Order Management Project (OMP) OMP Post-CareLink Analysis of Medication Turnaround Time in UMHS Final Report. December 12 th, 2008

Transcription

1 Order Management Project (OMP) OMP Post-CareLink Analysis of Medication Turnaround Time in UMHS Final Report December 12 th, 2008 Client: Brian Callahan, PharmD Assistant Director, Inpatient Pharmacy Services Project Coordinator: Bobby Beasley Management Engineer Fellow, POA Team Members: Andrew Rolph, IOE 481 Student Patrick Traynor, IOE 481 Student Alex Vanderkaay, IOE 481 Student

2 Table of Contents Page Executive Summary 4 Introduction 6 Background 6 Goals and Objectives 78 Project Scope 8 Support Provided From Operating Entities 89 Methodology 89 Phase 1 Staff Surveys Process Mapping Phase 2 Phase 3 Delivery Study 24 Hour Study Analysis and Results 112 Phase 1 Staff Surveys Process Mapping Literature Search Phase 2 Phase 3 Delivery Study 24 Hour Study Conclusions and Recommendations 23 Appendices 256 2

3 Charts/Tables/Graphs Table of Contents Page Figure 1 Process Flow Chart 78 Figure 2 Unavailable Data Flow Chart 101 Table 2 Log Sheet Example 101 Table 1 Literature Study Times 123 Figure 3 - Useable Data for all STAT/NOW Orders July September Figure 4-Orders June-September 2008 Exponentially Distributed 134 Table 3- Descriptive Statistics of June-September 2009 Data 145 Figure 5 - Floor Stock Medications Have Shorter TAT 156 Figure 6 - CVC & Mott have the shortest TAT stat 167 Figure 7 - Percentage of Orders by Satellite Pharmacy 167 Figure 8 -Turn around Times are Longest from 7am to 12 pm 178 Figure 9 -- CVC has the Most Floor Stock Orders 189 Figure 10 UH 6 th Floor Pharmacy Capability during off hours 1290 Table 4- Time Critical Medications (a/k/a Super-Stat ) 201 Figure 11 - Time Critical Medications have Longer TAT 201 Figure 12 Late Orders Pie Chart 223 3

4 Executive Summary Background CareLink is an electronic prescription ordering system that has replaced the paper based system used at the University of Michigan Health System (UMHS). A recent post implementation study confirmed that medication turnaround times (TAT) have improved significantly in the Mott Children s Hospital as a result of the new system. Prior to this report, the effect of CareLink on TAT stat in the University Hospital and Cardiovascular Center were not quantified. In addition, sources of variation in the prescription ordering process throughout UMHS were unstudied. Therefore, the primary object of this project was to quantify TAT stat throughout UMHS, and study sources of variation in the prescription ordering process. The scope of the study was limited to orders placed for inpatient nursing units, and to medications needed immediately by patients. Such medications are commonly referred to as STAT or NOW. By UMHS standards, STAT orders are to be administered within 20 minutes, and NOW orders within 90 minutes. To prescribe a medication, physicians enter order information in the CareLink system, which is then sent to the pharmacy for validation and further processing. Once verified by a pharmacist for safe use, a nurse can obtain the medication in one of four ways. The fastest method by which nurses can acquire medications is through the Omni-Cell Machines. Omni Cells are essentially vending machines filled with frequently used medications, and are located throughout UMHS. Prescriptions not available in the Omni-Cells are filled in the pharmacy and transported back to the patients via a pharmacy technician, nurse, or through the pneumatic tube system. The post implementation study conducted at the Mott Children s Hospital indicated that medications were ready for delivery within 15 minutes of being ordered in CareLink, but that TAT stat were well over 60 minutes. With this information the project was directed towards studying how medications are delivered to patients, and to sources of delays in medication administration. Methodology The project studied the medication administration process in three phases: (1) process exploration, (2) historic data analysis, and (3) semi-live data analysis. During phase 1, pharmacy & nursing staff members were interviewed, and detailed process flow charts were developed. In phase 2, over 30,000 STAT & NOW orders were analyzed from the summer months of In phase 3, the team collected data to quantify the amount of time medications spend in processing at the pharmacies, and studied the causes of delays in medication administrations. Findings The project determined that the average TAT for STAT/NOW medications is 77 minutes with 63% of all these orders being administered on time. Variation in TAT stat existed across pharmacies, order priorities, nursing units, delivery methods, and times of the day. However, such factors explained only about 15% of the TAT variation, and the remaining portion was inherent to the process. The data collected in phase 3 of the study indicated that once a prescription is ordered, it takes an average of 17.5 minutes for the prescription to exit the pharmacy. Finally results from the semi-live data analysis suggested that up toat least 50% of delays in medication administration can be attributed to poor electronic communications between the nurses and pharmacies. 4

5 Conclusions Approximately 30% of the total TAT for all STAT/NOW orders within UMHS is located within the pharmacy. The remaining time, roughly 57 minutes, is due to inefficient delivery and administration methods. Up to 50% of late orders can be attributed to poor communication between the pharmacy and nursing units. Pharmacists and nurses have limited knowledge of an orders status throughout the medication ordering process. Nurses may not be aware of existing orders for their patients, and. ppharmacists have limited knowledge of the status of an orders status after verification. The project found that a major cause of delays is lack of communication between nurses and the pharmacy. The nnurses have no no way of determining how medications are delivered This is contributing to the large, and many nurses do not know which medications are available in the Omni-Cells. Finally the project discovered that there is little knowledge of an order's status once in leaves the pharmacy. Finally, ppharmacy technicians do not know if an order is STAT/NOW unless the pharmacist verbally informs them of the priority. Therefore, the level of communication and knowledge of STAT/NOW orders could be improvedis poor within the satellite pharmacies. Recommendations The team developed the following recommendations from this project: Provide a method to inform nurses when & how a medication has left the pharmacy Send alerts to nurses and pharmacists when medications have not been administered on time. Educate hospital staff on time standards for STAT/NOW orders Advance towards a mostly Omni-Cell model similar to that of the Cardiovascular CenterCVC Provide STAT/NOW notification on labels printed to pharmacy technicians Improving communications between the nursing staff and pharmacy staff will significantly reduce several sources of variation contributing toof long TAT. Implementing a method to inform nurses of the mode of order transportation of an order, and alerting them once their order has left the pharmacy will likely reduceimprove TAT across UMHS.floorstock and pneumatic tube orders. The nurse, or nursing staff clerk, will be notified to that an order has been verified and that the medication is readily available within the nursing unit, effectively maximizing TAT. Formatted: Indent: Left: 0.2", Hanging: 0.3", Bulleted + Level: 1 + Aligned at: 0" + Tab after: 0.25" + Indent at: 0", No widow/orphan control, Don't adjust space between Latin and Asian text, Don't adjust space between Asian text and numbers In addition, ssending alerts to the pharmacists and nurses responsible forassociated with a late prescriptions order will help to reduce the frequency of excessively long TAT.. For example, aa simple alert windowalert within CareLink couldan informalert the pharmacists and nurses of the tardy orders so that proper, and timely action could be taken.en to administer the medication. 5

6 Educating hospital staff on the UMHS time requirements for different types of medication orders will improve overall quality with the pharmacies. Employees will have a broader knowledge of the process and its applications. This in turn will allow all employees to prioritize orders within their schedule to ensure they are being administered on time. Total TAT was the lowest for medications located within the Omni-Cells. A great potential within UMHS exists to reduce facility TAT by implementing a nearly all Omni-Cell (Cartless) model wherever applicable. Improving the communication within each satellite pharmacies will also reduce total TAT. This could be achieved by iimplementing a separate label specific to STAT/NOW orders will allow the pharmacy technicians to clearly identify and process the orders of highest priority. Introduction From 2005 to 2008, the computer system called CareLink was implemented into various hospitals in the University of Michigan Healthospital System (UMHS). CareLink is an electronic patient prescription order system that has replaced the paper system used in the past. Many hospital employees agreed that the previous order system was inefficient and out of date. A post CareLink study was conducted in 2007 strictly on Mott Children Hospital of the system s effectiveness and confirmed medication turnaround time (TAT) has improved significantly. However, there are still uncertainties of CareLink s efficiency throughout the different satellite pharmacies in UMHS. Therefore, the Inpatient Pharmacy Service Manager would like to know how to best utilize CareLink throughout the satellite pharmacies and make it most effective. Our team has been asked to analyze this new system more closely to determine if CareLink can be improved or used more efficiently throughout the satellites. Also the team has been asked to analyze CareLink at different times and satellite locations. This report provides a detailed analysis of CareLink throughout UMHS as well as recommendations to improve the system. Background Physicians enter prescription order data for patients into a CareLink computer on/off site. After the data is sent into CareLink, the pharmacy can access it by running a batch review. When the pharmacist retrieves and validates the order, the pharmacy technician builds the prescription and placesuts it with the other outgoing medication. If the prescription is a STAT/NOW, that order has priority over the others. In UMHS, the time definition of a STAT order is the medication being administered within 20 minutes and NOW is within 90 minutes, which implies that a STAT order has priority over a NOW order. 6

7 The filled prescriptions are transported back to the patients by physical delivery via the technician, nurse pickup, or pneumatic tube system, which may depend on the urgency of the medicine. Another quick option nurses can use to obtain more commonly needed medications is through the Omni-Cell Machines (or Omni Machines), which are basically vending machines filled with frequently used medications. The nurses or physicians can access these medications by entering the patient and their access information into the machine. CareLink has beenis implemented into these machines also, which is convenient for pharmacy approval and records collection. A visual display of the process can be seen in figure 1 and a more detailed flow chart is in Appendix B. Figure 1 Process Flow Chart Order is delivered Nurse picks up order Prescription is entered into UM-CareLink Pharmacist verifies order Pharmacy Technician builds order Medication is pulled from Omni-Cell Medication is administered to patient Order sent via pneumatic tube This new process, using CareLink, has decreased prescription turnaround time significantly from the paper system, as confirmedseen byin prpreviousior studies. Prior to CareLink, the physicians wrote the prescriptions on paper slips that were placed in outgoing prescription bins. These bins were picked up once every hour. The pharmacist read the handwriting on the slip, which could be difficult sometimes, and then filled the prescription. This area of the process is mainly where CareLink has made the most impact. The rest of the process has remained fairly the same. CareLink has also improved the accuracy, record keeping, and completeness of the prescription order system. The order data can now be accessed easily through CareLink. Goals and Objectives The goals and objectives of this project were to make recommendations for improvement in the CareLink prescription ordering system tohat will: 7

8 Improve STAT/NOW order compliance Decrease process variations between satellite pharmacies Implement best practices established in studies Continue process improvement by providing a means to analyze system failures Project Scope This project has focused on the TAT of STAT/NOW medication prescriptions processed by the satellite pharmacies in the University Hospital (UH), Mott Children s Hospital (Mott), and the Cardiovascular Center (CVC). Three satellite pharmacies are located in UH, one in CVC, and one in Mott. Specifically, the project has focused on the effectiveness of CareLink and identified and recommended how to correct any discrepancies between pharmacies. The project data and analysis consists of information for the 24-hour day, including weekends. Prescriptions not categorized as STAT or NOW have not been included in this project. Furthermore, the central pharmacy (B2), the soon-to-open emergency room pharmacy, and the operating room pharmacies are not included in the projects scope. Support Provided from Operating Entities The following operating entities supplied the team with the needed project information for the project: Brian Callahan, PharmD, Assistant Director of Inpatient Pharmacy Services - Proper contact info within pharmacy department - Observational study capabilities - Enforcement of staff participation - Sending s to participants - Weekly update meetings Bobby Beasley, Project Coordinator, Management Engineer Fellow, POA - Weekly update meetings - Project Guidance Chris Zimmerman, PharmD - Historical Data CareLink spreadsheet for July-September - Semi-live order data on a daily basis via - Unique names of nurses and pharmacists Pete Link, UMHS MCIT Clinical Support - Semi-live floorstock data on daily basis via Katie Coldren, Nurse Manger - Cooperation of nursing staff with semi-live data study 8

9 Methodology The primary purpose of the project has been to examine and analyze the medication order process, specifically the TAT for the satellite pharmacies within UMHS. The primary personnel involved wasare UMHSCareLink pharmacists, pharmacy technicians, and nursing staff. To achieve the goals and objectives stated in this document, the team has studied the process in three phases: process exploration, historic data analysis, and semi-live data analysis. Below is a detailed explanation of the phases. Phase 1 Process Exploration The team has documented the current state of CareLink in the satellite pharmacies within UH, CVC, and MOTT using the following tools: Staff surveys The team developed and delivered to the manager of inpatient pharmacy services an online survey using zoomerang.com. The manager dispersed the survey link via to pharmacists and pharmacy technicians in each satellite pharmacy. Pharmacists and pharmacy technicians were given a week to complete the online survey. The purpose of the surveys was to discover the following: The response differences between pharmacies GThe general feeling towards CareLink and its implementation The llevel of workload The llevel of STAT/NOW order efficiency The time definition difference between STAT and NOW orders The uutilization of Omni-Cell machines Process mapping Team members spent time in each pharmacy observing the process and developing a flow chart analysis of the overall process. Below are some of the main process components the team observed: - Utilization of Omni-Cell machines - Medication order delivery methods - Task differences between pharmacists, technicians and nurses Through observations and data collected in phase 3, the team has also created a value stream map for the order process (Appendix C). The purpose of this is to provide a high level overview of the processes current state. The value stream map illustrates processes that are essential to complete any prescription order. The culmination of these essential times, or process time, is indicated on the top tier of the time series at the bottom of the figure. The other times indicated in this figure, on the lower tiers, are not essential to complete the prescription order, and are designated as non-value added. NThis non-value added times are typically waiting times between processes and are significant sources of delay in the overall turnaround time. The team 9

10 will focused to reduce these non-value added times in order to improve the overall turnaround time. Phase 2 Historic data analysis The team has reviewed CareLink data from July and September, Data was provided in the form of a single Excel spreadsheet. Database fields included: Order placement, verification, and administration date & times Hospital and nursing unit Order priority (i.e. STAT/ NOW) A complete list of the nursing units studied, categorized by related pharmacy can be found in Appendix D. The team used this list, along with the particular nursing unitstaff responsible order, to deductively determine which pharmacy was responsible for the order. The team has developed an Excel macro to format the data for analysis. This process proved to be time consuming because algorithms for determining which pharmacy processed each order had to be developed. Once this was complete, the team used regression analysis, general linear models, and other statistical tools, to analyze and quantify the collected data. These findings have guided subsequent investigations and recommendations for process improvement. Phase 3-Semi Live Data Analysis The project s initial goal was to conduct a live-study of CareLink using a computer program. The program would have issued a computerized notification to the pharmacist when an order was deemed late. Unfortunately, access to the CareLink database was unavailable. Therefore the team developed a semi-live study. This study and its two parts, delivery study and 24-hour CareLink study, are detailed below. Delivery Study: While analyzing past reports, the team found that the time and method of medication departure from the pharmacy was unknown. The following diagram depicts the unknown information: Figure 2 Unavailable Data Flow Chart 10

11 The team collected time data on when and how prescriptions leave the pharmacy on a daily bases for 1 week (11/9-11/15). A delivery log-sheet was placed in each satellite pharmacy. The pharmacists were required to input the delivery time and method of each STAT/NOW order during the day. The following table is an example of the delivery-log sheet. Table 2 Log Sheet Example Example Formatted: Left, Indent: Left: 0" The data that was logged was then correlated with the data that was being sent daily. From this, the team determined the unavailable data times of medication departure from the pharmacy. 24 Hour CareLink Data Study: The team received daily CareLink data through . The data was analyzed and STAT outliers were determined. The team decided that any order over 150 minutes required examination. Each day of the study returned an average of 25 to 30 of these tardy orders. Then s were sent to the nurse and pharmacist responsible for administration and verification for the late order. The asked for information regarding reasons why the order was delayed. To complete this study the following was done: Determine outliers based on historical data Analyze data each day for 1 week (11/9-11/15) nurses and pharmacist to determine cause of administration delay Compile responses and analyze 11

12 Analysis and Results The team gathered information through the methods previously discussed. The detailed analysis and results from each of these phases is shown below. Phase 1 Process Exploration Staff surveys SThe staff survey results indicated at least 75% of pharmacists and pharmacy technicians surveyed have been employed by UMHS prior to CareLink implementation and they feel that CareLink has had a positive impact on the TAT of STAT orders. Pharmacists and pharmacy technicians estimated that half of all STAT orders leave the pharmacy within 15 minutes of order. Pharmacists and pharmacy technicians did not agree with the consistency of STAT orders throughout a shift. Both estimated that the frequency of STAT orders was not consistent, but it was not sporadic; pharmacists estimated that the frequency was more consistent. Pharmacists and pharmacy technicians feel that they have an overwhelming amount of work less than half the time; pharmacy technicians feel overwhelmed more often than pharmacists with their workload. To improve the TAT of STAT orders, several pharmacists and pharmacy technicians suggested the following: Increase the number of available pneumatic tubes available to send orders Identify medication orders as STAT on the printed pharmacy label Increase the number pharmacy technicians Flow Chart The flow chart is in Appendix B. This flow chart gives a general visual description of the medication order process. By looking at the flow chart, most of the tasks required to administer medication occur in the pharmacy, which is the pink area. From the research conducted in previous studies of Mott, the pharmacy accounts for the smallestlargest time allocation in the medication order todelivery and a administration process time, as shown in thewhich is the blue area. Although this flow chart gives a good representation of the process, there can be variation. For example, the delivery method may differ depending on the situation. In addition, tthere is no standardization for delivery method, it is subjective. A value stream map can be seen in Appendix D. Literature Search A previous study of CareLink STAT TAT stat was conducted in 2007 strictly on the Mott pharmacy. The team has referenced the following data from the report: Table 1 Literature Study Times Task Average Time 12

13 (Min) Order Placement 3.5 Pharmacist Notification 1.8 Order Verification 4.7 Delivery and Administration of Medication 49 A graphical representation of this data can be seen in Appendix A. The average time to deliver and administer the medication was 49 minutes, which is high. This specific piece of information persuaded the team to design a study in order to understand why it is high. This study iswill be described in phase three of the methodology section. Phase 2 Historic data analysis To gain a better understanding of current medication TAT performance the team requested CareLink & WORx data for a 90 day period. Initial review of the data revealed that approximately 11.5% of all STAT/NOW orders have incomplete data in the CareLink system. The team investigated these orders further and determined the vast majority were Discontinued before being administered to the patient. The team also found that approximately 5.75% of all STAT/NOW orders were recorded as being administered prior to pharmacist verification. In some life threatening situations it is possible that these medications were administered prior to verification, and were not recording errors on the nurses parts. To understand the nature of these orders the team requested a list of prescriptions from the client that are generally considered time critical, and urgently needed by patients (See Appendix G1) These medications equate to about 14% of all STAT/NOW orders, but only 6% of the orders administered prior to verification, as seen in Figure 3. For this reason the team concluded that the orders recorded as being administered prior to verification were errors on the nurses parts, and should be excluded from analysis. Figure 3 - Useable Data from all STAT/NOW Orders July September

14 All STAT/NOW orders recorded in CareLink from July-September 2008 Number of observations ~31,000 A relative frequency histogram was created for TAT in the useable date. As seen in Figure 4, these medication TATs are approximately exponentially distributed. Figure 4-Orders June-September 2008 Exponentially Distributed Mean N Frequency Turn around Time (Minutes) STAT/NOW orders recorded in CareLink from July-September 2008 EXCLUDING orders o With missing data in CareLink (n= 3609) o For operating rooms, or from operating room pharmacies o Recorded as being administered to the patient prior to pharmacist verification (n=1784) Number of observations ~26,000 14

15 Likewise the team created a relative frequency histogram of the amount of time pharmacists take to verify orders and noticed a similar distribution. Comparing this information to the observed mean and standard deviation shown in Table 3, it can be noticed that the exponential distribution is only an estimate of the Turn-Around/ Verification time data since the means are not equal to the standard deviations. Table 3- Descriptive Statistics of June-September 2009 Data Observed Statistic TAT [min] Verification Time[min] Mean Median Standard Deviation STAT/NOW orders recorded in CareLink from July-September 2008 EXCLUDING orders o With missing data in CareLink system (n= 3609) o For operating rooms, or from operating room pharmacies o Recorded as being administered to the patient prior to pharmacist verification (n=1784) Number of observations ~26,000 According to Michigan Health System standards, STAT orders are supposed to be administered within 20 minutes, and NOW orders within 90 minutes. With this in mind, 33% of all STAT orders, and 75% of NOW orders are administered on time. STAT orders make up 28% (NOW orders the remaining portion), therefore only around 63% of all medications are administered on time. Without a doubt, process improvements are needed for the Health Systems standards to become a reality. Before analyzing the data further, the team noticed that many outliers existed in the June- September data. In fact, some medications were recorded as taking up to 24 hours to administer. Using the statistical estimates developed above, the team determined that 3.5% of the data was more the 3 standard deviations from the mean. Such observations had TAT s in excess of 5 hours. From conversations with hospital staff the following were deemed as common reasons for such delays: The patient was in a procedure or unavailable The medication was intended to be taken with food, or prior to some event The medication was ordered as part of a regimen, and for use later in the day The patient was unable to take the medication at the time due to possible drug complications While multiple other reasons may exist for these excessive delays, the team and client are of the opinion that such orders are not truly representative of the STAT/NOW order population. As such, all observations with TAT s greater than 5 hours have been excluded from further analysis. In addition, the team noticed that several nursing units had very few observations (less than 30), and that there presents in the data would cause complications with certain statistical analyses. To abate these difficulties, the team decided to remove these nursing units from the data. Of 15

16 note, the total number of observations eliminated as a result of this decision consisted of much less than 1% of the total number of observations. For reference, a list of the excluded nursing units can be found in Appendix G2. From process observations the team quickly identified the order priority (STAT or NOW), and the medication location (Omni-Cell or Pharmacy) as potential sources of variation in TAT stat. Medications available in the Omni-Cell machines are referred to as Floor Stock items, and medications only available through the pharmacy are referred to as Dispensed items. Figure 5 - Floor Stock Medications Have Shorter TAT STAT/NOW orders recorded in CareLink from July-September 2008 EXCLUDING orders o With TAT greater than 5 hours (n= 833) o With missing data in CareLink system (n= 3609) o From nursing units with less than 30 orders (n=184) o For operating rooms, or from operating room pharmacies o Recorded as being administered to the patient prior to pharmacist verification (n=1784) Number of observations ~25,000 As is seen in the figure 5above, Floor Stock medications have substantially lower TAT s. In addition, STAT orders have shorter TAT s for both floor stock and dispensed medications. There is a 17% difference between NOW/STAT orders for Dispensed medications, and a 6% difference for NOW/STAT orders of Floor Stock medications. The team attributes these discrepancies to the fact that Floor Stock medications are available for administration much more quickly (immediately after verification) than Dispensed medications. Next the team examined the individual pharmacies as potential sources of variation. As seen in the figure below CVC & Mott have the shortest TAT s. In addition, the CVC pharmacy have the longest verification time, due impart to the fact that it only has one pharmacist of staff. 16

17 Figure 6 - CVC & Mott have the shortest TAT stat STAT/NOW orders recorded in CareLink from July-September 2008 EXCLUDING orders o With TAT greater than 5 hours (n= 833) o With missing data in CareLink system (n= 3609) o From nursing units with less than 30 orders (n=184) o For operating rooms, or from operating room pharmacies o Recorded as being administered to the patient prior to pharmacist verification (n=1784) Number of observations ~25,000 A one-way analysis of variance model (ANOVA) confirms that at least one pharmacy has a significantly different average TAT s (P-Value = 0.00). Construction of 95% confidence intervals for the average TAT s by pharmacies results in the conclusion that all pharmacies are statistically different from one another. In addition, a similar test was conducted to determine if the order verification time was different between pharmacies. As with the TAT s, all pharmacies have significantly different verification times. The results of these tests are located the Appendix G3 & G4. Formatted: Font color: Auto Figure 7 - Percentage of Orders by Satellite Pharmacy 17

18 STAT/NOW orders recorded in CareLink from July- September 2008 EXCLUDING orders o With TAT greater than 5 hours (n= 833) o With missing data in CareLink system (n= 3609) o From nursing units with less than 30 orders (n=184) o For operating rooms, or from operating room pharmacies o Recorded as being administered to the patient prior to pharmacist verification (n=1784) Number of observations ~25,000 As seen in the figures 6 & 7, the 6 th floor pharmacy handles almost half of all orders at the University Hospitals, while the slower 8 th and 5 th floor pharmacies account for less than 30% of all orders. The team was interested in understanding the reasons for these differences, but realized that some unaccounted for factors may be represented in the data. Therefore, the flowingfollowing list of reasons for differences between pharmacies was developed: The pharmacy hours of operation The number of staff members work in a pharmacy The nursing units serviced by a pharmacy For example, it was mentioned by the client that the 8 th floor pharmacy is responsible for chemotherapy nursing units, and that these units require much more time and attention. Such a special cause explains why the 8 th floor pharmacy has the longest TAT & longer verification times. The figure below shows TAT s by the hour of day as well as the number of orders during those hours. From this figure it is concluded that TAT s are longest during the 7am to 12pm period of the day, and shortest in the early morning. Figure 8 -Turn Aaround Times are Longest ffrom 7am to 12 pm 18

19 STAT/NOW orders recorded in CareLink from July-September 2008 EXCLUDING orders o With TAT greater than 5 hours (n= 833) o With missing data in CareLink system (n= 3609) o From nursing units with less than 30 orders (n=184) o For operating rooms, or from operating room pharmacies o Recorded as being administered to the patient prior to pharmacist verification (n=1784) Number of observations ~25,000 Given that the 5 th and 8 th floor pharmacies are normally closed after 12pm, it is understandable why the TAT s of these pharmacies are longer, because they do not operate during the early morning. Although this explanation is suitable for the 5 th & 8 th floor pharmacies, it does not explain why the CVC pharmacy had the second fastest TAT in spite of the fact that it is not open during the early morning. Figure 9 -- CVC has the Most Floor Stock Orders 19

20 STAT/NOW orders recorded in CareLink from July-September 2008 EXCLUDING orders o With TAT greater than 5 hours (n= 833) o With missing data in CareLink system (n= 3609) o From nursing units with less than 30 orders (n=184) o For operating rooms, or from operating room pharmacies o Recorded as being administered to the patient prior to pharmacist verification (n=1784) Number of observations ~25,000 Using the information in figure 9, it can be seen that the CVC pharmacy has the largest percentage of STAT/NOW orders administered from Floor Stock (Omni-Cell) inventory. Given that floor stock items have substantially shorter TAT s, the average TAT s of the CVC pharmacy can be reasonable expected to be lower. However, the question of why the Mott pharmacy had the shortest TAT, and least number of Floor-Stock items was also asked. From observations the team noticed that the Mott pharmacy had staffing similar to that of the UH 6 th Floor pharmacy, but processed less than half the orders. The client informed the team that many medications for children are administered in liquid form, and are prescribed as a percentage of the patient s weight (example milligrams of medication per kilogram). As a result, additional staffing is required at the Mott pharmacy to prepare medications in a reasonable amount of time. The team next studied the effect of nursing units on TAT. It was found that the PICU, PCTU & CVC4 units had significantly shorter TAT s (<40 minutes), while UB1C & BICU had significantly longer TAT s (>85 minutes). The PCTU is a pediatric intensive care unit located near the Mott pharmacy, thus explaining is shorter TAT s. In addition the PTCU, has a dedicated pharmacy technician assigned to it. The team suspects that CVC4 has shorter TAT s, because of the large amount of medications available via the Omni-Ccells. UB1C & BICU unitsnites are located in the Emergency Department & Trauma Burn Center. In November of 2008 a new pharmacy was opened to help provide better service to the emergency department, but as of this time the Trauma Burn Center is still serviced by the 6 th floor pharmacy. Interviews with pharmacy staff revealed that medications are often pre-ordered in anticipation of an incoming patient in the Trauma Burn Center, therefore explaining delays in medication administration. For reference the statistical model used to statistically reach these conclusions can be found in Appendix I. When the CVC, UH 5 th Floor, and UH 8 th floor pharmacies are closed, all orders from UH & CVC hospitals become the responsibility of the UH 6 th pharmacy. The client mentioned that concern existed among hospital staff members regarding the capabilities of the 6 th Floor ppharmacy at times when it is responsible for the other pharmacies orders. As seen in figure 10, TAT stat decreased in all cases when the 6 th floor pharmacy was responsible for other pharmacies orders. Figure 10 UH 6 th Floor Pharmacy Capability during off hours 20

21 STAT/NOW orders recorded in CareLink from July-September 2008 EXCLUDING orders o With TAT greater than 5 hours (n= 833) o With missing data in CareLink system (n= 3609) o From nursing units with less than 30 orders (n=184) o For operating rooms, or from operating room pharmacies o Recorded as being administered to the patient prior to pharmacist verification (n=1784) Number of observations ~25,000 Finally, the client requested that the team correlate the list of time critical prescriptions (Table 4) with the June- September data. These orders are commonly referred to as Super-Stat by pharmacy & nursing staff due to their urgency. As seen in the Figurecart 11below, all of the Super-Stat orders have higher than average TAT stat when compared to the general population of STAT/NOW orders. 21

22 Table 4- Time Critical Medications (a/k/a Super-Stat ) Super-Stat Medications Diltiazem Magnesium sulfate Dobutamine infusion Metoprolol Dopamine infusion Nitroglycerine infusion Epinephrine Norepinephrine infusion Esmolol Phytonadione Factor IX Potassium phosphate Factor VII Protamine Fosphenytoin Sodium phosphate Gentamicin Sodium polystyrene sulfonate Hydralazine Tobramycin Labetolol Vancomycin Figure 11 - Time Critical Medications have Longer TAT STAT/NOW orders recorded in CareLink from July-September 2008 EXCLUDING orders with medications not listed in Table 4 Number of observations ~3500 The client noted that many of the prescriptions listed in Table 4 must be dispensed forom the pharmacy, and can take a fair amount of time to make. However, the team is unsure of why the Floor-Stock, Super-Stat medications are tookaking longer to administer than that of the general population of STAT/NOW orders. Phase 3-Semi Live Data Analysis 22

23 The semi-live data analysis includes the delivery study and the 24-hour CareLink study. These studies provide information of previously unavailable data. Detailed analysis of these studies is shown below. Delivery Study From entering an order into CareLink to being picked up at the pharmacy, medications tookhere was an overall averagemean time of 17.3 minutes (95% confidence interval, 14.39min min). The data suggests that the orders are leaving the pharmacy with a maximum average time of 20 minutes. A box-plot of this data can be viewed in Figure F1 of Appendix F. Figure F2 in Appendix F shows a box plot of the mean time the order takes to leave for each pharmacy. The data shows that Mott has the shortest mean time with and CVC has the longest. A histogram (Figure F3 in Appendix F) was made to represent the number of observations that were recorded for each of the four delivery methods. The largest recorded method was the pneumatic tubes and the smallest was nurse pick up, which means that a majority of orders leave the pharmacy via pneumatic tubes. Means and confidence intervals for the time to leave the pharmacy were calculated for each delivery method as shown in Appendix F, tables F1-F4. These means that the faster times were delivered via pneumatic tubes and the longest times were via technician delivery. 24-hour CareLink Study Figure 12 shows the results of pharmacist and nurses as sources for the late TAT. These sources are categorized into six general sources. Floor -Sstock - Accounted for 24.4% of late orders. Prescription orders that the medication is located in an Omni Ccell. Prescriptions located in an Omni Ccell and administered late indicate a lack of communication between the pharmacist and the nurse. Medication complications - Accounted for 11.5% of late orders. Sources of delay for medication complications included patient allergic to prescriptions and prescriptions delivered intravenously (IV), but the patient does not have an IV accessible Within Pharmacy - Accounted for 10.3% of late orders. Sources of delay within a pharmacy included scheduling conflicts, lunch breaks, and mishandling orders between two different shifts. Pneumatic tubes - Accounted for 7.7% of late orders. Pharmacists noted that the order was transported via pneumatic tubes and complications with the tubes, or retrieving the order from the tube was the main source of tardiness. Unknown - Accounted for 24.4% of late orders. The pharmacist or nurse could not recall any information concerning the particular order. Miscellaneous - Accounted for 21.8% of late orders. Other sources of delay with little or no repeated instances. 23

24 Figure 12 Late Orders Pie Charts Number of observations ~80 Data collected from November 10 th -14 th, 2008 The results of this study reveal that a lack of communication between pharmacists and nurses is responsible for tardy STAT orders. Of the six sources of delay, floorstock (24.4%), within pharmacy (10.3%), and divisions of both unknown (24.4%) and miscellaneous (21.8%) are a direct result of poor communication. Tardy prescriptions located in an Omni Ccell (floor stock) are verified on-time, however the nurse doesn t know that the medication is located in the omnicell or is unaware of the order altogether. Implementing a simple page program to notify the appropriate nurse of an order, and it s availability in an Omni Ccell would virtually eliminate this source of delay. This implementation would result in the elimination of nearly a quarter of the observed late STAT orders administered within the defined time interval. 24

25 Conclusions and Recommendations Conclusions The projectteam determined that the average TAT for STAT/NOW orders was 77 minutes with 63% of all these orders being administered on time. Variation in TAT stat existed across pharmacies, order priorities, nursing units, delivery methods, and times of the day. However, such factors explained only about 15% of the TAT variation, and the remaining portion was inherent to the process. From data collected in phase 3 of the study, it was found that 17.5 minutes is the average amount of time taken froorm when a prescription is ordered to when it exits the pharmacy. Results from the semi-live data analysis suggested that at least 50% of delays in medication administration can be attributed to poor electronic communications between the nurses and pharmacies. Often nurses claimed to be unaware of new orders because they were not present when the orders were placed, or because they had yet to review their patient s medication administration records. Nurses also noted that they were frequently unaware of how their orders were delivered from the pharmacies. Such issues are to be expected given that no form of electronic indication is given to the nurses when orders are complete, or of how they were delivered. Communications within satellite pharmacies also showed room for improvement. Survey responses indicated that pharmacy technicians are not always aware of STAT/NOW orders. In fact, one particular late order during the 24-hour study occurred because the pharmacy technician was unaware of the orders priority. Currently, there is only verbal communication between the pharmacist and the technician as to the priority of orders. Finally, many pharmacists and pharmacy technicians are not well informed on UMHS standards for STAT/NOW orders. Survey responses indicated 30% of pharmacists and 40% of pharmacy technicians doid could not know thatcorrectly identify that a STAT orders needds to be administered within 20 minutes. Additionally, 74% of pharmacists and 90% of pharmacy technicians incorrectly identified the time requirements under UMHS policy for NOW orders. Recommendations The team developed the following recommendations to reduce TAT. Provide a method to inform nurses when & how a medication has left the pharmacy Send alerts to nurses and pharmacists when medications have not been administered on time. Educate hospital staff on time standards for STAT/NOW orders Advance towards a mostly Omni-Cell model similar to that of the CVC 25

26 Provide STAT/NOW notification on labels printed to pharmacy technicians Communication between the pharmacy and nurse units must improve to reduce TAT. Providing a method to inform the nursing staff vital information for STAT/NOW orders will significantly reduce sources of late TAT. From the 24 hour study, floor stock and pneumatic tube orders were late because the nurse did not know to retrieve the medication. These orders accounted for 32% of the total sources of late TAT. From the same study, pharmacists and nurses could not recall any information for 24% of all late STAT/NOW orders. These sources of tardiness couldwill be nearly eliminated with an automated alert systemmost affect by increased communication between the pharmacy and nursing staff. Improving the communication between these two is critical. Communication within the pharmacy can be improved by physically differentiating labels for STAT/NOW orders. The team recommends a specific label for each STAT/NOW order and that each satellite pharmacy utilizes a separate printer for STAT/NOW orders. Educating hospital staff on UMHS time policies for STAT/NOW orders will improve quality within the hospital. Educated and conscientious employees will direct their attention to orders they know are behind or at risk of becoming late. Total TAT was the lowest for medication located within an Omni-Cell. A There is great potential within UMHS to exists togreatly reduce facility TAT by implementing Omni-Cells wherever applicable. Although this has a high overhead cost, there is possibility for return on the investment. Improving the communication within each satellite pharmacies will reduce total TAT. Implementing a separate label specific to STAT/NOW orders will allow the pharmacy technicians to clearly identify and process the priority of current orders. 26

27 Appendix A- Literature Search Data Average Time Task (Min) Order Placement 3.5 Pharmacist Notification 1.8 Order Verification 4.7 Delivery and Administration of Medication 49 27

28 Appendix B Flow Chart Detailed Process Flow Chart of UH, Mott, and CVC Pharmacies: Yes Drug type and dosage of orders are verified by pharmacist (order verification) Is the order in Omni-Cell machine? No Prescription Label prints in Pharmacy Physician enters order Into CareLink Pharmacist runs a Batch Review/ Query In CareLink Is the medication a STAT/NOW order? Order waits in queue until STAT/ NOW orders are verified Yes Yes Is the medication prepacked? No No Technician gathers dosage Technician makes medication Nurse chooses medication order from patient s profile in Omni- Cell Pharmacist verifies medication against label = Preformed in Pharmacy Nurse removes medication from Omni-Cell machine Pharmacist decides on medication delivery type = Preformed in Patients Room = Preformed at Omni-Cell Machine Nurse picks up medication Medication is sent through pneumatic tube delivery system Technician delivers Medication to unit immediatly Medication is delivered by Technician on hourly run Nurse administers Medication 28

29 Appendix C: High Level Value Stream Map 29

30 Appendix D: Nursing Units by Pharmacies Pharmacy Hours UH 6 th Floor- 24/7 UH 5 th Floor UH 8 th Floor (M-F) and (S Su) CVC Mott 24/7 Mott UH 6 th Floor UH5 th Floor CVC UH 8 th Floor 4WEB 1A 4A CATH 8A 4WEM 1OBS 4AS CPU 8B 4WWB 6A 4B CPU 8C 4WWM 6AC 4BC CV2A 8D 5E 6AR 4CB1 CVC IR 7 9C 5EMC 6B 4CBI CVC IR 9 9D 5W2 6C 4CI CVC PACU A 6AP 6D 4CNU CVC PACU B 6M 7A 4DN CVC2A 7E 7A1 4DNI CVC4 7M 7B 4DS CVC5 7WEB 7C 5A 7WEM 7DN 5B BMU 8DS7 5C DIS AOBS 5D FADM B1IP 8D5 MB1C BAC 8DB1 MC2 BICU MPUA UADM MPUP UB1C NADM NICU NOBS PARU PCAT PCIU PCTU PDC PDI PICU PXRA 30

31 Appendix E: Operational Definitions Turn-around-Time (TAT) PRN o Total elapsed time between when a Rx order is placed by a physician and when the Rx is administered to the patient o Physician orders the medication As needed Order Verification o Pharmacists OK s Rx to be dispensed to a patient Floor Stock Prescription o Rx stocked in Omni-Cell machines, available immediately after verification Dispensed Prescription o Rx s available through satellite pharmacies only, made after verification STAT/NOW o No real difference (For pharmacists) Super-Stat o Time critical prescription, considered urgently needed 31

32 Appendix F: Phase 3 log sheet data Figure F1 400 Boxplot of TimeToMedPickup 300 TimeToMedPickup Hourly Run Nurse Pickup Pneumatic Tube Method Technician Delivery Figure F2 400 TimeToMedPickup [min] CVC Mott UH 5th Floor UH 6th Floor UH 8th Floor 32

33 Figure F # of Observations Hourly Run Nurse Pickup Pneumatic Tube Technician Delivery Tables F1 One-Sample T: TimeToMedPickup_ PHEUMATIC TUBE Variable N Mean StDev SE Mean 95% CI TimeToMedPickup (11.989, ) Table F2 One-Sample T: TimeToMedPickup_NURSE PICK UP Variable N Mean StDev SE Mean 95% CI TimeToMedPickup (10.76, 24.76) Table F3 One-Sample T: TimeToMedPickup_Tech Delivery Variable N Mean StDev SE Mean 95% CI TimeToMedPickup (7.25, 40.57) Table F4 One-Sample T: TimeToMedPickup Variable N Mean StDev SE Mean 95% CI TimeToMedPickup (14.39, 20.19) 33

34 Appendix G Historic Data Analysis Table G1 Super STAT Prescriptions Super-STAT Medications Diltiazem Magnesium sulfate Dobutamine infusion Metoprolol Dopamine infusion Nitroglycerine infusion Epinephrine Norepinephrine infusion Esmolol Phytonadione Factor IX Potassium phosphate Factor VII Protamine Fosphenytoin Sodium phosphate Gentamicin Sodium polystyrene sulfonate Hydralazine Tobramycin Labetolol Vancomycin Table G2 - Nursing Units excluded from analysis for having too few observations Nursing Unit # of Observation 1A 5 4WEB 5 4WWB 21 6AP 11 7WEB 1 7WEM 11 CATH 12 CV2A 15 CVC IR 7 1 CVC IR 9 4 CVC PACU A 17 FADM 2 MB1C 13 MPUA 5 NADM 6 PARU 29 UADM 26 34

35 Figure G3 - ANOVA TAT by Pharmacy Note: The TAT data was transformed in order to meet the assumptions of a 1way-ANOVA model. Figure G4 - ANOVA Verification Time by Pharmacy 35

36 Appendix H - ANOVA TAT by Nursing unit Note: The TAT data was transformed in order to meet the assumptions of a 1way-ANOVA model 36