Synchronizing Nursing and Pharmacy Workflows: A ...

2 downloads 0 Views 532KB Size Report
municate complex healthcare processes. ... Complex healthcare models can be built using this nota- tion. ..... available from Cardinal Health, 2008. 2. Battleman ...
Synchronizing Nursing and Pharmacy Workflows: A case study Amar Ramudhin École de Technologie Supérieure Montréal Canada, H3C1K3 [email protected]

ABSTRACT: The synchronization of nursing and pharmacy workflows are essential for rapid and safe initiation of medication therapy. This study describes the situation in a Canadian hospital where the time to initial dose was considered to be too high although both nursing and pharmacy personnel worked hard to ensure the standards and quality of their services. After an in-depth review of the medication ordering workflows in four care units and the analysis of processes in the pharmacy the causes of problem were identified and remedied. The study revealed significant process variations and differences in the metrics used by nursing and pharmacy in measuring turnover times. Workflow enhancements, synchronization of the metrics, matching demand with supply and increased communications resulted in a 66% reduction in the time to initial dose bringing the latter under control. This translated into faster and safer initiation of medication therapy for patients and better morale in both nursing and pharmacy personnel.

KEYWORDS: Time to initial dose, nursing workflow, medication dispensing and administration, pharmaceutical services, process modeling, simulation

1

INTRODUCTION

The time to initial dose (TTID) is the time between when a medication is prescribed for a patient and the time when the medication is administered to the patient. TTID is an important factor in patient outcomes as not only does it affect the length of stay of patients but can lead to serious complications and possible death if medication therapy is not initiated in time [2,6]. The prescription of a new medication order initiates a complex series of interrelated supply chain and workflow processes aimed at providing the medication as quickly, efficiently, accurately and cost-effectively as possible. The medication order is first sent to the pharmacy where it is reviewed and entered in the pharmacy information system. Depending on the medication type it is dispensed and delivered to the care unit according to one of several pathways. The medication on the care unit is administered to the patient by a nurse according to the rules of the 5Rs (Right medication, Right patient, Right time, Right dose and Right route). The number of medication orders prescribed in a hospital varies considerably based on its type and size and can be more than 2500 orders per day for a 500 bed hospital. The knowledge that each order is one of a kind and must be treated as efficiently as possible creates a lot of pressure on pharmacists, pharmacy personnel and nursing staff. No other interdepartmental relationship within a

hospital demands this many daily synchronization points [1] 2

CHALLENGES

Because TTID has a significant impact on the quality of care, the hospital under consideration had recently issued a policy under which nursing and pharmacy should ensure that, on the average, patients get their first dose within a 2 hour period. During follow-up meetings, nurses complained that pharmacy did not dispense the newly prescribed medications in a timely manner for administration to patients and hence they could not meet the target. They stated that it normally took between 2 to 4 hours before the medication would be available and that the TTID could be as high as 6 hours or more in some cases. Pharmacy managers on the other hand stated that, according to their metrics, the pharmacy turnaround time was less than 2 hours and hence met with the hospital’s target on TTID. The pharmacy director complained of delays in obtaining the medication orders for processing and dispensing. Also, according to pharmacy management, any further reduction in the turnaround time would require an increase in pharmacy personnel and hence would require more head counts. After discussions with the hospital management both nursing and pharmacy accepted to participate in a study to investigate nursing and pharmacy workflows with the

intent to identify the factors and or practices that contribute to the delays in the timely availability of initial doses.

3

METHODOLOGY

The first step in this study was to understand the hospital objectives and the demographics. Next was the selection of the units and processes to be observed in order to understand the dynamics of the various departments. The following processes were observed and modeled: 1.

1

This study was conducted according to the medBPM® (medical Business Process Modeling) methodology which consists of the following phases: 1. 2. 3. 4. 5.

Preparation and planning of study On-site observation, modeling and validation of models Off-site analysis Presentation and discussions of results Consensus building, development of the to-be models and improvement strategies

medBPM® is a software based framework that was specifically designed to model, measure, analyze, and communicate complex healthcare processes. The medBPM® tool was developed to capture the many activities of specialists, clinical professionals, information systems, equipment and materials in a straightforward fashion that is easy for clinicians to understand [3]. Figure 1 gives an example of a medBPM® model where nodes represent activities, delays or storage points and where object or agent flows are represented by arcs. Complex healthcare models can be built using this notation. The various pathways that must be synchronized to deliver a service can then be easily identified and quantified as each node may be assigned different metrics (time, cost, % non-value added, risks, etc) and the corresponding pathway metrics generated accordingly. In the example of Figure 1, there are three pathways namely, a pathway for the provider, a pathway for the medication and a pathway for the patient. The total time for the nurse to administer the medication is 17 mins 10 secs at a cost of $7.15 and this pathway has the nurse waiting for 8 minutes which potentially could be eliminated.

2. 3.

Medication ordering on the care units and transmission of orders to the pharmacy Pharmacy order review, dispensing and delivery of newly ordered medications Nursing administration of medications to patients.

In addition to the pharmacy, four care units were selected for the observations of medication ordering and administration processes. They are • • • •

27-bed paediatric cardiology 19-bed cardiology unit 23-bed neurosurgical unit 18-bed general surgery and urology unit

The hospital management felt that these four units provided a good sample of the variations in practices depending on the levels of care in the hospital. A first site visit was conducted to see the layouts of the units and the pharmacy. Meetings were scheduled with both pharmacy and nursing personnel during which the objectives of the study and the methodology were explained. Pharmacy and nursing had to fill a questionnaire that allowed for a better understanding of the units, the work strategies, issues, metrics, workflows and schedules. In each nursing unit, orders were written by the physicians during each of their two rounds (morning and afternoon). Medication was administered to patients during each of the three scheduled medication administration times (morning, afternoon and evening) and as prescribed for their therapy. The morning physician rounding process and the morning nursing medication administration workflow were observed and modeled in each unit during a 4 day period. The time for each activity in the model was obtained either through direct observations, deduced from information contained in reports and other information systems or obtained through interviews. It took a team of three people three weeks to observe, develop, validate and analyze the three models. A detailed presentation of these models is beyond the scope of this paper but the complexity of these models can be inferred from the data presented in Table 1.

Figure 1: Example of a medBPM® model showing three different pathways that, when combined together, deliver a service. 1

medBPM® is a trademark of BlueSail Solutions (www.bluesailsolutions.com).

Table 1: Complexity of models: nodes and resources

4

ANALYSIS

Analysis of the pharmacy model allowed the identification of two main medication dispensing pathways. Approximately 70% of all medications were dispensed via Automated Dispensing Cabinets (ADC) that are located on the care units and therefore at the point of use. The non-ADC pathway consisted of IVs and patient specific medications that must be dispensed from the pharmacy and delivered to point of use. ADCs are medication storage devices or cabinets that electronically dispense medications in a controlled fashion and track medication use. Their principal advantage lies in permitting nurses to obtain medications for inpatients at the point of use when properly stocked. These systems require user identifiers and passwords, and internal electronic devices track nurses accessing the system, track the patients for whom medications are administered, and provide usage data to the hospital's financial office for billing [6]. The overall process of ordering, dispensing and administration of new orders is similar to the one described in [1] and consists of the following detailed steps: 1. 2. 3. 4. 5. 6. 7. 8.

9. 10. 11. 12. 13.

Medication orders are prescribed on the care units by physicians The orders wait to be scanned to the pharmacy by the unit clerk Orders are scanned to the pharmacy Orders wait for processing in pharmacy A pharmacy technician enters the order in the pharmacy information system Orders wait to be verified by a pharmacist A pharmacist verifies and approves the order Medications stored in Automated Dispending Cabinets (ADC) on the care units are immediately available for administration Non-ADC medication are prepared in the pharmacy Non-ADC medication are delivered to the care units Nurse retrieves medication from ADC Nurse gathers other medications and supplies from other locations Nurse walks to patient room, administers and documents medication administration in the medication administration report (MAR).

All medication prescriptions are hand written and it is required that the physician date, time and sign the orders. The orders are then scanned to the pharmacy using Pyxis® Connect, a routing package that keeps track of when an order was scanned, the time it was entered in the pharmacy information system (PIS) and the time at which the order was ultimately verified and approved by the pharmacist. At this point, if the ADC is profiled and the medication stocked, the medication is available for nursing administration. In non-profiled ADCs, medications can always be retrieved by nurses prior to pharmacist verification which can lead to safety concerns. Removal times for medication orders are tracked by the ADCs and the administration times are documented in the patient’s medication administration record (MAR) after each medication administration. For the example shown in Figure 2, the medication order was written at 17:45, scanned at 20:17 pm, entered in the PIS at 21:02 pm and verified and approved by a pharmacist at 21:34 pm. Hence the total write-to-scan time on the care unit is 2 hrs 32 mins. This means that the order reached the pharmacy 2 hrs 32 mins after it was written. In the pharmacy, the order waited 45 mins before it was entered in the PIS and was finally verified and approved 32 mins later. At this point, if the order is for an ADC based medication, the latter can be administered on the care unit. Non-ADC medications must be prepared and delivered from the pharmacy to the point of use before it can be administered. The total time between when the order was written and the time it was verified in the pharmacy is 3 hrs 49 mins for this example. The next sections present the results and the recommendations for this case study. 5

RESULTS

5.1 Medication ordering and processing workflows Figure 3 shows the main sources of variations in the medication ordering workflow. It can also be observed that no two units have the same processes. Minimizing TTID means minimizing the write-to-scan time of an order and the latter is heavily dependent on steps 3, 4 and 5 of Figure 3. Best practices would dictate that:

Steps 1-3 are part of medication ordering model and this process varies by unit. Steps 4-10 are part of the pharmacy new order model processing model. Steps 10 – 13 are part of the nursing medication administration model and again this process varies by nurse and by department. The pathway metrics obtained from the models were validated by comparing the metrics to the results of the reconstructing the flow of a collection of 386 medication orders through the system. This was done by obtaining each medication order’s time-stamp as they passed through various stages in the system as described below and illustrated in Figure 2.

• • •

the patient charts containing new orders are delivered immediately to the unit clerk (3.1 in Figure 3) the unit clerk processes the charts as they arrive (4.1) medication orders from all charts are scanned first (5.1).

As can be seen from Figure 3 none of the care units implemented these three components of the best practice. Table 2 gives the average write-to-scan for the four care units studied. The times are given in the two formats, h:mm:ss and hrs, for better readability. The fastest write-

to-scan time is 0.69 hrs or 0:41:28 for U1 and slowest is 2.95 hrs or 2:57:16 for U4. Even though the average write-to-scan time for the 4 units is 1.74 hrs (with a standard deviation of 0.998), it suffered from very high variability: from a minimum of less than 5 minutes to a maximum of over 8 hours.

a unit is 1 hr 06 mins. Hence the pharmacy turnaround times3 are: •

3:06:00 for ADC-based medication



4:35:00 for non-ADC based medication.

Mean write-to-scan time Unit h:mm:ss 0:41:28 2:57:16 1:13:01 2:06:54 1:44:40

hrs U1 0.69 U2 2.95 U3 1.22 U4 2.12 Average 1.74 Std Dev 0.998 Table 2: Mean write-to-scan times by care units

5.1.2 TTID Table 4 summarizes the times for the various steps for an initial dose. From this table it follows that the TTID for an ADC-based medication is 4:59:274 and 6:28:27 for non-ADC medications. Since 70% of all medications are dispensed through ADCs, the overall average TTID is 5:26:09. Again the variability observed was significant as actual TTIDs observed ranged from less than 5 minutes to greater than 10 hours. Step

5.1.1 Pharmacy order review and verification Once an order is scanned to the pharmacy, it waits for a pharmacy technician to enter it in the Pharmacy Information System (PIS). Then the order waits for a pharmacist to review, verify and approve it. Table 3 gives the average and median times for order entry, order review and verification and the total time required. It is worth noting that the actual processing times required for entering an order or to review and verify the order is less than one minute on the average for each activity. This means that orders spend 99% of the total flow time in queues waiting to be processed. Order Entry

Order Review & Verify

Total Time

Average Time

1:23:00

1:43:00

3:06:00

Median Time

0:36:00

1:07:00

1:43:00

Table 3: Time to review and verify an order in the pharmacy After the order is entered and verified, the medication is dispensed according to its type: • ADC-based medications do not have to be dispensed2 from the pharmacy because they are already available in the cabinets on the units. • Non-ADC based medications must be dispensed and delivered from the pharmacy to the care unit in order to be administered. Analysis of the models showed that the average time to dispense a non-ADC based medication is 23 minutes and the average time to deliver the medication to

Unless there is a stock-out in the cabinet, in which case the pharmacist must decide if a first dose is sent or the stock-out can wait until the scheduled refill time.

1

Write-to-Scan orders

2

Enter, review and verify order Sub total

3 4

5

Dispensing of non-ADC orders Delivery of non-ADC orders Sub total Medication administration Total

Average Time h:mm:ss 1:48:27 3:06:00 4:54:27 0:23:00 1:06:00 6:23:27 0:05:00 6:28:27

Table 4: Time by process step From a nursing point of view, computation TTID started when the new medication order was prescribed. For pharmacy, the turn around time started after the medication order was received in the pharmacy. Since the median time for order entry and review is 1:43:00 (see Table 3) pharmacy assumed its turnaround to be under the hospital’s target of 2 hours since only 30% of the medications (non-ADC based) needed to be dispensed and delivered. This mismatch in metrics was the cause of many disputes. 5.2 Workflow Improvements

5.2.1 Nursing By modifying the medication ordering models to comply with the best practices (i.e. patient charts containing new orders are delivered to the unit clerk immediately after a patient visit and having the clerk scans all new medication orders before processing the rest of the charts) the minimum attainable write-to-scan for each unit was generated from the revised medBPM® models. As can be 3

2

Process

From the time the order was scanned to the time that the medication is available to be administered at the point of use.  4 4:54:27 plus the medication administration time (0:05:00).

seen from Table 5, the average write-to-scan times now ranged between 3.5 minutes to 7 minutes depending on the care unit. This corresponds to a 90% to 98% reduction in write-to-scan times when compared to the actual.

Unit

Current average write-toscan times

U1 U2 U3 U4

0:41:28 2:57:16 1:13:01 2:06:54

Average write-to-scan times based on ‘Best Practices’ 0:03:43 0:03:35 0:07:00 0:03:00

1. 2.

% reduction

91% 98% 90% 98%

Table 5: Write-to-scan times based on current and best Practices

After discussions with hospital management and nursing leadership, the changes in the workflows were made to conform to best practices but the new target for the average write-to-scan times was fixed at 20 minutes so as to account for interruptions (e.g. phone calls, responding to emergencies, etc.) and variations in processes. This target was deemed appropriate and attainable. 5.2.2 Pharmacy One of the main issues contributing to the delays in pharmacy processing was the large backlog of orders to be processed when the pharmacy opened at 8:00 am. The daily operating hours of the pharmacy is 8:00 am to 11:00 pm and medication orders received during the night had to wait for the next shift for processing. When the pharmacy opened at 8:00 am, there was an average of 350 orders waiting to be processed for a workload backlog of 6 hours. Hence, the pharmacy staff spent most of the morning to process the backlog of orders which resulted in longer turnaround times for all orders. Furthermore, the rate at which medication orders are scanned to the pharmacy varies during the day. Because most medication orders are written during physician rounds and because there are two rounds (morning and afternoon) the rates vary according to a bi-modal distribution. By analyzing the patterns of when the orders are scanned, the schedules of the order entry technicians and pharmacists can be adjusted to match the peaks and valleys in demand so as to achieve the optimal level of performance. The problem now becomes one of finding a desired level of performance given the current level of head count. The desired average turnover time for pharmacy order entry and review was set to one hour for those orders scanned during opening hours of the pharmacy. Fixing that value to 1 hour and varying i) the pharmacy opening times, 2) the daily processing rates (in increments of 30 minutes), the most convenient operating procedure was found to consist of the following:

A new schedule for technicians and pharmacists as described in Table 6 ; that pharmacists help in entering orders to get rid of the morning backlog as follows; a. between 07:00 – 07:30: as many as 3 of the 4 pharmacists assist in order entry; b. between 07:30 – 08:00: as many as 2 of the 4 pharmacists assist in order entry; c. between 08:00 – 09:00: as many as 1 of the 4 pharmacists assists in order entry.

Shift

Number of Order Entry Technicians

Number of Pharmacists available to verify orders

07:00 – 15:15

3

4

9:00 – 17:15

0.6

14:45 – 23:00

3

3

Table 7: New daily schedule for technicians and pharmacists These operating procedures were validated through simulation and the results are shown in Figure 4 and Figure 5. As can be seen, the above operating rules are enough to get rid of the morning backlog by 9:00 am and that the two peaks of orders are resolved in a timely manner (see Figure 5). Implementation of the above recommendations guarantees an average time to complete order entry and review of 1 hour. The consequences are: • •

1 hour turnaround time for the 70% of the medications distributed through ADC 2 hrs 29 mins for the remaining 30% of nonADC medications as these still need to be dispensed and deliverd.

With these changes, the projected overall average pharmacy turnaround time will be 1:26:42. Given the target time of 20 minutes for medication ordering and an average allowance of 5 minutes for administration of the medication, the average TTID should be less than the hospital target of 2 hours. Furthermore, to simplify implementation of the changes and to ensure the 1 hour limit on order entry and review in the pharmacy, a self-rectifying dynamic rule was tested and recommended. The rule is to monitor and control the order entry queue as follows: in the event the queue length exceeds 75 orders once the morning backlog is resolved, all available pharmacy technicians and pharmacists must attend to the order entry queue and empty it. Pharmacists entering the orders will enter and verify the orders in one single step. This simple rule limits the backlog of work building in the queue to be less than 30 minutes.



5.3 Conclusion This study showed how process analysis combined with simulation could be used to streamline nursing and pharmacy processes. The processes were modeled in medBPM®, a business process modeling tool specialized for healthcare. The models were useful to capture the details of the processes, variations and metrics. Its graph based visual representation helped nurses and pharmacy personnel in validation of the activities and metrics. The visualization of workflow also facilitated communication between nursing and pharmacy. Once the actual metrics were obtained, changes and enhancements were made to nursing and pharmacy processes in agreement with the stakeholders. Simulation was used to fine-tune demand with supply in the pharmacy and the work schedule was changed to match the workload profile. Having the ability to develop the detailed processes and performing simulation on the same models saved a lot of time and effort during the execution of the project. Time to initial dose (TTID) was drastically reduced as can be seen from Table 8. This table summarizes the actual and target times for each step in the medication ordering, dispensing and administration process. TTID was reduced by 72% for ADC-based medications and 55% for Non-ADC based medication. The overall average target value for TTID is 1:51:42 which represents a 66% reduction from the current TTID value of 5:26:09. Process

Actual Time h:mm:ss

Target h.mm.ss

% Reduction

1

Write-to-Scan orders

1:48:27

0:20:00

82%

2

Enter, review and verify order

3:06:00

1:00:00

68%

Sub total

4:54:27

1:20:00

73%

0:23:00

0:23:00

0%

1:06:00

1:06:00

0%

Sub total

6:23:27

2:49:00

55%

Medication administration

0:05:00

0:05:00

0%

Total

6:28:27

2:54:00

55%

4:59:27

1:25:00

72%

6:28:27

2:54:00

55%

3 4

5

Dispensing of nonADC orders Delivery of non-ADC orders

TTID for ADC-based medications TTID for Non- ADC based medications

Table 8: Current and new targeted TTIDs The following three phased approach was suggested to management for implementing the changes •

Phase I: Pilot new pharmacy and nursing practices in at least 4 units o Educate personnel as to new practices and expected results o Run pilot for 2 weeks o monitor and discuss results daily with nursing and pharmacy personnel



Phase II: Review results and adjust plans for hospital wide deployment (1 week) o Adjust targets and schedule based on results of pilot Phase III: Deployment to other units and continuous monitoring o Monitor results daily for 1 week o Debrief after 1 week and adjust practices according o Continue monitoring and send results to nursing and pharmacy management on a weekly basis

In conclusion, it must be said that the significant gains in TTID were made without any additional investment in personnel or equipment but by enhancements in the processes, better communication between the two departments and by synchronization of supply with demand. Further reduction in TTID can be achieved by a dynamic stock replenishment policy where the number of medications and quantity stored in an Automated Dispensing Cabinet is dynamically monitored and changed to match the demand profile in each unit. This will be the subject of another paper. REFERENCES 1. Baker, Draves and Ramudhin, “Analysis of the Medication Management in Seven Hospital“, white paper available from Cardinal Health, 2008. 2. Battleman, D., M. Callahan, H.T. Thaler, “Rapid Antibiotic Delivery and Appropriate Antibiotic Selection reduce Length of Stay of Patients with Community-Acquired Pneumonia“, Arch Intern Med. 2002, 162:682-688. 3. Chan E., A. Ramudhin, “An Evaluation of Various Business Process Modeling Techniques Applied to Healthcare“, Proc of ISEM 07, Beijing, May 2007. 4. Elganzouri et al., “Medication administration time study: Nursing staff performance of medication administration,” JONA, 2009, 39(5):204-210. 5. Houcka P.M., D.W. Bratzlerb, “Administration of first hospital antibiotics for community acquired pneumonia: does timeliness affect outcomes? “,Curr Opin Infect Dis, 2005 18:151–156. 6. Murray M., “Automated Medication Dispensing Devices“, Chapter 11 of Making Health Care Safer A Critical Analysis of Patient Safety Practices Evidence, Technology Assessment, No. 43, Prepared for the Agency for Healthcare Research and Quality, http://www.ahrq.gov/clinic/ptsafety/chap11.htm. 7. Pedersen, Schnieder and Scheckelhoff, “ASHP National Survey of Pharmacy Practice in Hospital Settings: Dispensing and Administration – 2008“, Am J Health-System Pharm, 2009, vol.66, 926-946.

Figure 2: Reconstruction of the history of a medication order

Figure 3: Variations in medication ordering/processing workflows

Figure 4: Backlog is resolved by 9:00 as pharmacist help in order entry

Figure 5: Pharmacist capacity is for order review is enough to resolve morning and afternoon peaks