Advertisement

Cost–benefit analysis comparing trough, two-level AUC and Bayesian AUC dosing for vancomycin

  • Brian V. Lee
    Affiliations
    School of Pharmacy, University of Southern California, Los Angeles, CA, USA
    Search for articles by this author
  • Gary Fong
    Affiliations
    School of Pharmacy, Chapman University, Irvine, CA, USA

    Department of Pediatrics, The Lundquist Institute, Torrance, CA, USA

    Harbor-UCLA Medical Center, Torrance, CA, USA
    Search for articles by this author
  • Michael Bolaris
    Affiliations
    Department of Pediatrics, The Lundquist Institute, Torrance, CA, USA

    Division of Pediatric Infectious Diseases, Harbor-UCLA Medical Center, Torrance, CA, USA
    Search for articles by this author
  • Michael Neely
    Affiliations
    Division of Infectious Diseases, Children's Hospital Los Angeles, Los Angeles, CA, USA

    Department of Pediatrics, Keck School of Medicine, University of Southern California, Los Angeles, CA, USA

    Laboratory of Applied Pharmacokinetics and Bioinformatics, The Saban Research Institute, Los Angeles, CA, USA
    Search for articles by this author
  • Emi Minejima
    Affiliations
    School of Pharmacy, University of Southern California, Los Angeles, CA, USA
    Search for articles by this author
  • Amy Kang
    Affiliations
    School of Pharmacy, Chapman University, Irvine, CA, USA

    Department of Pediatrics, The Lundquist Institute, Torrance, CA, USA

    Harbor-UCLA Medical Center, Torrance, CA, USA
    Search for articles by this author
  • Grace Lee
    Affiliations
    Harbor-UCLA Medical Center, Torrance, CA, USA
    Search for articles by this author
  • Cynthia L. Gong
    Correspondence
    Corresponding author:C.L. Gong, Children's Hospital Los Angeles, 4650 Sunset Boulevard, MS #31, Los Angeles, CA 90027, USA.
    Affiliations
    Department of Pediatrics, Keck School of Medicine, University of Southern California, Los Angeles, CA, USA

    Fetal & Neonatal Institute, Division of Neonatology, Children's Hospital Los Angeles, Los Angeles, CA, USA

    Schaeffer Center for Health Policy & Economics, University of Southern California, Los Angeles, CA, USA
    Search for articles by this author
Published:November 19, 2020DOI:https://doi.org/10.1016/j.cmi.2020.11.008

      Abstract

      Objectives

      Area under the time–concentration curve (AUC) -guided dosing provides better estimates of exposure than vancomycin trough concentrations. Though clinical benefits have been reported, the costs of AUC-guided dosing are uncertain. The objective of this study was to quantify the costs of single-sample Bayesian or two-sample AUC strategies versus trough-guided dosing.

      Methods

      A cost–benefit analysis from the institutional perspective was conducted using a decision tree to model the probabilities and costs of acute kidney injury (AKI) associated with vancomycin administered over 48 hours up to 21+ days. Costs included vancomycin concentrations, Bayesian software and AKI hospitalization costs, and probabilities were obtained from primary literature. Robustness was assessed via both one-way and probabilistic sensitivity analyses.

      Results

      In the base-case model, two-sample AUC versus trough dosing saved an average of US$ 846 per patient encounter, and single-sample Bayesian AUC versus trough dosing saved an average of US$ 2065 per patient encounter. This translates into annual cost-savings of US$ 846 810 and US$ 2 065 720 for two-sample and single-sample Bayesian methods versus trough dosing, respectively, assuming 1000 vancomycin-treated patients per year. Assuming a budget of US$ 100 000 per year for Bayesian software, an institution would need to treat ≥41 patients with vancomycin for at least 48 hours to break even.

      Conclusions

      There are significant institutional cost benefits using two-sample AUC or single-sample Bayesian methods over trough dosing, even after accounting for the annual costs of Bayesian programs. The potential to decrease rates of AKI, improve clinical outcomes and reduce costs to the institution strongly warrants consideration of improved dosing methods for vancomycin.

      Graphical abstract

      Keywords

      Introduction

      Vancomycin remains the drug of choice for methicillin-resistant Staphylococcus aureus (MRSA) infections and is often used empirically when concern for MRSA infection exists. Older guidelines recommended the use of single trough samples (hence referred to simply as ‘troughs’) for vancomycin dosing [
      • Rybak M.J.
      • Lomaestro B.M.
      • Rotschafer J.C.
      • Moellering R.C.
      • Craig W.A.
      • Billeter M.
      • et al.
      Vancomycin therapeutic guidelines: a summary of consensus recommendations from the infectious diseases Society of America, the American Society of Health-System Pharmacists, and the Society of Infectious Diseases Pharmacists.
      ]. However, using single trough samples as a means for targeting a therapeutic range of 15–20 mg/L is associated with increased rates of nephrotoxicity while minimally improving outcomes [
      • Hidayat L.K.
      • Hsu D.I.
      • Quist R.
      • Shriner K.A.
      • Wong-Beringer A.
      High-dose vancomycin therapy for methicillin-resistant Staphylococcus aureus infections: efficacy and toxicity.
      ,
      • Jeffres M.N.
      • Isakow W.
      • Doherty J.A.
      • Micek S.T.
      • Kollef M.H.
      A retrospective analysis of possible renal toxicity associated with vancomycin in patients with health care-associated methicillin-resistant Staphylococcus aureus pneumonia.
      ,
      • Neely M.N.
      • Youn G.
      • Jones B.
      • Jelliffe R.W.
      • Drusano G.L.
      • Rodvold K.A.
      • et al.
      Are vancomycin trough concentrations adequate for optimal dosing?.
      ,
      • Kullar R.
      • Davis S.L.
      • Rybak M.J.
      Impact of vancomycin exposure on outcomes in patients with methicillin-resistant Staphylococcus aureus bacteremia: support for consensus guidelines suggested targets.
      ]. Subsequently, troughs were shown to be poorly predictive of AUC [
      • Rybak M.J.
      • Le J.
      • Lodise T.P.
      • Levine D.P.
      • Bradley J.S.
      • Liu C.
      • et al.
      Therapeutic monitoring of vancomycin for serious methicillin-resistant Staphylococcus aureus infections: a revised consensus guideline and review by the American Society of Health-System Pharmacists, the Infectious Diseases Society of America, the Pediatric Infectious Diseases Society, and the Society of Infectious Diseases Pharmacists.
      ]. The revised 2020 vancomycin consensus guidelines published in Clinical Infectious Diseases now recommend Bayesian-derived area under the curve (AUC)/MIC ratios to accurately achieve targets for clinical efficacy while improving patient safety. Though these recommendations exist, implementation of either AUC-guided dosing or Bayesian-guided dosing is not widespread [
      • Kufel W.D.
      • Seabury R.W.
      • Mogle B.T.
      • Beccari M.V.
      • Probst L.A.
      • Steele J.M.
      Readiness to implement vancomycin monitoring based on area under the concentration–time curve: a cross-sectional survey of a national health consortium.
      ].
      Newer dosing methods, such as the use of a two-sample AUC-guided dosing strategy or Bayesian programs, may also provide clinical benefit, but with uncertain costs and implementation [
      • Meng L.
      • Wong T.
      • Huang S.
      • Mui E.
      • Nguyen V.
      • Espinosa G.
      • et al.
      Conversion from vancomycin trough concentration–guided dosing to area under the curve–guided dosing using two sample measurements in adults: implementation at an academic medical center.
      ]. Bayesian programs, including BestDose, DoseMeRx, InsightRx and PrecisePK, are able to use Bayesian priors derived from specific pharmacokinetic models developed from vancomycin data to optimize dosing based on the collection of one or two vancomycin concentrations [
      DoseMeRx Vancomycin Page (Internet).
      , ,
      InsightRx Products Page (Internet).
      ,
      • Meng L.
      • Wong T.
      • Huang S.
      • Mui E.
      • Nguyen V.
      • Espinosa G.
      • et al.
      Conversion from vancomycin trough concentration–guided dosing to area under the curve–guided dosing using two sample measurements in adults: implementation at an academic medical center.
      ]. The cost of these programs vary based on degree of electronic medical record integration, size of institution, extent of functions, dosing capabilities for other drugs, contracting and other variables. Besides cost, implementation challenges also include clinician training and adaptation, unique patient populations, validity testing and other concerns.
      The objective of this study was to quantify the cost benefits of using single-sample Bayesian or two-sample AUC-guided versus trough-guided dosing for patients treated with vancomycin. This study takes into account the recommendations from the 2020 consensus guidelines and adds a cost–benefit perspective for practical institutional considerations. In addition, we assessed the feasibility of various Bayesian programs for a spectrum of institutional budgets and identified the options that would provide the most cost benefit for each.

      Materials and methods

      We performed a cost–benefit analysis using a decision tree to model the clinical course of a hospitalized individual requiring vancomycin for infection. The analysis was performed from the institutional perspective, assessing costs to a health system from patient admission to discharge. Using relevant cut-offs obtained from literature, we defined several time periods over the course of hospitalization: 0–48 hours of initial vancomycin administration, 48 hours to 7 days, 7–14 days and 14–21 days and beyond [
      • Aljefri D.M.
      • Avedissian S.N.
      • Rhodes N.J.
      • Postelnick M.J.
      • Nguyen K.
      • Scheetz M.H.
      Vancomycin area under the curve and acute kidney injury: a meta-analysis.
      ,
      • Lodise T.P.
      • Patel N.
      • Lomaestro B.M.
      • Rodvold K.A.
      • Drusano G.L.
      Relationship between initial vancomycin concentration-time profile and nephrotoxicity among hospitalized patients.
      ,
      • Wong-Beringer A.
      • Joo J.
      • Tse E.
      • Beringer P.
      Vancomycin-associated nephrotoxicity: a critical appraisal of risk with high-dose therapy.
      ]. We assumed that once vancomycin therapy was started, it was continued for at least 48 hours regardless of dosing method, at which point the patient progressed into one of three different outcomes: continue therapy without acute kidney injury (AKI), experience AKI, or exit the model because of discharge, death or discontinuation of vancomycin. The same decision-tree model of events was used for each dosing strategy investigated (trough, two-sample AUC and one-sample Bayesian AUC dosing), with strategy-specific probabilities and costs obtained from the literature. Trough-based dosing is described as using measured ‘trough’ concentrations to dose adjust towards a goal therapeutic range, after which they are obtained less frequently. Two-sample AUC and Bayesian AUC dosing strategies, however, use target concentration interventions, using pharmacokinetic–pharmacodynamic principles to estimate individual parameters to target an AUC goal, and calculate the most appropriate next dose [
      • Holford N.
      • Ma G.
      • Metz D.
      TDM is dead. Long live TCI!.
      ]. Two-sample AUC requires greater upfront concentrations compared with trough dosing, where Bayesian dosing requires the fewest concentrations of the three methods (see Fig. 1).
      Fig. 1
      Fig. 1Decision-tree model—clinical course of a patient requiring and treated with vancomycin, for at least 48 hours. Patients in the model will experience no AKI, experience AKI, or leave the model because of discontinuation of vancomycin, discharge or death. The decision-tree branches for the non-Bayesian approach arm and Bayesian software arm are the same as the trough-based standard of care arm.
      To standardize the probabilities of AKI obtained from literature for the three strategies, we used data from Wong-Beringer et al., which reported the occurrence of AKI by duration of vancomycin therapy [
      • Wong-Beringer A.
      • Joo J.
      • Tse E.
      • Beringer P.
      Vancomycin-associated nephrotoxicity: a critical appraisal of risk with high-dose therapy.
      ]. We then calculated the AKI frequency by duration of therapy for each dosing method [
      • Meng L.
      • Wong T.
      • Huang S.
      • Mui E.
      • Nguyen V.
      • Espinosa G.
      • et al.
      Conversion from vancomycin trough concentration–guided dosing to area under the curve–guided dosing using two sample measurements in adults: implementation at an academic medical center.
      ,
      • Neely M.N.
      • Kato L.
      • Youn G.
      • Kraler L.
      • Bayard D.
      • van Guilder M.
      • et al.
      Prospective trial on the use of trough concentration versus area under the curve to determine therapeutic vancomycin dosing.
      ]. For example, in the prospective trial evaluating Bayesian dosing by Neely et al., rates of nephrotoxicity were at most 8% in Year 1 (trough) and 2% in Year 3 (Bayesian) [
      • Neely M.N.
      • Kato L.
      • Youn G.
      • Kraler L.
      • Bayard D.
      • van Guilder M.
      • et al.
      Prospective trial on the use of trough concentration versus area under the curve to determine therapeutic vancomycin dosing.
      ]. These data were adjusted to the incidence of AKI in trough-based dosing (6.4%) on day 5 as per Wong-Beringer et al. [
      • Wong-Beringer A.
      • Joo J.
      • Tse E.
      • Beringer P.
      Vancomycin-associated nephrotoxicity: a critical appraisal of risk with high-dose therapy.
      ], so the incidence of AKI using Bayesian dosing would be 6.4% × (2%/8%) = 1.58%. Probabilities of death, discharge and lengths of stay for each strategy in Table 1 were also derived from literature [
      • Kim N.H.
      • Koo H.L.
      • Choe P.G.
      • Cheon S.
      • Kim M.S.
      • Lee M.J.
      • et al.
      Inappropriate continued use of empirical vancomycin in a hospital with a high prevalence of MRSA.
      ,
      • Olson J.
      • Stockmann C.
      • Hersh A.L.
      • Anderson C.
      • Zobell J.
      • Thorell E.
      Optimizing vancomycin prescribing through a pharmacist driven monitoring intervention at a children’s hospital.
      ,
      • Minejima E.
      • Choi J.
      • Beringer P.
      • Lou M.
      • Tse E.
      • Wong-Beringer A.
      Applying new diagnostic criteria for acute kidney injury to facilitate early identification of nephrotoxicity in vancomycin-treated patients.
      ,
      • Wang H.E.
      • Munter P.
      • Chertow G.M.
      • Warnock D.G.
      Acute kidney injury and mortality in hospitalized patients.
      ]. To avoid overestimating the benefits of newer dosing methods, we used the most conservative ranges of probabilities available, and adjusted them as above, to the probabilities in trough-based literature. In our model, we compared two-sample AUC with trough, and Bayesian dosing with trough, as the most appropriate way to use the probabilities available in the literature.
      Table 1Parameters (probabilities and costs)
      Trough arm
      Parameter%Min %Max %Misc. cost (US$)AKI cost (US$)Reference
       αNo AKI, LOS >48 hours

      (1 – AKI – Hospital discharge or death)
      38.730.9646.44
       βAKI, LOS >48 hours6.30.0213.3[
      • Jeffres M.N.
      • Isakow W.
      • Doherty J.A.
      • Micek S.T.
      • Kollef M.H.
      A retrospective analysis of possible renal toxicity associated with vancomycin in patients with health care-associated methicillin-resistant Staphylococcus aureus pneumonia.
      ,
      • Holford N.
      • Ma G.
      • Metz D.
      TDM is dead. Long live TCI!.
      ,
      ]
      γHospital discharge or death55.039.068.0780[
      • Neely M.N.
      • Kato L.
      • Youn G.
      • Kraler L.
      • Bayard D.
      • van Guilder M.
      • et al.
      Prospective trial on the use of trough concentration versus area under the curve to determine therapeutic vancomycin dosing.
      ,
      • Kim N.H.
      • Koo H.L.
      • Choe P.G.
      • Cheon S.
      • Kim M.S.
      • Lee M.J.
      • et al.
      Inappropriate continued use of empirical vancomycin in a hospital with a high prevalence of MRSA.
      ,
      • Olson J.
      • Stockmann C.
      • Hersh A.L.
      • Anderson C.
      • Zobell J.
      • Thorell E.
      Optimizing vancomycin prescribing through a pharmacist driven monitoring intervention at a children’s hospital.
      ,
      • Minejima E.
      • Choi J.
      • Beringer P.
      • Lou M.
      • Tse E.
      • Wong-Beringer A.
      Applying new diagnostic criteria for acute kidney injury to facilitate early identification of nephrotoxicity in vancomycin-treated patients.
      ]
      δNo AKI

      (1 – AKI – Hospital discharge or death)
      20.916.7225.1
      εAKI21.120.024.6[
      • Jeffres M.N.
      • Isakow W.
      • Doherty J.A.
      • Micek S.T.
      • Kollef M.H.
      A retrospective analysis of possible renal toxicity associated with vancomycin in patients with health care-associated methicillin-resistant Staphylococcus aureus pneumonia.
      ,
      • Lodise T.P.
      • Patel N.
      • Lomaestro B.M.
      • Rodvold K.A.
      • Drusano G.L.
      Relationship between initial vancomycin concentration-time profile and nephrotoxicity among hospitalized patients.
      ]
       θDischarge or death58.046.469.61180[
      • Neely M.N.
      • Youn G.
      • Jones B.
      • Jelliffe R.W.
      • Drusano G.L.
      • Rodvold K.A.
      • et al.
      Are vancomycin trough concentrations adequate for optimal dosing?.
      ,
      • Olson J.
      • Stockmann C.
      • Hersh A.L.
      • Anderson C.
      • Zobell J.
      • Thorell E.
      Optimizing vancomycin prescribing through a pharmacist driven monitoring intervention at a children’s hospital.
      ]
       κAKI

      (1 – discharges or death)
      26.020.831.2
       λDischarge or death74.059.288.831412 176[
      • Neely M.N.
      • Youn G.
      • Jones B.
      • Jelliffe R.W.
      • Drusano G.L.
      • Rodvold K.A.
      • et al.
      Are vancomycin trough concentrations adequate for optimal dosing?.
      ,
      • Olson J.
      • Stockmann C.
      • Hersh A.L.
      • Anderson C.
      • Zobell J.
      • Thorell E.
      Optimizing vancomycin prescribing through a pharmacist driven monitoring intervention at a children’s hospital.
      ,
      • Minejima E.
      • Choi J.
      • Beringer P.
      • Lou M.
      • Tse E.
      • Wong-Beringer A.
      Applying new diagnostic criteria for acute kidney injury to facilitate early identification of nephrotoxicity in vancomycin-treated patients.
      ]
       μAKI30.0243643114 206[
      • Jeffres M.N.
      • Isakow W.
      • Doherty J.A.
      • Micek S.T.
      • Kollef M.H.
      A retrospective analysis of possible renal toxicity associated with vancomycin in patients with health care-associated methicillin-resistant Staphylococcus aureus pneumonia.
      ,
      • Lodise T.P.
      • Patel N.
      • Lomaestro B.M.
      • Rodvold K.A.
      • Drusano G.L.
      Relationship between initial vancomycin concentration-time profile and nephrotoxicity among hospitalized patients.
      ]
       πNo AKI

      (1 – AKI)
      70.056842740
      ρLOS >14 days45.0365462721 308[
      • Neely M.N.
      • Youn G.
      • Jones B.
      • Jelliffe R.W.
      • Drusano G.L.
      • Rodvold K.A.
      • et al.
      Are vancomycin trough concentrations adequate for optimal dosing?.
      ]
       χLOS >14 days45.0365474533 485[
      • Neely M.N.
      • Youn G.
      • Jones B.
      • Jelliffe R.W.
      • Drusano G.L.
      • Rodvold K.A.
      • et al.
      Are vancomycin trough concentrations adequate for optimal dosing?.
      ]
      σDischarge or death55.0446647014 206[
      • Neely M.N.
      • Youn G.
      • Jones B.
      • Jelliffe R.W.
      • Drusano G.L.
      • Rodvold K.A.
      • et al.
      Are vancomycin trough concentrations adequate for optimal dosing?.
      ]
       ψDischarge or death55.0446658826 382[
      • Neely M.N.
      • Youn G.
      • Jones B.
      • Jelliffe R.W.
      • Drusano G.L.
      • Rodvold K.A.
      • et al.
      Are vancomycin trough concentrations adequate for optimal dosing?.
      ]
      Two-sample AUC arm
       αNo AKI, LOS >48 hours41.2232.97649.46[
      • Meng L.
      • Wong T.
      • Huang S.
      • Mui E.
      • Nguyen V.
      • Espinosa G.
      • et al.
      Conversion from vancomycin trough concentration–guided dosing to area under the curve–guided dosing using two sample measurements in adults: implementation at an academic medical center.
      ,
      • Golestaneh L.
      • Alvarez P.J.
      • Reaven N.L.
      • Funk S.E.
      • McGaughey K.J.
      • Romero A.
      • et al.
      All-cause costs increase exponentially with increased chronic kidney disease stage.
      ]
       βAKI, LOS >48 hours3.782.05.38
      γHospital discharge or death55.0044661570
      δNo AKI28.9023.1234.68
      εAKI13.105.0018.03[
      • Golestaneh L.
      • Alvarez P.J.
      • Reaven N.L.
      • Funk S.E.
      • McGaughey K.J.
      • Romero A.
      • et al.
      All-cause costs increase exponentially with increased chronic kidney disease stage.
      ]
       θDischarge or death58.0046.469.62740
       κAKI26.0020.831.2
       λDischarge or death74.0059.288.835312 176
       μAKI19.097.0025.6051014 206[
      • Golestaneh L.
      • Alvarez P.J.
      • Reaven N.L.
      • Funk S.E.
      • McGaughey K.J.
      • Romero A.
      • et al.
      All-cause costs increase exponentially with increased chronic kidney disease stage.
      ]
       πNo AKI80.9164.72897.0923140
      ρLOS >14 days45.00365466621 308
       χLOS >14 days45.00365482333 485
      σDischarge or death55.00446651014 206
       ψDischarge or death55.00446666626 382
      Bayesian AUC arm
       αNo AKI, LOS >48 hours43.42534.7452.11
       βAKI, LOS >48 hours1.57501.12.00[
      • Holford N.
      • Ma G.
      • Metz D.
      TDM is dead. Long live TCI!.
      ]
      γHospital discharge or death55.004466390
      δNo AKI36.7329.38444.076
      εAKI5.2754.226.33[
      • Holford N.
      • Ma G.
      • Metz D.
      TDM is dead. Long live TCI!.
      ]
       θDischarge or death58.0046.469.6780
       κAKI26.0020.831.2
       λDischarge or death74.0059.288.815712 176
       μAKI7.506.09.031414 205[
      • Holford N.
      • Ma G.
      • Metz D.
      TDM is dead. Long live TCI!.
      ]
       πNo AKI92.50741.111960
      ρLOS >14 days45.00365435321 308
       χLOS >14 days45.00365439233 485
      σDischarge or death55.00446627414 206
       ψDischarge or death55.00446631426 382
      Abbreviations: AKI, acute kidney injury; LOS, length of stay.
      Table includes: probabilities related to each step in the decision-tree analysis, fixed costs in terms of ‘AKI Cost’ and marginal costs in terms of ‘Misc Cost’ due to personnel related to serum concentrations and interpretation. Min and Max % for sensitivity analysis are ±20% or otherwise obtained from literature. Parameters that have been crossed through are included in a larger variable. For example, the probability of ‘Hospital Discharge’ is included in the ‘Discharge or Death’ total probability.
      The direct costs associated with nephrotoxicity during hospitalization were based on Healthcare Cost and Utilization Project cost data, which include national estimates of direct costs associated with management of the reported primary diagnosis [
      ]. We included International Classification of Diseases, ninth revision (ICD-9) codes N14.1 (Nephropathy Induced by Drug), N17.9 (Acute Kidney Injury), and calculated the average cost per day of hospitalization using these two codes. Hospitalization costs per day and due to vancomycin-associated AKI (V-AKI) were validated with data from a cost-minimization model by Patel et al., which found that costs associated with management of vancomycin-associated AKI range from US$ 9379 to US$ 20 467, in 2018 US dollars [
      • Patel N.
      • Huang D.
      • Lodise T.
      Potential for cost saving with iclaprim owing to avoidance of vancomycin-associated acute kidney injury in hospitalized patients with acute bacterial skin and skin structure infections.
      ].
      Other costs included labor costs for pharmacists and phlebotomists, obtained from the U.S. Bureau of Labor Statistics. These wages were used in conjunction with the number of troughs associated with each dosing strategy over different time periods, verified with physician and pharmacist guidance. The three strategies differed in the number of troughs, where two-sample AUC required greater upfront and similar follow-up concentrations to trough-based dosing, and Bayesian required the fewest concentrations overall. We assumed it would take a maximum of half an hour for a pharmacist or a phlebotomist to perform all actions necessary to manage a vancomycin trough concentration (hourly wages costing US$ 61.83/hour and US$ 16.58/hour, respectively) [
      Bureau of Labor StatisticsU.S. Department of Labor
      Occupational Outlook Handbook.
      ,
      Bureau of Labor StatisticsU.S. Department of Labor
      Occupational Outlook Handbook.
      ]. These marginal personnel costs related to concentrations are reflected in the Misc Cost category in Table 1, and differ based on parameter. Finally, costs of Bayesian dosing programs were obtained from company-provided quotes, as well as website information, available in Table 2. All costs in this model were adjusted using the medical component of the Consumer Price Index to 2019 US dollars [
      • Neumann P.J.
      • Sanders G.D.
      • Russell L.B.
      • Siegel J.E.
      • Ganiats T.G.
      Cost-effectiveness in health and medicine.
      ].
      Table 2Program costs
      Approximate prices are variable and subject to change, and may differ from what is stated here. Prices are only used for threshold analysis and post-modelling analysis.
      ProgramCost (US$)NotesReference
      BestDose0Not for commercial sale.
      InsightRx83 100/year[
      InsightRx Products Page (Internet).
      ]
      DoseMeRx12 000–30 000/year[]
      PrecisePK10 000/year[]
      a Approximate prices are variable and subject to change, and may differ from what is stated here. Prices are only used for threshold analysis and post-modelling analysis.
      The primary outcome of our analysis was the comparative cost–benefit of the total cost associated with each strategy or dosing method. We conducted one-way sensitivity analysis using the decision-tree model to determine the relative influence of each model parameter by varying one parameter at a time, and probabilistic sensitivity analysis to further evaluate uncertainty. We also conducted a secondary analysis to include costs for a small subset of patients who may later develop chronic kidney disease (CKD) to provide insight into the potential societal costs associated with each strategy. Finally, we estimated how many patients would need to be treated in an institution using Bayesian software to break even for the costs of the respective program. All calculations were performed using Microsoft Excel, 2019.

      Results

      In the base case, we identified that the costs of managing AKI for a vancomycin-treated patient with trough, two-sample AUC and single-sample Bayesian AUC dosing methods, were US$ 2982, US$ 2136 and US$ 917, respectively (Table 3). Hence, there is an incremental net benefit favouring two-sample AUC over trough of US$ 846 per patient, and Bayesian over trough of US$ 2065 per patient. Given this finding, we estimated that if an institution treated 1000 patients/year with vancomycin, it would save US$ 846 810 in AKI-associated costs by using two-sample AUC dosing over trough, and US$ 2 065 720 by using single-sample Bayesian dosing over trough. One thousand patients treated with vancomycin for ≥48 hours was used as a conservative estimate, assuming a hospital with ≥100 beds.
      Table 3Base case outcomes
      Dosing methodTrough (US$)Two-sample AUC (US$)Bayesian (US$)
      Additional AKI treatment cost per patient29822136917
      Incremental cost benefit vs trough per patient8462065
      Incremental cost benefit for 500 vancomycin patients/year vs trough423 0001 032 500
      Incremental cost benefit for 1000 vancomycin patients/year vs trough846 8102 065 720
      Bayesian AUC program thresholds (no. of patients needed to break-even program cost)
      BestDose00
      PrecisePK125
      InsightRx9941
      DoseMeRx Year 1156
      DoseMeRx Year 22912
      DoseMeRx Year 33615
      Abbreviation: AKI, acute kidney injury.
      Costs of Bayesian dosing programs vary significantly, ranging from costless (BestDose research tool) to as low as ~ US$ 10 000/year and more than ~ US$ 50 000/year. A threshold analysis was performed to test the feasibility of Bayesian programs for an institution, with a 1-year perspective. In a break-even analysis, assuming the costliest Bayesian program (approaching US$ 100 000/year), an institution would need to treat at least 40 patients with vancomycin for ≥48 hours to break-even for the costs of using the Bayesian program versus trough dosing. Similarly, it would need to treat at least 68 patients with vancomycin for ≥48 hours to break-even for the costs of using a single-sample Bayesian method versus two-sample AUC dosing.
      One-way sensitivity analysis showed that the probability of AKI at 48 hours to 7 days and the probability of discharge or death at 48 hours to 7 days were the most sensitive parameters (Fig. 2). In all one-way simulations, there was an incremental net benefit to using two-sample AUC dosing versus trough, as well as using single-sample Bayesian dosing versus trough.
      Fig. 2
      Fig. 2One-way sensitivity analysis. The centre lines indicate the base-case net benefits. In 100% of the one-way simulations, there was a net benefit to using two-sample AUC dosing versus trough, ranging from US$ 72 to US$ 1707. Similarly, in 100% of one-way simulations, there was a net benefit to using Bayesian dosing versus trough, ranging from US$ 1291 to US$ 2926.
      In 10 000 simulations of two-sample AUC versus trough, the probabilistic sensitivity analysis yielded a median cost benefit of ~US$ 1300 per patient, with all simulations yielding cost benefit (ranging from US$ 300 to US$ 2550) (Fig. 3). These findings were similar with single-sample Bayesian versus trough, with a cost benefit ranging from US$ 800 to US$ 3150.
      Fig. 3
      Fig. 3Probabilistic sensitivity analysis (likelihood of incremental net benefit). There is a 100% probability that two-sample AUC and Bayesian methods will have an incremental net cost benefit over trough dosing.
      For our secondary analysis, we investigated the impact of individuals with CKD on final costs in our model. For modeling purposes, we excluded the small percentage of patients who would enter the model with pre-existing CKD, given that the likelihood would be the same between the three dosing methods [
      • Rangaswamy D.
      • Sud K.
      Acute kidney injury and disease: Long-term consequences and management.
      ]. Using data available in the literature, there is a documented incidence of 3%–15% for end-stage renal disease (ESRD) and 20%–50% for CKD resulting from an occurrence of reversible AKI in a hospital stay, and a hazard ratio of 1.91 for de novo CKD in AKI versus control [
      • Rangaswamy D.
      • Sud K.
      Acute kidney injury and disease: Long-term consequences and management.
      ,
      • Bucaloiu I.D.
      • Kirchner H.L.
      • Norfolk E.R.
      • Hartle II J.E.
      • Perkins R.M.
      Increased risk of death and de novo chronic kidney disease following reversible acute kidney injury.
      ]. Using these average probabilities and reported medical costs of Medicare patients with CKD Stages 2, 3, 4–5 and ESRD (excluding dialysis), the total respective costs in managing both V-AKI and resulting CKD increases dramatically, with an incremental cost benefit of US$ 1845 per patient when using two-sample AUC over trough, and US$ 7965 per patient when using single-sample Bayesian over trough [
      • Golestaneh L.
      • Alvarez P.J.
      • Reaven N.L.
      • Funk S.E.
      • McGaughey K.J.
      • Romero A.
      • et al.
      All-cause costs increase exponentially with increased chronic kidney disease stage.
      ]. Hence, the secondary addition of CKD costs post-AKI in the model only adds net benefit to two-sample AUC and single-sample Bayesian methods over trough.

      Discussion

      To our knowledge, no studies have empirically shown the cost benefits associated with AUC dosing through reduced rates of AKI and related costs. According to a 2019 multicentre survey (n = 124), 23.1% of academic medical centres performed AUC-based vancomycin monitoring, and the remainder of academic institutions performed trough-based vancomycin monitoring [
      BestDose Products Page (Internet).
      ]. Of the centres who were unsure or not planning to transition to AUC-based monitoring within the next year, the most significant barriers included pharmacist/provider unfamiliarity (73%) and training requirements (45%). A 2019 study estimated potential cost savings associated with Bayesian versus two-sample AUC solely through fewer vancomycin concentrations, which exceeded US$ 49 000 per year [
      • Aljefri D.M.
      • Avedissian S.N.
      • Rhodes N.J.
      • Postelnick M.J.
      • Nguyen K.
      • Scheetz M.H.
      Vancomycin area under the curve and acute kidney injury: a meta-analysis.
      ]. However, this comparison is unable to account for cost savings from reductions in AKI and/or durations of hospital stay.
      Our results quantify that the use of two-sample AUC dosing or single-sample Bayesian dosing for vancomycin yields significant cost benefits compared with existing methods of trough dosing. The potential to decrease rates of AKI and reduce costs to the institution reinforce the recommendations in the new vancomycin guidelines for use of AUC-based dosing. A majority of this benefit is derived from avoidance of AKI, with additional benefit from reduced monitoring and personnel costs.
      Our base case results were robust in all sensitivity analyses and consistent with the cost–benefit expectations and probabilities from literature. In sensitivity analyses, the large impact of AKI at 48 hours to 7 days on costs is consistent with our clinical expectations, as AKI occurring early within hospitalization would have compounding effects on duration of hospitalization, clinical outcomes and overall health-care costs. Threshold analysis demonstrates that at any level of adaptation and permissible budget, there are cost benefits for an institution, on top of potential clinical benefits.
      We acknowledge various limitations to our model and analysis. First, this economic model derives probabilities from diverse literature, with varying types of infections, patient populations, institutional practices and definitions of nephrotoxicity (e.g. AKIN [Acute Kidney Injury Network Criteria] versus RIFLE [Risk, Injury, Failure, Loss, ESRD Criteria]). Furthermore, creatinine-defined nephrotoxicity in the literature does not distinguish between creatinine increases with no evidence of clinical sequelae, as opposed to AKI with severe consequences—the former costing significantly less than the latter. To reconcile this, we used average national hospitalization costs of diagnosis codes (N14.1 and N17.9) for our costs to encompass the entirety of this spectrum for our model. We did not account for empirical versus definitive therapy and chose the 48-hour cut-off as a surrogate, given that most empirical therapy is stopped at that time. Similarly, the model is unable to account for provider-specific behaviour such as de-escalation, or therapy-switching (e.g. vancomycin to daptomycin). However, for the purpose of this model, we believe that these behaviours would have marginal differences between strategies. Finally, although there is an abundance of vancomycin nephrotoxicity literature, there is little real-world reported incidence for two-sample AUC and Bayesian dosing. In addition, there is little comparative data on the number of individuals with V-AKI who progress to acute renal failure needing additional interventions. Our probabilities are not flawless—even when comparing the data used for two-sample AUC, there are differences in methodologies that potentially affect rates of nephrotoxicity, such as the exclusion of concomitant piperacillin-tazobactam use in Finch et al. [
      • Meng L.
      • Wong T.
      • Huang S.
      • Mui E.
      • Nguyen V.
      • Espinosa G.
      • et al.
      Conversion from vancomycin trough concentration–guided dosing to area under the curve–guided dosing using two sample measurements in adults: implementation at an academic medical center.
      ,
      • Finch N.A.
      • Zasowski E.J.
      • Murray K.P.
      • Mynatt R.P.
      • Zhao J.J.
      • Yost R.
      • et al.
      A quasi-experiment to study the impact of vancomycin area under the concentration–time curve-guided dosing on vancomycin-associated nephrotoxicity.
      ].
      Implementation fees such as electronic medical record integration, contracting and training are not factored in, which underestimates the overall costs. Also, we cannot account for the imperfect use of Bayesian methods or the two-sample AUC strategy. In addition, the costs in this model were generated from ICD codes and national averages. For this reason, this model does not factor the percentage of patients who enter the model with pre-existing CKD or ESRD, given the unknown costs of CKD management and/or dialysis. Instead, we performed a secondary analysis to look at changes in cost benefit in the patients who may develop CKD or ESRD after a reversible incidence of AKI during the treatment duration. Our model demonstrated that factoring in these probabilities and costs from the literature would provide even further incremental cost benefit of using AUC or Bayesian dosing over trough.
      Though not directly demonstrated by our model, we believe that there are also soft cost savings after implementation. As an example, the accumulation of population pharmacokinetic data would add to the robustness of Bayesian priors, providing further accuracy and ease of dosing, even in more complex clinical scenarios. Improved ease and trust of utilization may free up time for other clinical tasks for providers. It may be plausible to state that there are potential cost avoidances in rehospitalizations from AKI, or in long-term costs from the small number of patients who will develop CKD.
      We believe our results provide the first compelling evidence for the significant cost benefits of two-sample AUC and single-sample Bayesian dosing over the existing trough methodology. Though adoption of these newer practices may be challenging or not yet feasible for some institutions, the combination of available literature and this economic model provides useful data demonstrating significant clinical and economic benefits to AUC-based dosing, even more so with Bayesian methodology.

      Author contributions

      BL is the lead author and CG is the corresponding author. BL and CG wrote the original draft; BL, GF, MB, MN, EM, AK, GL and CG reviewed and edited the manuscript. Conceptualization was by GF, MB and CG; investigations by BL and CG; methodology by BL, GF, MB, MN, EM, AK, GL and CG; and formal analysis by BL and CG. Project administration was by GF and CG.

      Transparency declaration

      MN receives license fees from InsightRx for use of BestDose. No other authors have any conflicts of interest to report. Funding: No external funding was received.

      References

        • Rybak M.J.
        • Lomaestro B.M.
        • Rotschafer J.C.
        • Moellering R.C.
        • Craig W.A.
        • Billeter M.
        • et al.
        Vancomycin therapeutic guidelines: a summary of consensus recommendations from the infectious diseases Society of America, the American Society of Health-System Pharmacists, and the Society of Infectious Diseases Pharmacists.
        Clin Infect Dis. 2009; 49: 325-327
        • Hidayat L.K.
        • Hsu D.I.
        • Quist R.
        • Shriner K.A.
        • Wong-Beringer A.
        High-dose vancomycin therapy for methicillin-resistant Staphylococcus aureus infections: efficacy and toxicity.
        Arch Intern Med. 2006; 166: 2138-2144https://doi.org/10.1001/archinte.166.19.2138
        • Jeffres M.N.
        • Isakow W.
        • Doherty J.A.
        • Micek S.T.
        • Kollef M.H.
        A retrospective analysis of possible renal toxicity associated with vancomycin in patients with health care-associated methicillin-resistant Staphylococcus aureus pneumonia.
        Clin Ther. 2007; 29: 1107-1115
        • Neely M.N.
        • Youn G.
        • Jones B.
        • Jelliffe R.W.
        • Drusano G.L.
        • Rodvold K.A.
        • et al.
        Are vancomycin trough concentrations adequate for optimal dosing?.
        Antimicrob Agents Chemother. 2014; 58: 309-316https://doi.org/10.1128/AAC.01653-13
        • Kullar R.
        • Davis S.L.
        • Rybak M.J.
        Impact of vancomycin exposure on outcomes in patients with methicillin-resistant Staphylococcus aureus bacteremia: support for consensus guidelines suggested targets.
        Clin Infect Dis. 2011; 25: 9755-9781https://doi.org/10.1093/cid/cir124
        • Rybak M.J.
        • Le J.
        • Lodise T.P.
        • Levine D.P.
        • Bradley J.S.
        • Liu C.
        • et al.
        Therapeutic monitoring of vancomycin for serious methicillin-resistant Staphylococcus aureus infections: a revised consensus guideline and review by the American Society of Health-System Pharmacists, the Infectious Diseases Society of America, the Pediatric Infectious Diseases Society, and the Society of Infectious Diseases Pharmacists.
        Am J Health-System Pharm. 2020; 77: 835-864
        • Kufel W.D.
        • Seabury R.W.
        • Mogle B.T.
        • Beccari M.V.
        • Probst L.A.
        • Steele J.M.
        Readiness to implement vancomycin monitoring based on area under the concentration–time curve: a cross-sectional survey of a national health consortium.
        Am J Health Syst Pharm. 2019; 76: 889-894
      1. BestDose Products Page (Internet).
        (Available from:)
      2. DoseMeRx Vancomycin Page (Internet).
        (Available from:)
      3. PrecisePK Pricing (Internet).
        (Available from:)
      4. InsightRx Products Page (Internet).
        (Available from:)
        • Meng L.
        • Wong T.
        • Huang S.
        • Mui E.
        • Nguyen V.
        • Espinosa G.
        • et al.
        Conversion from vancomycin trough concentration–guided dosing to area under the curve–guided dosing using two sample measurements in adults: implementation at an academic medical center.
        Pharmacotherapy. 2019; 39: 433-442https://doi.org/10.1002/phar.2234
        • Aljefri D.M.
        • Avedissian S.N.
        • Rhodes N.J.
        • Postelnick M.J.
        • Nguyen K.
        • Scheetz M.H.
        Vancomycin area under the curve and acute kidney injury: a meta-analysis.
        Clin Infect Dis. 2019; 69: 1881-1887https://doi.org/10.1093/cid/ciz051
        • Lodise T.P.
        • Patel N.
        • Lomaestro B.M.
        • Rodvold K.A.
        • Drusano G.L.
        Relationship between initial vancomycin concentration-time profile and nephrotoxicity among hospitalized patients.
        Clin Infect Dis. 2009; 49: 507-514https://doi.org/10.1086/600884
        • Wong-Beringer A.
        • Joo J.
        • Tse E.
        • Beringer P.
        Vancomycin-associated nephrotoxicity: a critical appraisal of risk with high-dose therapy.
        Int J Antimicrob Agents. 2011; 37: 95-101https://doi.org/10.1016/j.ijantimicag.2010.10.013
        • Holford N.
        • Ma G.
        • Metz D.
        TDM is dead. Long live TCI!.
        Br J Clin Pharmacol. 2020; (Online Open Access. Available from:)https://doi.org/10.1111/bcp.14434
        • Neely M.N.
        • Kato L.
        • Youn G.
        • Kraler L.
        • Bayard D.
        • van Guilder M.
        • et al.
        Prospective trial on the use of trough concentration versus area under the curve to determine therapeutic vancomycin dosing.
        Antimicrob Agents Chemother. 2018; 62 (pii:e02042-17)https://doi.org/10.1128/AAC.02042-17
        • Kim N.H.
        • Koo H.L.
        • Choe P.G.
        • Cheon S.
        • Kim M.S.
        • Lee M.J.
        • et al.
        Inappropriate continued use of empirical vancomycin in a hospital with a high prevalence of MRSA.
        Antimicrob Agents Chemother. 2015; 59: 811-817https://doi.org/10.1128/AAC.04523-14
        • Olson J.
        • Stockmann C.
        • Hersh A.L.
        • Anderson C.
        • Zobell J.
        • Thorell E.
        Optimizing vancomycin prescribing through a pharmacist driven monitoring intervention at a children’s hospital.
        Open Forum Infect Dis. 2016; 3https://doi.org/10.1093/ofid/ofw172.1472
        • Minejima E.
        • Choi J.
        • Beringer P.
        • Lou M.
        • Tse E.
        • Wong-Beringer A.
        Applying new diagnostic criteria for acute kidney injury to facilitate early identification of nephrotoxicity in vancomycin-treated patients.
        Antimicrob Agents Chemother. 2011; 55: 3278-3283https://doi.org/10.1128/AAC.00173-11
        • Wang H.E.
        • Munter P.
        • Chertow G.M.
        • Warnock D.G.
        Acute kidney injury and mortality in hospitalized patients.
        Am J Nephrol. 2012; 35: 349-355https://doi.org/10.1159/000337487
      5. Healthcare Cost and Utilization Project (HCUP), National Inpatient Sample (NIS) [Internet]. Agency for Healthcare Research and Quality (AHRQ), 2014 (Available from:)
        • Patel N.
        • Huang D.
        • Lodise T.
        Potential for cost saving with iclaprim owing to avoidance of vancomycin-associated acute kidney injury in hospitalized patients with acute bacterial skin and skin structure infections.
        Clin Drug Investig. 2018; 38: 935-943https://doi.org/10.1007/s40261-018-0686-5
        • Bureau of Labor Statistics
        • U.S. Department of Labor
        Occupational Outlook Handbook.
        Pharmacists (Internet). 2019; (Available from:)
        • Bureau of Labor Statistics
        • U.S. Department of Labor
        Occupational Outlook Handbook.
        Phlebotomists (Internet). 2019; (Available from:)
        • Neumann P.J.
        • Sanders G.D.
        • Russell L.B.
        • Siegel J.E.
        • Ganiats T.G.
        Cost-effectiveness in health and medicine.
        2nd ed. Oxford University Press, New York2016
        • Rangaswamy D.
        • Sud K.
        Acute kidney injury and disease: Long-term consequences and management.
        Nephrology. 2018; 23: 969-980https://doi.org/10.1111/nep.13408
        • Bucaloiu I.D.
        • Kirchner H.L.
        • Norfolk E.R.
        • Hartle II J.E.
        • Perkins R.M.
        Increased risk of death and de novo chronic kidney disease following reversible acute kidney injury.
        Kidney Int. 2012; 81: 477-485https://doi.org/10.1038/ki.2011.405
        • Golestaneh L.
        • Alvarez P.J.
        • Reaven N.L.
        • Funk S.E.
        • McGaughey K.J.
        • Romero A.
        • et al.
        All-cause costs increase exponentially with increased chronic kidney disease stage.
        Am J Manag Care. 2017; 23: S163-S172
        • Finch N.A.
        • Zasowski E.J.
        • Murray K.P.
        • Mynatt R.P.
        • Zhao J.J.
        • Yost R.
        • et al.
        A quasi-experiment to study the impact of vancomycin area under the concentration–time curve-guided dosing on vancomycin-associated nephrotoxicity.
        Antimicrob Agents Chemother. 2017; 61 (pii: e01293-17)https://doi.org/10.1128/AAC.01293-17

      Linked Article