB) and performed in their POCT BGA analyzer during the last trimester of 2017 (POCT). Data from patients older than 18 years with at least 1 venous BGA sample obtained in daily clinical practice during 2016 or 2017 were collected from the laboratory information system. The details of the recruitment process are shown in Figure 1. The hospital's Research Ethics Committee approved the study, and informed consent was not required.

BGAs were performed using ABL90 Flex Plus blood gas analyzers (Radiometer) installed in either the laboratory (LAB) or the nephrology unit (POCT). Renal function tests were performed using a Dimension Vista 1500 system (Siemens Healthineers, Erlangen, Germany). Conversion factors from conventional units to International System units are listed in Table 1. Patients' data were collected from our laboratory information system and from medical records. These data included sociodemographic features, renal function tests (creatinine, glomerular filtration rate), date, time, and clinical setting (LAB or POCT) of the test request. Critical values established in our laboratory for pH, sodium, potassium, chloride, ionized calcium, glucose, lactate, and hemoglobin were evaluated.

Differences between the POCT and LAB turnaround time (TAT) were calculated by means of a subset of 20 patients within the POCT period. Requests included both BGAs performed as POCT and laboratory tests. Therefore, the difference between the BGA measurement and the reception time of the other laboratory tests was calculated. Records of preanalytical errors (nonhomogen or insufficient material) were collected from the POCT data management system, AQURE (Radiometer).

Variables to calculate the global cost per process were obtained from studies made previously in other clinical settings,¹²  in collaboration with warehouse registries of direct costs of consumables and the Official State Gazette (Boletín Oficial del Estado) for cost per hour of health professionals' work (laboratory specialist, nephrologist, nurse, laboratory technician, and nursing assistant). The time taken for each step performed by experienced health professionals was estimated by observation.

First, the homogeneity between patient populations during the LAB and POCT periods was assessed to detect any baseline differences in sex, age, and renal function. Linear models in accordance with data distribution were applied. Quantitative data followed a normal distribution and were described by average and standard deviation. Qualitative data were described by absolute frequencies and percentages.

A linear mixed model with fixed effect for BGA location and a random effect for participant was used to analyze the BGA measurands over different episodes of care (described as a 24-hour period) within-between patients. The estimation method used restricted maximum likelihood. Unstructured covariance among repeated measures was applied based on model diagnostics. Least squares means from each BGA measurand by location were obtained and compared. All statistical tests were considered bilateral, and P < .05 was considered significant; 95% CIs were estimated. All statistical analyses were performed using SAS 9.3 software (SAS Institute Inc., Cary, NC, USA). Mean differences obtained for each measurand were also evaluated, taking into account the analytical performance specifications established by the laboratory (Table 2). Mean differences in TAT between periods were evaluated by a t-test of paired samples. Differences in total number of BGA critical results and number of preanalytical errors in the LAB and POCT periods were compared by a X²  test and a phi coefficient for magnitude of effect (φ). Finally, total costs per process were estimated and represented in a bar graph.

The homogeneity test showed no significant differences in age (P = .31), creatinine (P = .97), or glomerular filtration rate (P = .81) (Figure 2, A–D). However, significant differences in sex distribution were detected, given men were more prevalent during the POCT period (P < .001). There was a statistically significant increase during the POCT period in oxyhemoglobin and glucose, and a decrease in partial pressure of carbon dioxide (pCO2), carboxyhemoglobin, lactate, and sodium. Furthermore, glucose, pCO2, lactate, and sodium differences exceeded the analytical performance specifications set in our laboratory (see Table 2). Data displayed in Table 3 show the BGA results in both periods and the differences found in the study.

A total amount of 61 of 1564 (3.9%) critical values during LAB and 43 of 1564 (2.6%) during POCT were detected (Table 4). This difference was statistically significant (P = .04), although the magnitude of effect was small (φ= −0.08). The critical value specifications of our laboratory and the number of BGAs that exceeded those limits are shown in Table 5. The higher number of critical results during the LAB period was especially remarkable for glucose (<45 mg/dL) and lactate (>45 mg/dL). There were 3 samples with critical values for both lactate and glucose, all collected during the LAB period. Although the number of critical results for potassium remained the same in both periods, maximum elapsed time between the critical result and the analytical evidence of its correction (read as a noncritical result in a subsequent sample) decreased from 1 week in LAB to 24 hours in POCT.

Regarding the difference in TAT between LAB and POCT, the observed mean was 89.5 minutes (95% CI 65.5–93.75), with a range between 25 and 120 minutes (Table 6). Elapsed time between requests was evaluated in 106 patients, who had 2 or more BGAs performed in a 24-hour period. The mean elapsed time was 511 minutes (8.5 hours) for LAB and 286 minutes (4.8 hours) for POCT. The difference found was statistically significant (P < .001).

Of 26 409 BGAs measured during the LAB period, 680 (2.61%) did not show a result due to a preanalytical error. During the POCT period, 2407 BGAs were measured in the nephrology department and 101 (4.2%) did not show a result. This increase in preanalytical errors was statistically significant (P < .001), although the magnitude of effect was small (φ = −0.027). Another comparison was made 1 year later, in June 2018. It showed 103 preanalytical errors of 3708 (2.78%) in the laboratory blood gas analyzers and 13 of 569 (2.28%) in the nephrology department, with no statistically significant difference (P = .58) (Table 7).

Total cost per process was estimated at 6.58 for each BGA taken during the LAB period and 4.76 during POCT, with a reduction of 1.82 (27.72%) per BGA, as shown in Figure 3. Although there were 91 (5.86%) additional BGA requests (1643 BGAs during POCT versus 1552 during LAB), the total costs decreased from 10 814 in LAB to 7815 in POCT, meaning a reduction of 2999 (23.49%). The time spent by staff members for every step of the BGA process is shown in Figure 4.

LAB and POCT populations had a homogeneous distribution regarding age, creatinine, and glomerular filtration, essential data for establishing basal renal function.^10,13,17  Taking this into account, we considered that differences found during the study were not due to basal variation among LAB and POCT patients. There was only heterogeneity in sex distribution, which might affect some of the differences obtained. Some population studies have shown that partial pressure of oxygen is higher in men,¹⁸  whereas others did not find this difference but detected a higher pCO2 and lower pH, which could be explained by women's faster breathing rate.¹⁹  This had to be considered during the evaluation of differences in BGA results of these measurands.

During the evaluation of differences in BGA results, attention was especially paid to those that, in addition to being statistically significant, exceeded the analytical performance specifications established by our laboratory (see Table 2) given those changes could impact patient care in a greater way. These measurands were glucose, pCO2, lactate, and sodium. The decrease detected in mean pCO2 was predictable because of the smaller time frame between sample collection and measurement, lowering both penetrance of environmental gas through the BGA syringe and cellular consumption.⁶  The higher number of male patients in the POCT period could contribute to this result. However, given that the clinical implications of a variation of blood gases in venous blood are complex due to intra- and interindividual variation, these results should be carefully taken into account.

Analytical performance specifications for sodium concentration were based on biological variation minimum systematic error, a strict but reasonable criterion due to the high clinical impact of a minimal variation.²⁰  These specifications could explain the small differences found in our study that were considered clinically significant, although they could also be caused by the current level of technological development.^19,21  However, our hospital has implemented a monthly monitoring program for inaccuracy in blood gas analyzers, and none of the nephrology or laboratory analyzers surpassed the established limits; thus, it can be assumed the differences found were because of extra-analytical factors. The increase in mean concentrations of oxyhemoglobin and glucose, in contrast to the decrease of carboxyhemoglobin and lactate, can probably be explained by lower oxygen consumption by blood cells.⁶  These changes could contribute to the higher pH found during the POCT period.

Hemoglobin concentration can be a useful measurement to assess the sample management skills of professionals, given it would be altered by poor sample homogenization.^21,22  In our case, although our analytical performance specifications for this parameter were high, there were no statistically significant differences found, which indicates similar sample management and processing in the nephrology department compared with the laboratory, and that the differences found were related to the patients' condition or the time until processing rather than to operator-dependent factors. At a clinical level, hemoglobin concentration is important to detecting hidden blood loss in postoperative patients, and for the monitoring of anemia associated to chronic kidney disease.^5,23 

The number of critical results decreased after the incorporation of a POCT blood gas analyzer in the nephrology department. The greater number of critical results during the LAB period for glucose and lactate probably meant that a longer TAT meant more spurious results due to sample degradation during transport to the laboratory.²⁴  Subsequently, the POCT approach was more accurate and had less false-positive critical results. Having faster and more accurate results was especially relevant for glucose, given critical hypoglycemia could be managed by nurses immediately after being detected rather than having to wait for the physician to complete the laboratory request. Faster detection of critical potassium levels is relevant to establishing adequate treatment for its correction, thus contributing to an operative improvement.

The differences in TAT were similar to those calculated in previous studies performed in other environments¹² ; thus, although there was some dispersion these results were considered acceptable. Hospitalized nephrology patients could benefit from a rapid response in the presence of hyponatremia or critical potassium levels, which are important prognostic markers.²⁰  Metabolic acidosis monitoring is also crucial in order to decrease mortality rates in these patients.⁹  Further studies with larger samples and more thorough analyses of the implications of the TAT in the department daily practice and its integration with medical interventions or nursing would be necessary to achieve more accurate results in nephrology.

Regarding the elapsed time, easier access to BGA resulted in greater use and a shorter time frame between BGAs for each patient. Given patient populations are considered homogeneous, this variation cannot be explained by more severe pathologies; therefore, patients had closer monitoring during the POCT period. This closer monitoring is especially advisable during dialysis, given that oxygenation and electrolytic imbalances can occur abruptly and an immediate response considerably improves short- and long-term prognosis.^25,26  Electrolyte and pH levels are sufficient to prescribe an urgent hemodialysis and to adjust fluids in an emergency situation.^27,28  BGA can also be used to evaluate arteriovenous fistulas²⁹  and to monitor ionized calcium levels when citrate-dependent anticoagulants are applied to membranes and calcium needs to be restored after dialysis.²⁸  Postoperative transplanted patients also need close follow-up, which can be adequately performed by POCT blood gas analyzers.

However, we did not evaluate such operative variables as timeframe until the establishment of an etiologic treatment, the length of stay of the patient in the hospital, or number of hospital-acquired infections; thus, all these operative benefits could only be inferred. Studies in other environments showed that rapid results can have an impact in the establishment of etiologic treatments and better clinical outcomes for patients.^2–4  More research would be needed to statistically assess these benefits in nephrology.

Because of the short time between the POCT analyzer installation and the beginning of the current study, we observed a number of samples with preanalytical errors during POCT within the acceptable range established by the laboratory that were higher than in the laboratory. Although all professionals of the nephrology department who could be involved in the POCT BGA measurements received initial training before the implementation of the BGA analyzer, there was still a learning curve that could explain that preanalytical errors were higher during the POCT period.^21,22  However, a close follow-up of this indicator as a part of the laboratory's monthly monitoring program during the next year showed that preanalytical errors had been reduced and were within acceptable ranges over time. Monitoring during the same 1-year period after analyzer installation showed that the learning curve resulted in the same amount of errors as those detected in the laboratory.

Global costs per process were remarkably lower in the POCT period, a finding that is in line with previous studies in other environments.^7,12  The fact that total costs were lower in the POCT period even when there were more BGAs performed, makes us conclude that POCT BGA allows a closer supervision of patients with no additional costs. In summary, based on this study, the implementation of a point-of-care BGA in a nephrology department has a positive impact on patient care in clinical, operative, and economic terms.