HEALTHCARE

Hospital Reduces Using DMAIC

by Yani Benitez, Leslie Forrester, Carolyn Hurst and Debra Turpin

rying to improve quality by eliminating process steps might appear to be a contradiction. But this was not the case when a 200-plus full service hospital in Illinois targeted medication error reduction as a 2005 goal. In fact, the Joint Commission on Accreditation of

Healthcare Organizations (JCAHO) accredited Alton Memorial Hospital achieved a more than 90% improvement in process performance as a result of the project.

Background
Alton Memorial, part of the St. Louis based BJC HealthCare, naturally supports a culture of patient safety as part of its mission. The Agency for Healthcare Research & Quality estimates the incident rate for an adverse drug event (ADE) at between two and seven per 100 hospital admissions nationwide, with a mean cost of $4,685 per event.1 At only 0.02%, Alton Memorials medication error rate was low compared to those nationwide statistics. But the potential cost savings to the institution that would result from even a small improvement in ADEs could not be ignored. Of those medication errors reported in the hospitals risk management database, 43% were caused by transcription errors. Transcription involves the copying of a physician order for the purpose of processing. The strict guidelines that define a transcription error are based on the National Coordinating Council for Medication Error Reporting and Prevention (NCC MERP) error category index.2

In 50 Words Or Less
An Illinois hospital reduced work while improving process performance 90% during a medication error reduction project. A Black Belt helped a multidisciplinary team use Six Sigmas define, measure, analyze, improve and control methodology. The team then used quality function deployment to design and develop the process functions.

38

I JANUARY 2007 I www.asq.org

Medication Errors and QFD

Information copied incorrectly or omitted is considered a transcription error. This includes missing, inaccurate or only partially provided special instructions with a physician order. An example of a special instruction is: Give only if pain level greater than five one hour after Tylenol. A multidisciplinary team was formed at the hospital to find ways to reduce the hospitals medication errors. Team members included representatives from pharmacy, nursing, clinical information systems, nursing management, performance improvement and medication safety. Hospital management requested the help of a Six Sigma Black Belt (BB) to guide the team through the process. A BB was assigned to this team in a facilitator and trainer role. (EU), because of the immediate nature of its needs. With these two goals, the team expected to improve order entry accuracy related to transcription errors by 50%. In the measure phase, the team flowcharted the current medication order entry processes. The team quickly realized each unit had developed its own medication ordering process. Most used two documents to review each order: the medication administration record (MAR) and a chronological sheet, a tool devised and used only by nursing staff. The intent of the chronological sheet was twofold: Provide a historical list of medications received throughout a patient stay. Act as a repeat check against the MAR each shift. The purpose of this double check system was to help ensure order accuracy. Although the MAR information should have been verified against the physician orders, the staff deemed this step too cumbersome and labor intensive and created the chronological sheet for nurses to rewrite every medication order by hand.

Methodology
Initially, the team followed a define, measure, analyze, improve and control (DMAIC) methodology. In the define phase the team realized hospital management mandated two distinct goals: Reduce the defect rate of the current process with quick hit initiatives. Develop a standardized process that works for all hospital units except the emergency unit

QUALITY PROGRESS

I JANUARY 2007 I 39

HEALTHCARE

Eliminating a Document
It became clear that data needed to be collected to compare the accuracy of the MAR and the chronological sheet. A small baseline study was performed in the third quarter of 2004. The pharmacists order entry accuracy in the MAR was 95% the first time through, yet once the order was verified by the pharmacist, the pharmacy software and the nursing staff, the MAR accuracy increased to 99.98%. On the other hand, the chronological sheet was itself fairly inaccurate (10% error rate). Unfortunately, nurses had developed a false sense of security with

These changes improved the medication order entry process, yet the issue of the multiple ways staff processed medication errors remained.

the chronological sheets, assuming they were correct. Furthermore, this tool was used during the entire patient stay, so mistakes could carry through multiple days. Concurrently, interpretations of orders, abbreviations and the definition of the policy guidelines had differences between the two forms. These variances were contributing to the frequency and type of transcription errors. The team concluded the chronological sheet had to be retired because it added no value to medication order entry accuracy. Eliminating it represented an average seven minute workload reduction per patient per day.

orders to be sent to pharmacy before 6 a.m. daily so they could be prepared early. Establishing a missing medication sheet for the clinical units to fax instead of phoning their questions and concerns to the pharmacy. In the pharmacy, a technician would review the request and assess whether it could be handled by someone other than a pharmacist. After completing the order entry for a given patient, the pharmacist then would address the waiting requests prior to starting the order entry for the next patient. Training pharmacists on a consistent medication order entry format. Another cause of transcription errors was the illegibility of the physician order. There was not enough space on the form for physicians to write medication orders, forcing them to write outside the assigned box or in tiny script. This caused multiple issues such as illegible orders, eliminated words and missing information. Ten more lines were added to the form. Margins were darkened to reduce the possibility of writing outside the faxable area. The guidelines inside the writing area were made to be invisible when faxed to pharmacy. Finally, fax machines were tested and adjusted to improve quality. These changes improved the medication order entry process, yet the issue of the multiple ways staff processed medication errors remained.

Standardization
The team shifted its focus to the second goal: designing one standard medication order process for all hospital units to use, except the EU. The new process had to function flawlessly in such diverse areas as critical care, where patient medications change constantly, and long-term care where patients tend to have a long list of medications. To achieve this goal, the team decided to follow a design for Six Sigma methodology, quality function deployment (QFD), to link the needs of the customers with the design and development of the process functions. QFD helps organizations identify both spoken and unspoken needs, translate these into actions and designs, and focus various business functions toward achieving this common goal. The first question to answer was: Who is our customer in the medication order entry process?

Reducing Pharmacist Interruptions

By analyzing cases, the team identified disruptions to the pharmacist during the order entry process as a repeating cause of mistakes. Actions taken to reduce disruptions included: Establishing a process for all intravenous fluid
40

I JANUARY 2007 I www.asq.org

TABLE 1 Hierarchy of Nurses Needs Even though the ultimate customer is the patient, the primary Nurses need Defined as: Hierarchy and more immediate customer was determined to be the nurse in Available immediately for printing (less than one minute). 2% Quick access charge of the patient. Less than three steps to access the document. 2% The team then set about cap38% Quick pharmacy Less than two hours. turing the voice of the customer turnaround time Fax of order available in real time. 7% (VOC), getting insights into the Able to see all medications for the entire patient stay. 2% Provide history nurses needs through interviews. of meds Able to see medication start and stop dates. 2% The top needs identified were: Portable/mobile Nurse should not need to go back to station for document. 2% Quick access to medication Serve to double check all orders in medication order information. It should 12% Check on MAR administration record (MAR). take no more than one Medications sort in the same way as MAR. 7% minute and three steps to access a patients medication At least 95% accurate the first time. 19% Trustworthy information. Consistency in order entry by all pharmacists. 7% Quick pharmacy turnMAR = medicaton administration record. around time. Orders should be entered into the MAR system within 120 minutes. A Because each unit followed a different order entry real-time copy of the order should be available process, its evaluation of the current process perwhile the order is being entered into the MAR formance was relative to its own application of it. system. The nurses needs then were converted into The process must provide a history of paquantitative measures of success with established tient medications. Nurses should be able to metrics and targets. These would become the see all medication orders, including start and design requirements for the new process. Each stop times for the entire patient stay. design requirement was rated against the customer Information should be portable and mobile. (nurse) needs in terms of the strength of its correlaNurses should not need to return to the nurstion. The resulting QFD matrix is shown in Figure 1. ing station to learn what medications are due to be administered and when. Analyzing Options The new process should provide double Now that the design requirements had been checking capability for MAR information. defined and the nurses needs measured, the next Order entry needs to be trustworthy. It step was to analyze options. In this phase the team should be at least 95% accurate the first time generated alternative concepts of processes that they are entered in the MAR. Pharmacists could be designed. These concepts then were evalshould be consistent in the way they enter uated against customer (nurse) needs from Table 1. the order. Some of the concepts considered were: A hierarchical value then was assigned to each 1. Use the existing patient care activity record customer need, representing the relative impor(PCAR) system as the medication list, with tance the need had with respect to other needs. some minor enhancements. PCAR uses the For instance, a hierarchical value of 2% can be MAR information to generate its medication interpreted as meaning only 2% of the customers listone source of information. This process considered it important. Table 1 summarizes the already was scheduled to be replaced by an results. electronic medication bar coding system The team analyzed the current processes and the called the medication administration checking perceptions nurses from all units had of each system (MAK), slated for implementation in processs effectiveness. The medication order entry the fall of 2005. process flowchart for each unit also was revisited.

QUALITY PROGRESS

I JANUARY 2007 I 41

HEALTHCARE

FIGURE 1

Quality Function Deployment Matrix

+ + + + + + Fax to MAR line Desktop to document MD order availability + + + Process to enter medication Availability printed copy of document First pass percentage orders incomplete First pass number of defects Customer rating

Correlation Matrix ++

Direction of improvement Design requirements Importance

Desktop to document clicks (steps)

Q MD order availability

Q Process to enter medication orders

Q Availability printed copy of document

First pass percentage orders incorrect

First pass number of defects

Fax to MAR time

1 = worst

Quick access

Customer requirements Available immediately for printing Less than three steps to access the document

3 3

9 3 9 3 3 3

3 9 1

1 1 1 3 1 1 1 9 9 9 3 3 3 3 1 1 1 1 1 9 9 271 8 529 15 3 3 1 1 1 1 1 9 9 529 15 @ @ # # x @x # @x # @ x # x# @x# x x

@x # @ x#

50 Quick pharmacy Quick TAT turn around Facsimile of order available time (TAT) 10 in real time Able to see all meds for 3 Provide history the entire patient stay of medications Able to see med start 3 and stop dates Nurse should not need to go Portable/mobile 3 back to station for document Serve to double check 15 all orders in MAR Check on MAR Medications sort in the same 10 way as MAR At least 95% accurate 25 the first time Trustworthy Consistency in order entry 10 by all pharmacists Absolute Technical importance Relative (percentage)

3 9

9 3 9 9

1 1 1

3 3 1 3

x# @# @ @

3 3 3 3 714 21

1 1 145 4

1 1 344 10

3 9 930 27

MAR = medication administration record, @ = in-patient unit, x = labor and delivery, # = long-term care.
42

I JANUARY 2007 I www.asq.org

5 = best

2. Deploy the MAK system FIGURE 2 Pugh Selection Matrix early. MAK would detect any discrepancy between Concepts the medication and the order at the bedside. It was scheduled for implementation in the fourth quarter of 2005. This proposal called for extra resources to expedite the systems launch by the spring of 2005. 3. Use optical character Available immediately for printing. S recognition (OCR) techLess than three steps to access the document. S S nology. OCR equipment Quick turnaround time. S S scans images of medicaFacsimile of order available in real time. S S tion orders as they are Able to see all medications for the entire patient stay. faxed to the pharmacy. It Able to see medication start and stop dates. S S converts handwriting into Nurse should not need to go back to station for document. S S electronic lists of medications stored in a folder that Serve to double check all orders in MAR. S S is accessible with a mouse Medications sort in the same way as in MAR. + + click. At least 95% accurate the first time. + + A Pugh selection matrix was Consistency in order entry by all pharmacists. + + used to evaluate these concepts Total + 3 3 against customer (nurse) Total 1 2 needs.3 The Pugh matrix compares multiple concepts against + = Better than datum a baseline model in terms of = Worse than datum S = Same as datum how well they address VOC. The preferred concept can be either the one with the highest number of plus signs minus the number of minus signs or a Going Live new concept that incorporates superior characterisThe team discussed verification plans, including tics of the proposed concept ideas. a pilot program. A go live date of April 4 was Based on the Pugh matrix results in Figure 2, the selected because of the usual low patient census on team decided to implement the PCAR option Sunday nights. Mandatory training sessions were because it had the lowest number of minus signs, programmed throughout the month of March to same number of plus signs as MAK, yet was very educate nursing staff on the new process. easy to test and could be implemented quickly. On April 3, the team performed a second hospiOption two, MAK, would have required signifitalwide baseline study to reassess the accuracy of cant hardware and structural changes to the way the current process. For simplicitys sake, the metmedication is ordered and delivered. More lead ric used was errors per bed instead of the more time was required than was available. The OCR representative errors per order. The resulting basetechnology requires the purchase of hardware. line value was 0.4 errors per bed, a value similar to There also were reliability and capability concerns the results of the initial baseline study conducted at regarding its ability to read physicians handwritthe beginning of the project, which yielded 90% ing.
Datum (baseline) Criteria
QUALITY PROGRESS

Current process (chronology) Existing patient care activity record Medication administration checking system (MAR) Optical character recognition S S S S S + + + 3 3

I JANUARY 2007 I 43

HEALTHCARE

FIGURE 3

Control Chart of Medical Transcription Errors per Bed

Errors/chart Mean 1 sigma 2 sigma 3 sigma

0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0.00

Defects/bed

May 7, 05

May 14, 05

May 28, 05

April 17, 05

April 30, 05

June 6, 05

April 3, 05

July 11, 05

July 18, 05

June 13, 05

June 20, 05

Baseline Study

Date

accuracy per order or 0.5 errors per bed. At midnight, the chronological sheet was removed from all patient charts and the PCAR process was put into place. During the day the team surveyed all units, answered questions, gathered feedback and supported the nurses throughout the transition. The new process uses the physician orders as the double check for the MAR accuracy. When a new medication order is written, the nurse keeps the information in the PCAR. Once pharmacy enters the order into the MAR, the nurse verifies it against what was in the PCAR and signs off the orders as entered in the MAR. In case of discrepancy, the nurse goes back to the physician for verification. Any changes are recorded in the MAR and corrected in the PCAR on the next printout. Of course, the keystone of the PCAR initiatives success is for staff to actually verify new medication orders in the computer before signing off on them. This process was not happening with the chronological sheet, thus mistakes could remain
44

undetected for multiple days. The PCAR process is quite similar to the process required by the new MAK. Starting it now would allow for a seamless transition to MAK later.

Results
A control plan was developed to collect data on the current process and ensure the improvements are maintained over the long term. The plan addresses issues regarding the type of random audits to perform, by whom, how often and the reaction plan if the process goes out of control. These random audits are conducted weekly by the medication safety officer. An average of 30 chart audits are completed per week based on individual nursing unit census. The results of the audits are shown in a control chart (see Figure 3). The control chart allows the process owner to monitor changes in medication errors by distinguishing random day-to-day variation from variation caused by a significant change in process performance. In the control chart the average

I JANUARY 2007 I www.asq.org

June 27, 05

July 25, 05

Aug. 8, 05

July 4, 05

weekly errors per bed is plotted against a series of lines representing the overall mean and standard deviations (+/ 1, 2 and 3 sigmas). The control chart data show the error reduction teams goal of 50% reduction was surpassed. Audits revealed the percentage of order entry errors consistently improved by 90% to less than 0.04 errors per bed every month for four months after the process changes.

International Symposium on Quality Function Deployment, Union of Japanese Scientists and Engineers, Tokyo, March 1995. 5. Yang, Kai, and Basem El-Haik, Design For Six Sigma: A Roadmap for Product Development, McGraw Hill, 2003, pp: 49-68 and 173-196.

YANIRA BENITEZ is business process leader and Six Sigma

Power of Information
Nurses now ask the team, Why did we do all that extra work? when referring to the former medication order entry process. The new process is simpler, works on all units and is more accurate. But the power of information is what will keep the staff from digressing back to old habits. When audits showed an increase in transcription errors during the third week after implementation of the initiative, the Alton Memorial knew exactly what to do. The reaction plan devised by the team was put into place, a root cause analysis was done, the source of error was removed and the process went back in control in no time. As an Alton Memorial administrator said: Finding out we had a problem even before we knew it and knowing exactly what to do to fix it was priceless.

Black Belt for BJC HealthCare in St. Louis. She earned a masters degree in industrial engineering Pennsylvania State University and is pursuing a doctorate at the University of Missouri. Benitez is a member of ASQ.

LESLIE FORRESTER is a registered nurse and clinical

information specialist/educator for Alton Memorial Hospital, Alton, IL. She has a bachelors degree in nursing from McKendree College, Lebanon, IL.

CAROLYN HURST is a registered nurse and medication safe-

ty officer for Alton Memorial Hospital. She has an associates degree in nursing from Lewis & Clark Community College, Godfrey, IL.

DEBRA TURPIN is a registered nurse and lead clinical

REFERENCES

1. Medication Errors & Patient Safety, www.ahrq.gov / qual/errorsix.htm. 2. National Coordinating Council for Medication Error Reporting and Prevention, www.mccmerp.org. 3. Pugh Matrix, www.isixsigma.com/dictionary/ glossary. asp.

information specialist at Alton Memorial Hospital. She has a bachelors degree from McKendree College.

BIBLIOGRAPHY

1. BMG University, Transactional Design for Six Sigma Course Manual, BMG, 2004, sections two and seven, www.bmgu.com. 2. Crow, Kenneth, Performing QFD Step by Step, www.npd-solutions.com/qfdcons.html, 2005. 3. Evans, James, and William Lindsay, The Management & Control of Quality, sixth edition, 2005, pp 568-578. 4. Mazur, Glenn, Jeff Gibson and Bruce Harries, QFD Applications in Health Care and Quality of Worklife,"

Please comment
If you would like to comment on this article, please post your remarks on the Quality Progress Discussion Board at www.asq.org, or e-mail them to [email protected].

QUALITY PROGRESS

I JANUARY 2007 I 45