Proficiency-based progression intern training to reduce critical blood sampling errors including ‘wrong blood in tube’ [version 1; peer review: 2 approved with reservations]

Background: Blood sampling errors including ‘wrong blood in tube’ (WBIT) may have adverse effects on clinical outcomes. WBIT errors occur when the blood sample in the tube is not that of the patient identified on the label. This study aims to determine the effect of proficiency-based progression (PBP) training in phlebotomy on the rate of blood sampling errors (including WBIT). Methods: A non-randomised controlled trial compared the blood sampling error rate of 43 historical controls who had not undergone PBP training in 2016 to 44 PBP trained interventional groups in 2017. In 2018, the PBP training programme was implemented and the blood sampling error rate of 46 interns was compared to the 43 historical controls in 2016. Data analysis was performed using logistic regression analysis adjusting for sample timing. Results: In 2016, 43 interns had a total blood sample error rate of 2.4%, compared to 44 interns in 2017, who had error rate of 1.2% (adjusted OR=0.50, 95% CI 0.36-0.70; <0.01). In 2018, 46 interns had an error rate of 1.9% (adjusted OR=0.89, 95% CI 0.65-1.21; p=0.46) when compared to the 2016 historical controls. There were three WBITs in 2016, three WBITs in 2017 and five WBITs in 2018. Conclusions: The study demonstrates that PBP training in phlebotomy has the potential to reduce blood sampling errors. Trial registration number: NCT03577561


Introduction
Errors during sampling, labelling and transport to the laboratory (pre-analytical errors) are a common problem in the laboratory and account for up to 70% of all laboratory mistakes 1 . Frequent errors occurring in the pre-analytical phase include: i) identification errors, ii) errors in request procedures, iii) over-or under-filling of the specimen tube, iv) empty or missing tubes, v) contradictory demographic information on the tube and the request, (vi) 'wrong blood in tube' (WBIT) 2 . The most serious error, WBIT errors, occur when the blood is taken from the wrong patient but labelled with the intended patient's details (mis-collected) or blood is taken from the intended patient and labelled with the wrong patient's details (mislabeled samples) 3 . Guidelines exist on the correct practice of phlebotomy 4-6 . An observational study in 12 European countries demonstrated that compliance with phlebotomy procedures was low, with patient identification and tube labelling being the most critical steps requiring immediate action 7 .
The Testing the Utility of Collecting Blood Electronically (TUBE) study in the USA 8 , found a lower incidence of WBIT in electronically labelled samples than manually labelled samples (1:3046 compared to 1:14,606, respectively). Furthermore, the sample rejection rate for samples deviating from labelling policy was 1:67 samples, with 1:26 of mislabelled samples being possible WBIT events, therefore justifying policies which do not use mislabelled samples for analysis or cross match.
At our university hospital, the laboratory has been monitoring and recording details of the occurrence of WBIT and other sampling errors that result in the rejection of blood samples since 2010. Video recordings of doctors performing phlebotomy identified practices with the potential to lead to incorrect labelling of blood specimens 9 . Previous efforts to reduce mislabeling errors have included educational sessions with the haemovigilence officer, a zero-tolerance policy for transfusion samples, and educational campaigns to inform staff. These efforts failed to reduce the 'baseline rate' of sampling errors. Since the introduction of an online request clinical management system (isoft Clinical Manager, ICM) in the hospital, documented identification errors have increased despite standard education and training. This is a major concern.
Training and medical education can reduce the occurrence of pre-analytical errors including WBIT 3,10-12 but is not sufficient to eradicate WBIT. This study proposes proficiency-based progression (PBP) simulation training as a solution to reduce pre-analytical errors including WBIT. PBP is an approach demonstrated to be more effective than traditional training models in procedural skills [13][14][15] . The study aims to compare the blood sampling error rate of interns who commenced work in Cork University Hospital (CUH) in July 2016 and did not receive PBP training in phlebotomy (historical controls) to PBP trained interventional groups who commenced work in July 2017 (intervention group 2017) and July 2018 (intervention group 2018) over three months.

Study design
This non-randomised controlled trial involved two phases, the first to develop the PBP training programme and the second to determine the effectiveness of the PBP programme to reduce blood sampling errors, primarily WBIT.
Phase 1: proficiency-based progression training programme development 9 . To design a new training programme, phlebotomy procedure metrics were characterised, guided by the methodological design outlined by Gallagher et al. 16 . We identified and defined 11 phases of the phlebotomy procedure. These 11 phases had 77-steps (metrics) for safe phlebotomy performance. The procedure characterisation focused on the correct procedure performance, patient safety and on identifying critical steps to avoid errors, including pre-analytical phase blood specimen errors and WBIT. These phases and metrics were then presented to a multidisciplinary Delphi panel of procedure experts, who unanimously concurred that they represented a comprehensive, quantifiable depiction of the procedure. Following the Delphi panel, the metrics demonstrated construct validity (mean inter-rater reliability 0.91). An expert panel established the proficiency benchmark at a minimum observation of 69 steps, with no critical errors and no more than 13 errors in total. A list of the 11 phases and the defined critical errors are illustrated in Table 1 9 .
Phase 2: controlled trial. The second phase of the study aimed to determine if this bespoke PBP training programme could reduce the blood sampling and labelling error rate, including WBIT. The PBP training programme was delivered to the incoming interns in July 2017 and in 2018. The blood sampling error rate was monitored over three months and compared to interns commencing work in July 2016 (historical controls).
Qualitative analysis took place during mentorship of the interns while performing clinical duties to investigate which factors were contributing to blood sampling errors and to inform the training programme for the next phase of the study in July 2018.

Setting and participants
The study took place in a university teaching hospital with 800 beds. Participants were interns who commenced work in July of each year for a three-month rotation, recruited at induction training in the hospital.
Control group. The control group was comprised of 45 interns who were commencing work for the first-time following graduation in July 2016. The control group data had been collected prospectively as part of routine clinical practice. This group had received traditional phlebotomy training as medical students in their third medical year on two occasions and once in the final year of medical school, comprising a phlebotomy guide, training videos, and a practical training session in the clinical skills laboratory 2017 pilot study group and 2018 follow-on study group. The intervention group consisted of 45 interns who were commencing work for the first-time following graduation in July 2017 and 46 interns commencing work in July 2018. All the interns commencing work in the hospital in these years consented to participate. They were invited to the training as part of their induction but participation in the trial was optional. The intervention, in the form of a PBP training programme, was given to the interns on the commencement of their employment in the hospital. The group gave written informed consent for enrolment into the controlled trial.
The intervention in July 2017 and July 2018: proficiencybased progression training programme in phlebotomy The interns who were commencing work in July 2017 first completed an online training module, comprising of a video of the correct process of performing bloods in the hospital. In the second component, the interns attended face-to-face training. This consisted of a short motivational introductory talk from a consultant haematologist and a laboratory scientist to outline the importance of following the correct procedure and the consequences of errors. The interns worked in groups of three, with one person acting as a patient, a second person marking according to the metric, and a third person taking bloods. A tutor was assigned to each team. Each person had to perform phlebotomy on model IV arms on a simulated ward to the proficiency standard of performing at least 69 steps with no more than 13 errors occurring and no critical errors allowed, to graduate from the course. The third phase of training occurred once the interns had commenced work. The interns were observed performing phlebotomy on patients on the wards to ensure they continued to achieve the proficiency benchmark in real time.
All 124 interns were trained in July 2018 and due to work in the hospital during the 2018/2019 rotation, which led to the 2018 group having fewer tutors available; the intern-to-tutor ratio was therefore much higher than in the 2017 intervention Release tourniquet while last blood tube is filling before removing needle from arm Once all blood tubes collected -disconnects last blood tube before removing needle

IX
Fixes patient up after procedure and labelling of blood tube 60 61 Writes down patient's name and date of birth or patient identification number onto the blood tubes using a pen before leaving bedside If mobile patient label printer is available at bedside, prints label and checks it against the wristband.

Computer 72 73
Prints off patient's labels for blood tubes after blood collection Checks name and date of birth/ patient identification number on labels against patient's details written on the blood tubes if applicable.

Tidies up and sends bloods off
group. The ratio changed from three students per tutor in 2017 to 6-12 students per tutor for some sessions in 2018. Additionally, in 2018, each intern was provided with feedback on any error that occurred in the transfusion or haematology laboratory at the end of each month.
Quantitative and qualitative data collection and analysis Cognos was used to interrogate the laboratory information system, APEX, for pre-analytical phase blood specimen errors and WBIT. The electronic ordering software ICM system records the person who prints the label placed on the tube after taking the test. A search on the ICM system provided a list of each blood test performed in the three-month period, including the healthcare practitioner who performed the test and the patient identifier number, to link to the rejected samples in APEX. By matching the searches, this provided a list of the persons who obtained the rejected samples and a list of how many blood tests were taken by the interns over three months. Descriptive statistics were performed on the type and rate of errors. WBIT was the primary outcome. Secondary outcomes were all sampling errors (including WBIT). These included over-or under-filling of the tube, clotted samples, haemolysed samples, incorrect tube type received, no specimen received, and miscellaneous errors. Logistic regression analysis was used to examine the association between the odds for rejects in the 2016 control group to the intervention groups in 2017 and 2018 using Statistical Programme for Social Sciences (SPSS V24, IBM Corporation, 2016). To adjust for potential confounding factors, the month of the test and whether the test was taken on call or during normal working hours were included in the analysis. In a post-hoc analysis, we examined the potential clustering effect by intern in the logistic regression models. A p-value <0.05 was considered significant.
A process evaluation took place during mentorship on the wards. This comprised of ethnographic notes recorded by the investigator during blood taking, observing ease of access to equipment, the patient, the computer, the label printer, and any interruptions which occurred. Steps of the phlebotomy metric which required assistance or that were omitted were recorded. The interns were invited to give feedback on the training and any obstacles or challenges since work commenced. Comments were reviewed using a theoretical domain analysis framework 17 by two reviewers. The results of this qualitative analysis were used to inform changes to the training programme in 2018.

Results
The baseline characteristics of the 2017 pilot study and 2018 follow-on study groups are provided in Table 2. Descriptive statistics are not available for the 2016 control group as they were not working in the hospital at the time the study started. Figure 1 provides a flow chart to illustrate the recruitment of interns and the analysis of blood tests which they performed in the trial.
There appeared to be an increase in the primary outcome, WBIT, (although the numbers detected are very small) from 0.7 per 1,000 in 2016 (three WBITs) and 0.66 per 1,000 in 2017 (three WBITs), increasing to 1.3 per 1,000 in 2018 (five WBITs). The absolute numbers make an interpretation of trends difficult. Each of the instances of WBIT was identified by the laboratory, but one of the WBITs in 2018 was identified when the doctor rang the laboratory to self-report that they had mislabeled the tube. It is possible that this type of event would have been undetected in 2016, unless there was a discrepancy with previous results available in the laboratory.
There were 4,016 blood tests performed by interns in the control group who did not receive PBP phlebotomy training from July 11 th , 2016, to September 10 th , 2016; 96 (2.4%) of the blood samples were rejected. For the same period in 2017, 4,560 tests were taken by PBP trained interns and 55 (1.2%) of the blood samples were rejected. In 2018, 3,724 tests were taken by PBP-trained interns and 72 (1.9%) were rejected. Table 3 describes the breakdown of errors that occurred.
Logistic regression analysis (   11% reduction in the odds of a blood sample being rejected in the PBP trained group in comparison to 2016 control group, but this was not statistically significant ( Environmental context and resources: Written instructions given to the interns on the ward were noted to be poor with only one patient identifier provided. Time delays were commonly caused by label printers not working (seven cases) and unavailability of computers (eight cases). Essential equipment such as Azowipes®, tourniquets or blood tubes, were frequently missing (21 cases). Interns reported occasional instances of phlebotomy in patients who were not wearing an ID band.
Emotion: Nursing staff support on the wards contributed to calm and safe phlebotomy performance, while there were frequent interruptions by bleeps and time constraints contributed to stress, and multiple technical errors. Several interns expressed nervousness while somebody was watching them perform phlebotomy.
Knowledge: Interns had an average of 5.8 errors. They required prompting to ensure the steps of the metric were followed correctly.

Social influences:
The interns appeared to be influenced by their senior colleagues, some of whom felt that labelling at the bedside and printing labels at the computer located away from the patient after taking bloods was time consuming and unnecessary extra work.

Discussion
This study demonstrated that interns who received PBP training in 2017 had significantly fewer samples rejected than in the control group. There was a 50% reduction in errors in 2017. A reduction in errors was recorded in 2018 (compared to historical controls), amounting to a 11% reduction in odds for a rejected sample which was not statistically significant. The diminished effectiveness is possibly multifactorial, including the lack of safety culture around mislabeling on the wards and environmental stressors. Due to difficulty recruiting and the large number of interns trained in July 2018, the ratio of tutors to students increased, with one tutor attempting to teach 6-12 students for some sessions. The enthusiasm around PBP training in the hospital was not as heightened in 2018, possibly due to familiarity and perceived lack of reduction in WBITs.
This study contrasts with previous research indicating the benefits of PBP training 13-15 . It is a novel technique using technology-enhanced learning and simulation. Previous educational strategies tend to follow didactic-style teaching to improve phlebotomy technique and often rely on self-reported questionnaires to determine effect 18 . Educational strategies have been shown to reduce but not eliminate WBIT 10,11 . Bar code systems that scan the patient's wristband demonstrate a reduction in labelling errors and improve positive patient identification practices 8,19,20 . Many organisations, however, do not have sufficient resources to deploy these devices; the device is aimed primarily at transfusion sampling and requires proper training to be effective. Multiple interventions and feedback are likely to be more effective than single interventions, but the sustainability of improvements is not certain from previous research 11 . WBIT rates in mislabelled samples are estimated at 1.4% 8 , which is much higher than in correct samples, indicating that rejection of the sample due to any mislabelling event is indicative of an error-prone phlebotomy process that could have led to WBIT errors, and justifies the investigation of all blood sample errors to represent instances where there was a high risk of WBIT errors. This study demonstrates that, despite the introduction of comprehensive PBP training in phlebotomy, if the environment and process is error-prone, with the potential for shortcuts that can increase the risk of error, it is not possible to eliminate the risk of WBIT and other blood sampling errors even with optimal electronic systems. Healthcare organisations must adapt their systems to reduce distractions and tendency towards shortcuts which can lead to harm.
PBP training has robust evidence demonstrating a 40%-60% improvement in procedural performance [13][14][15]21 . This study examines the effectiveness of the intervention for a three-month period over two years and gives a clear insight into the sustainability of the project. The development and design of the project was multidisciplinary and involved all stakeholders in developing a training programme that was highly relevant.
While in many cases the use of historical data is not ideal, the thorough, systematic nature of this data, which has been documented systematically since 2010, allows us to be confident that the 2016 historical control group is representative.
The study has several limitations. The study did not have a sufficient sample size in order to examine the effect of the intervention on WBITs, the primary outcome of interest. One doctor in the 2017 group did not attend the final assessment and three of the doctors in the study were discovered to not be using their own ICM login, and therefore these participants could not continue the study.
Interns may have been negatively influenced by mentors who had not undergone PBP training and this could have undermined and weakened the potential impact of PBP training.
There was a concern that, although the interns were trained to proficiency, they did not always follow the process when unsupervised as it took longer to perform.
The qualitative element of the study identified a number of environmental factors which could potentially increase the risk of WBIT, including patients not wearing identification wristbands, difficulty accessing essential equipment, insufficient hardware and stress. These factors were persistent in all years; however, it was not possible to measure the level of these over the three years.
Given that the project was heavily promoted in the laboratory, it is possible that there was an increased awareness of WBIT, and this could have led to an increased detection rate amongst laboratory as well as ward staff. This detection bias has been described in previous studies involving WBIT, where errors increased despite the introduction of quality improvement initiatives 22 .
In conclusion, PBP training in phlebotomy can reduce blood sampling errors, but must take place in an environment that clearly acknowledges the importance of the training and the quality of the training must be properly resourced.
This study was unable to demonstrate if PBP training in phlebotomy influences the incidence of WBIT due to an inadequate sample size and possible detection bias.

Main messages
• PBP training in phlebotomy can reduce blood sampling errors, but barriers to following the correct procedure on the wards must be considered and removed if possible.
• Educational interventions alone are insufficient to reduce blood sampling, including WBIT, if the environment does not allow for a safe and efficient phlebotomy process, including the availability of appropriate hardware (especially bedside label printers).
• This bespoke PBP training programme in phlebotomy can be adapted for use in other healthcare facilities to train healthcare practitioners at commencement of employment, following external validation.

Current research question
• Future studies should examine the effect of bedside label printers coupled with PBP training in phlebotomy to examine the effect on blood sampling errors including WBIT.

Data availability
The underlying data (anonymised) will be made available on request for bona fide researchers, including researchers outside of UCC. Approval for data sharing was not sought at ethics approval stage nor was it included in the information sheets and consent forms provided to participants. Also, the sample size is small, and this raises the risk of potential reidentification of participants. To request access to the data please email the corresponding author, Dr Noirin O' Herlihy-100312258@umail.ucc.ie

Acknowledgements
We thank all the participating volunteers for their efforts and contributions including the laboratory at CUH and the healthcare practitioners who participated in the organisation and delivery of the PBP training programme in phlebotomy. We acknowledge Dr Patrick Kiely's role in the development of the online eLearning programme.

Contributors
NOH collected data, performed formal analysis, investigation, data curation, writing original draft, visualization, project administration, SG revised the manuscript, PH writing review and editing, investigation, RG writing review and editing, investigation, AK formal analysis and writing review and editing, MR data curation, writing review and editing, AG conceptualization, funding and acquisition, supervision. methodology, formal analysis writing review and editing, MC conceptualization, funding and acquisition, supervision. methodology, formal analysis writing review and editing, The authors are to be commended for systematically investigating the impact of their educational intervention. The main limitation of this study is the use of historical controls: this study design was practical but did not allow the investigators to prove a causal relationship between the intervention and reduction in rejected samples. Nevertheless, the data are suggestive of a beneficial effect of PBP training. This paper would be improved by providing some additional details, as suggested below.   training in phlebotomy, if the environment and process is error-prone, with the potential for shortcuts that can increase the risk of error, it is not possible to eliminate the risk of WBIT and other blood sampling errors even with optimal electronic systems."

Methods
I agree in principle that it is not possible to completely eliminate errors such as WBITs even with optimal electronic systems, as long individuals participating in the sample collection process have the ability to cut corners/use electronic systems improperly. But I don't think that this study demonstrates this point -"optimal electronic systems" were not studied directly here. It's also not clear to me that shortcuts were specifically documented in the data set.

Are sufficient details of methods and analysis provided to allow replication by others? Partly
If applicable, is the statistical analysis and its interpretation appropriate? I cannot comment. A qualified statistician is required. The authors describe a novel method of practical phlebotomy training for junior medical staff and show that this was associated with a reduction in sample rejections (similar interactive and simulation training for transfusion has been introduced for final year medical students in Wales).
Although not the subject of this paper, correct patient identification and sampling are particularly important in transfusion, since patients have died as a result of misidentification. However, the same standards should apply to all phlebotomy. Positive patient identification, asking the patient to say their full name and date of birth, is not included in the standards and should be.
I suggest the paper should specifically reference 'human factors' and the need for training in this science. The factors are there in the paper, but an additional paragraph would be of benefit.

Introduction:
There is appropriate description of the errors made with pre-analytical variables. It is also the focus in the results.
I am also surprised that there is no mention of human factors science as this is increasingly recognised as key in what we do.
(Note mis-spelling of haemovigilance officer in third paragraph -currently has 'e' where there should be an 'a').
It is very worrying that 'since the introduction of an online request system...documented identification errors have increased'. There is no further discussion of this in this paper and I would recommend further explanation. In my experience where patients are listed alphabetically it is easy to select a patient with similar or same last name. How has this been mitigated?

Methods:
The paper referencing the background work for this project is not open access which is unfortunate. It is clear that procedures for blood sampling through to return of results and any resulting changes in patient management are many and complex, so this is an interesting project for analysis of the steps and training to reduce the errors.
Similar work was done in Scotland by Pickup et al. (2017) 3 . Using human factors methodology they analysed 50 observations, 15 interviews and 12 months of incident data from all Scottish hospitals noting the influence of working environment, equipment, clinical context, work demands and staff resources. There are many reasons why 'work as done' is not 'work as imagined' (i.e. departures from the correct or recommended procedures).
The authors identified 11 phases and a total of 77 steps with the phases shown in Table 1. I was very surprised that under phase IV positive patient identification (i.e. asking the patient to state their name and date of birth) was not mentioned at all as this is critical to avoid 'wrong blood in tube'. Under phase IX the description 'fixes patient up' could perhaps be better expressed -what exactly does it mean?
The control group: We are told that these interns had received 'traditional phlebotomy training…'. It would be helpful to know what that was, particularly what they were taught about positive patient identification and labelling at the patient's side. These two points have been identified regularly by the Serious Hazards of Transfusion haemovigilance scheme (SHOT), where I am affiliated, as causes of error since its inception in 2006.

Results:
There was a progressive reduction in rejected samples with successive cohorts (mainly reduction in clotted and over/under filled tubes) although there was an increase in WBIT. There was a reduction in total errors from 24 per 1000 samples in 2016 to 12.1 in 2017 but 19.3 in 2018 (still a reduction compared the control group).
The observations of the interns once on the wards show the importance of environment and working conditions, and impact of senior colleagues who tended to discourage some of the correct practices. These are all 'human factors'.

Conclusions:
Are mostly justified by the results. The authors show that the training had some effect in reducing errors. The reasons for the reduced effectiveness in 2018 are well described. It would be good to see further discussion of 'lack of safety culture around mislabelling on the wards' because this is worrying and links to the lack of information about positive patient identification in the training discussed above.
The sentence 'this study contrasts with previous research' is unclear -do the authors mean the contrast is in the methods used rather than the effectiveness?
The sentence at the top of page 8 is very long and would benefit from being divided into shorter multiple sentences. Short cuts are mentioned here but not elsewhere in the paper. What shortcuts were identified?
I am not sure that I agree with the statement 'it is not possible to eliminate…errors even with optimal electronic systems'. This may be true, but the full vein-to-vein electronic systems have not been discussed or demonstrated in this project. I am convinced that they improve safety for blood