The Impact of a Best Practice Advisory for Shifting to Oral Medications: A Quality Improvement Project

Abstract

We developed a Best Practice Advisory (BPA) to encourage switching from intravenous (IV) to oral (PO) formulations of antimicrobials with equal efficacy, to decrease costs and IV usage. We conducted a prospective study analyzing our BPA’s effect on prescribing practices of antimicrobials in the inpatient setting of Stanford Hospital in Palo Alto, CA, over a 22-month period. Our BPA achieved a 1,526/6,585 (23.2%) acceptance rate and led to decreased use of IV antimicrobials, with substantial direct cost savings of at least $66,976. Future efforts will focus on expanding our BPA to non-antimicrobials given ongoing IV fluid shortages.

Problem:

High and increasing prescription costs drive up the cost of health care in the United States, a cost born by patients, payers and health systems.¹ Switching from expensive intravenous (IV) formulations to less expensive oral (PO) formulations of equal efficacy has been suggested as a potential avenue to decrease costs in the inpatient setting given that oral medications can be 10-fold cheaper and with equal efficacy to the IV formulation especially for antimicrobials.^2–5 Moreover, given the recent IV shortages due to factories damaged by hurricanes, now resolved, it is paramount to have effective non-IV options to deliver necessary medications safely to patients. Shifting from intravenous to oral medications can reduce the risk of line-related infections, nursing time, and lead to earlier discharges, further reducing the costs of the hospital stay.² Best practice advisories (BPA) can encourage prescribers to engage in certain practices, through the appearance of messages in the electronic health record (EHR) when certain criteria such as demographics, diagnoses, or medications prescribed are met. The efficacy of BPAs has been evaluated and are most effective when actively presented to the prescriber and do indeed alter provider behavior.⁶ Previously, our group conducted a pilot study from 2017–2019 focusing on using cost transparency, the act of displaying the price of medications when ordered, to prompt prescribers to shift from IV to PO formulations, which yielded $50,000 in saving over 12 months due to over 25% decrease in IV prescriptions.⁷ However, our prior efforts did not link the cost transparency with any suggested medication switches.

Specific Aim statement:

We sought to build on our prior work by creating a BPA to track precisely if educating providers regarding equal efficacy of IV and oral antimicrobials, and suggesting order changes, would lead to antimicrobial formulation switches.

Innovation:

Our study took place in the inpatient setting of Stanford Hospital in Palo Alto, CA, which is a tertiary referral hospital with 613 beds, along with our associated community hospital Stanford Heath Care Tri-Valley, but results are only reported for Stanford Hospital to more accurately represent how the best practice advisory functions when trainees are responsible for the majority of placing orders and responding to advisories. The demographics and admitting diagnoses for the Best Practice Advisory patients and Stanford Hospital admissions during the study period are reported in Table 1. As a quality improvement project, our project was determined to be exempt from Human Subjects Research approval by the Stanford Institutional Review Board, Protocol 80985. T-tests, Chi-square tests, and Fishers Exact Tests were performed in R.⁸

Table 1:

Demographic and hospital admission characteristics for BPA and overall Stanford Hospital, December 1, 2021-September 30, 2023

BPA patients (n = 3,458)		Overall Patients (n = 450,667)
Demographic Characteristic	Value	Demographic Characteristic	Value	P-value
Median (IQR) age at admission, years	61 (48–70)	Median (IQR) age at admission, years	60 (43–71)	<0.001
Sex (n patients, %)				<0.001
Female	1657 (47.92%)	Female	248193 (55.07%)
Male	1801 (52.08)	Male	202275 (44.88%)
Unknown	0	Unknown	199 (0.04%)
Race (n patients, %)				<0.001
White	1725 (49.88%)	White	207443 (46.03%)
Black	161 (4.65%)	Black	17887 (3.97%)
Asian	630 (18.22%)	Asian	91061 (20.21%)
Pacific Islander	61 (1.76%)	Pacific Islander	4148 (0.92%)
Native American	19 (0.55%)	Native American	1806 (0.40%)
Other	827 (23.92%)	Other	99539 (22.09%)
Unknown	35 (1.01%)	Unknown	28783 (6.39%)
Ethnicity (n patients, %)				<0.001
Hispanic-Latino	798 (23.08%)	Hispanic-Latino	76308 (16.93%)
Not Hispanic-Latino	2624 (75.88%)	Not Hispanic-Latino	343400 (76.20%)
Unknown	36 (1.04%)	Unknown	30959 (6.87%)
BPA Admissions (n= 3,730)*		Overall Admissions (n = 94,589)*
Admission characteristic	Value	Admission Characteristic	Value	P-value
Median (IQR) length of hospital stay, days	9 (4 – 22)	Median (IQR) length of hospital stay, days	3(1–6)	<0.001
30-day readmission rate	120 (3.22%)	30-day readmission rate	14212 (15.03%)	<0.001
Primary diagnoses (n diagnoses, %) Some patients with multiple primary diagnoses per admission. Primary diagnosis based on ICD-10 codes letter assignments.
Behavioral health	397 (10.46%)	Behavioral health	5865(6.20%)	<0.001
Circulatory system	1960 (52.55%)	Circulatory system	36285(38.36%)	<0.001
COVID-19	240 (6.43%)	COVID-19	3137(3.32%)	<0.001
Digestive system	1828 (49.01%)	Digestive system	20410 (21.58%)	<0.001
Endocrine/Metabolic	599 (16.06%)	Endocrine/Metabolic	8421(8.90%)	<0.001
Eyes, ears, nose, and throat	70 (1.88%)	Eyes, ears, nose, and throat	1044(1.10%)	<0.001
Genitourinary system	1037(27.80%)	Genitourinary system	12738(13.47%)	<0.001
Infectious diseases	620 (16.62%)	Infectious diseases	4076(4.31%)	<0.001
Injuries	726 (19.46%)	Injuries	12967(13.71%)	<0.001
Malignancy and Hematology	1704 (45.68%)	Malignancy and Hematology	20871(22.06%)	<0.001
Nervous system	817 (21.90%)	Nervous system	12199(12.90%)	<0.001
Other	3416 (91.58%)	Other	71483(75.57%)	<0.001
Pregnancy and perinatal conditions	3 (0.08%)	Pregnancy and perinatal conditions	1657 (1.75%)	<0.001
Respiratory system	1876 (50.29%)	Respiratory system	19874 (21.01%)	<0.001
Skin and musculoskeletal system	910 (24.40%)	Skin and musculoskeletal system	18080 (19.11%)	<0.001
Unspecified	1754 (47.02%)	Unspecified	30308 (32.04%)	<0.001

^*For those admitted for over 24 hours

At any time during the hospitalization, when a provider entered the order for an IV antimicrobial with an oral equivalent and the patient had enteral access through a diet order or feeding tube order, an active BPA would appear in a force-function mode, requesting the prescriber switch to an alternative or override the recommendation with a specific reason (Supplemental Figure 1). The detailed rules for the BPA are detailed in Supplementary Table 1. The BPA initially went live on December 1, 2021, with a significantly improved iteration that went live on July 1, 2022 that additionally showed the diet order as evidence the patient could receive enteral medications. Our study period extends from the initial BPA rollout on December 1, 2021, through September 30, 2023.

Impact:

The primary metric of our study was the percentage of BPAs accepted after the go-live on December 1, 2021. Secondary outcomes included the percentage of BPAs accepted by specific medication, hospital location, and provider type, as well as the reasons prescribers did not accept the BPA, and cost savings. We defined BPA acceptance as the percentage of BPAs with a linked oral equivalent order. The BPA fired for 3,730 admissions, out of a total of 94,589 admissions for over 24 hours. The BPA fired 6,585 times during the study period. After the BPA go-live of December 1, 2021, our BPA was accepted in 1,526 (23.2%) of cases.

Azithromycin was the most accepted BPA in 364/898 (40.5%) of cases, with the other medications ranging in acceptance from 9–29%. Floor Wards had the highest rate of BPA acceptance with 1,156/4,532 (25.5%), followed by the Intensive Care Unit (ICU) with 260/1423 (18.3%). Attending physicians also had the highest acceptance rate with 263/993 (26.5%) while pharmacists had the lowest acceptance rate at 64/593 (10.8%) (Table 2).

Table 2:

BPA Acceptance by Medication, Hospital Location, and Provider Type

Medication	BPA accepted	Total BPAs	BPA acceptance (%)
Azithromycin	364	898	40.5%
Doxycycline	64	224	28.6%
Rifampin	10	36	27.8%
Metronidazole	543	2,348	23.1%
Ciprofloxacin	140	700	20.0%
Levofloxacin	126	627	20.1%
Linezolid	55	288	19.1%
Isavuconazole	31	174	17.8%
Fluconazole	87	545	16.0%
Posaconazole	61	382	16.0%
Moxifloxacin	4	26	15.4%
Trimethoprim-Sulfamethoxazole	35	270	13.0%
Voriconazole	6	67	9.0%
Hospital location	BPA accepted	Total BPAs	BPA acceptance (%)
Floor Wards	1,156	4,532	25.5%
ICU	260	1,423	18.3%
Peri-Op	108	623	17.3%
Infusion Center	2	7	28.6%
Provider type	BPA accepted	Total BPAs	BPA acceptance (%)
Attending	263	993	26.5%
Resident	643	2,521	25.5%
Fellow	207	858	24.1%
Advanced Practice Provider	349	1,620	21.5%
Pharmacist	64	593	10.8%

In the cases where the BPA was not accepted, a reason was given in 3,531/5,059 cases (69.8%). These were analyzed for pre-set answers (e.g., Infectious Diseases recommendation) as well as free-text answers when “Other” was selected. For these cases, the most common reasons included that the patient was NPO in 2,016/3,531 (57.1%), the patient had poor PO tolerance/absorption in 527/3,531 (14.9%), and the Infectious Diseases consult team recommended IV treatment in 307/3,531 (8.7%).

We calculated cost savings based on the number of BPA orders accepted for each antimicrobial, the cost differential between the IV and PO formulations, and the assumption that the patient at minimum received 2 days of the PO formulation instead of the IV formulation. We estimated that we saved at least $66,976.60 in direct costs over the BPA period (December 1, 2021, through September 30, 2023), as shown in Supplemental Table 2. The cost savings were driven both antimicrobials with a high-cost differential between IV and PO formulations, especially posaconazole ($311.62 cost differential) and isavuconazole ($279.16 cost differential), as well as antimicrobials with high volumes.

Length of stay among the BPA population (median 9, IQR 4–22) was longer than overall hospital admissions (median 3, IQR 1–6) (Table 1). However, the BPA group differed significantly from overall hospital patients and admissions in every aspect. BPA patients were initially sicker on admission relative to the hospital group as a whole, e.g., 50.3% of BPA patients had a primary diagnosis involving the respiratory system, versus 21.0% of hospital patients overall. This group had more opportunities for IV to oral conversion given how sick they were, first requiring IV medications and with a longer hospital length of stay. We did not hear any anecdotal reports of adverse events related to the BPA, e.g., need to reinitiate IV therapy due to failure of PO therapy. In addition, the 30-day readmission rate for the BPA patients of 3.2% was lower than the overall hospital readmission rate of 15.0%.

Takeaways:

BPAs can automate teaching and implementation regarding effective and safe medication switches, and our work provides a roadmap for further optimization. Our BPA’s acceptance rate of 23.2% overall is similar to previous reports of acceptance rates of active BPAs.⁶ This is the first publication to our knowledge which describes a BPA for converting from IV to oral antimicrobials. There is longstanding literature regarding the effectiveness of oral antimicrobials compared to IV formulations, and how to encourage providers to switch formulations.^2–5,7

The most successful BPAs, with the highest acceptance rates, were those used for common infections, such as azithromycin at 40.5% which used for atypical pneumonia coverage; doxycycline at 28.6% which is used for skin and soft tissue infections; and metronidazole at 23.1% which is used in intra-abdominal infections. In contrast, the azole antifungals had some of the lowest rates of BPA acceptance at 17.8% for isavuconazole, 16.0% for fluconazole, and 16.0% for posaconazole, reflecting potentially sicker patients with fungal infections, with more concerns about oral tolerance and absorption, as well as specialist input. In addition, the Wards had the highest rate of acceptance (25.5%) versus the ICU (18.3%), again likely reflecting a difference in patient acuity and specialist involvement. Further optimization of this effort would concentrate on education of Wards providers, as their patients have the potential to benefit the most from the conversion in terms of earlier discharge. However, given current IV shortages, focus groups with ICU providers might be warranted. The pharmacist workflow can also supplement the BPA intervention. Pharmacists can receive the BPA when entering verbal orders on behalf of treating team during rounds. However, there is not another BPA which activates when pharmacist review medications as part of their patient monitoring responsibilities, though conversion from IV to PO when warranted is part of optimal medication management. This, too, would provide the opportunity for increasing the rate of conversion from IV to PO.

Ultimately, through this project and our follow-on efforts, we are seeking to refine our BPA to give an active alert that is both highly sensitive and highly specific, as many providers declined the BPA due to concerns about enteral absorption despite the patient having a diet or enteral access. Our BPA has tried to follow the five principles of Clinical Decision support: “The right information to the right person, in the right format, through the right channel, at the right time in the workflow.”⁹ Perhaps as a first step, this would consist of excluding patients who had an anti-emetic dispensed within the past 24 hours. We might also consider turning off the alert for oral antifungals in the Intensive Care Unit. Having an optimized BPA is critical as over-alerting providers has the potential to decrease the likelihood they will be accepted over time.¹⁰

One limitation of our study is that 30.2% of the BPA’s were declined without a reason given, making it more difficult to fully understand the reasons that the BPA was not accepted; however, based on oral feedback we received, those cases likely fall into the categories delineated above (e.g., concern about oral absorption, subspecialist recommendations, etc.). In addition, this BPA did not explore converting antimicrobials without an obvious oral equivalent, like an IV carbapenem, and further work is needed to delineate the role of antimicrobial stewardship efforts to optimize broad-spectrum antibiotic use.

Future work will seek to evaluate how the optimization of this BPA affects its acceptance. Additionally, we are exploring expanding our BPA to non-antimicrobial medications, provided that the efficacy is similar between formulations, the non-antimicrobial is used in high volume within the inpatient setting, and there is a cost differential between the IV and oral versions of the medication.

Keys to Innovation:

• Best Practice Advisories (BPA) can be an effective intervention for encouraging the transition from intravenous to oral antimicrobials.

• The most successful BPAs were for common antibiotics and the least successful included the azole antifungals.

• Higher uptake in certain units suggests that future targeting can lead to increased optimization.

• Education initiatives focusing on the benefits of switching from IV to oral medications can be used to enhance conversation rates.

Data availability:

Data available upon request.