Automated parenteral chemotherapy dose-banding to improve patient safety and decrease drug costs

Introduction

Traditional anticancer therapies, primarily cytotoxic chemotherapy, have been historically dosed based on patient-specific factors, including patient body weight.^1–3 After it was discovered that body surface area (BSA) correlated with several other measurable physical parameters between species, BSA-based dosing strategies were developed for many chemotherapy agents. These dosing strategies have not been consistently directly correlated with a patient’s therapeutic outcomes and BSA-based dosing fails to account for a variety of other patient-specific factors that may lead to variability in drug distribution. These factors include age, renal and hepatic function, enzymatic activity, and tumor cell drug resistance. ^1–3

Newer anticancer therapies include agents such as monoclonal antibodies that are more targeted than traditional cytotoxic chemotherapy. While the therapeutic index for these medications is not as narrow and they lack the same dose–response relationship as cytotoxic chemotherapy, their dosing is also often weight-based and relies on estimation. Evidence published to-date has found that outcomes in patients who receive dose-rounded chemotherapy are non-inferior to those whose therapies are not rounded.^4–8 The rounding of prescribed doses of chemotherapy agents to a safe measurable unit has been found to be associated with fewer calculation and measuring errors.^9–11 Rounding practices improve the safety of injectable chemotherapy prescribing, preparation, dispensing, and administration. Chemotherapy dose rounding has also been found to lead to a reduction in drug waste by preventing the disposal of partially used vials during preparation. ⁵

Monoclonal antibodies are significantly more expensive than traditional cytotoxic chemotherapy and reduced drug waste is associated with decreased drug costs—both environmental and monetary—related to the disposal of unused chemotherapy products. Additionally, if doses of chemotherapy are rounded down, clinical benefit can be achieved while utilizing less drug product. The practice of rounding chemotherapy doses within 10% has been endorsed by professional organizations, with a recommendation to automate the dose-rounding processes within the electronic health record (EHR) whenever possible.^12,13

Since 2014, UW Health has had a collaborative practice guideline recommending the rounding of cytotoxic chemotherapy doses either up or down to a vial size or measurable unit within 5% and biologics within 10%. Pharmacists have been delegated the authority to manually round chemotherapy orders as outlined in the guideline. For manual dose-rounding, upon verification of a chemotherapy order, the first verifying pharmacist must recognize that the dose is capable of being rounded to a vial size within the 5% or 10% parameter, manually calculate the rounded dose, update the chemotherapy dose and dosing units, and document in the administration instructions. This manual process was usually not seen by providers as it occurred after providers signed the order and was not occurring at every site throughout our health system.

Many institutions have implemented manual chemotherapy dose-rounding, but automated dose-banding uses the EHR to remove the human factor. This automated process improves patient safety while maintaining the benefits of chemotherapy dose-rounding. Dose-banding standardizes weight or BSA-based chemotherapy doses by selecting pre-determined weight or BSA ranges and standardized administered doses that correspond to those ranges and fall within a certain rounding parameter. ⁹ Since the standardized dose-banding tables are predetermined and built into the EHR, pharmacists do not have to perform manual calculations nor manipulate the medication order, and the medication doses administered to patients are standardized to measurable doses.

Implementing dose-banding has been found to reduce medication dosing errors for antibiotic medications for pediatric patients as well as high-dose methylprednisolone for patients with acute spinal cord injuries.^14,15 Automated dose-banding can further reduce medication errors by eliminating the opportunity for human calculation or transcription errors, while also increasing overall drug savings. Total drug exposure from dose-banded chemotherapy as measured by area under the curve has been found to not differ significantly from traditional dosing strategies in pediatric patients. ¹⁶ This dosing strategy has been implemented as a way to increase pharmacy efficiency, while also helping reduce medication errors as well as unnecessary drug waste in European hospitals, most notably throughout England’s National Health Services.^17,18 There has been little published to date about the implementation of automated dose-banding strategies, so the purpose of our project was to incorporate automated dose-banding within our EHR in order to increase drug cost savings while improving patient safety.

Methods

In order to automate dose-banding throughout our health system, a workgroup consisting of one student pharmacist, seven pharmacists and one informatics analyst was formed. The pharmacists participating in the workgroup represented a variety of practice areas, including four oncology pharmacists as well as one pharmacist each practicing in informatics, leadership, and research. We recognized that dose-banding would have a system wide effect, so we also engaged our research group and regional cancer center affiliates that utilize the institutional EHR. The initial scope of this project was limited to three chemotherapy agents utilized throughout our 698-bed academic health system and comprehensive cancer center. All treatment protocols and orders are signed within our EHR and chemotherapy is prepared on-demand. The workgroup reviewed medication errors reported to the institution’s event reporting system from the previous 12 months associated with manual rounding of parenteral chemotherapy doses to assess the type of errors that were occurring and monitored our event reporting system for reported medication events related to automated dose banding in the post-intervention time period. Through direct observation, the workgroup calculated the minimum number of clicks and keystrokes it took pharmacists to verify an order both before and after automated dose-banding was implemented. This project received exemption from our institutional review board for quality improvement.

The workgroup determined that automation of the dose-banding could occur by incorporating pre-specified dose-banding tables (see Table 1 for an example) into each drug’s specific mediation record within the EHR. For each medication, a dose-banding table was built that adhered to the 5% or 10% maximum rounding parameter as defined by UW Health’s clinical practice guideline. Tables were constructed so that all ordered doses of the selected medications would be automatically rounded to the nearest banded dose and preferentially rounded down, and the lower and upper bounds of the tables were determined based on doses administered throughout the health-system in the previous two years. Once the tables were implemented, prescribers were able to see the dose-banding logic for an order before signing it. The first three parenteral chemotherapy agents, bevacizumab, rituximab, and trastuzumab, were selected for automated dose-banding implementation based on being both high-cost and high-use agents throughout our health-system in order to maximize project impact. Prior to implementation, dose-banding tables were vetted and approved through both internal and external stakeholders representing a variety of different disciplines.

OPEN IN VIEWER

Table 1.

Selected excerpt of automated dose-banding parameters for bevacizumab incorporated into the electronic health record.

Standardized dose (mg)	Standard dose lower bound (mg)	Standard dose upper bound (mg)	Lower bound rounding (%)	Upper bound rounding (%)
500.00	500.01	555.55	0.002	9.999
550.00	555.56	611.11	1.001	10.000
600.00	611.12	666.66	1.820	9.999
650.00	666.67	722.22	2.500	10.000
700.00	722.23	777.77	3.078	9.999
750.00	777.78	833.33	3.572	10.000
800.00	833.34	888.88	4.001	9.999
850.00	888.89	944.44	4.375	10.000

Post-implementation retrospective chart reviews were performed to measure total medication savings in both milligrams and dollars, calculated from the wholesale acquisition cost (WAC) of the three medications, as well as the number of unique doses administered for each medication. The intervention went live in our EHR on 12 September 2017 and we defined our post-implementation period as 12 September 2017 through 31 December 2017. Doses that were administered during the post-implementation time period, and thus automatically dose banded, were compared to doses administered during a pre-intervention time period, and thus manually rounded, of 12 September 2016 through 31 December 2016. Net drug savings (milligrams) were calculated for each administration during both the pre-implementation and post-implementation time periods by subtracting the administered dose from the originally ordered dose. Cost savings were calculated by multiplying net milligram drug savings by the WAC of the drug. Drug milligram savings and corresponding cost savings from manual dose-rounding were compared to savings associated with automated dose-banding.

Results

Of the error reporting submissions that were reviewed from the year prior to implementation, there were four medication errors associated with this manual rounding process that included pharmacists calculating the rounded dose incorrectly, mistyping the calculated dose when updating the order, and not changing the dosing unit from milligrams per meter squared to milligrams, resulting in the automatic calculation of a new, higher dose within our EHR. The workgroup determined that manual chemotherapy dose rounding introduced the opportunity for transcription errors, while pharmacists were manipulating the order for the purposes of rounding the chemotherapy dose. Dose-banding eliminated the opportunity for these types of errors to occur. Throughout the post-implementation period, no medication errors were submitted to the internal medication-error reporting system regarding the medications that were dose-banded.

During the pre-implementation period, there were 729 total administrations of bevacizumab, rituximab, and trastuzumab within our health system. Pharmacists manually rounded 264 (36.2%) doses during verification of the medication order within the EHR, utilizing at least 6,561 additional clicks and 8,019 additional keystrokes. At the medication-specific level, orders were manually rounded 17.5% of the time for bevacizumab, 84.8% for rituximab, and 0.7% for trastuzumab, respectively. During the post-implementation period, 100% of the 1,033 total administration of bevacizumab, rituximab, and trastuzumab doses were automatically dose-banded, saving pharmacists at least 9,297 additional clicks and 11,363 additional keystrokes.

During the pre-implementation period, manual rounding was associated with a net drug savings of 994 mg (3,833 mg annualized or 1.36 mg per administration). Manual rounding by pharmacists saved 686 mg of bevacizumab, 248 mg of rituximab, and 60 mg of trastuzumab. The post-implementation period had a total of 24,571 mg (84,243 mg annualized or 23.79 mg per administration) of drug savings. Automatic dose-banding saved 7,106 mg of bevacizumab, 13,414 mg of rituximab and 4,051 mg of trastuzumab, respectively (Figure 1).

OPEN IN VIEWER

Figure 1.

Total drug savings associated with both manual (pre-implementation) and automated (post-implementation) rounding by implementing dose-banding.

Using WAC pricing, the drug savings associated with manual rounding translated to an estimated cost savings of $8,420 ($32,870 annualized) associated with manual rounding and $224,746 ($770,556 annualized) associated with automated dose-banding. In the pre-implementation period, the net medication-specific cost savings for bevacizumab was $5,467, rituximab $2,330, and trastuzumab $623. Automated dose-banding during the implementation period for bevacizumab was $56,631, rituximab $126,027 and trastuzumab $42,088 (Figure 2).

OPEN IN VIEWER

Figure 2.

Total cost savings associated with both manual (pre-implementation) and automated (post-implementation) rounding by implementing dose-banding.

Discussion

Implementing dose-banding within our EHR reduced variation in rounded chemotherapy doses, increased drug and monetary savings, and reduced medication safety risks associated with manual manipulation of the chemotherapy orders. By standardizing the rounding parameters and the administered doses of select chemotherapy agents at our institution, we ensured all doses were rounded consistently and safely. Since our dose-banding intervals preferred rounding doses down whenever possible within a 5% or 10% parameter, we were also able to increase net drug savings and, by association, cost savings. Additional savings were realized by implementing automated dose-banding in administration locations within our health system that were not manually rounding chemotherapy doses although this represents only a small volume of our total chemotherapy orders

Utilizing pre-determined rounding logic and incorporating pre-specified dose-banding tables within our EHR allowed us to select the desired administered doses of all medication orders for bevacizumab, rituximab, and trastuzumab, and standardize the rounding parameters for each administered dose. Because the intervals we selected for our administered doses were smaller than the full 10% rounding parameter allowed by our delegation protocol and collaborative practice guideline, we were able to configure the dose-banding tables so that we could preferentially round down whenever possible. This allowed us to minimize variability in our rounding practices for parenteral chemotherapy while also achieving a net drug, and thus cost savings.

There are several limitations to our analysis. The first is that when calculating drug and cost savings, we assumed that the entire contents of a medication vial were used before the vial became outdated. This is normally the case at our institution; however, there are still some instances where vials reach their beyond-use date and are discarded with unused drug remaining. Our cost-savings calculations did not account for any changes in WAC during the one-year gap between the beginning of our pre-intervention period and the beginning of our post-intervention period. Since our actual acquisition cost differed from the WAC, this analysis overestimates the actual cost savings realized by our institution.

Not all of the selected medications had been manually rounded to the same degree. When the delegation protocol and collaborative practice guideline were first implemented, rituximab was the first agent targeted and may have been more heavily ingrained within the pharmacists as a target for dose-rounding. Trastuzumab, in comparison, was another one of the first agents we started to automatically round; however, pharmacists were not routinely manually rounding trastuzumab orders until right before automation went live, when the trastuzumab vial size changed from a 450-mg multiuse vial to the 150-mg single use vial.

Another limitation to our analysis is that one of the ways in which we designed and analyzed our intervention was through review of medication errors submitted to our medication event reporting system, which was likely affected by underreporting of medication errors by front-line staff. If underreported, the true number of medication errors relating to manual dose rounding in the year prior to implementation was likely higher than the four that were identified.

Following successful implementation of automatic dose-banding within our EHR for bevacizumab, rituximab, and trastuzumab, our institution has worked to extend this automation for additional agents. In addition, incorporating dose-banding is now a required step in the process of building medication records for any new chemotherapy agent being added to our formulary and EHR.

Future directions of this project include building dose-banding tables and incorporating automated dose-banding for the remaining chemotherapy agents we currently have on formulary and exploring the possibility of incorporating automated dose-banding into non-chemotherapy medications that our pharmacists have been rounding to nearest vial sizes to reduce waste, such as infliximab. The data collected throughout the implementation and quality assurance stages of this project and future dose-banded medications could be used for determining optimal vial sizes for serving the needs of our patients while minimizing medication waste. This information could then be shared with manufacturers to potentially reduce some of the waste and costs realized by health systems due to oversized single dose vials. We are also working to update our dose-banding tables to increase the maximum rounding percentage for cytotoxic agents from 5% to 10% now that 10% rounding on all chemotherapy has been endorsed by numerous professional organizations.

The process by which we implemented dose-banding within our EHR by incorporating pre-determined dose-banding tables so that the rounding logic is predefined could be utilized to accomplish additional dosing modifications. Dosing tables could be built to implement dose-capping of chemotherapy doses above a certain total dose, body weight, or BSA. Further modifications could be made to increase the dose-banding parameter, by increasing the maximum rounding percentage up to 10% for cytotoxic chemotherapy. Despite the many benefits of dose-banding, one limitation is the flat dosing of anticancer therapies. While all three of the medications initially selected for dose-banding were monoclonal antibodies, there is limited utility in dose-banding biologics that have flat doses, as the dose is not dependent on any patient-specific factors.

Conclusion

Implementing automated banding of parenteral chemotherapy doses for bevacizumab, rituximab, and trastuzumab resulted in improved safety within our chemotherapy ordering processes. Additionally, net drug savings were associated with this automated dose-banding. By reducing unnecessary drug waste and increasing drug savings, we were also able to increase overall cost-savings.