https://orcid.org/0000-0001-8846-3664
Abstract
Objective
Contrast-sparing strategies have been developed for percutaneous coronary intervention (PCI) patients at increased risk of contrast-induced acute kidney injury (CI-AKI), and numerous CI-AKI risk prediction models have been created. However, the potential clinical and economic consequences of using predicted CI-AKI risk thresholds for assigning patients to contrast-sparing regimens have not been evaluated. We estimated the clinical and economic consequences of alternative CI-AKI risk thresholds for assigning Medicare PCI patients to contrast-sparing strategies.
Methods
Medicare data were used to identify inpatient PCI from January 2017 to June 2021. A prediction model was developed to assign each patient a predicted probability of CI-AKI. Multivariable modeling was used to assign each patient two marginal predicted values for each of several clinical and economic outcomes based on (1) their underlying clinical and procedural characteristics plus their true CI-AKI status in the data and (2) their characteristics plus their counterfactual CI-AKI status. Specifically, CI-AKI patients above the predicted risk threshold for contrast-sparing were reassigned their no CI-AKI (counterfactual) outcomes. Expected event rates, resource use, and costs were estimated before and after those CI-AKI patients were reassigned their counterfactual outcomes. This entailed bootstrapped sampling of the full cohort.
Results
Of the 542,813 patients in the study cohort, 5,802 (1.1%) had CI-AKI. The area under the receiver operating characteristic curve for the prediction model was 0.81. At a predicted risk threshold for CI-AKI of >2%, approximately 18.0% of PCI patients were assigned to contrast-sparing strategies, resulting in (/100,000 PCI patients) 121 fewer deaths, 58 fewer myocardial infarction readmissions, 4,303 fewer PCI hospital days, $11.3 million PCI cost savings, and $25.8 million total one-year cost savings, versus no contrast-sparing strategies.
Limitations
Claims data may not fully capture disease burden and are subject to inherent limitations such as coding inaccuracies. Further, the dataset used reflects only individuals with fee-for-service Medicare, and the results may not be generalizable to Medicare Advantage or other patient populations.
Conclusions
Assignment to contrast-sparing regimens at a predicted risk threshold close to the underlying incidence of CI-AKI is projected to result in significant clinical and economic benefits.
Introduction
Contrast-induced acute kidney injury (CI-AKI) is a common and costly complication of percutaneous coronary intervention (PCI)Citation1, resulting in increased morbidity, mortality, medical resource use, length of stay, and costCitation2–8. Consequently, contrast-sparing strategies, including those with intravascular ultrasound (IVUS) imaging, have been developed for patients considered at increased riskCitation9–11. However, in the current fiscal environment, where budgetary constraints play an important role in the uptake of new medical technology and programs, widespread adoption of contrast-sparing strategies in PCI may depend on the extent to which additional costs to both providers and payers are offset by savings associated with the clinical and economic benefits of reducing CI-AKI.
Numerous prediction models for CI-AKI in PCI have been developed, and one practical application involves assigning PCI patients to a contrast-sparing strategy based on their predicted risk. Thus far, evaluation of individual prediction models, and comparison of their relative performance, has largely been performed by estimating the area under the receiver operating characteristic curve (ROC)Citation12–16, with most CI-AKI risk models producing an area under the ROC of between 0.7 and 0.9Citation12,Citation13. While this method of comparison provides insight into the relative accuracy of alternative models and the tradeoffs between sensitivity and specificity along a continuum of predicted risk thresholds for CI-AKI, it provides little guidance for how to select the most efficient and effective risk threshold, i.e. the one at which the safety, effectiveness, efficiency, and cost-effectiveness of assignment to a contrast-sparing strategy would be optimized across the PCI population. And, although a recent study using the National Cardiovascular Data Registry compared CI-AKI rates at alternative risk thresholds based on the Cath PCI risk prediction modelCitation17, and another conducted at a single center estimated the cost savings associated with assignment to contrast-sparing based on the Cath PCI modelCitation18, no study has directly explored the medical resource, cost, and long-term clinical consequences of selecting alternative CI-AKI risk thresholds for assignment to contrast-sparing.
In this context, it may be more useful for decision-makers to consider how many true positive, false positive, true negative, and false negative CI-AKI cases prediction models produce at any given decision threshold, because each of these “results” can be linked to clinical and economic outcomes specific to that model. For instance, true positives, a function both of model sensitivity and the underlying incidence of CI-AKI, represent an opportunity to prevent CI-AKI, its outcomes, and costs, if these patients are assigned to a contrast-sparing strategy.
Large administrative and medical claims databases containing both clinical and economic information may be particularly useful for this sort of investigation. In these data, real-world clinical and economic outcomes of the condition predicted can be estimated directly from the data used to develop the prediction model. Furthermore, these databases contain enough patients to investigate the possible clinical and economic consequences of adopting alternative predicted risk thresholds for medical decision-making.
The objective of this study, therefore, was to estimate the clinical and economic consequences of assigning patients undergoing PCI to a contrast-sparing regimen based on their predicted risk of CI-AKI.
Methods
Study design and patient selection
Medicare claims and administrative data were used to identify patients who underwent inpatient PCI between 1 January 2017 and 1 June 2021. Patients were included if they had at least 12 months of continuous Medicare fee-for-service (FFS) enrollment before the index PCI admission (baseline period). PCI was identified using International Classification of Diseases, 10th Revision, Procedure Coding System (ICD-10-PCS)Citation19 procedure codes (Supplemental Materials) in the Medicare claims. Since ICD-10-PCS procedure codes are used for classifying procedures in hospital inpatient settings, those undergoing outpatient PCI were excluded.
Prediction model for contrast-induced acute kidney injury
Overview
The first step of the evaluation entailed developing a risk prediction model for CI-AKI, which was defined as the presence of International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) diagnosis code N14.1 “Nephropathy induced by other drugs, medicaments and biological substances” during the index PCI admission. Independent variables considered for the prediction model included patient demographic characteristics, underlying clinical conditions, e.g. stage of chronic kidney disease (CKD) (available from ICD-10-CM diagnosis codes N18.1-N18.6 and N18.9), and lesion/procedural characteristics of the index PCI. These were identified using ICD-10 diagnosis and procedure codes in the Medicare claims from both the baseline period and the index admission (Supplemental Materials).
Selection of predictors
Patients meeting the study inclusion criteria were randomly assigned 1:1 to a development or a validation cohort. The development cohort was used to identify relevant factors and calibrate the predicted risk of CI-AKI from the observed occurrence of CI-AKI in patients included in the development cohort. This model was later applied to the validation cohort to calculate how well the model correctly predicted CI-AKI in those patients. We report prediction metrics in both cohorts for completeness. In the development cohort, logistic regression with backwards stepwise elimination (at p ≤ 0.1) was used to identify patient factors statistically significantly associated with CI-AKI. In instances where candidate predictor variables contained more than two levels, e.g. stage of chronic kidney disease, the entire variable was forced to be considered for inclusion during stepwise elimination.
Calibration and validation
Assessment of the initial calibration of the model was performed by sorting patients from lowest to highest predicted probability of CI-AKI (pCI-AKI), generating intervals each containing 100 patients, and plotting the mean observed (i.e. the actual percent of patients who experienced CI-AKI) versus the mean pCI-AKI in each interval using a locally estimated scatterplot smoothing (LOESS) curve (Supplemental Materials). The deviance of the LOESS curve from the line representing equality of observed and mean values was examined to detect the under- or over-prediction of the model. Re-calibration to correct detected overprediction at higher observed CI-AKI probabilities was performed by first calculating the logit (ln[pCI-AKI/1 – pCI-AKI]) of pCI-AKI as for each patient, using pCI-AKI from the initial model, and then re-running the logistic regression on CI-AKI with the logit(p) polynomial among the predictor variables to generate a re-calibrated pCI-AKI for each patient. Using the new pCI-AKI values, the steps for initial calibration were repeated (Supplemental Materials) and the goodness-of-fit was then evaluated using both the Pearson and the Hosmer-Lemeshow tests. ROC curves with C-statistics (area under the ROC) for both the development and validation samples were generated (Supplemental Materials) to assess the validity of the model and its predictive power.
Sensitivity, specificity, and number needed to treat
To further describe the performance of the model, its sensitivity, specificity, and number needed to treat (NNT), defined as the number of patients needed to prevent one CI-AKI within a hypothetical contrast-sparing strategy that prevented all CI-AKI, were plotted across predicted risk thresholds above which patients would be included in that contrast-sparing strategy.
Clinical and economic outcomes of CI-AKI
Outcomes variables
The second step of this analysis entailed constructing clinical and economic outcome variables for each patient in the cohort. Clinical outcomes consisted of all-cause mortality and readmission for myocardial infarction (MI) during one year after the index PCI. Economic outcomes included the index PCI length of stay (LOS), index PCI cost to Medicare, and cumulative total cost to Medicare during one year after the index PCI (including the cost of the index admission). Costs were defined as Medicare paid amounts for facility and professional services. However, since the Medicare 100% files do not contain most professional claims for office-based services, these were not included in the totals. All costs were inflated to 2022 USD using the Medical Care Component of the Consumer Price IndexCitation20.
Estimation and assignment of outcomes
Multivariable logistic regression was used to estimate the predicted probability of death and readmission for MI at one year for each patient in the cohort. Multivariable Poisson regression was used to estimate the predicted index admission LOS, and multivariable generalized linear modeling with a gamma distribution and log link was used to estimate the index admission total cost and the cumulative cost at one year. Each patient was assigned two marginal predicted values for each outcome: the first was based on their unique vector of underlying clinical and procedural characteristics plus their actual observed CI-AKI status in the data; and the second was based on their unique characteristics plus their “counterfactual” CI-AKI status, e.g. no CI-AKI in a patient with observed CI-AKI. The relevance of assigning two of each value to each patient is clarified below.
Simulated outcomes of contrast-sparing
The third step consisted of estimating the clinical and economic outcomes of selecting alternative CI-AKI risk thresholds to assign patients to a contrast-sparing strategy.
Selection of predicted risk thresholds for assignment
First, a series of risk thresholds for CI-AKI ranging from 0 (all patients assigned to contrast-sparing) to 1 (no patients assigned to contrast sparing) were selected, and all patients with a predicted probability of CI-AKI exceeding that threshold were assigned to a contrast-sparing strategy. Selection of the risk thresholds was based on an examination of the sensitivity, specificity, and NNT curves from the CI-AKI prediction model. This examination concentrated on the region where the sensitivity and specificity curves crossed and where the decline in NNT associated with higher risk thresholds began to level off. This also coincided with the overall incidence of CI-AKI in the study population.
Simulated benefits of contrast-sparing
Second, for each predicted risk threshold above, 100 bootstrapped random samples of 100,000 patients (with replacement) each were selected from the study population. To simulate the clinical and economic benefits of contrast-sparing, all patients above that risk threshold who also had CI-AKI were re-assigned their counterfactual outcomes values. In re-assigning all CI-AKI patients their counterfactual outcomes, it was assumed the contrast-sparing strategy would prevent all cases of CI-AKI. After this reassignment, event predicted probabilities, LOS, and costs were summed within each of the bootstrapped samples (at each risk threshold), and then the average sums (with 95% confidence intervals) for each outcome were calculated across the 100 samples. Average sums at each risk threshold were compared to the base-case scenario in which no patients were reassigned to contrast-sparing.
Secondary analyses
In addition to the primary analysis, which selected the bootstrapped samples from the full study cohort and incorporated the definition of CI-AKI described above, secondary analyses repeating the first three steps were conducted in which the definition of CI-AKI was expanded to include “Acute Kidney Failure, not otherwise specified,” and in which patients aged <65 years and those with stage 5 CKD or dialysis-dependent CKD (those with an ICD-10-CM diagnosis code of N18.5 “CKD stage 5” or N18.6 “End stage renal disease”) were excluded from the study cohort prior to development of the prediction model and subsequent steps.
All analyses were conducted using STATA v17 (StataCorp. 2021; College Station, TX, USA).
Results
Patients
There were 542,813 patients in the study cohort, of whom 5,802 (1.1%) were diagnosed with CI-AKI. Overall, the mean age at PCI was 72.1, 14.9% were age <65, 62.4% were male, 33% had an ElixhauserCitation21 score ≥5, and 32.3% had underlying CKD. Those with CI-AKI tended to be older, to have underlying CKD, and had one or more indicators of greater procedural complexity, including calcification or chronic total occlusion (Supplemental Materials).
Prediction model
The final prediction model included a combination of patient demographic, baseline clinical, and lesion/procedural characteristics, among which those with the highest odds ratios of CI-AKI included age, CKD, heart failure, MI, and several indicators of procedural complexity (). Initially, the model over-predicted CI-AKI (Supplemental Materials) in higher observed probability segments of the population. However, this was largely corrected by re-calibration (Supplemental Materials) and confirmed by a Hosmer–Lemeshow goodness-of-fit test with 10 intervals (p = 0.20). C-statistics for the development and validation samples were 0.806 and 0.803, respectively, demonstrating strong predictive power (Supplemental Materials), and within the range reported in the literature. A plot of the sensitivity, specificity, and NNT per CI-AKI avoided suggests the optimal predicted risk threshold for contrast-sparing might lay in the region of 0.01 to 0.02, below which the NNT increased rapidly ().
Figure 1. Prediction model for contrast-induced acute kidney injury.
Figure 2. Performance of the prediction model.
Outcomes of contrast-sparing
Primary analysis
The results of the primary analysis, which entailed bootstrap sampling from the entire cohort and the “restricted” definition of CI-AKI are presented in and (deaths and readmission for MI) and Citation4 (index LOS, index cost, and cumulative one-year cost). As shown in the upper panel of , in the absence of any enrollment in contrast sparing regimens (base-case), there were 592,367 hospital days for the PCI admission, $2.65 billion in hospital costs for the PCI admission, $5.30 billion in cumulative one-year costs (including the PCI admission), 12,392 (12.4%) readmissions for MI, and 15,430 (15.4%) deaths at one year (all per 100,000 PCI patients). Lowering the risk threshold for contrast sparing resulted in reduced deaths and MIs () and lower resource use and costs (). In general, these benefits were not realized until the decision threshold for assignment to contrast sparing was below a threshold of 0.05 (meaning that the risk of CI-AKI was 0.05 or greater) ( and ), which is consistent with the cumulative distribution of the predicted probabilities where very few patients had values above 0.05 (not shown).
Figure 3. Projected impact of contrast sparing on (a) mortality and (b) MI readmission by CI-AKI risk threshold for assignment.
Figure 4. Projected impact of contrast sparing on (a) length of stay, (b) index admission cost and (c) cumulative one-year cost by CI-AKI risk threshold for assignment.
Table 1. Projected clinical and economic outcomes of contrast-sparing.
Lowering the predicted risk (decision) threshold for enrollment in a contrast-sparing protocol to 0.02 (i.e. the risk of CI-AKI is 0.02 or greater) resulted in 121 fewer deaths, 58 fewer MI readmissions, 4,303 fewer PCI days hospitalized, $11.3 million lower hospital costs for the PCI admission, and $25.8 million lower cumulative one-year costs per 100,000 patients (, Lower Panel). As the predicted risk threshold was lowered, the proportion of patients enrolled in contrast-sparing increased faster than the reduction in the number of deaths, resulting in an increase in the NNT per death avoided to 149 at a risk threshold of 0.02.
Secondary analyses
Results from the secondary analyses followed similar patterns to those in the primary analyses, with benefits of assignment to contrast-sparing beginning only below a risk threshold of CI-AKI of 0.05 or greater (Supplemental Materials). Changing the definition of CI-AKI had little impact on the findings. However, applying the additional exclusion criteria to the study cohort prior to developing the prediction model had the net effect of reducing the number of deaths and MI readmissions prevented, and resource and cost savings across the entire range of risk threshold for assignment to contrast-sparing (Supplemental Materials). It also increased the NNTs for preventing death at each risk threshold.
Conclusions
More than 30 separate models have been developed for predicting the risk of CI-AKI in individual patients who undergo PCICitation12. The conventional approach to evaluating a model’s performance, and to comparing among models, has been to estimate its ROC curve and attendant C-statistic. These methods depict the tradeoff between sensitivity and specificity – or true positive and false positive rates – across the continuum of probability thresholds above which patients are considered positive according to this “test.” While a necessary step, such an analysis is not sufficient for decision-making, because measures of model accuracy provide little insight into where or when an alternative treatment modality is prescribed. In the context of CI-AKI prediction models in PCI, pragmatically this applies to selecting the risk threshold above which patients are assigned to a contrast-sparing strategy. Others have recognized this limitation and recommended a broader approach to evaluating the performance of prediction models using a decision-analytic frameworkCitation22 in which clinical and economic outcomes are attached to the test results, e.g. the true positive patients represent potential cases avoided with clinical and economic benefits if they are assigned to an effective prevention strategy, and the tradeoffs in these outcomes are then evaluated across a range of plausible decision thresholds.
In this study, real-world data from a large cohort of Medicare beneficiaries who underwent inpatient PCI were used to estimate the clinical and economic benefits of assigning patients to a contrast-sparing strategy across a range of plausible risk thresholds for predicted CI-AKI. Model accuracy was adequate, with an area under the ROC > 0.80 for both development and validation samples, in line with previous risk-prediction modelsCitation12,Citation13. The findings indicate that, compared with no assignment to a contrast-sparing strategy, selecting a predicted CI-AKI risk threshold of >0.02 for assignment to contrast-sparing could result in significant resource and cost savings, and deaths avoided. The results further suggest that selecting a risk threshold above >0.05 is unlikely to result in appreciable benefits and selecting one below 0.01 likely would result in diminishing benefits relative to the number of patients assigned to contrast-sparing.
Extending risk-prediction modeling by considering clinical and economic benefits provides a more actionable framework for clinicians than risk scores alone. Furthermore, it facilitates a more patient-centered approach within PCI. Others have successfully employed patient-centered strategies to prevent CI-AKI within PCI patientsCitation17,Citation18, demonstrating that personalizing risks is feasible and effective within this population. Such strategies are also cost-saving; Amin et al., after implementing a process to assess AKI risk before PCI and adjust contrast use accordingly, observed an average savings in hospital costs of $2,171 (2016 USD) across all PCI patients, and of $5,697 among patients at a high risk of AKICitation18. Given the authors reported mean hospital costs of approximately $20,000 for all patients and $25,000 for high risk patients, these savings reflect a 10% to 20% reduction in hospital costs.
The strengths of this study include the use of a large dataset and the use of re-calibration in the development of the prediction model. Additionally, the robust simulation methods to estimate the economic benefits of contrast-sparing regimens. However, the results should be viewed in light of the limitations as well. First, while claims data allow for the identification of procedures and costs, they rely on the accuracy of coding and the information available on the claims. For example, while we report a CI-AKI incidence of 1.1%, another claims-based analysis (that uses more ICD-10-CM diagnosis and procedure codes to identify CI-AKI) observed an incidence of 3.1%Citation4, and other studies – including those that use lab-based definitions – have reported much higher incidencesCitation8,Citation23. Furthermore, the sensitivity and specificity of the ICD-10-CM code N14.1 to identify CI-AKI has not been established, and its description (“Nephropathy induced by other drugs, medicaments and biological substances”) is inclusive of more than just contrast. However, when limited to a cohort of individuals receiving PCI, it seems reasonable to assume that contrast would be the most likely “other drug, medicament, or biologic substance.” The dataset also lacked professional claims for office-based services, meaning that some costs may have been omitted from the analysis. Additionally, the claims reflect only those enrolled in FFS Medicare, so the results may not be generalizable to the Medicare Advantage population. For the sake of the simulation, the theoretical model assumes that a contrast sparing regimen completely removes the risk of CI-AKI, which may not reflect reality. Additionally, we did not attempt to estimate the amount of reduction that would be appropriate for individual patients, and we refer to “contrast sparing” as reflecting physician-determined reduction based on individual patient risk profiles. It should also be noted that the model we present includes as an input anatomy which is knowable only after contrast has been given. Without this information, the accuracy of pre-procedural prediction would be reduced. Multiple contrast sparing strategies exist in practice, such as using a safe contrast limit, using intravascular ultrasound to minimize contrast use, or staging procedures when performing multivessel PCI. While the goal is to perform procedures with better contrast sparing, we recognize the potential risks associated with contrast sparing strategies. Occasionally staging procedures could lead to incomplete revascularization and continued symptoms of angina and dyspnea among patients, as well as risk of cardiovascular events such as myocardial infarction while waiting for staged procedures. Occasionally, intravascular imaging is used to supplant contrast, which has a slightly higher risk of coronary dissections as compared to angiography using contrast. Finally, the analysis limits its examination to total costs, without consideration of changes to specific cost sources. For example, cost-sparing regimens may incur additional costs, such as those related to the use of IVUS imaging. However, since the total cost of the index hospitalization is included, these additional costs are captured in the aggregate, so the cost savings identified in the results accounts for these costs.
Using predicted risk thresholds combined with expected event rates, resource use, and costs to assign patients to cost-sparing regimens may produce significant clinical and economic benefits. Employing similar methods may assist clinical decisions regarding patients undergoing PCI who are at risk for CI-AKI.
Transparency
Declaration of financial/other relationships
At the time of writing, RIG was a full-time employee of, and shareholder in, Boston Scientific. AB, AMM, and LMH are full-time employees of, and shareholders in, Boston Scientific. CAS is a consultant to Boston Scientific. At the time of writing, APA was Associate Professor of Medicine, Geisel School of Medicine at Dartmouth. APA is Section Chief, Interventional Cardiology, Rush University Medical Center, Interventional Cardiology Fellowship Program Director, Associate Professor of Medicine/Cardiology, Rush College of Medicine. APA was not compensated for his participation in this study. Peer reviewers on this manuscript have no relevant financial or other relationships to disclose. Two Editorial Board Members helped with adjudicating the final decision on this paper.
Author contributions
All authors meet the International Committee of Medical Journal Editors authorship guidelines and have reviewed and agree with the content of the article.
Supplemental Material
Download MS Word (2.1 MB)Additional information
Funding
References
- Nash K, Hafeez A, Hou S. Hospital-acquired renal insufficiency. Am J Kidney Dis. 2002;39(5):930–936. doi: 10.1053/ajkd.2002.32766.
- Abellás-Sequeiros RA, Raposeiras-Roubín S, Abu-Assi E, et al. Mehran contrast nephropathy risk score: is it still useful 10 years later? J Cardiol. 2016;67(3):262–267. doi: 10.1016/j.jjcc.2015.05.007.
- Araujo GN, Wainstein MV, McCabe JM, et al. Comparison of two risk models in predicting the incidence of contrast-induced nephropathy after percutaneous coronary intervention. J Interv Cardiol. 2016;29(5):447–453. doi: 10.1111/joic.12315.
- Aubry P, Brillet G, Catella L, et al. Outcomes, risk factors and health burden of contrast-induced acute kidney injury: an observational study of one million hospitalizations with image-guided cardiovascular procedures. BMC Nephrol. 2016;17(1):167. doi: 10.1186/s12882-016-0385-5.
- Chong E, Poh KK, Liang S, et al. Risk factors and clinical outcomes for contrast-induced nephropathy after percutaneous coronary intervention in patients with normal serum creatinine. Ann Acad Med Singap. 2010;39(5):374–380. doi: 10.47102/annals-acadmedsg.V39N5p374.
- Chou RH, Huang PH, Hsu CY, et al. CHADS2 score predicts risk of contrast-induced nephropathy in stable coronary artery disease patients undergoing percutaneous coronary interventions. J Formos Med Assoc. 2016;115(7):501–509. doi: 10.1016/j.jfma.2015.12.008.
- Lian D, Liu Y, Liu YH, et al. Pre-Procedural risk score of contrast-induced nephropathy in elderly patients undergoing elective coronary angiography. Int Heart J. 2017;58(2):197–204. doi: 10.1536/ihj.16-129.
- Caruso M, Balasus F, Incalcaterra E, et al. Contrast-induced nephropathy after percutaneous coronary intervention in simple lesions: risk factors and incidence are affected by the definition utilized. Intern Med. 2011;50(9):983–989. doi: 10.2169/internalmedicine.50.4976.
- Mehran R, Faggioni M, Chandrasekhar J, et al. Effect of a contrast modulation system on contrast media use and the rate of acute kidney injury after coronary angiography. JACC Cardiovasc Interv. 2018;11(16):1601–1610. doi: 10.1016/j.jcin.2018.04.007.
- Almendarez M, Gurm HS, Mariani J Jr, et al. Procedural strategies to reduce the incidence of contrast-induced acute kidney injury during percutaneous coronary intervention. JACC Cardiovasc Interv. 2019;12(19):1877–1888. doi: 10.1016/j.jcin.2019.04.055.
- Azzalini L, Laricchia A, Regazzoli D, et al. Ultra-Low contrast percutaneous coronary intervention to minimize the risk for contrast-induced acute kidney injury in patients with severe chronic kidney disease. J Invasive Cardiol. 2019;31(6):176–182.
- Allen DW, Ma B, Leung KC, et al. Risk prediction models for contrast-induced acute kidney injury accompanying cardiac catheterization: systematic review and meta-analysis. Can J Cardiol. 2017;33(6):724–736. doi: 10.1016/j.cjca.2017.01.018.
- Silver SA, Shah PM, Chertow GM, et al. Risk prediction models for contrast induced nephropathy: systematic review. BMJ. 2015;351:h4395. doi: 10.1136/bmj.h4395.
- Duan C, Cao Y, Liu Y, et al. A new preprocedure risk score for predicting contrast-induced acute kidney injury. Can J Cardiol. 2017;33(6):714–723. doi: 10.1016/j.cjca.2017.01.015.
- Gurm HS, Seth M, Kooiman J, et al. A novel tool for reliable and accurate prediction of renal complications in patients undergoing percutaneous coronary intervention. J Am Coll Cardiol. 2013;61(22):2242–2248. doi: 10.1016/j.jacc.2013.03.026.
- Mo H, Ye F, Chen D, et al. A predictive model based on a new CI-AKI definition to predict contrast induced nephropathy in patients with coronary artery disease with relatively normal renal function. Front Cardiovasc Med. 2021;8:762576. doi: 10.3389/fcvm.2021.762576.
- Malik AO, Amin A, Kennedy K, et al. Patient-centered contrast thresholds to reduce acute kidney injury in high-risk patients undergoing percutaneous coronary intervention. Am Heart J. 2021;234:51–59. doi: 10.1016/j.ahj.2020.12.013.
- Amin AP, Crimmins-Reda P, Miller S, et al. Reducing acute kidney injury and costs of percutaneous coronary intervention by patient-centered, evidence-based contrast use. Circ Cardiovasc Qual Outcomes. 2019;12(3):e004961.
- World Health Organization. ICD-10: international statistical classification of diseases and related health problems: tenth revision. 2nd ed. Geneva: World Health Organization; 2004.
- US Bureau of Labor Statistics. Measuring price change in the CPI: medical care. 2022; [accessed 2022 Apr 15]. Available from: https://www.bls.gov/cpi/factsheets/medical-care.htm
- Elixhauser A, Steiner C, Harris DR, et al. Comorbidity measures for use with administrative data. Med Care. 1998;36(1):8–27. doi: 10.1097/00005650-199801000-00004.
- Steyerberg EW, Pencina MJ, Lingsma HF, et al. Assessing the incremental value of diagnostic and prognostic markers: a review and illustration. Eur J Clin Invest. 2012;42(2):216–228. doi: 10.1111/j.1365-2362.2011.02562.x.
- van der Molen AJ, Reimer P, Dekkers IA, et al. Post-contrast acute kidney injury - part 1: definition, clinical features, incidence, role of contrast medium and risk factors: recommendations for updated ESUR contrast medium safety committee guidelines. Eur Radiol. 2018;28(7):2845–2855. doi: 10.1007/s00330-017-5246-5.