ABSTRACT AJCP /ORIGINAL ARTICLE

Transcription

1 AJCP /ORIGINAL ARTICLE Neonatal Intensive Care Unit Quality Initiative Identifying Preanalytical Variables Contributing to Specimen Hemolysis and Measuring the Impact of Evidence-Based Practice Interventions Nicole V. Tolan, PhD, 1,3 Erin J. Kaleta, PhD, 1,4 Jennifer L. Fang, MD, 2 Christopher E. Colby, MD, 2 William A. Carey, MD, 2 Brad S. Karon, MD, PhD, 1 and Nikola A. Baumann, PhD 1 From the 1 Department of Laboratory Medicine and Pathology and 2 Division of Neonatal Medicine, Mayo Clinic, Rochester, MN; 3 Department of Pathology and Laboratory Medicine, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA; and 4 Sonora Quest Laboratories, Tempe, AZ. Key Words: Specimen hemolysis; Lipid emulsion infusion; Sample transport; Arterial catheter draws; Neonatal ICU; Clinical chemistry; Quality improvement; Clinical practice intervention Am J Clin Pathol July 2016;146: DOI: /AJCP/AQW086 ABSTRACT Objectives: Blood specimen hemolysis is a major cause of sample recollection in the neonatal intensive care unit. We aimed to reduce the hemolysis rate from 6.3% at baseline to less than 4% within the 9 months duration of the study. Methods: Intravenous infusion of lipid emulsion during sample collection, sample collection site, and blood sample transportation methods were investigated as possible contributors to hemolysis. Subsequently, two practice improvements were implemented: pausing lipid emulsion infusion prior to collection and slowing withdrawal rates through arterial catheters. Results: Samples were more likely to be hemolyzed if they were collected during lipid infusion and subsequently transported by pneumatic tube or collected through an arterial catheter. Retrospective analysis demonstrated a decreased number of tests cancelled due to specimen hemolysis (3.5%) after our interventions. Conclusions: We identified three variables contributing to hemolysis and instituted two clinical practice interventions to significantly reduce test cancellations due to hemolysis. In vitro hemolysis, the lysis of RBCs and release of intracellular contents and hemoglobin into the serum or plasma portion of separated whole blood, is a major source of laboratory errors and cause for specimen rejection. 1,2 The need for subsequent recollection may create delays in patient treatment, admission, or discharge. Excessive collection also increases the risk for iatrogenic anemia, which is of particular concern in the neonatal population. In these patients, it is critical to minimize the number of recollections and total volume of blood taken each day during their hospitalization. 3 A great deal of time and effort has been spent by hospitals and manufacturers of blood collection products alike, in an attempt to identify and mitigate contributors to specimen hemolysis. In vitro hemolysis is a result of preanalytical collection factors such as vein/needle size, difficult venous access, residual alcohol on the collection site, underfilled tubes, excessive mixing, jarring transportation methods, lipid content, temperature extremes, and pressure differences in connections of intravenous catheters. 4-6 Increased rates of hemolysis have been observed in the acute care setting, which further complicates critical testing A large number of studies have looked at rates of specimen hemolysis and ways to reduce it; however, these studies have most often excluded children, who typically have more limited venous access and much smaller veins. A single pediatric study performed at a level I trauma center demonstrated a 13% hemolysis rate for all blood draws. 11 Many acute care settings find it challenging to meet the 2% American Society for Clinical Pathology, All rights reserved. Downloaded For from permissions, please journals.permissions@oup.com on 03 July Am J Clin Pathol 2016;146: DOI: /ajcp/aqw086

2 Tolan et al /REDUCING SPECIMEN HEMOLYSIS IN THE NEONATAL ICU hemolysis rate benchmark identified as best practice by the American Society for Clinical Pathology. 8 As a result, reducing the rate of hemolysis has been the driving force behind a number of hospital and clinical laboratory quality improvement initiatives. In an effort to maintain high quality and accurate results, our clinical laboratory requests recollection of specimens if the degree of hemolysis exceeds a threshold that will affect the accuracy of the test results. Together, the Departments of Neonatal Medicine and Laboratory Medicine recognized a high rate of blood sample rejection from the clinical laboratory due to sample hemolysis beyond the degree acceptable for laboratory testing. The aim of our quality improvement (QI) project was to reduce the rate of specimen rejection due to hemolysis in our neonatal intensive care unit (NICU) population to less than 4% within the 9 months duration of our study. We used the DMAIC (Define, Measure, Analyze, Improve, and Control) approach to identify the greatest contributors to hemolysis, implement two targeted practice interventions based on this information, and subsequently measured the impact of these interventions. Materials and Methods This study involved the review of data obtained from a 9-month QI project, which was determined to not constitute research, as defined by the institutional review board at Mayo Clinic. The Mayo Clinic NICU in Rochester, Minnesota, is a 26-bed level IV regional referral center. Approximately 45% of the 300 to 350 NICU admissions per year are born outside Mayo Clinic, and most patients are admitted for prematurity or for respiratory distress unrelated to preterm birth. While there is an on-site stat laboratory, routine testing samples are analyzed in a core laboratory located 1.5 miles from the hospital, and specimens are regularly transported through a pneumatic tube system connecting the two locations. Samples collected from indwelling catheters are obtained by NICU nursing staff, whereas venipunctures are performed by trained laboratory phlebotomists. Planning Intervention When concerns arose about an unusually high number of NICU blood specimen recollections, neonatal nurses began manually recording incidences of sample rejection as a quality measure. From these data (n ¼ 85), it was determined that hemolysis was the leading cause of unacceptable specimens (58%, n ¼ 50) compared with sample contamination with intravenous fluids (12%, n ¼ 10), sample clotting (6%, n ¼ 5), inadequate sample volume (5%, n ¼ 4), or other causes (19%, n ¼ 16). The major collection site associated to hemolysis was indwelling arterial catheters (68%, 34/50) compared to collection via venous catheter (16%, 8/50), venipuncture (10%, 5/50), or capillary heel stick (6%, 3/50). In addition, the nursing logs indicated that continuous intravenous infusion of lipid emulsion was more highly associated with hemolysis (74%, 37/50) than saline alone (16%, 8/ 50). Determination of Contributors to Hemolysis On the basis of the nursing log results, we systematically measured the baseline rates of hemolysis and the variables associated with specimen collections. Over a period of 9 months, we defined and monitored the following preanalytical variables that can be controlled for: collection site, intravenous infusion(s) at the time of sample collection, and the subsequent specimen handling and transportation to the clinical laboratory. During this measurement portion of the study, specimen transport was alternated by day, as samples were either hand carried or sent through the 1.5-mile-long pneumatic tube system. Colored specimen collection cards were used to record information regarding each collection: laboratory tests ordered, sample collection site, lipid emulsion infusion status, and transportation method. Only samples collected from NICU patients with orders for routine chemistry tests were included in the study to mitigate the risks of delaying critical testing. Samples collected for routine laboratory tests performed in the core laboratory were included in this study. For these tests (total and direct bilirubin, magnesium, phosphorus, and C-reactive protein), the Roche Cobas 6000 Modular Analyzer Series system (Roche Diagnostics, Indianapolis, IN) was used to determine the hemolysis index (H-index), defined by the manufacturer as a semiquantitative measure of hemolysis for each sample (each unit of the H-index is approximately equivalent to 1 mg/dl free hemoglobin). The degree of specimen hemolysis was analyzed in the context of the associated preanalytic variables. The number of hemolyzed or not hemolyzed samples was also defined by the number of samples with an H-index greater than 50 (H-index >50), based on the most stringent hemolysis interference threshold, that of direct bilirubin. The degree of specimen hemolysis was reported as median H- index, range, interquartile range (IQR ¼ 75th percentile 25th percentile), and the absolute number and percentage of specimens considered hemolyzed with an H-index >50. Statistical analysis was performed to determine the significance between preanalytical factors contributing to either the median degree of hemolysis (Student t test) or categorical assignment of the specimen as hemolyzed or not hemolyzed (v 2 test). A two-tailed P value less than.05 was considered statistically significant. 114 Am J Clin Pathol 2016;146: American Society for Clinical Pathology 114 DOI: /ajcp/aqw086 Downloaded from

3 AJCP /ORIGINAL ARTICLE The measurement arm of the study was continued throughout two practice interventions in the NICU: (1) reducing the duration of lipid emulsion infusion to allow for a pause in the infusion prior to sample collection, termed lipid cycling, and (2) providing education on the best practice for the rate of sample collection through indwelling catheters, including an audit before and after educational intervention. Design of Lipid Cycling Intervention The nursing logs indicated that samples were more often hemolyzed when they were collected during intravenous lipid emulsion infusion. We suspected that the lipid emulsion within the sample could increase the fragility of cellular membranes, potentially by way of increased osmotic stress, glycol content, or other mechanisms. As a result, we implemented a pause in the lipid infusion for at least 1 hour prior to sample collection. The duration of lipid emulsion infusion was standardized for all NICU patients and reduced from continuous over 24 hours to cycled over 20 hours. The pause occurred from 1:00 AM to 5:00 AM to coincide with routine morning blood draw scheduling. This practice change was communicated to staff via medical/nursing practice committees, divisional newsletters, and notices posted throughout the unit. As a balancing measure during the intervention phase, we also monitored the length of the lipid pause as documented in the electronic medical record (EMR). If the pause was too short, this could minimize its effect on sample hemolysis. Conversely, if the pause was too long, this could potentially compromise the nutritional status of these critically ill neonates. Design of Audit and Educational Intervention of Withdrawal Rates Another preanalytical variable also contributing to hemolysis was the collection of samples through an indwelling arterial catheter. It was hypothesized that reducing the specimen collection withdrawal rate through the arterial line could reduce the sheer stress imposed on circulating blood cells and reduce the rate of hemolysis. In response, the second intervention included audits of the rate of blood collection through arterial catheters and education of the bedside providers on the optimal rate of collection, following the practice recommendation outlined in the quality initiative of Gordon et al. 12 The withdrawal rate audit was performed by NICU nursing quality coaches and included the observation of 33 arterial line collections. To provide education to bedside providers, the NICU nursing education specialist developed a slide deck that discussed the recommended arterial withdrawal rates and the results of the audit to provide education and feedback for the improvement process. A poster with similar content was displayed on the NICU Quality bulletin board, and reminder cards were placed by each patient s bedside where blood collection materials were stored. Retrospective Analysis of Intervention Effects While the initial evaluation of the nursing logs provided an estimate of the rate of specimen rejection, a manual entry system for quality analysis is limited in its accuracy and prone to subjectivity. Retrospective analysis of laboratory data reported through query of the laboratory information system (LIS) was performed to assess the validity of the measurement phase of the study and determine the true and objective number of samples rejected. All specimen collections from the NICU before and after each intervention were extracted from the LIS, filtering by NICU location-specific collection codes and patients aged less than 6 months, with tests performed in the core laboratory. These data were sorted into those tests that were cancelled due to hemolysis and those that were acceptable for testing, using test-specific H-index thresholds established by the laboratory. Hemolysis rates were calculated as a percentage of tests that exceeded the test-specific H-index threshold out of total tests performed. To ensure that ordering practices had not changed, which in itself could have given rise to a change in the rejection rates due to the testspecific hemolysis thresholds, we verified test frequencies to be comparable across each practice intervention. Results During the measurement phase, 384 samples were collected by venipuncture (n ¼ 110, 29%), venous catheter (n ¼ 125, 33%), arterial catheter (n ¼ 147, 38%), arterial puncture (n ¼ 1, 0.3%) or heel stick (n ¼ 1, 0.3%). These samples were collected from 132 unique NICU patients (63 males and 69 females) ranging in age from 0 to 169 days (median, 6 days). As a result of only a single heel stick sample and a single arterial puncture being collected, these two points were excluded from our data analysis. For 2 months during the intervention phase, the lipid emulsion pause duration was reviewed in the EMR for 36 infants for a total of 233 total parenteral nutrition (TPN) days to assess the compliance with the lipid cycling initiative. The lipid pause was the correct 4-hour duration for 89% (n ¼ 208) of the TPN days, with 4% being too short (defined as <3.5 hours, n ¼ 9) and 7% too long (defined as >4.5 hours, n ¼ 16). The range of the lipid pause during this period was 2.5 to 6 hours. During the control phase of the study, errors relating to the lipid emulsion pause were monitored using our hospital incident reporting system. For American Society for Clinical Pathology Am J Clin Pathol 2016;146: Downloaded 115 from DOI: /ajcp/aqw086

4 Tolan et al /REDUCING SPECIMEN HEMOLYSIS IN THE NEONATAL ICU the subsequent 9 months, only three errors were reported; two were failure to resume, and one was a failure to pause. No significant effect on blood sample hemolysis was observed as a function of lipid infusion status alone Table 1. An increase in the number of samples with an H- index >50 was observed for those transported via pneumatic tube compared with samples that were hand carried (33% vs 17%, P <.05), but no significant increase in the median H- index was observed (median H-index of 29 vs 14, P >.4). However, most important, compounding these variables resulted in a significant increase in hemolysis rates. Samples collected during lipid infusion then transported via pneumatic tube were more likely to have an H-index >50 (50% vs 6%, P ¼.002) and an increased median H-index (45 vs 9, P <.04) compared with the samples collected without a lipid infusion prescription that were hand carried to the clinical laboratory. Retrospective analysis demonstrated a reduction in the number of hemolyzed samples from 6.4% to 4.8% with the lipid cycling intervention Table 2. However, this did not reach statistical significance (P >.4), likely due to the limited number of samples collected during this period. When analyzing hemolysis rates by blood sample collection site, samples collected through indwelling arterial catheters had a significantly greater number of samples with an H-index >50 (41%, P <.007) and also the greatest degree of hemolysis (median H-index of 40, P ¼.03) compared with samples collected by venipuncture (15%, median H-index of 18) or through venous catheters (13%, median H-index of 15) Figure 1A. It was determined through the audit of arterial catheter withdrawal rates that while the average withdrawal rate was acceptable at 1.84 ml/min (range, ml/min), 30% (10/33) of samples were collected faster than the recommended rate of 2.0 ml/min. Providing education to NICU nursing staff on the optimal rate reduced this number to 20% (7/35). This translated to a significant reduction in the number of arterial collections received with an H-index of more than 50 (21%, P <.02) and reduced the degree of hemolysis (median H-index of 17, P ¼.008) compared with arterial catheter collection prior to audit and educational intervention (41%, median H-index of 40) Figure 1B. Retrospective analysis (Table 2) also demonstrated that this significantly decreased the number of samples rejected due to hemolysis from 4.8% to 2.8% during the draw-rate audit (P ¼.002). Hemolysis rates remained low at 3.5% throughout the posteducation phase. Altogether, retrospective analysis of data showed a significant reduction in the rate of hemolysis from the baselinerateof6.4%to3.5%(p ¼.0004) with the implementation of both lipid cycling and arterial collection rate interventions. Table 1 Median H-Index and Percentage of Specimens With an H-Index of More Than 50 for Samples Collected and Transported Under Varying Conditions a % Samples With Median H-Index, H-Index >50 Range (IQR; Q3-Q2) (No./Total No.) Lipid status No lipids prescribed 18 (0-412; 42) 24 (11/45) Cycling pause 29 (5-1,128; 1,123) 24 (13/54) Receiving lipids 25 (2-1,670; 71) 38 (8/21) Transportation Hand carried 14 (0-1,128; 32) 17 (9/54) Pneumatic tube 29 (5-1,670; 51) 33 b (27/81) Lipid status and transportation Hand carried No lipids 9 (0-93; 23) 6 (2/33) Cycling pause 16 (2-1,128; 32) 14 (6/43) Receiving lipids 24 (2-68; 43) 22 (2/9) Pneumatic tube No lipids 23 (0-412; 48) 27 (14/51) Cycling pause 28 (0-302; 64) 27 (22/81) Receiving lipids 45 c (7-1,670; 101) 50 d (6/12) H-index, hemolysis index; IQR, interquartile range; Q, quartile. a Lipid status describes the lipid emulsion infusion status (no lipids prescribed, cycling pause, receiving lipids) at the time of specimen collection. Transportation refers to the method of transport of samples to the laboratory (hand carried or transported through a 1.5-mile-long pneumatic tube system). The combinatorial effects of lipid emulsion infusion and transportation method are shown by significant P values. b P <.05. c P <.04. d P ¼.002. Table 2 Analysis of the Rate of Hemolysis During the Evaluation Phase, as Measured by Tests Ordered, for Each Phase of the Study Intervention: Baseline (1 Year Prior to Lipid Cycling), Post Lipid Cycling, Arterial Catheter Draw Rate Audit Phase, and Posteducation of Draw Rate a Hemolysis Rate of Tests, Intervention Phase % (No./Total No.) Baseline 6.4 (54/848) Post lipid cycling b 4.8 (28/519) Audit phase c 2.8 (52/2,056) Post-education d 3.5 (45/1,415) a There is a significant decrease in the overall rate of hemolysis from 6.4% to 3.5% over the 9-month quality improvement initiative. b P >.5 (lipid cycling). c P <.002 (arterial withdrawal rate). d P ¼.0004 (overall significance of interventions). Discussion Blood sample hemolysis is the most common reason for specimen rejection in the clinical laboratory. The overall rate of rejection due to specimen hemolysis is affected by the susceptibility of the laboratory s method to interference from hemolysis, the frequency that tests with stringent hemolysis rejection criteria are ordered, and how prone the patient population being tested is to hemolysis. As a result, 116 Am J Clin Pathol 2016;146: American Society for Clinical Pathology 116 DOI: /ajcp/aqw086 Downloaded from

5 AJCP /ORIGINAL ARTICLE Tolan A B 2,000 1, Hemolyzed (H 50) Not hemolyzed (H < 50) 2,000 1, Hemolyzed (H 50) Not hemolyzed (H < 50) H-Index H-Index Venipuncture Venus Line Arterial Line 1 Baseline Audit/Posteducation Figure 1 A, Hemolysis indices (H-indices) plotted for each sample collection site (venipuncture, venous catheter, and arterial catheter) for samples collected during the pre lipid cycling and post lipid cycling interventions of the measurement phase of this study. Both the median degree of specimen hemolysis (P ¼.03) and the number of samples with an H-index of more than 50 (P <.007) are significantly increased for arterial line collections compared with that for samples collected by venipuncture or through venous catheter. B, H-indices plotted for arterial line collections from baseline (pre lipid cycling and post lipid cycling phases) and audit/ posteducation phases. Here, the degree of specimen hemolysis (P ¼.008) and the number of specimens with an H-index of more than 50 (P <.02) are significantly decreased during withdrawal rate audit and educational intervention phases. laboratorians are faced with the need to produce highquality and accurate results while also providing timely testing and reducing unnecessary recollections, all in an effort to provide the best patient care possible. The largest contributor to hemolysis in our NICU population was the site of sample collection, where arterial catheter collections had the greatest number of samples with an H-index >50 and were hemolyzed to the greatest extent. The combination of intravenous lipid infusion at the time of sample collection and transportation through the pneumatic tube also resulted in a greater degree of hemolysis and number of hemolyzed specimens. By targeting our interventions to reduce sample collection during intravenous lipid emulsion infusion and provide education on optimal arterial catheter sample withdrawal rate, we observed a significant reduction in the number of hemolyzed specimens, achieving our goal of less than 4% of samples rejected due to hemolysis. Reducing the rate of hemolysis in the NICU not only improves turnaround time but also reduces the total blood volume collected and the risk of iatrogenic anemia. Institution of a lipid pause prior to sample collection allows for samples to be more efficiently transported via the pneumatic tube system, rather than incurring the additional cost and delay of transporting specimens by courier. When collection of specimens through an indwelling catheter is preferred over venipuncture, slowing the rate at which samples are withdrawn can reduce the rate and degree of specimen hemolysis. We targeted our interventions toward reducing the number of samples with an H-index >50, reflecting the threshold for direct bilirubin measurement, which is the most stringent hemolysis threshold in the laboratory. However, these efforts also significantly reduced the number of samples with gross hemolysis (ie, H-index >250). Specifically slowing the withdrawal rate of specimens American Society for Clinical Pathology Am J Clin Pathol 2016;146: Downloaded 117 from DOI: /ajcp/aqw086

6 Tolan et al /REDUCING SPECIMEN HEMOLYSIS IN THE NEONATAL ICU through indwelling arterial catheters after a pause in the lipid infusion not only reduced the number of samples hemolyzed for direct bilirubin measurement, as indicated by the lowered audit/posteducation median H-index (Figure 1B), but also reduced the number of grossly hemolyzed samples, as demonstrated by the reduced range of H- indices after intervention. While these practice interventions were targeted at a method very sensitive to hemolysis, the benefits were also notable in a potentially more critically ill patient population that may be predisposed to gross in vitro hemolysis due to their clinical condition. This QI initiative demonstrated the ability to systematically identify the greatest contributors to specimen hemolysis and target practice interventions to reduce specimen recollection in the NICU, a process that is generalizable within and across institutions, and could easily be applied to other aspects of laboratory medicine and clinical care. We were able to capture hemolysis rates before and after each intervention based on the quantitative degree of hemolysis through prospective measurement and also validate these findings through retrospective analysis for all samples collected in the NICU, using analyte-specific hemolysis cutoffs for rejection. With the QI project now in the control phase, specimen rejection and indications for requesting recollections continue to be monitored using the NICU incident reporting system in conjunction with the Department of Laboratory Medicine and Pathology. Corresponding author: Nikola A. Baumann, PhD, Dept of Laboratory Medicine, Mayo Clinic, 220 First St SW, Rochester, MN 55905; baumann.nikola@mayo.edu. Acknowledgments: We thank the many hospital staff participating in this study. Starting first with the nursing staff, who identified the increased prevalence of rejected specimens in the NICU and collected preliminary data that ultimately identified ways to improve practice. Second, we thank the phlebotomy team who was committed to recording the prospective data regarding blood collection during early morning rounds while laboratory personnel arranged for couriers to transport these specimens. Finally, we acknowledge the hard work and dedication of the clinical NICU team involved in conducting the audit of arterial withdrawal rates and subsequent educational interventions. This was a successful practice intervention that included clinical care and laboratory medicine specialties within our institution, which made a lasting impact on NICU patient care. References 1. Lippi G, Blanckaert N, Bonini P, et al. Haemolysis: an overview of the leading cause of unsuitable specimens in clinical laboratories. Clin Chem Lab Med. 2008;46: Jones BA, Calam RR, Howanitz PJ. Chemistry specimen acceptability: a college of American pathologists Q-Probes study of 453 laboratories. Arch Pathol Lab Med. 1997;121: Nexø E, Christensen NC, Olesen H. Volume of blood removed for analytical purposes during hospitalization of low-birthweight infants. Clin Chem. 1981;27: Bush V, Mangan L. The hemolyzed specimen: causes, effects, and reduction. BD Vacutainer Syst Preanalytical Solut. 2003;2003: Jaben EA, Koch CD, Karon BS. Lipid emulsion solution: a novel cause of hemolysis in serum and plasma blood samples. Clin Biochem. 2011;44: Streichert T, Otto B, Schnabel C, et al. Determination of hemolysis thresholds by the use of data loggers in pneumatic tube systems. Clin Chem. 2011;57: Burns ER, Yoshikawa N. Hemolysis in serum samples drawn by emergency department personnel versus laboratory phlebotomists. Lab Med. 2002;33: Lowe G, Stike R, Pollack M, et al. Nursing blood specimen collection techniques and hemolysis rates in an emergency department: analysis of venipuncture versus intravenous catheter collection techniques. J Emerg Nurs. 2008;34: Ong ME, Chan YH, Lim CS. Reducing blood sample hemolysis at a tertiary hospital emergency department. Am J Med. 2009;122:1054.e Dugan L, Leech L, Speroni KG, Corriher J. Factors affecting hemolysis rates in blood samples drawn from newly placed IV sites in the emergency department. J Emerg Nurs. 2005;31: Bush RA, Mueller T, Sumwalt B, et al. Assessing pediatric trauma specimen integrity. Clin Lab Sci. 2010;23: Gordon M, Bartruff L, Gordon S, et al. How fast is too fast? a practice change in umbilical arterial catheter blood sampling using the Iowa model for evidence-based practice. Adv Neonatal Care. 2008;8: Am J Clin Pathol 2016;146: American Society for Clinical Pathology 118 DOI: /ajcp/aqw086 Downloaded from