Lean in Healthcare
Impact of Lean on Lead Time Performance in Microbiologic Laboratories Master thesis, MSc Supply Chain Management
University of Groningen, Faculty of Economics and Business
June 21, 2013
Tian Yan
Student number: 2297213 E-mail: t.yan.1@student.rug.nl
Supervisor/ university prof. dr. ir. C.T.B. (Kees) Ahaus
Co-assessor/ university O.P. Roemeling, MSc
ABSTRACT ... 2
INTRODUCTION ... 2
THEORETICAL BACKGROUND ... 7
Lean Thinking ... 7
Lean in Healthcare ... 8
The Importance of Timeliness in the Laboratory ... 10
Comparison in Time within a Laboratory ... 10
METHODOLOGY ... 11 Data Sources ... 11 Study Design ... 14 Research Method ... 14 Data Cleaning ... 15 Statistical Analysis ... 19 RESULTS ... 25
Question 1 Differences in Lead Time Performance between the Lean Laboratory and Non-lean Laboratories ... 25
Question 2 Differences in Lead Time Performance between Three Phases of the Lean Laboratory ... 27
Question 3 Material(s) that Have Been Improved by Lean Implementation in the Lean Laboratory ... 31
DISCUSSION ... 33
Question 1 Differences in Lead Time Performance between the Lean Laboratory and Non-lean Laboratories ... 33
Question 2 Differences in Lead time Performance between Three Phases of the Lean Laboratory ... 35
Question 3 Material(s) that Have Been Improved by Lean Implementation in the Lean Laboratory ... 38
APPENDIX A: Methodology ... 41
APPENDIX B: Statistical analysis... 44
APPENDIX C: Results ... 46
ABSTRACT
To meet the increasing demands on the timeliness of results reporting, the microbiologic laboratory expects to improve the lead time performance through the adoption of lean thinking. A six-year mixed case study is performed on one lean microbiologic laboratory and three non-lean microbiologic laboratories of Dutch hospitals. A combination of longitudinal retrospective and cross-sectional approach is used to make comparisons of the lead time performance. We adopt the Mann-Whitney U test to check the difference in performance between laboratories. The Kruskal-Wallis test is used to inspect the difference in performance between different phases (the pre-lean period, the start-up lean period and the mature lean period). The material lead times and productivity ratios are measured. The statistical results show that in the lean laboratory the productivity ratios increase sharply in the mature lean period. Meanwhile, in the lean laboratory the lead time performance has been improved after the lean implementation. However, the most remarkable improvement appears in the start-up lean period rather than the mature lean period. Compared with the negative materials tested in one comparable non-lean laboratory, the negative materials in the lean laboratory do not show any outperformance.
Key words: lean thinking; lead time; microbiologic laboratory; Mann-Whitney U test; Kruskal-Wallis test
INTRODUCTION
lean has enabled public service providers to ‘do more with less’. Lean is pointed out to be the most influential management revolution of the past century (Liker, 2004). Now lean thinking is also applied in service industries, including healthcare (Manos, Sattler & Alukal, 2006). Nevertheless, lean is relatively new to healthcare and had not entered the healthcare sector until the early 2000s (Young and McClean, 2008).
Nowadays healthcare is meeting several challenges. These challenges include increasing costs, complex regulatory environments, increasing error rates, labor shortages in key departments, and the aging baby boomer population (Jimmerson, Weber and Sobek, 2005). Medical errors have a great impact on patient outcomes (Stankovic, 2004). Inadvertent errors in the delivery of medical care are regarded as a main cause of inpatient morbidity and mortality (Awad at al., 2005). Life expectancy is lengthening almost linearly in most developed countries. However, there is evidence suggesting that the prevalence of diseases, including heart disease, arthritis, and diabetes in the elderly population has generally increased over time (Christensen, Doblhammer, Rau & Vaupel, 2009). This phenomenon would lead to a growth in the demand for healthcare and in national health spending (Reinhardt, 2003). Meeting these challenges, lean has been one of the latest management imports to the healthcare sector (Mazzocato, Savage & Brommels, 2010). The aim of lean thinking in healthcare is to eliminate anything that does not add value to the patient’s care (Womack & Miller, 2005). Growing evidence demonstrates that lean thinking has potential impacts on healthcare performance (Radnor and Boaden, 2008). The reported tangible benefits are focused around a reduction in waiting time, space and cost with improved quality through the reduction of errors (Silvester, Lendon, Bevan, Steyn and Walley, 2004). Radnor and Walley (2008) examine the the potential intangible benefits, which include a rise in employee motivation and increased customer satisfaction.
laboratory represents a small percentage of medical central costs, it influences a majority of all critical decisions. The laboratory results provide essential information for patient diagnosis and treatment (Howanitz and Howanitz, 2001). Even though there only exists sparse data, timeliness in reporting laboratory results undoubtedly affects clinician and patient satisfaction as well as length of hospital stay (Howanitz and Howanitz, 2001). In a laboratory, lead time is the time required for one material (e.g. blood, urine, etc.) to travel through the entire stream of all the processing steps performed on that material (Buesa, 2009). With all of its complexities, exploring ways to shorten lead times and streamline the work environment without compromising safety has become a huge topic of discussion in the laboratory (Herasuta, 2007).
Lean thinking has been used successfully as a powerful intervention technique in a wide variety of healthcare settings (Mazzocato et al, 2010). At the early time, the value of lean implementation was confirmed in core clinical and supporting services throughout the hospital (Ben-Tovim, Bassham, Martin, Dougherty, Szwarcbord & Bolch, 2007). Lean implementation in healthcare has been largely applied to emergency departments and the acute care hospital management (Dickson, Anguelov, Vetterick & Singh, 2009). As Zarbo and D’Angelo (2006) indicate that since laboratories exhibit similarities to industrial production lines, lean principles seemingly could have major impacts on improving lead time performance. However, there is a limited amount of studies focusing on the influence of lean implementation on lead time performance in a laboratory setting. One interesting example is of Raab et al. (2008) describing the positive effect of lean implementation in the histopathology section of an anatomical pathology laboratory. Another article is written by Campos (2012) investigating whether eliminating institutionalized inefficiencies can lead to better lead time performance in a microbiologic laboratory at the Children’s National Medical Centre (CNMC) in Washington, DC.
these laboratories as the ‘non-lean laboratories’ in the following content. From the information provided by the lean laboratory, we know two major events happened during the researching period. One major event is the adoption of lean thinking in 2008. We would like to name the period before the lean implementation as the ‘pre-lean period’ (also known as the ‘non-‘pre-lean period’). Another major event is the purchase and the installation of a laboratory automation system in April 2010. This major event is accompanied by an investment of more than one million euros and a reduction of 1.5 FTEs (Full Time Equivalents). We call the phase between two major events as the ‘start-up lean period’. In all the articles published previously, the phases of lean implementation are divided into the pre-lean period and the post-lean period. However, within lean there is no final product and no end game; lean is a journey that needs to start strong and never ends (Bhasin, 2012). ‘Post-’ has a meaning of ‘over’ which goes against the lean journey. We would like to use the term ‘mature lean’ rather than ‘post-lean’. It is because ‘mature lean’ can stand for a further lean implementation based upon the start-up lean period. Thus the phase after the second major event is the mature lean period.
In the thesis, we will focus on the lead time performance of the materials tested in the microbiologic laboratories. The link between lean thinking and the performance of a laboratory through decreasing lead times will be the research theme. Therefore, the research question to be investigated is:
Will the lead time performance in a microbiologic laboratory be improved by introducing lean thinking principles?
Sub-questions:
What are the differences in lead time performance between the lean laboratory and non-lean laboratories?
What are the differences in lead time performance between different phases (the pre-lean period, the start-up lean period and the mature lean period) of the lean laboratory?
This research aims to provide insight into changes in the lead time performance as a consequence of applying lean thinking principles in a microbiologic laboratory. Adoption of lean principles may provide large efficiency and effectiveness gains to the microbiologic laboratory through a reduction of lead times.
The use of lean thinking principles in laboratories is still very new and the effects on lead times have not been well researched yet. Moreover, it is a novelty to make comparisons of lead time performance between laboratories, during different periods of lean implementation as well as between different materials. Raab et al. (2008) focus on one lean histopathology section and one non-lean histopathology section in a laboratory over four years. Campos (2012) concentrates to the material of which the growth is sufficient to permit bacteria grow-up (refer to ‘the positive material’) over two-year lean period. In our study we broaden the scope looking at a period of six years and compare the lean laboratory with four different non-lean laboratories. Meanwhile, we do not limit our study to the positive materials. The material which does not show bacteria growth (referred to ‘the negative material’) is also included in the research. These are quite addition to both Raab et al.’s and Campos’s studies. This study therefore adds valuable knowledge to the existing academic literature on the effect of lean implementation, specifically in microbiologic laboratories for both the positive materials and the negative materials.
THEORETICAL BACKGROUND
To understand what the previous scholars think about lean in healthcare and applying lean in the laboratory, the PubMed database was searched systematically. We focus on the articles published during the past thirty years, namely from 1 January 1983 up to 20 June 2013. Key terms were: Lean laboratory, Lead time, Turnaround time (TAT), Performance, Laboratory, Hospital and Comparison. From this electronic database, we identified more than 1837 papers with the word lean laboratory in the abstract. To obtain more journals relevant to our study, we centered on several combinations of research words, including ‘Lean lead time laboratory’, ‘Lean TAT laboratory’, ‘Lean performance hospital’, ‘Lean performance laboratory’ and so on. In the first assessment round, duplicated publications as well as non-English articles were discarded. Articles that reported hybrid approaches (such as ‘Lean Six Sigma’) and confusing concepts (such as ‘Obesity’ which is opposite to ‘lean’, but has no relation with our study) were excluded. Review papers were filtered and kept for further research. In a second assessment round, to add relevant articles, a snowball approach was conducted by tracing references cited in the previously selected articles. Abstracts were evaluated to find out whether or not these papers really fitted with our research objectives. From all the retrieved titles, we initially selected a subset of 48 papers. In order to ensure that only papers reporting on the relationship between lean and lead time performance were involved, we have examined all these 48 papers in greater detail and focused on the studies of the following four parts.
Lean Thinking
identify seven types of wastes in an organization which hinder the overall development, namely transportation, motion, inventory, waiting, over processing, overproduction and defects. By eliminating variability in the system, a process will intrinsically be more efficient, since it will only consist of value adding steps (Chiodo et al., 2012). In order to simplify the lean worldview, Womack and Jones (1996) define the following five basic principles of lean thinking:
Clear definition of the customer’s perception of product value;
Identification of the components in production that add to product value with elimination of all other non-valuable components;
Streamlining of the sequence of the remaining steps to allow for a smoother work flow;
Building a system that is driven by the pull of the customer’s requirements rather than the push of the manufacturer;
Pursuit of perfection through continuous re-evaluation and improvement.
The principles of lean thinking challenge the current circumstances in favor of simplified and standardized methods of performing tasks (Wong, Levi, Harigopal, Schofield & Chhieng, 2012). For both Toyota and the healthcare setting, the basic idea is to provide what is needed, at the time it is needed and where it is needed.
Lean in Healthcare
As a unit in healthcare, the laboratory is under sustained pressure to improve its service. The mantra of ‘faster, cheaper, better’ is heard in the laboratory world as loudly as in other sectors of society. The bottleneck for the whole laboratory is its central processing area and core laboratory (routine chemistry, hematology, coagulation, and urinalysis). Because these areas are influenced by increased demand and are more impressionable to making mistakes under pressure (Rutledge and Simpson, 2010). Adopting traditional management models does not lead to the required increases in quality and productivity, or error reduction (Stankovic and DiLauri, 2008). Lim et al. (2010) show that laboratory efficiency can be improved with the increased usage of automatic laboratory instrumentation and/or modifying an existing protocol. In addition, the efficiency may also be improved with the application of lean thinking to raise productivity through eliminating wastes (Downey, 2005). A number of laboratories from all the pathology disciplines have used lean to shorten lead times, improve quality (reduce errors) and improve productivity (Clark et al., 2013). Gubb (2009) proves a decreased average lead time in pathology from over 24 hours to two to three hours with less space and fewer resources by staff at the Royal Bolton NHS (National Health Services) Foundation Trust (RBH), UK. Now RBH has become an example of lean implementation (Bohmer & Ferlins, 2006; Ben-Tovim et al., 2007).
The Importance of Timeliness in the Laboratory
Due to the increasing pressures to provide more care at lower costs along with the increasing burden of documentation, clinicians are insisting on rapid lead times (Howanitz and Howanitz, 2001). The assurance of timely reporting is especially significant in a hospital setting where the tests are done for inpatients who are waiting for a decision about their treatment and procedures (Buesa, 2009). Only with the appropriate information available can laboratory results be explained to the patient, thus increasing clinician efficiency and patient satisfaction. Therefore, enhancing timeliness of results reporting is essential to laboratory quality improvement. However, it is not an easy task to improve lead times, because improving lead times requires education of individuals, long-term planning, and completion of innumerable jobs (Howanitz and Howanitz, 2001). Luckily, through using different methods, several laboratories have reported improvements in performance measures, including lead times, specimen flow, technologist productivity, quality, etc. (Persoon, Zaleski & Frerichs, 2006; Melanson et al., 2009).
Comparison in Time within a Laboratory
performance between the pre-lean period and the post-lean period. Compared with the studies of Hassell et al. (2010), Raab et al. (2008) and Campos (2012), we find there are similarities in research questions and given conditions.
METHODOLOGY
Data SourcesFIGURE 1
Timeline of Lean Implementation in Lab L
All the data was obtained from the laboratory database systems. The same type of laboratory database system is in use in all the studied laboratories. Using the same system allows us to have the same measurement points and largely equal processing steps. The data is transformed by an ICT specialist who is also a former analyst. The analyst is familiar with the processes and understands which kind of information is suitable for our questions. All the factors have helped to decrease the opportunity of obtaining incorrect information from the laboratory database, thus to improve the reliability of our dataset. Moreover, Nunnally and Bernstein (1994) present that the precision required in measurement theory cannot tolerate the large doses of sampling error. Possibilities for sampling error can be descended by ascending the size of the sample (Adams, 1951). When the sample size approaches the population size, chances for error become smaller. Our study benefits from the sample that has included the entire population, so there is almost no error.
Process steps for materials delivered to the microbiologic laboratory are shown in Figure 2. Initially materials with unique identifier numbers are delivered to the laboratory at the front desk. Once delivered to the bacteriology department, the materials are labeled and then grafted at the insert station. Later on the materials will be incubated in incubators. These incubators maintain a certain environment pleasant to bacteria to enable bacteria growth which allows for further testing. Every night the pictures of materials are made by the laboratory automation system. However, in the pre-lean laboratory, the start-up lean laboratory or the non-lean laboratory, almost all culture reading is handled by people during the daytime. After inspecting these pictures, an analyst determines if there is actual bacteria growth. Once the positive
2007 Pre-lean 2008-2009 Start-up lean 2010-Present Mature lean
First phase implementation lean
result (bacteria growth) is determined, tests are being performed to investigate which bacteria are present. If it is read as negative (no bacteria growth), it is important to check if the predetermined protocol time has passed. For example, if the protocol dictates that a certain material should incubate four days, the material will be returned into the incubator. If the material has already passed its protocol time, then it can be discarded as a negative result. The test result is automatically confirmed by test machinery when a result leaves the lab. Therefore, it is possible that a test result can have the same timestamp as the confirmation. Eventually, a micro-biologist authorizes the final result to be available to the initial client or applicant. Due to the automatic authorization, confirmation and authorization might have the same timestamp.
FIGURE 2
Basic Process Steps for the Material Delivered to the Laboratory
The datasets provide several definitions of processing steps for materials delivered to the laboratory, including identifier, material type, several measurement points, etc. These measurement points consist of ‘Label’, ‘Incubation’, ‘Test’, ‘Confirmation’ and ‘Authorization’ in sequence. Based on the measurement points, we would calculate the lead time of the whole process. In this study we focus on the research of lead time performance from ‘Label’ till ‘Confirmation’.
No
Yes No Protocol passed? Material Label Incubator
Confirmation
Authorization Positive?
Study Design
FIGURE 3
Design of the Study to Test the Effect of Lean on Lead Time Performance
The design of the study is outlined in Figure 3. The lead time performance of four microbiological laboratories within the past 6 years will be investigated. Based upon our conceptual model we expect a positive relationship between lean thinking and lead time performance in laboratories. We attempt to compare the lead time performance during three lean phases, including the pre-lean, the start-up lean period and the mature lean period, in the lean laboratory. Additionally, the comparisons between the lean laboratory and three non-lean laboratories will be made. Since all laboratories process mostly the same materials, it is possible to compare the lead time performance of each material. Outcomes of this comparison then allow for a decision to be made on which material(s) improvement efforts should be focused.
Research Method
Voss et al. (2002) proposed that the case study is especially applicable for the studies that have little or no existing theories. Even though several scholars have investigated the impact of lean implementation in the laboratory, however, the link between lean thinking and lead time performance in a microbiologic laboratory is still unclear. In our research, a mixed case study approach is planned to assess the influences of lean thinking on lead times during different phases of lean implementation and how it is different from non-lean laboratories. We use quantitative data provided by laboratory database systems to answer the research question and sub-questions. The multiple
Pre-lean period
Mature period
Start-up period Lean Lead Time Materials
Conceptual model
cases will offer a strong basis for theory building and increase the validity of research findings (Yin, 2008). Moreover, the usage of multiple cases allows for triangulation in the research to some degree (Easterby-Smith, Thorpe & Lowe, 2002).
To understand the changes of lead time performance starting from January 2007 in Lab L, a longitudinal retrospective approach is to be adopted. Based upon the rich continuous data, we can get a deep view on what types of changes occurred along with when and how these different types of changes occurred during the past years. In the meantime, we will adopt a cross-sectional comparative approach to make comparisons between the lean laboratory and non-lean laboratories. With purely cross-sectional data, we can only compare instantaneous snapshots among four laboratories. With longitudinal data we are only able to know what has changed during the phases of lean implementation. Therefore, a combination of longitudinal retrospective and cross-sectional approach will be used to answer the research question and sub-questions (Table 1). By using this combined method, Jacke et al. (2013) have successfully compared health quality on various healthcare levels.
TABLE 1
Research Methodology
Questions Research methodology Research question Multiple-case
study
Cross-sectional and longitudinal retrospective case studies
Sub-question 1 (Between lean and non-lean
laboratories)
Multiple-case study
Cross-sectional case studies Sub-question 2
(During different phases)
Single case study Longitudinal retrospective approach Sub-question 3
(Different materials)
Multiple-case study
Longitudinal retrospective approach
Data Cleaning
finishing processes before starting, which is impossible and points towards data error. Commonly the identifier number of a material is unique. However we can still find the same identifier multiple times in the database. It is because one single material can be used for various tests, or because of an error in identical records. Whilst Excel holds the possibility to remove the duplicate values, this does not help with tests that have different lead times. When one single material is used for various tests, the lead times of earlier tested materials can therefore include waiting times before final outcomes could be confirmed. In this case, the test with the lowest (quickest) lead time is the correct record. We pick out the identifiers which are not unique. Among these non-unique identifiers, the waiting time is defined as the difference between ‘Confirmation’ and ‘Test result’. For one identifier, only the data with the quickest waiting times is kept and others are removed. As the process steps introduced previously, the ‘Confirmation’ is operated before or at the same time with ‘Authorization’. Thus, the values ‘Authorization’ minus ‘Confirmation’ should not be less than zero. At this stage of data cleaning, we get the initial cleaned data (Table 2).
TABLE 2
Number of Data after Data Cleaning
2007 2008 2009 2010 2011 2012 2013 Total Lab L 45197 49134 48728 50132 49773 44863 25 287852
Lab F 10864 11845 13330 14228 14684 14070 79021
Lab M 38666 39391 46014 47923 46988 44033 263015
Lab Z 8489 9002 8968 9839 10303 10080 56681
in the lean lab and in non-lean labs, as well as between different phases of the lean laboratory. For comparability it would be better to unify the values of both the lower boundary and the upper boundary in all the laboratories. Meanwhile, it is necessary to make sure a majority of data are within the boundaries.
The laboratory automation system is possible to take a picture of the material after 16 hours, even if the picture is taken at night. However, in non-lean laboratories, the pre-lean laboratory and the start-up pre-lean laboratory, there are no automatic processes. The materials have to be in the incubators for at least one night (12 hours) until people come to inspect the material by hand. Therefore, based upon a conservative approach, the lower boundary is set as 12 hours.
using moving average is more obvious when we divide the materials into positive materials and negative materials in the following study. This is because there are more missing dates. Anyhow, adopting a one-week moving average we could get the continuous moving average lead time from January 1, 2007 till December 25, 2012. With the moving average lead times of both positive and negative materials, we draw the frequency and cumulative percentage tables (Appendix A). Based upon these tables, the moving average lead time of positive materials is focusing on 3.5 days and 4 days in Lab L, Lab M and Lab Z, as well as 3 days in Lab Z. For negative materials, the moving average lead time is concentrating on 2.5 days in Lab F, Lab M and Lab Z, along with 2.5 days and 3 days in Lab Z (Table 3). Table 3 shows the main moving average lead time values and the percentages of these days to the six-year period (2192 days). It is manifest that within six years, lead times are focusing on two or three values.
TABLE 3
Main Lead Time Values and Percentages
Basea (days) % POSb (days) % NEGc (days) %d Lab L 3 and 3.5 89.781% 3.5 and 4 72.035% 2.5 and 3 88.732% Lab F 2.5, 3 and 3.5 94.843% 3.5 and 4 82.784% 2 and 2.5 90.522% Lab M 3 and 3.5 72.278% 3.5, 4 and 4.5 71.043% 2.5 and 3 88.650% Lab Z 2.5 and 3 83.797% 3 and 3.5 78.022% 2.5 and 3 86.133%
aInitial cleaned data bThe positive materials cThe negative materials
dThe percentage of days with main lead time values to the six-year period
From Table 4, it is noticed that in some laboratories around 15% of the raw data is moved out. However, there are gaps between Table 4 and Appendix A in the percentage part. Appendix A shows that 5-day lead time has contained more than 90% or even reaching 99% of all the moving average data in these laboratories. As we know the moving average has taken the lead times of all the materials tested within one week into consideration. These lead times include outliers, such as the lead time with hundreds of days. Even though the moving average looking one week backward has reduced the influence of the outliers, however, it cannot promise 100% accuracy in evaluating the daily mean lead times. The lead times are focusing on between 2.5 days and 4 days. Because of the small value, the lead times with less than 2.5 days rarely have influence on the moving average. The value of moving average lead time might be enlarged due to the big values (hundreds of days). Therefore, the percentage gap between Table 4 and Appendix A can be explained by the outliers. The larger the gap is, the more outliers are existing in raw data. Moving out these outliers through setting boundaries helps us to enhance the reliability of our study. The datasets within the lower boundaries of 12 hours and the upper boundary of 5 days are used in the following statistical analysis section.
TABLE 4
Number and Percentage of Records Focused on Analyses
BASE POS NEG
Original Within
boundaries %
a Original Within
boundaries % Original Within boundaries % Lab L 287852 253058 87.913% 132131 113391 85.817% 155721 139667 89.691% Lab F 79021 69301 87.699% 33373 27826 83.379% 45648 41475 90.858% Lab M 263015 234204 89.046% 94890 79984 84.291% 168125 154220 91.729% Lab Z 56681 50034 88.273% 22287 20365 91.376% 34394 29669 86.262%
aThe percentage of researched data to initial cleaned data
Statistical Analysis
there is a length of time required to train the staff in new work methods and to stabilize processes. According to the run chart (Figure B1, Appendix B) of moving average lead times for both positive materials and negative materials in the lean laboratory, there are two adaptive periods with sharp decreasing fluctuation and increasing fluctuation. The former period is from 1-4-2008 till 1-6-2008 due to the initial implementation of lean. Another is from 1-4-2010 to 1-6-2010 with the purchase of the laboratory automation system. After the adaptive period, the laboratory would reach a stable environment that is amenable to additional improvements (Rutledge and Simpson, 2010). Eventually, the three periods of the lean implementation in the lean laboratory will be from 1-1-2007 till 31-3-2008 (the pre-lean period), from 2008 till 31-3-2010 (the start-up lean period) and from 1-6-2010 till present (the mature lean period). The distinction of three phases is also proved reliable by the XmR charts (referred to Figure 5 and Figure 6).
Data is analyzed using the statistical package SPSS 20. Before we make a decision on the suitable statistic method, the dataset is identified by the Kolmogorov-Smirnov (K-S) test to check if it is normally distributed. In our study, all statistical tests used in the analyses are nonparametric because of the non-normal distributions of lead times in sampling units (Altman, 1991).
Commonly the test process will be finished within one week. In such a short time period, the opportunity for external influence will be much less, whereas in longer period of time it is difficult to control external factors. This factor has helped to strengthen the reliability of the study. If changes occur in the four laboratories at the same time period, it is to be expected that these changes are caused by external factors. On the contrary, if during one period, changes only appear in one laboratory, it is quite possible that the internal factor leads to the changes. In our study, we do not compare the exact lead times of Lab F and Lab Z with the lead time performance of the lean laboratory. However, the datasets provided by Lab F and Lab Z could help us judge whether it is the external factor or the internal factor that leads to weird fluctuation or peaks.
The Mann-Whitney U test, which is also known as the Wilcoxon rank sum test (Mann & Whitney, 1947), is a non-parametric rank-based test for identifying a difference between populations with respect to their medians. This test is frequently used for continuous data that do not approximate a normal distribution after log transformation (Shepperd et al., 1998) and it is relatively sensitive to the homogeneity of the variance of sample data (Daniel, 1978). Even when the underlying distributions are normal, the Mann-Whitney test is about 96% as efficient as the t-test for moderately large sample sizes (Gibbons and Chakraborti, 2003). In our study, the Mann-Whitney test is performed to test the difference between Lab L and Lab M, a p<.05 is treated as significant. Moreover, median and mean values, as well as standard deviation values help us to judge the exact differences between laboratories.
The productivity ratio which is defined as the number of records divided by the number of personnel FTEs (Raab et al., 2008). The productivity ratio depends on the capacity of the system. While comparing between the laboratories, the productivity ratios can provide the information of the system capacity in each laboratory. Besides that, the productivity ratio per year describes the changes of system capacity.
Question 2 Differences in lead time performance between three phases of the lean
laboratory. Based upon the pivot table in Excel (Table B1, Appendix B), we obtain
materials have been tested for thousands of times. We would like to focus on the majority of all the materials. Therefore, the materials that have appeared during all the three phases and have been tested for more than 1000 times within six years are investigated. With this additional limitation, we still study 82.44% of all the positive materials and 87.30% of all the negative materials in Lab L (Table 5). The cleaned data within boundaries is used.
TABLE 5
The Majority of Materials in Lab L
POS Count NEG Count
catheter 1935 catheter 3164
genitaal 4356 genitaal 10255
luchtweg 21101 keel 2017
oor 1991 Keel/neus/perineum 18549
urine 60849 liquor 4470
vocht, weefsel 5489 luchtweg 11248
water 1061 neus 2091 wond 12152 oor 1156 Total 108934 urine 63151 % 82.44% vagina\anus 1245 vocht, weefsel 9429 water 3162 wond 6014 Total 135951 % 87.30%
*Count: the number of the material
significance levels using a Bonferroni correction. The Bonferroni correction multiplies the single test p-value by the number of independent tests to achieve an expected error rate (Thomas et al., 2006). We chose the multiple comparison procedure of Dunn (1964) which is appropriate for the use with the Kruskal-Wallis test (Neave and Worthington, 1988). The pairwise comparisons calculated as in Dunn (1964) rely on the data as a whole. This is unlike running multiple Mann-Whitney tests, which only rely on the data of the two groups they are comparing.
Mohammed (2004) points out that there is considerable interest in the use of statistical process control in healthcare. As a part of an overall philosophy of continual improvement, the implementation of statistical process control usually requires the production of control charts (Mohammed, Worthington & Woodall, 2008). The primary objective of the control chart is to distinguish between common (chance) and special (assignable) causes of variation (Mohammed et al., 2008). There are various control charts. The Moving Range Chart (referred to as the XmR chart) has the advantage that it does not assume a normal distribution in the data. Nevertheless, it is for a single observation per time period and these observations are independent of each other (Hovor & Walsh, 2007). Because the XmR chart is based on averages of absolute consecutive differences (Shewhart & Deming, 1939), an XmR chart can be easily affected by an extreme value (e.g., an outlier). Such extreme values enlarge the distance between the control limits and reduce the ability of the control chart to identify changes in the data. In our case, this problem has been solved by setting upper and lower boundaries (Note: in the XmR chart, the word ‘limit’ is adopted, and the ‘boundary’ is only used for cleaning data). Meanwhile, while testing a large number of data points, the extreme observation is averaged with many other observations and the impact of the outliers on the control limits is mitigated. These factors benefit the reliability of our study.
of materials tested per day from 1-1-2007 till 3-1-2013. It is clear that the number of materials tested at weekends is relative small and is mostly less than 30. Even if the data has been limited by the upper and lower boundaries, however, the mean lead time of the material tested for only few times might lead to an outlier in the control chart and distort the test results. Thus, the dates with materials tested daily less than 30 times are excluded.
Due to the two major events in the lean laboratory, it is necessary to decide from which period the control limits should be calculated. And extending this period to others can judge if the phases differ. Selecting the period with the least variability would produce the tightest control limit1. Thus, we examine the variability within
three phases by calculating the difference between the maximum value and the minimum value in each phase. Consequently, the period with the lowest variability will be chosen as the baseline.
Control limits in the XmR chart are calculated from a moving range (mR). The moving range is based on the absolute value of consecutive differences in observations. According to Wheeler (1995), the numbers of periods corresponds to a unique constant E value which makes sure 99% of the data fall within the control limits. Typically two consecutive differences are adopted. The correction factor E for two consecutive periods is 2.660 (Wheeler, 1995). The upper control limit (UCL) is the average of the observations plus a constant E times the average moving range. The lower control limit (LCL) equals to the average of the observations minus a constant E time the average moving range (Hovor & Walsh, 2007). Therefore, in our study
UCL = Average of observations + 2.66 * Average of moving range LCL = Average of observations – 2.66 * Average of moving range
Then we would like to find out whether the daily mean lead times stay between the control limits.
Question 3 Material(s) that have been improved by lean implementation in the lean
laboratory. Based on the table (Table B1, Appendix B) used in Question 2, we get the
number of every material tested in Lab L. Then we determine which material(s) we should focus on to fulfill the improvement research. In Lab L there are 31 kinds of
1
materials tested. However, some materials are rare (Biopt, Blaasvocht, maagbiopt, Pleura, sinusuitstrijk, and uitstrijk), and some are tested only during one phase (Biopt, luchtstrips, uitstrijk and scopen). There are only positive or negative results for some materials (scopen). Thus, these 8 kinds of materials are moved out while making comparisons among materials.
We would like to view the lead time changes of every material during different phases. The Kruskal-Wallis test is performed on each material separately. If the Kruskal-Wallis test is statistically significant (p < .05), post-hoc tests would be investigated to determine where exactly the differences lie within three phases. The mean and median values help to determine the percentage of improvement or deterioration.
RESULTS
Question 1 Differences in Lead Time Performance between the Lean Laboratory and Non-lean Laboratories
The results of the Mann-Whitney U test are shown in Table 6. For positive materials, only the lead time performance during the pre-lean period does not show a statistically significant difference between Lab L and Lab M (p=.295). While for negative materials, there are significant differences in lead time performance between Lab L and Lab M during three phases (p=.000).
TABLE 6
Between-lab Comparison of Lead Time Performance in Lab L and Lab M
Phase p-POS p-NEG Pre-lean 0.295 0.000 Start-up lean 0.000 0.000 Mature lean 0.000 0.000
period, the performance in Lab L is better than that in Lab M. However, in the mature lean period, there shows no outperformance in Lab L. The average lead time in Lab L is 3.77% higher than that in Lab M.
TABLE 7
Median Values, Mean Values and Standard Deviations in the Pre-lean Period
Lab N Mean Median Std. Deviation Lab L-POS 22392 2.886 2.859 1.102
Lab M-POS 14163 2.891 2.883 1.044 Lab L-NEG 27578 2.302 2.047 1.074 Lab M-NEG 29323 2.042 1.899 1.028
TABLE 8
Median Values, Mean Values and Standard Deviations in the Start-up Lean Period
Lab N Mean Median Std. Deviation Lab L-POS 37521 2.607 2.221 0.977
Lab M-POS 25642 2.740 2.754 0.971 Lab L-NEG 43860 2.097 1.960 1.076 Lab M-NEG 46833 1.977 1.925 0.989
TABLE 9
Median Values, Mean Values and Standard Deviations in the Mature Lean Period
Lab N Mean Median Std. Deviation Lab L-POS 47595 2.727 2.773 1.050
Lab M-POS 35988 2.628 2.171 0.893 Lab L-NEG 60612 2.109 1.935 1.141 Lab M-NEG 69772 1.771 1.865 0.843
FIGURE 4
Productivity Ratios per Year in Four Laboratories
Question 2 Differences in Lead Time Performance between Three Phases of the Lean Laboratory
The results of the Kruskal-Wallis tests present that there are significant differences in lead time performance between different phases in Lab L. The differences are available for both the positive materials (p=.000) and the negative materials (p=.000). Pairwise comparisons for both positive materials and negative materials manifest that lead time performance is statistically significantly different between any two phases (Table 10 and Table 11).
TABLE 10
Median and Mean Values for Positive Materials during Phases
Phase-POS N Mean Median Std. Deviation Pre-lean 21348 2.879 2.852 1.106
Start-up lean 35987 2.590 2.213 0.972 Mature lean 45933 2.715 2.765 1.046
TABLE 11
Median and Mean Values for Negative Materials during Phases
Phase-NEG N Mean Median Std. Deviation Pre-lean 26614 2.300 2.052 1.079
Start-up lean 42743 2.088 1.958 1.079 Mature lean 59191 2.098 1.932 1.142
Based upon the analyses above, we obtain the following results. For both positive materials and negative materials the lead time performance in the start-up lean period
is superior to that in pre-lean and mature lean periods. While regarding the pre-lean period as the baseline, for positive materials the performance in the start-up lean period and mature lean period has improved by 10.06% and 5.70% separately. The percentages of improvement are 9.2% and 8.76% for negative materials. However, the data indicates that there is a change for the worse between the start-up lean period and the mature lean period. From the start-up lean period to the mature lean period, the lead times increase by 4.83% for the positive materials and increase by 0.48% for the negative material.
To decide the baseline of the XmR charts, the maximum, minimum and variability values of three phases for both the positive materials and the negative materials are shown in Table 12 and Table 13. The mature lean period and start-up lean period are chosen as the baselines for the positive material and the negative material separately.
TABLE 12
Variability between Maximum Value and Minimum Value for Positive Materials
Phase-POS Max Min Variability Pre-lean 4.363 1.921 2.442 Start-up lean 4.212 1.828 2.385 Mature lean 4.162 1.962 2.199
TABLE 13
Variability between Maximum Value and Minimum Value for Negative Materials
Phase-NEG Max Min Variability Pre-lean 3.344 1.497 1.847 Start-up lean 3.250 1.586 1.664 Mature lean 3.400 1.532 1.867
TABLE 14
UCLs and LCLs for both Positive Materials and Negative Materials
Average of observation Average of moving range UCL LCL
POS (Mature lean) 2.707 0.342 3.617 1.797
NEG (Start-up lean) 2.077 0.233 2.695 1.458
FIGURE 5
Control Chart for Positive Materials Depicting Lead Time Performance within 6-year Period
FIGURE 6
Control Chart for Negative Materials Depicting Lead Time Performance within 6-year Period
Question 3 Material(s) that Have Been Improved by Lean Implementation in the Lean Laboratory
The Kruskal-Wallis test is performed on 23 kinds of specimens separately, to check if there are differences in lead time performance between three phases. For some specimens that appear in only two phases are performed by the Mann-Whitney U test. According to the test results (Table 15), we are able to recognize the lead times of which materials are significantly different in three phases.
TABLE 15
Results of the Kruskal-Wallis Test for Each Material
Material p-POS p-NEG
BAL 0.000 bloed 0.122 0.101 catheter 0.058 0.000 epidemiologisch 0.032 0.951 genitaal 0.000 0.000 GO 0.262 0.000 keel 0.584 0.000 Keel/neus/perineum 0.001 0.000 liquor 0.004 0.000 luchtrips 0.000 0.000 luchtweg 0.000 0.000 mond 0.171 0.009 neus 0.009 0.014 omgeving 0.995 0.077 oog 0.338 0.195 oor 0.000 0.000 perineum 0.404 0.036 urine 0.000 0.000 vagina\anus 0.032 0.000 vloeistof\filters 0.070 0.475 vocht, weefsel 0.411 0.000 water 0.000 0.000 wond 0.000 0.000
values would be influenced by the specific date, such as, the holiday and the weekend. Meanwhile, commonly the mean values would be impacted by the outliers in the datasets. In our study the raw data has been cleaned by setting boundaries. The influence of the outlier will be eliminated. Therefore, the mean values are used to calculate the percentage of improvement and the median values are treated as reference. It is possible the percentage results of the median and the mean show a difference. If the difference is small, we refer to the percentage result of mean values. While there are significant differences between the outcomes of the mean and median values, the whole range will be adopted. Table 16 presents the materials with improved lead time performance as well as the percentages of improvement. 72.58% of all the positive materials and 74.85% of all the negative materials show an increasing in lead time performance in the mature lean period.
TABLE 16
Materials with the Improvement
POS NEG Count-POS Count-NEG
DISCUSSION
Question 1 Differences in Lead Time Performance between the Lean Laboratory
and Non-lean Laboratories
Considering the positive materials, in the pre-lean period there is no difference between Lab L and Lab M in the lead time performance (p=0.295). In the start-up lean period, Lab L outperforms Lab M by 5.1% (p=.000). However, the productivity ratios in the pre-lean period and in the start-up lean period have no big changes with 7 FTEs in both Lab L and Lab M. Figure 7 shows the moving average lead times in Lab L and Lab M during the start-up lean period. Before May 2009, the lead times in Lab L are much lower than those in Lab M. Furthermore, before the peak points (Christmas holiday in 2008), the pattern in Lab L is opposite to the pattern in Lab M. After May 2009, there are many overlaps in two lines. In Figure 8, which presents the patterns of Lab L and Lab F in the start-up lean period, mostly the lead times in Lab L are lower than those in Lab F before May 2009. Thus it can be concluded that before May 2009, the improvement of lead time performance in Lab L might be caused by some internal factor. It is plausible that this internal factor concerns the lean implementation. Combining Figure 7 with Figure 9, we find that after May 2009 the lead times in Lab M decrease slightly. While in Lab L and Lab F the lead time performance fluctuates at a stable level during the periods before and after May 2009. Therefore, it might deduce that Lab M adopted some change method to improve the performance in May 2009.
FIGURE 7
Moving Average Lead Times in Lab L and Lab M for Positive Materials during the Start-up Lean Period
2.3 2.8 3.3
FIGURE 8
Moving Average Lead Times in Lab L and Lab F for Positive Materials during the Start-up Lean Period
FIGURE 9
Moving Average Lead Times in Lab M and Lab F for Positive Materials during the Start-up Lean Period
In the mature lean period, Lab M outperforms Lab L by 4.56% (p=.000), while the productivity ratio in Lab L is around 30% more than that in Lab M. We have shown productivity improvement similar to those demonstrated by Raab et al. (2008) in their multiple-year implementation of lean methods across the histology laboratory. In 2010, Lab L moved out 1.5 FTEs, while in Lab M there were still 7 FTEs. Although not directly quantified, improved efficiency correlates with a decreased level of expenditure due to higher production per FTE. However, no information provided to us shows the balance between the cost of lean implementation and the reduction of FTEs. In the further research, the trade-offs between the investment and the reduction of human inputs can be investigated. In the laboratory, there are many processes operated by machinery automatically, especially after the installation of the laboratory automation system. Despite that, there still needs considerable human inputs, such as reading the pictures made by the laboratory automation system. To some degree, the decreasing of FTEs leads to an increasing in the lead times. The reduction of FTEs can explain why Lab L underperforms Lab M in the mature lean period. Meanwhile, comparing 3.77% more in lead times with a reduction of 1.5 FTEs, we cannot
conclude that lean implementation has a negative effect on the performance in Lab L.
Nevertheless, for negative materials, Lab L does not outperform Lab M and Lab F during all the three phases (Figure 10). Within six years, the pattern of Lab F shows a steady fluctuation. As with the positive materials, the lead times of negative materials in Lab M reduced after May 2009. And then from 2010 the patterns of Lab F and Lab M overlap mostly. This factor enhances the probability that Lab M adopted certain change process around May 2009. For negative materials, the lead time performance in Lab L has no obvious advantage after the lean implementation and the purchase of the laboratory automation system. This result has been beyond our expectation. Moreover, based upon Table 7, Table 8 and Table 9, the number of positive materials tested in Lab L is larger than that in Lab M, while the situation for negative materials is opposite. However, the larger number of the positive materials does not influence the lead time performance in Lab L. It is strange that with fewer negative materials needed to be tested, Lab L still underperforms Lab M in all the phases. The processing steps introduced in methodology section mention the predetermined protocol time for the negative materials. It is possible the protocol time in Lab L is longer than that in Lab M, and then influences the overall lead time performance. It is also possible that in Lab L there are more breaks in the afternoon which leads to delaying the processing steps till next morning thus prolonging the lead times. These explanations are based on our assumptions without robust proof, which would be a limitation in our study. Therefore, further research can be done to investigate why lean implementation does not show a positive influence on lead time performance of the negative materials.
FIGURE 10
Moving Average Lead Times in Lab L, Lab M and Lab F for Negative Materials
Question 2 Differences in Lead time Performance between Three Phases of the Lean Laboratory
lean period and the mature lean period have shown outperformance in lead times. Based upon Figure 4, the productivity ratios in 2007, 2009 and 2009 are almost the same with 7 FTEs. Meanwhile, according to Figure 7, Figure 8 and Figure 10, the patterns of Lab F and Lab M do not show any strange changes in the start-up lean period. Therefore, the reduction of lead times in Lab L is caused by an internal factor. It is plausible that this internal factor concerns the lean implementation in 2008.
Compared with the productivity ratios in the pre-lean period and the start-up lean period, the productivity ratios have increased by 30.5% in the mature lean period. Meanwhile, Figure 4 shows a decrease in productivity ratio in 2012 and this reduction also happens in other two laboratories. Therefore, this reduction might be induced by some external factor which leads to a decrease in the number of materials delivered to the laboratories.
While regarding the pre-lean period as the baseline, for positive materials the performance in the start-up lean period and mature lean period has improved by 10.06% and 5.70% separately. The percentages of improvement are 9.2% and 8.76% for negative materials. The period with much more remarkable (9%-10%) improvement is the start-up lean period rather than the mature lean period after the purchase of the laboratory automation system. Compared with the lead times in the start-up lean period, the lead times increase by 4.83% and 0.48% in the mature lean period for positive materials and negative materials separately. After the purchase of the laboratory automation system, Lab moved out 1.5 FTEs in 2010. The introduction of a laboratory automation system and the simultaneous reduction of human inputs may be the cause leading to worse lead time performance. Compared with the positive materials, the negative materials are less influenced by a reduction of FTEs.
reach a stable environment that is amenable to additional improvements.
From April 2008 along with lean implementation, there is an obvious performance improvement. As Zarbo and D’Angelo (2006) point out that lean principles seemingly could have major impacts on improving lead time performance. Raab et al.’s (2008) findings indicate that the implementation of lean processes in a histopathology section decreased specimen lead times. Our results further their assumptions and findings that lean implementation has a positive influence on lead time performance of both the positive materials and the negative materials in one laboratory.
Table 17 shows the dates of peaks appearing in the XmR chars during the start-up lean period and mature lean period. Referring to the calendars in the Netherlands2, some of the peaks can be explained by the common causes, such as the public holidays. For positive materials all the peaks happened in the start-up lean period are during public holidays or one day before public holidays. Moreover, it is noticeable that among peaks that were not appearing on holidays, most of these peaks came up on Friday. Further research can be done to analyze which assignable causes lead to the peaks beyond the limits.
2
TABLE 17
Dates of Peaks in the XmR Charts during the Start-up lean Period and the Mature Lean Period
Start-up lean Mature lean
POS 24-12-2008 Christmas holiday 24-9-2010 Friday 11-4-2009 Easter holiday 22-10-2010 Friday
29-5-2009 31-5 White Sunday 24-12-2010 Christmas holiday 24-12-2009 Christmas holiday 30-12-2010 1-1 New year 31-12-2009 1-1 New year 4-3-2011 Friday
22-4-2011 Good Friday
29-4-2011 30-4 Queen's birthday 1-6-2011 2-6 Ascension Day 30-9-2011 Friday
NEG 2-10-2008 Thursday 30-7-2010 Friday 21-11-2008 Saturday 6-8-2010 Friday 24-12-2008 Christmas holiday 3-9-2010 Friday 26-3-2009 Tuesday 10-9-2010 Thursday 11-4-2009 12-4 Easter Day 23-9-2010 Thursday
15-5-2009 Friday 24-12-2010 Christmas holiday 20-5-2009 21-5 Ascension Day 31-12-2010 1-1 New year
24-12-2009 Christmas holiday 29-4-2011 30-4 Queen's birthday 30-9-2011 Friday
19-7-2012 Thursday 12-10-2012 Friday
Question 3 Material(s) that Have Been Improved by Lean Implementation in the Lean Laboratory
laboratory. Furthermore, not all the main materials present an improvement in lead time performance. The materials that do not show an improvement include catheter, luchtweg, vocht and weefsel for positive materials, along with Keel/neus/perineum, liquor, luchtweg and vocht weefsel for negative materials. As Table 16 indicates that the percentage of improvement can reach as high as 25%, while for some materials the improvement is not that obvious. These situations present a correlation between the lean implementation and lead time improvement. Further research can be focused on why some main materials do not show an improvement and how to improve these materials in order to gain better performance.
Setting boundaries is also a limitation in our study. While cleaning the datasets, 5-day is treated as the upper boundary, the lead time with a value higher than 5 was moved out. Even though we still included between 85% and 90% of all the data, it is may be considered to try setting the upper boundary as 6 days or 7 days. If the statistical analysis with a higher upper boundary shows the same results as what we get now, it would enhance the reliability of the conclusion.
APPENDIX A: Methodology
Table A1
Frequency and Cumulative Percentage Tables of Moving Average Data in Lab L
Lab L Base POS NEG
Lead time Frequency Cumulative % Frequency Cumulative % Frequency Cumulative %
0 0 0.00% 0 0.00% 0 0.00% 0.5 0 0.00% 0 0.00% 0 0.00% 1 0 0.00% 0 0.00% 0 0.00% 1.5 0 0.00% 0 0.00% 0 0.00% 2 0 0.00% 0 0.00% 0 0.00% 2.5 3 0.14% 0 0.00% 636 29.01% 3 953 43.61% 224 10.22% 1309 88.73% 3.5 1015 89.92% 945 53.33% 233 99.36% 4 186 98.40% 634 82.25% 7 99.68% 4.5 25 99.54% 235 92.97% 4 99.86% 5 8 99.91% 107 97.86% 3 100.00% 5.5 2 100.00% 25 99.00% 0 100.00% 6 0 100.00% 1 99.04% 0 100.00% 6.5 0 100.00% 1 99.09% 0 100.00% 7 0 100.00% 17 99.86% 0 100.00% Others 0 100.00% 3 100.00% 0 100.00% Table A2
Frequency and Cumulative Percentage Tables of Moving Average Data in Lab F
Lab F Base POS NEG
Lead time Frequency Cumulative % Frequency Cumulative % Frequency Cumulative %
Table A3
Frequency and Cumulative Percentage Tables of Moving Average Data in Lab M
LAB M BASE POS NEG
Lead time Frequency Cumulative % Frequency Cumulative % Frequency Cumulative %
Table A4
Frequency and Cumulative Percentage Tables of Moving Average Data in Lab Z
LAB Z BASE POS NEG
Lead time Frequency Cumulative % Frequency Cumulative % Frequency Cumulative %
APPENDIX B: Statistical analysis
FIGURE B1
Moving Average Lead Times in Lab L
TABLE B1
Number of the Material in Lab L
Material Count POS Count NEG
APPENDIX C: Results
TABLE C1
Median, Mean and the Percentage of Improvement for Each Material
POS-Median NEG-Median POS-Mean NEG-Mean
REFERENCES
Adams, L. (1951). Confidence limits for the Petersen or Lincoln Index used in animal population studies. The Journal of Wildlife Management, 13-19.
Altman, D. G. (1991). Practical statistics for medical research (Vol. 12). Chapman & Hall/CRC.
Awad, S. S., Fagan, S. P., Bellows, C., Albo, D., Green-Rashad, B., De La Garza, M., & Berger, D. H. (2005). Bridging the communication gap in the operating room with medical team training. The American Journal of Surgery, 190(5), 770-774.
Ben-Tovim, D. I., Bassham, J. E., Bolch, D., Martin, M. A., Dougherty, M., & Szwarcbord, M. (2007). Lean thinking across a hospital: redesigning care at the Flinders Medical Centre. Australian Health Review, 31(1), 10-15.
Bhasin, S. (2012). An appropriate change strategy for lean success. Management Decision, 50(3), 439-458.
Bohmer, R. M., & Ferlins, E. M. (2006). Virginia Mason Medical Center. Boston: Harvard Business School.
Buesa, R. J. (2009). Adapting lean to histology laboratories. Annals of Diagnostic Pathology, 13(5), 322-333.
Campos, J. (2012). Lean lab in action. Microbiology, 2, 2
Carey, R. G. (2002). How do you know that your care is improving? Part II: Using control charts to learn from your data. The Journal of Ambulatory Care Management, 25(2), 78-88.
Christensen, K., Doblhammer, G., Rau, R., & Vaupel, J. W. (2009). Ageing populations: the challenges ahead. The Lancet, 374(9696), 1196-1208.
Clark, D. M., Silvester, K., & Knowles, S. (2013). Lean management systems: creating a culture of continuous quality improvement. Journal of Clinical Pathology.
Daniel, W. W. (1978) Applied Nonparametric Statistics, pp. 211-214. Houghton Mifflin Company, Boston.
de Souza, L. B., & Pidd, M. (2011). Exploring the barriers to lean health care implementation. Public Money & Management, 31(1), 59-66.
Dickson, E. W., Anguelov, Z., Vetterick, D., Eller, A., & Singh, S. (2009). Use of lean in the emergency department: a case series of 4 hospitals. Annals of emergency medicine, 54(4), 504-510.
Downey, L. (2005). Instrumentation in the Cytogenetics Laboratory. In The Principles of Clinical Cytogenetics (pp. 113-130). Humana Press.
Dunn, O. J. (1964) Multiple comparisons using rank sums. Technometrics 7, 241-252.
Easterby-Smith, M., Thorpe, R. and Lowe, A. (2002), Management Research: An Introduction, 2nd ed., Sage, London.
Forsman, R. W. (1996). Why is the laboratory an afterthought for managed care organizations?. Clinical Chemistry, 42(5), 813-816.
Gibbons, J. D., & Chakraborti, S. (2003). Nonparametric statistical inference (Vol. 168). CRC press.
Guthrie, B., Love, T., Fahey, T., Morris, A., & Sullivan, F. (2005). Control, compare and communicate: designing control charts to summarise efficiently data from multiple quality indicators. Quality and Safety in Health Care, 14(6), 450-454.
Ham, C., & Dickinson, H. (2008). Engaging Doctors in Leadership: What We Can Learn from International Experience and Research Evidence?. NHS Institute for Innovation and Improvement.
Hassell, L., Glass, C., Yip, C., & Eneff, P. (2010). The combined positive impact of Lean methodology and Ventana Symphony autostainer on histology lab workflow. BMC Clinical Pathology, 10(1), 2
Herasuta, M. (2007). A “Lean” Laboratory. Lab Medicine, 38(3), 143-144.
Hilborne, L. H., Oye, R. K., McArdle, J. E., Repinski, J. A., & Rodgerson, D. O. (1989). Evaluation of stat and routine turnaround times as a component of laboratory quality. American Journal of Clinical Pathology, 91(3), 331.
Hovor, C., & Walsh, C. (2007). Tutorial on monitoring time to next medication error. Quality Management in Healthcare, 16(4), 321-327.
Howanitz, P. J., Cembrowski, G. S., Steindel, S. J., & Long, T. (1993). Physician goals and laboratory test turnaround times. Archives of Pathology & Laboratory
Medicine, 117, 22-8.
Howanitz, J. H., & Howanitz, P. J. (2001). Laboratory results timeliness as a quality attribute and strategy. American Journal of Clinical Pathology, 116(3), 311-315.
Jimmerson, C., Weber, D., & Sobek, D. K. (2005). Reducing waste and errors: piloting lean principles at Intermountain Healthcare. Joint Commission Journal on Quality and Patient Safety, 31(5), 249-257.
Keagle, M. B., & Gersen, S. L. (2005). The principles of clinical cytogenetics. Springer.
Krafcik, J. F. (1988). Triumph of the lean production system. Sloan Management Review, 30(1), 41-52.
Kruskal, J. B., Reedy, A., Pascal, L., Rosen, M. P., & Boiselle, P. M. (2012). Quality Initiatives: Lean Approach to Improving Performance and Efficiency in a Radiology Department. Radiographics, 32(2), 573-587.
Liker, J. K. (2004). The Toyota way: 14 management principles from the world's greatest manufacturer (pp. 35-41). New York: McGraw-Hill.
Liker JK, Meier D. (2005) The Toyota Way Fieldbook. New York, NY, McGraw-Hill Companies
Lim, A. S., Chen, T. J., Lim, T. H., Tan, M., Lau, L. C., Lim, P., Lee, G. Y., Loo, L. E., Liaw, F. P. S., Chuah, C. T. H., Goh, L. E., Tien, S.L. (2010). Bone Marrow Cytogenetics Workup: Application of Lean Management System to Determine if Additional Cell Workup is Helpful and Necessary to Analysis. Annals Academy of Medicine Singapore, 39(9), 696.
Mann, H. B., & Whitney, D. R. (1947). On a test of whether one of two random variables is stochastically larger than the other. The Annals of Mathematical Statistics, 18(1), 50-60.
Manos, A., Sattler, M., & Alukal, G. (2006). Make healthcare lean. Quality Progress, 39(7), 24.
