SciELO - Scientific Electronic Library Online
 
vol.49 issue2Assessment of changes in the electrical activity of the brain during general anesthesia using portable electroencephalographyPediatric airway: What is new in approaches and treatments? author indexsubject indexarticles search 		Home Pagealphabetic serial listing  

Services on Demand

Journal

Article

Indicators

Related links

  • On index processCited by Google
  • Have no similar articlesSimilars in SciELO
  • On index processSimilars in Google

Share

Colombian Journal of Anestesiology

Print version ISSN 0120-3347On-line version ISSN 2256-2087

Rev. colomb. anestesiol. vol.49 no.2 Bogotá Apr./June 2021  Epub May 24, 2021

https://doi.org/10.5554/22562087.e944 

SYSTEMATIC REVIEW

Use of matching methods in observational studies with critical patients and renal outcomes. Scoping review

Henry Oliveros-Rodrígueza  b  * 
http://orcid.org/0000-0002-7632-906X

Giancarlo Bu¡tragoc 

Paola Castellanos-Saavedrad 

a PhD Program in Clinical Epidemiology, Department of Clinical Epidemiology and Biostatistics, Pontificia Universidad Javeriana. Bogotá, Colombia.

b Medical School, Universidad de la Sabana. Chía, Colombia.

c Clinical Epidemiology and Biostatistics Department, Pontificia Universidad Javeriana. Bogotá, Colombia.

d Medical School, Pontificia Universidad Javeriana. Bogotá, Colombia.

Abstract

Introduction

Although the use of matching techniques in observational studies has been increasing, it is not always adequate. Clinical trials are not always feasible in critically ill patients with renal outcomes, and observational studies are an important alternative.

Objective

Through a scoping review, determine the available evidence on the use of matching methods in studies involving critically ill patients and renal outcome assessments.

Methods

Medline, Embase, and Cochrane databases were used to identify articles published between 1992 and week 10 of 2020. Studies had to assess different exposures in critically ill patients with renal outcomes using propensity score matching.

Results

Most publications are cohort studies 94 (94.9%), with five (5.1%) reporting cross-sectional studies. The main pharmacological intervention was the use of antibiotics in seven studies (7.1%) and the main risk factor studied was renal injury prior to ICU admission in 10 studies (10.1%). The use of standardized means to assess balance of baseline characteristics was found in only 28 studies (28.2%). Most studies 95 (96%) used logistic regression to calculate the propensity score.

Conclusion

Major inconsistencies were observed regarding the use of methods and reporting of findings. Considerations related to the use of propensity score matching methods and reporting of findings are summarized.

Keywords: Critical illness; critical care; renal replacement therapy; acute kidney injury; propensity score

Resumen

Introducción

El uso de técnicas de emparejamiento en estudios observacionales ha ido en aumento y no siempre se usa adecuadamente. Los experimentos clínicos no siempre son factibles en los pacientes críticos con desenlaces renales, por lo que los estudios observacionales son una alternativa importante.

Objetivo

Mediante una revisión de alcance, determinar la evidencia disponible sobre la utilización de los métodos de emparejamiento en los estudios que incluyen pacientes críticamente enfermos y que evalúan desenlaces renales.

Métodos

Se utilizaron las bases de datos de Medline, Embase, y Cochrane para identificar artículos publicados entre 1992 y 2020 hasta la semana 10, que estudiaran diferentes exposiciones en el paciente crítico con desenlaces renales y utilizaran métodos de emparejamiento por propensión.

Resultados

La mayoría de las publicaciones (94) son estudios de Cohorte (94,9 %), cinco estudios (5,1 %) correspondieron a cross-secctional. La principal intervención farmacológica fue el uso de antibióticos en siete estudios (7,1 %) y el principal factor de riesgo estudiado fue la lesión renal previa al ingreso a UCI en diez estudios (10,1 %). El balance entre las características de base evaluado mediante medias estandarizadas se encontró solo en 28 estudios (28,2 %). La mayoría de los estudios 95 (96 %) utilizaron regresión logística para calcular el índice de propensión.

Conclusiones

Se observaron grandes inconsistencias en la utilización de los métodos y en el informe de los hallazgos. Se hace un resumen de los aspectos por considerar en la utilización de los métodos y reporte de los hallazgos con el emparejamiento por índice de propensión.

Palabras clave: Enfermedad crítica; cuidados intensivos; terapia de reemplazo renal; lesión renal aguda; puntaje de propensión

What do we know about this problem?

In observational studies, matching methods are increasingly being used to control for confounding variables derived from non-random assignment of individuals to exposure groups. Multivariate regression or matching algorithms are the methods most frequently used to address confounding risk.

What does this new study add?

The use of these methods, such as the propensity score, is not always adequate, as is also the case with the reporting of study findings. A scoping review was conducted in order to assess the use and reporting of these methods, describing the aspects that need to be taken into account in the report.

INTRODUCTION

A clinical trial is the quintessential design to determine, in most cases, causal associations between exposures and the outcomes that it seeks to establish. Random subject assignment, coupled with adequate sample size, allows for homogenous distribution of baseline characteristics among the individuals in the comparison groups. However, controlled clinical trials are not always feasible due to ethical reasons, costs, or other and, therefore, observational studies are an important option in clinical research, despite the higher probability of obtaining biased estimations 1. Hence the need to use methods designed to control for confounding variables.

Causal inference techniques have been developed over the past few years, propensity score matching (PSM) being one of the most frequently used to control confounding variables. The goal of matching using the propensity index is to achieve a balance of baseline characteristics in the comparison groups 2-6.

Success in achieving adequate balance of baseline variables depends on the correct specification of the propensity score (PS) model. This requires including all potential confounding variables, i.e., those that could modify both the probability of being exposed or receiving the treatment, as well as the occurrence of the outcome 7-9.

The boom of scientific papers that describe the use of these methods requires editors, peer reviewers and researchers in the area of critical medicine to be knowledgeable of how studies that include those methods should be reported. Publications using PSM have been reviewed in other clinical settings. In the urological literature, Nayan found deficient reporting 10 and, more recently, Yao 11 conducted a review of the application of PS methods in treated cancer patients assessing inconsistencies. However, no studies exploring the use of matching methods in critically ill patients with renal outcomes have been found to date.

In this work, an exploratory systematic review was conducted based on the available evidence in critical medicine of the studies that assessed the effects of different interventions and risk factors in renal outcomes, using matching techniques. The aim is to describe the methods employed and give recommendations regarding their adequate use.

METHODOLOGY

The protocol followed the PRISMA guidelines for Scoping Reviews.

Eligibility criteria

The studies included in the review had to assess different exposures in critically ill patients and to describe renal outcomes. PSM techniques should have been included in the analysis.

The Medline, Embase and Cochrane databases were queried for the period between week 1 of 1990 and week 10 of 2020, using the following search strategy:

((((renal dysfunction) OR (renal outcomes)) OR (acute renal failure)) AND ((((critical care patients) OR (intensive care unit)) OR (critical illness)) OR (critical ill))) AND ((((propensity score matched analysis) OR (propensity-matched)) OR (propensity score matching)) OR (propensity score)) Filters: Adult: 19+ years, Humans.

Study selection and classification

All observational cohort and cross-sectional studies were included. Based on titles and abstracts, two reviewers (HO and GB) working independently selected the analytical studies that met the criteria of critically ill patients, use of propensity score matching (PSM) and assessment of renal outcomes. In cases of disagreement, a third reviewer (PC) reviewed the full text in order to settle the discrepancy.

Study characteristics

The year of publication, journal, country of origin, treatments or exposures, and the criteria to evaluate the methodological quality of the matching methods described by authors like Austin, Nayan and Yao were taken into account 6,10,11:

  1. Publication year, author and country

  2. Exposures or treatments assessed

  3. Study type: cohort, cross-sectional

  4. Method used to estimate the propensity score (logistic, probit, Poisson, regression trees, neuronal networks)

  5. Description of the covariables included in the propensity score model

  6. Rationale for the use of covariables

  7. Matching algorithm description (nearest neighbor, Caliper, Greedy Matching)

  8. Description of subject coincidence method

  9. Assessment of covariable balance between exposure groups (standardized differences ofp value)

  10. Description of statistical methods designed to assess exposure effect on outcomes. Need to incorporate standard error adjustments due to matching-originated dependence.

Summary of the results

The full text of each of the studies was paired-reviewed by the authors (HO and GB) in order to assess the items required be reported in observational studies that use PSM, in accordance with Austin's recommendation. The information contained in the studies was entered into an Excel database according to the list of study characteristics, determining whether the relevant criterion was met or not, or whether it was not reported (Complementary content 1, Table 1).

TABLE 1 Items that need to be taken into account in observational studies that use PSM methods. 

Characteristic n (%)
Study type
Cohort 94(94.9)
Cross-sectional 5(5.1)
PS calculation method
Logistic regression 95(96.0)
Not reported 4(4.0)*
Matching method
Agreement matching 64(65.7)
Stratification 13(13.1)
Covariable in one model 12(12.1)
IPW (weighting) 3(3.0)
Stratification + covariable 1(1.0)
IPW + stratification 1(1.0)
Not reported 5(5.1) *
Closeness determination
Nearest neighbor 14(14.1)
Caliper 12(12.1)
Greedy Matching 12(12.1)
Nearest neighbor + caliper 10(10.1)
Not reported 51(51.5)*
Balance assessment
p value 54(54.5)
Standardized means 21(21.2)
C statistic 9(9.1)
ROC curve 7(7.1)
Standardized means+p value 4(4.0)
Standardized means + C statistic 2(2.0)
Standardized means + ROC value 1(1.0)
Not reported 1(1.0)*
Methods in the analysis
Student's t test 19(19.2)
Fisher's exact test-Mann-Whitney U test 17(17.2)
Fisher exact test-chi-square test 17(17.2)
chi-square test Wilcoxon rank 14(4.1)
Kaplan Meier-Log rank 7(7.1)
McNemar 3(3.0)
Mann-Whitney rank 3(3.0)
Cochran test-Fisher exact test. 2(2.0)
Cochran test-Fisher exact test. 2(2.0)
Not reported 17(17.2)*
Covariable description
Yes 97(97.9)
No/Not reported 2(2.0)

*Number of articles with no reporting of the assessed criterion. IPV: inverse probability weighting.

SOURCE: Authors.

RESULTS

The search strategy identified 123 studies 4,5,12-96. Of these, 99 were selected based on title and abstract reviews in accordance with the inclusion criteria (Figure 1).

SOURCE: Authors.

FIGURE 1 Flow diagram of the selection of studies assessing critically ill patients with renal injury using matching methods. 

The first study was published in 2002 and, as shown in Figure 2andComplementary content 2, the number of published studies has grown since then.

SOURCE: Authors.

FIGURE 2 Number of publications per year with adjusted propensity score for every 1000 publications with and without PS. 

The largest number of publications (35) corresponded to American journals (35.3%) and the journals with the largest number of publications were Critical Care with 13 studies (13.2%), Critical Care Medicine, 11 studies (11.2%) and Intensive Care Medicine, 7 studies (7.14%) (Complementary content 3and4).

In terms of exposures or treatments assessed in the selected publications, pharmacological interventions were the most frequent, in 33 studies (33.3%), followed by assessment of surgical and other invasive procedures in 20 studies (20.4%); risk factors were third, assessed in 19 studies (19.4%); 16 studies with dialysis therapy were fourth (16.3%); and, finally, 11 studies assessed administrative processes and biomarkers (11.2%) (Complementary content 5), For PS estimations, 95 studies (95.9%) used logistic regression, and the use of covariables in PS calculations was reported and justified in the majority of studies. Only two studies (2.0%) did not describe the rationale for the use of covariables.

Matching methodology reporting

In terms of the type of PSM, the most frequent was matching in 62 studies (64.3%), followed by the stratification method in 13 studies (13.2%). The propensity index was used as covariable in the multivariate analysis in 12 studies (12.2%); and 5% of the studies did not mention the matching algorithm. Regarding the selection of the proximity method in control assignments, it was found that the nearest neighbor method was used in 14 studies (14.2%), followed by caliper, Greedy Matching and nearest neighbor + caliper, with 12 (12.2%), 12 (12.2%) and 9 (9.2%), respectively. The majority of the studies, 51 (52.0%), did not report a particular method. When assessing baseline characteristic balancing through the calculation of standardized mean differences, it was reported only in 27 studies (27.6%). In terms of tests used for analyzing the effect of the exposure on outcome, parametric tests such as Student's t test for independent samples were used in 18 studies (18.4%). Only 10 of the studies reviewed (10%) reported how standard error calculation was approached.

Evolution of matching methods

Key to achieving balance of baseline characteristics is the PS method used. Figure 3 shows that matching was the most widely used, followed by stratification and the use of propensity index as a covariable in a multivariate analysis model; change of algorithms over time can also be observed, with initial predominance of the propensity index as a covariable, with incorporation of inverse probability weighting (IPW) having occurred only recently.

IPW: inverse probability weighting. SOURCE: Authors.

FIGURE 3 Evolution of the use of matching algorithms. 

DISCUSSION

This scoping review identified 98 studies published in the past 18 years that used the PSM methodology. It was found that 51 studies (52%) did not report the closeness algorithm used for matching, and the balance of baseline characteristics was inadequately assessed in 54 studies (55%) and consequently, adequate control of confounding variables could not be ascertained in half of the studies. On the other hand, matching was the predominant method, reported in 63 studies (64.3%), while more robust methods such as IPW were used only in 3 studies (3.1%). However, it was found that all studies published in the last three years reported the balance of baseline characteristics in the form of standardized differences; they were all cohort studies with adequate mention of the matching methods, which is evident in the 19 studies published between 2018 and 2020; and only 5 studies did not specify the closeness algorithm.

These findings are consistent with those described by Nayan and Austin 10,97, who assessed studies in urology and cardiovascular surgery. In these areas of medicine, significant inconsistencies are also found in the reports of the procedures used with the PSM methodology. These inconsistencies do not only compromise the estimated validity but could also introduce estimation bias, additional to inconsistencies in variance calculation and residual confounding due to omission of non-observed variables.

In terms of the balance of baseline characteristics, the consensus is that it should be done through comparison of differences in standardized means both in the exposed as well as the non-exposed groups and it is expected that these differences should not be higher than 10%. Balance assessment by p value only is not appropriate, given sample size influence on these values (Complementary content 6).

Why is it important to report the matching method employed?

The choice of the matching method can determine the presence of greater or smaller bias, greater or smaller variance.

In matching 98-100 it is unlikely that two subjects in a population will have equal score so, in practice, matching is usually done on the basis of propensity score "closeness" which means that a treated subject with an estimated Pr[T=1|X] probability of being treated of 0.67492, can be matched with a not treated subject with an estimated Pr[T=1|X] probability of 0.67510. Now then, there are several ways to define closeness, one of them being nearest neighbor distance or the use of a caliper based on 0.2 standard deviations from the logarithmic transformation of the propensity score 99,101. In matching, the definition of closeness implies balance between bias and accuracy: if the closeness criteria are too permissive, individuals with relatively different values will be matched among themselves, giving rise to a variable score distribution between the treated and the not treated population, and creating estimation risk 101,102; but if closeness criteria are too stringent, matching will exclude many subjects, giving rise to wider 95% confidence intervals and lower accuracy 103.

Additionally, the consequence of probability assumptions different from zero and lower than 1 1 in the PS, will determine that matching will only be done for scores greater than 0 and lower than 1 that coincide in treated and not treated subjects, restricting the population to the common support region 99. The consequence of limiting to the common support region will be a smaller population size and, therefore, lower accuracy 104(Figure 4).

SOURCE: Authors.

FIGURE 4 Common support region in relation to the PS in treated and not treated subject distributions. 

Third, sample restriction may result in a limited causal effect, i.e., a causal estimator being far from the initial study population.

Fourth, when the sample is restricted to the common support region, it could maintain the subjects that benefit the most from the intervention, excluding those who benefit the least, with the obvious result of biased estimates.

As for the matching by stratification method described by Rosebaum and Rubin 1, five strata are created and individuals exhibiting similar but not identical values are matched; they demonstrated up to 90% elimination of bias derived from observed confounding variables. Stratification also has limitations, given that differences in distributions between treated and not treated subjects are found in each of the strata, compromising subject interchangeability with the stratum 2,105. Finally, another method consists of using the propensity score as a continuous covariable together with other covariables in the multivariate analysis 106. With this approach, an indicator variable that shows treatment status and the estimated propensity score is included 6. Hernán and Robins 107 highlight that when the propensity score is included in the regression it is important to include an interaction term between the score and the set of covariables; the new limitation with the regression is that the parameters cannot be interpreted causality.

In view of the above considerations regarding the variations that may occur with the use of matching methods in terms of standard error calculation and the choice of statistical tests to assess exposure in relation to outcome, there is a need to adhere to the criteria reported in the literature, summarized in Table 2 6,10,97,108, so that they are considered in the analysis and reporting of the results of studies using PSM.

TABLE 2 Description of methods used in the studies that apply propensity scores in the literature that assesses renal outcomes in critically ill patients. 

Section Description Alternatives Interpretation
Title Identifies type of study and methodology to be used Cohort, cross-secctional One of the advantages of the propensity score is that it allows to separate the design from the analysis; cohort designs make PS estimation easier according to the identification of variables found temporarily before the exposure, whereas in cross-sectional and case-control designs establishing this temporality is not possible 1.
Abstract Provides information on study objective, methods, results and conclusions Design:

  1. Cohort, cross-sectional

  2. Participants

  3. Sample size

  4. Exposure and outcomes

  5. Estimates and their accuracy

 Failing to provide information in the summary will prevent the study from being adequately selected in preliminary searches.
Introduction Introduction to the context, data source datos Highlights the importance of the study, states study objective Provides information on the topic, states the background for the study, highlights the value and usefulness of the work; provides only the necessary information.
Methods Describes the method for propensity score estimation Models: logistic, probit, regression tress, neuronal networks Describing the method is important for the reproducibility of the of the results; machine learning methods can be more difficult to interpret and do not offer a clear advantage over conventional statistical methods 2.
Describes how the information was collected Administrative databases, electronic medical records Database validation, description of algorithms for diagnosis determinations, approach to missing data.
Describes the covariables included in the model and inclusion rationale Exposure-associated confounding variables Compliance of the conditional independence assumption. Selection bias is assumed to be due only to observable characteristics, requiring the inclusion of all the characteristics that cause selection bias 7,102,107.
Describes matching algorithm Matching, weighting, stratification and use ofpropensity score as covariable The choice of the matching algorithm is not based on principles and, therefore, it is mainly arbitrary. Most matching algorithms depend on the order of data observation. Non-matched observations are discarded, resulting in loss of accuracy and power. When propensity score is included in a regression, the fact that it can interact with treatment is often forgotten, hence the need to adequately specify the regression model and assess that interaction 6,101.
Describes closeness methods in exposed and non-exposed subjects according to propensity score Nearest neighbor, Kernel, caliper, Greedy Matching Closeness criteria condition a balance between bias and accuracy; therefore, very stringent closeness criteria result in the exclusion of many subjects, increasing variance, while very loose criteria match individuals with somewhat different characteristics, increasing bias 99,101.
Matching quality: describes methods used to assess balance between groups in terms of baseline characteristics Use of standardized mean differences, p value, graphic assessment The p value is independent from sample size; small differences are found as significant in large samples, standardized differences must be preferred for balance assessment 7.
Mentions the type of statistical analysis Uses tests such as McNemar, conditional logistic regression log Rank adjusted, Cox proportional risk models, matching-stratified Measurement dependencies must be taken into account; failure to do so will result in standard error inconsistencies when tests for independent samples are applied 103.
Results Describes participant population according to the percentage of matched individuals Common support region, trimming Limitation to this common support region requires considering that results cannot be interpolated to the study population; any effect of the limitation and trimming on the result must be verified; few verification techniques have been described.
Describes standard error calculation used to obtain confidence intervals Use of:

  1. Bootstrapping

  2. Robust errors

  3. Jacknife

  4. Analytically

 The variance due to propensity score estimation as well as the variance due to the matching method (with or without replacement) must be incorporated.
Discussion Discusses the limitations of the study Population proportion The more the population is restricted, the farther from the study population and other subsequent populations will the causal estimator be.
Residual confounding Residual confounding must be expected when confounding variables are not being observed. Confounding variables that may affect the result must be recognized and named; a sensitivity analysis may be included in the best scenario.
Accuracy Standard error calculation limitations must be taken into account.

SOURCE: Authors.

Not less important than the assessment of the above criteria in matching methods is the criterion used for selecting the variables when building the model for propensity index calculation. This work found that, although 97 studies (97.9%) reported the variables used in PS calculation, none of the studies included a thorough discussion of the relationship of each of the included variables with both the exposure as well as the outcome, considering that exposure-related variables included in the model will not contribute to the estimation but will actually increase the estimator standard error. Something similar can also happen with exposure-associated variables with just a weak association with the outcome, while variables associated only with the outcomes which are included in the model, would diminish the estimator standard error 109. Special attention must be paid to variables found in the causal pathway as intermediate variables because they not only do not contribute but, when included, would cancel out the sought effect which, in this case, is estimating the probability of being treated or exposed. Current observational studies aimed at establishing causality include directed acyclic graphs (DAGs) which are a valuable tool to clarify relationships of observed confounding variables and intermediate variables within the causal pathway.

CONCLUSIONS

Studies of critically ill patients that assess renal function outcomes and apply PS vary significantly in terms of reporting of the methods used for the analysis and treatment of exposure, outcome and confounding variables. Results must be interpreted in the light of the differences in the results derived from the use of matching algorithms, assessment of balance between the groups in terms in baseline characteristics, and the use of analytical methods for the exposure variable; this will result in the limitations discussed above, consisting mainly of the balance between bias and accuracy. Moreover, clinical judgement will have to be used as a starting point to determine the plausibility of the relationship between the exposures and outcomes under study, taking into account the great limitation imposed by the variables that will notbe observed.

The methodological criteria proposed in the literature which establish the correct use of methods when it comes to analyzing and reporting the results of studies that use PSM were applied in this review.

ETHICAL RESPONSIBILITIES

Human and animal protection

The authors declare that no human or animal experiments were carried out as part of this research.

Data confidentiality

The authors declare that the protocols of their center pertaining to patient data were followed.

Right to privacy and informed consent

The authors declare that no patient data appear in this article.

The authors obtained the informed consent of the patients and/or subjects referred to in this article. This document is kept by the corresponding author.

ACKNOWLEDGEMENTS

Authors contributions

HO, GB and PAC participated in question formulation and design conception; review of the articles; data analysis; writing of the manuscript; critical review; and final approval of the manuscript.

Appreciation

None declared.

REFERENCES

1. Rosenbaum PR, Rubin DB. Reducing bias in observational studies using subclassification on the Propensity Score. J Am Statist Assoc. 1984;79:516-24. doi: http://doi.org/10.2307/2288398. [ Links ]

2. Glass TA, Goodman SN, Hernan MA, Samet JM. Causal inference in public health. Annu Rev Public Health. 2013;34:61-75. doi: http://doi.org/10.1146/annurev-publhealth-031811-124606. [ Links ]

3. Hernan MA. A definition of causal effect for epidemiological research. J Epidemiol Community Health. 2004;58(4):265-71. [ Links ]

4. Liborio AB, Leite TT, Neves FM, Teles F, Bezerra CT. AKI complications in critically ill patients: association with mortality rates and RRT. CJASN. 2015;10(1):21-8. doi: http://doi.org/10.2215/cjn.04750514. [ Links ]

5. Dou L, Lan H, Reynolds DJ, Gunderson TM, Kashyap R, Gajic O, et al. Association between obstructive sleep apnea and acute kidney injury in critically Ill Patients: A propensity-matched study. Nephron. 2017;135(2):137-46. doi: http://doi.org/10.1159/000453367. [ Links ]

6. Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behavioral Research. 2011;46:399-424. doi: http://doi.org/10.1080/00273171.2011.568786. [ Links ]

7. Johnson SR, Tomlinson GA, Hawker GA, Granton JT, Feldman BM. Propensity Score methods for bias reduction in observational studies of treatment effect. Rheum Dis Clin North Am. 2018;44(2):203-13. doi: http://doi.org/10.1016/j.rdc.2018.01.002. [ Links ]

8. Garrido MM, Kelley AS, Paris J, Roza K, Meier DE, Morrison RS, et al. Methods for constructing and assessing propensity scores. Health Serv Res. 2014;49(5):1701-20. doi: http://doi.org/10.1111/1475-6773.12182. [ Links ]

9. Raghunathan K, Layton JB, Ohnuma T, Shaw AD. Observational research using Propensity Scores. Adv Chronic Kidney Dis. 2016;23(6):367-72. doi: http://doi.org/10.1053/j.ackd.2016.11.010. [ Links ]

10. Nayan M, Hamilton RJ, Juurlink DN, Finelli A, Kulkarni GS, Austin PC. Critical appraisal of the application of propensity score methods in the urology literature. BJU international. 2017;120(6):873-80. doi: http://doi.org/i0.1111/bju.1393Q. [ Links ]

11. Yao XI, Wang X, Speicher PJ, Hwang ES, Cheng P Harpole DH, et al. Reporting and Guidelines in Propensity Score Analysis: A systematic review of cancer and cancer surgical studies. J Natl Cancer Inst. 2017;109(8):djw323. doi: http://doi.org/10.1093/jnci/djw323. [ Links ]

12. Mehta RL, McDonald B, Gabbai F, Pahl M, Farkas A, Pascual MT, et al. Nephrology consultation in acute renal failure: does timing matter? Am J Med. 2002;113(6):456-61. [ Links ]

13. Mehta RL, Pascual MT, Soroko S, Chertow GM. Diuretics, mortality, and nonrecovery of renal function in acute renal failure. JAMA. 2002;288(20):2547-53. [ Links ]

14. Uchino S, Doig GS, Bellomo R, Morimatsu H, Morgera S, Schetz M, et al. Diuretics and mortality in acute renal failure. Crit Care Med. 2004;32(8):1669-77. [ Links ]

15. Cho KC, Himmelfarb J, Paganini E, Ikizler TA, Soroko SH, Mehta RL, et al. Survival by dialysis modality in critically ill patients with acute kidney injury. J Am Society Nephrol. 2006;17(11):3132-8. doi: http://doi.org/10.1681/asn.2006030268. [ Links ]

16. Liu KD, Himmelfarb J, Paganini E, Ikizler TA, Soroko SH, Mehta RL, et al. Timing of initiation of dialysis in critically ill patients with acute kidney injury. Clin J Am Soc Nephrol. 2006;1(5):915-9. doi: http://doi.org/10.2215/cjn.01430406. [ Links ]

17. Uchino S, Bellomo R, Kellum JA, Morimatsu H, Morgera S, Schetz MR, et al. Patient and kidney survival by dialysis modality in critically ill patients with acute kidney injury. Int J Artificial Organs. 2007;30(4):281-92. [ Links ]

18. Schortgen F, Girou E, Deye N, Brochard L. The risk associated with hyperoncotic colloids in patients with shock. Int Care Med. 2008;34(12):2157-68. doi: http://doi.org/10.1007/s00134-008-1225-2. [ Links ]

19. Bagshaw SM, Lapinsky S, Dial S, Arabi Y, Do-dek P, Wood G, et al. Acute kidney injury in septic shock: clinical outcomes and impact of duration of hypotension prior to initiation of antimicrobial therapy. Int Care Med . 2009;35(5):871-81. doi: http://doi.org/10.1007/s00134-008-1367-2. [ Links ]

20. Benedetto U, Luciani R, Goracci M, Capuano F, Refice S, Angeloni E, et al. Miniaturized cardiopulmonary bypass and acute kidney injury in coronary artery bypass graft surgery. Ann Thor Surg. 2009;88(2):529-35. doi: http://doi.org/10.1016/j.athoracsur.2009.03.072. [ Links ]

21. Billings FT, Pretorius M, Siew ED, Yu C, Brown NJ. Early postoperative statin therapy is associated with a lower incidence of acute kidney injury after cardiac surgery. J Cardiothor Vasc Anesth. 2010;24(6):913-20. doi: http://doi.org/10.1053/j.jvca.2010.03.024. [ Links ]

22. Chamchad D, Horrow JC, Nachamchik L, Sutter FP Samuels LE, Trace CL, et al. The impact of immediate extubation in the operating room after cardiac surgery on intensive care and hospital lengths of stay. J Cardiothor Vasc Anesth . 2010;24(5):780-4. doi: http://doi.org/10.1053/j.jvca.2010.04.002. [ Links ]

23. Engoren M. Does erythrocyte blood transfusion prevent acute kidney injury? Propensity-matched case control analysis. Anesthesiology. 2010;113(5):1126-33. doi: http://doi.org/10.1097/ALN.0b013e181f70f56. [ Links ]

24. Chou YH, Huang TM, Wu VC, Wang CY, Shiao CC, Lai CF, et al. Impact of timing of renal replacement therapy initiation on outcome of septic acute kidney injury. Crit Care (London, England). 2011;15(3):R134. doi: http://doi.org/10.1186/cc10252. [ Links ]

25. Le Manach Y, Ibanez Esteves C, Bertrand M, Goarin JP, Fleron MH, Coriat P et al. Impact of preoperative statin therapy on adverse postoperative outcomes in patients undergoing vascular surgery. Anesthesiology. 2011;114(1):98-104. doi: http://doi.org/10.1097/ALN.0b013e31820254a6. [ Links ]

26. Lee EH, Chin JH, Choi DK, Hwang BY, Choo SJ, Song JG, et al. Postoperative hypoalbuminemia is associated with outcome in patients undergoing off-pump coronary artery bypass graft surgery. J Cardiothor Vasc Anesth . 2011;25(3):462-8. doi: http://doi.org/10.1053/j.jvca.2010.09.008. [ Links ]

27. Shahin J, DeVarennes B, Tse CW, Amarica DA, Dial S. The relationship between ino-trope exposure, six-hour postoperative physiological variables, hospital mortality and renal dysfunction in patients undergoing cardiac surgery. Crit Care (London, England) . 2011;15(4):R162. doi: http://doi.org/10.1186/cc10302. [ Links ]

28. Attaran S, Saleh HZ, Shaw M, Ward A, Pullan M, Fabri BM. Does the outcome improve after radiofrequency ablation for atrial fibrillation in patients undergoing cardiac surgery? A propensity-matched comparison. Eur J Cardio-thoracic Surg. 2012;41(4):806-10; discussion 10-1. doi: http://doi.org/10.1093/ejcts/ezr107. [ Links ]

29. Azuma N, Uchida H, Kokubo T, Koya A, Akasaka N, Sasajima T. Factors influencing wound healing of critical ischaemic foot after bypass surgery: is the angiosome important in selecting bypass target artery? Eur J Vas Endovasc Surg. 2012;43(3):322-8. doi: http://doi.or-g/10.1016/j.ejvs.2011.12.001. [ Links ]

30. Clec'h C, Darmon M, Lautrette A, Chemouni F, Azoulay E, Schwebel C, et al. Efficacy of renal replacement therapy in critically ill patients: a propensity analysis. Crit Care (London, England) . 2012;16(6):R236. doi: http://doi.org/10.1186/cc11905. [ Links ]

31. Lee EH, Baek SH, Chin JH, Choi DK, Son HJ, Kim WJ, et al. Preoperative hypoalbuminemia is a major risk factor for acute kidney injury following off-pump coronary artery bypass surgery. Int Care Med . 2012;38(9):1478-86. doi: http://doi.org/10.1007/s00134-012-2599-8. [ Links ]

32. Vaara ST, Reinikainen M, Kaukonen KM, Pettila V. Association of ICU size and annual case volume of renal replacement therapy patients with mortality. Acta Anaesthesiol Scand. 2012;56(9):1175-82. doi: http://doi.org/10.1111/j.1399-6576.2012.02747.x. [ Links ]

33. Wu VC, Lai CF, Shiao CC, Lin YF, Wu PC, Chao CT, et al. Effect of diuretic use on 30-day postdialysis mortality in critically ill patients receiving acute dialysis. PloS One. 2012;7(3):e30836. doi: http://doi.org/10.1371/journal.pone.0030836. [ Links ]

34. Apel M, Maia VF! Zeidan M, Schinkoethe C, Wolf G, Reinhart K, et al. End-stage renal disease and outcome in a surgical intensive care unit. Crit Care (London, England) . 2013;17(6):R298. doi: http://doi.org/10.1186/cc13167. [ Links ]

35. Bayer O, Schwarzkopf D, Doenst T, Cook D, Kabisch B, Schelenz C, et al. Perioperative fluid therapy with tetrastarch and gelatin in cardiac surgery--a prospective sequential analysis. Crit Care Med . 2013;41(11):2532-42. doi: http://doi.org/10.1097/CCM.0b013e3182978fb6. [ Links ]

36. Ehrmann S, Badin J, Savath L, Pajot O, Garot D, Pham T, et al. Acute kidney injury in the critically ill: is iodinated contrast medium really harmful? Crit Care Med . 2013;41(4):1017-26. doi: http://doi.org/10.1097/CCM.0b013e318275871a. [ Links ]

37. Leite TT, Macedo E, Pereira SM, Bandeira SR, Pontes PH, Garcia AS, et al. Timing of renal replacement therapy initiation by AKIN classification system. Crit Care (London, England) . 2013;17(2):R62. doi: http://doi.org/10.1186/cc12593. [ Links ]

38. Costa e Silva VT, Liano F, Muriel A, Diez R, de Castro I, Yu L. Nephrology referral and outcomes in critically ill acute kidney injury patients. PloS One. 2013;8(8):e70482. doi: http://doi.org/10.1371/journal.pone.0070482. [ Links ]

39. Brewer R, Theurer PF, Cogan CM, Bell GF Prager RL, Paone G. Morbidity but not mortality is decreased after off-pump coronary artery bypass surgery. Ann Thor Surg . 2014;97(3):831-6. doi: http://doi.org/10.1016/j.athoracsur.2013.08.062. [ Links ]

40. Elmistekawy E, McDonald B, Hudson C, Ruel M, Mesana T, Chan V, et al. Clinical impact of mild acute kidney injury after cardiac surgery. Ann Thor Surg . 2014;98(3):815-22. doi: http://doi.org/10.1016/j.athoracsur.2014.05.008. [ Links ]

41. Iwagami M, Yasunaga H, Doi K, Horigu-chi H, Fushimi K, Matsubara T, et al. Postoperative polymyxin B hemoperfusion and mortality in patients with abdominal septic shock: a propensity-matched analysis. Crit Care Med . 2014;42(5):1187-93. doi: http://doi.org/10.1097/ccm.0000000000000150. [ Links ]

42. Raghunathan K, Shaw A, Nathanson B, Sturmer T, Brookhart A, Stefan MS, et al. Association between the choice of IV crystalloid and in-hospital mortality among critically ill adults with sepsis. Crit Care Med . 2014;42(7):1585-91. doi: http://doi.org/10.1097/ccm.0000000000000305. [ Links ]

43. Linder A, Fjell C, Levin A, Walley KR, Russell JA, Boyd JH. Small acute increases in serum creatinine are associated with decreased long-term survival in the critically ill. Am J Resp Crit Care Med . 2014;189(9):1075-81. doi: http://doi.org/10.1164/rccm.201311-2097OC. [ Links ]

44. Oh HJ, Lee MJ, Kim CH, Kim DY, Lee HS, Park JT, et al. The benefit of specialized team approaches in patients with acute kidney injury undergoing continuous renal replacement therapy: propensity score matched analysis. Crit Care (London, England) . 2014;18(4):454. doi: http://doi.org/10.1186/s13054-014-0454-8. [ Links ]

45. Picard W, Bazin F, Clouzeau B, Bui HN, Soulat M, Guilhon E, et al. Propensity-based study of aminoglycoside nephrotoxicity in patients with severe sepsis or septic shock. Antimicrobial Agents Chemother. 2014;58(12):7468-74. doi: http://doi.org/10.1128/aac.03750-14. [ Links ]

46. Vaara ST, Pettila V, Kaukonen KM, Bendel S, Korhonen AM, Bellomo R, et al. The attributable mortality of acute kidney injury: a sequentially matched analysis. Crit Care Med . 2014;42(4):878-85. doi: http://doi.org/10.1097/ccm.0000000000000045. [ Links ]

47. Wald R, Shariff SZ, Adhikari NK, Bagshaw SM, Burns KE, Friedrich JO, et al. The association between renal replacement therapy modality and long-term outcomes among critically ill adults with acute kidney injury: a retrospective cohort study. Crit Care Med . 2014;42(4):868-77. doi: http://doi.org/10.1097/ccm.0000000000000042. [ Links ]

48. Ding W, Ji Q Wei Q Shi Y, Ma R, Wang C. Prophylactic application of an intra-aortic balloon pump in high-risk patients undergoing off-pump coronary artery bypass grafting. Cardiology. 2015;131(2):109-15. doi: http://doi.org/10.1159/000377720. [ Links ]

49. Horkan CM, Purtle SW, Mendu ML, Moromi-zato T, Gibbons FK, Christopher KB. The association of acute kidney injury in the critically ill and postdischarge outcomes: a cohort study. Crit Care Med . 2015 43(2)354-64. doi: http://doi.org/10.1097/ccm.0000000000000706. [ Links ]

50. Leite TT, Macedo E, Martins Ida S, Neves FM, Liborio AB. Renal outcomes in critically Ill patients receiving propofol or midazolam. Clin J Am Society Nephrol . 2015;10(11):1937-45. doi: http://doi.org/10.2215/cjn.02330315. [ Links ]

51. Sang BH, Bang JY, Song JG, Hwang GS. Hypoalbuminemia within two postoperative days is an independent risk factor for acute kidney injury following living donor liver transplantation: A Propensity Score analysis of 998 consecutive patients. Crit Care Med . 2015;43(12):2552-61. doi: http://doi.org/10.1097/ccm.0000000000001279. [ Links ]

52. Shaw AD, Schermer CR, Lobo DN, Munson SH, Khangulov V, Hayashida DK, et al. Impact of intravenous fluid composition on outcomes in patients with systemic inflammatory response syndrome. Crit Care (London, England) . 2015;19:334. doi: http://doi.org/10.1186/s13054-015-1045-z. [ Links ]

53. Shen TC, Wang IK, Wei CC, Lin CL, Tsai CT, Hsia TC, et al. The risk of septicemia in end-stage renal disease with and without renal transplantation: A propensity-matched cohort study. Medicine. 2015;94(34):e1437. doi: http://doi.org/10.1097/md.0000000000001437. [ Links ]

54. Allen CJ, Ruiz XD, Meizoso JP Ray JJ, Livingstone AS, Schulman CI, et al. Is hydroxyethyl starch safe in penetrating trauma patients? Military Med. 2016;181(5 Suppl):152-5. doi: http://doi.org/10.7205/milmed-d-15-00132. [ Links ]

55. Bonnet V, Boisselier C, Saplacan V, Belin A, Gerard JL, Fellahi JL, et al. The role of age and comorbidities in postoperative outcome of mitral valve repair: A propensity-matched study. Medicine. 2016;95(25):e3938. doi: http://doi.org/10.1097/md.0000000000003938. [ Links ]

56. Carmona P, Paredes F, Mateo E, Mena-Duran AV, Hornero F, Martinez-Leon J. Is off-pump technique a safer procedure for coronary revascularization? A propensity score analysis of 20 years of experience. Interact Cardiovasc Thor Surg. 2016;22(5):612-8. doi: http://doi.org/10.1093/icvts/ivw005. [ Links ]

57. Thongprayoon C, Cheungpasitporn W, Harrison AM, Kittanamongkolchai W, Ungprasert P Srivali N, et al. The comparison of the commonly used surrogates for baseline renal function in acute kidney injury diagnosis and staging. BMC Nephrol. 2016;17:6. doi: http://doi.org/10.1186/s12882-016-0220-z. [ Links ]

58. Cipolle MD, Ingraham Lopresto BC, Pirrung JM, Meyer EM, Manta C, Nightingale AS, et al. Embedding a trauma hospitalist in the trauma service reduces mortality and 30-day trauma-related readmissions. J Trauma Acute Care Surg. 2016;81(1):178-83. doi: http://doi.org/10.1097/ta.0000000000001062. [ Links ]

59. Clark E, Kumar A, Langote A, Lapinsky S, Dodek P Kramer A, et al. Septic shock in chronic dialysis patients: clinical characteristics, antimicrobial therapy and mortality. Intens Care Med. 2016;42(2):222-32. doi: http://doi.org/10.1007/s00134-015-4147-9. [ Links ]

60. Hammond DA, Smith MN, Painter JT, Meena NK, Lusardi K. Comparative incidence of acute kidney injury in critically Ill patients receiving vancomycin with concomitant piperacillinta-zobactam or cefepime: A retrospective cohort study. Pharmacotherapy. 2016;36(5):463-71. doi: http://doi.org/10.1002/phar.1738. [ Links ]

61. Latour-Pérez J, Gómez-Tello V, de-Miguel-Balsa E, Llamas-Álvarez A, Carrillo-López A, Sánchez-Roman JA, et al. Routine invasive strategy in acute coronary syndrome patients with renal dysfunction. Results of the ARIAM-SEMI-CYUC registry. Med Int. 2016;40(5):280-8. doi: http://doi.org/10.1016/j.medin.2015.09.008. [ Links ]

62. Latour Joung KW, Kong YG, Yoon SH, Kim YJ, Hwang JH, Hong B, et al. Comparison of postoperative acute kidney injury between ileal conduit and neobladder urinary diversions after radical cystectomy: A propensity score matching analysis. Medicine. 2016;95(36):e4838. doi: http://doi.org/10.1097/md.0000000000004838. [ Links ]

63. Jun IG, Lee B, Kim SO, Shin WJ, Bang JY, Song JG, et al. Comparison of acute kidney injury between ABO-compatible and ABO-incompatible living donor liver transplantation: A propensity matching analysis. Liver Transplant. 2016;22(12):1656-65. doi: http://doi.org/10.1002/lt.24634. [ Links ]

64. Karamanos E, Dulchavsky S, Beale E, Inaba K, Demetriades D. Diabetes mellitus in patients presenting with adhesive small bowel obstruction: Delaying surgical intervention results in worse outcomes. World J Surg. 2016;40(4):863-9. doi: http://doi.org/10.1007/s00268-015-3338-4. [ Links ]

65. Lai WH, Rau CS, Wu SC, Chen YC, Kuo PJ, Hsu SY, et al. Post-traumatic acute kidney injury: a cross-sectional study of trauma patients. Scand J Trauma, Res Emerg Med. 2016;24(1):136. doi: http://doi.org/10.1186/s13049-016-0330-4. [ Links ]

66. Lin CY, Tsai FC, Chen YC, Lee HA, Chen SW, Liu KS, et al. Correlation of preoperative renal insufficiency with mortality and morbidity after aortic valve replacement: A Propensity Score matching analysis. Medicine. 2016;95(9):e2576. doi: http://doi.org/10.1097/md.0000000000002576. [ Links ]

67. Mazzeffi M, Galvagno S, Gammie JS, Tanaka K. Impact of aspirin use on morbidity and mortality in massively transfused cardiac surgery patients: a propensity score matched cohort study. J Anesthesia. 2016;30(5):817-25. doi: http://doi.org/10.1007/s00540-016-2213-2. [ Links ]

68. Penuelas O, Muriel A, Frutos-Vivar F, Fan E, Raymondos K, Rios F, et al. Prediction and outcome of intensive care unit-acquired paresis. JIntens Care Med . 2018;33(1):16-28. doi: http://doi.org/10.1177/0885066616643529. [ Links ]

69. Park JY, An JN, Jhee JH, Kim DK, Oh HJ, Kim S, et al. Early initiation of continuous renal replacement therapy improves survival of elderly patients with acute kidney injury: a multicenter prospective cohort study. Crit Care (London, England) . 2016;20(1):260. doi: http://doi.org/10.1186/s13054-016-1437-8. [ Links ]

70. Ruan SY, Wu HY, Lin HH, Wu HD, Yu CJ, Lai MS. Inhaled nitric oxide and the risk of renal dysfunction in patients with acute respiratory distress syndrome: a propensity-matched cohort study. Crit Care (London, England) . 2016;20(1):389. doi: http://doi.org/10.1186/s13054-016-1566-0. [ Links ]

71. Thongprayoon C, Cheungpasitporn W, Srivali N, Harrison AM, Gunderson TM, Kittanamongkolchai W, et al. AKI after transcatheter or surgical aortic valve replacement. J Am Society Nephrol . 2016;27(6):1854-60. doi: http://doi.org/10.1681/asn.2015050577. [ Links ]

72. Chan MJ, Lee CC, Chen SW, Tsai FC, Lin PJ, Fan PC, et al. Effect of different surgical type of coronary artery bypass grafting on kidney injury: A propensity score analysis. Medicine. 2017;96(45):e8395. doi: http://doi.org/10.1097/md.0000000000008395. [ Links ]

73. Dos Santos TOC, Oliveira MAS, Monte JCM, Batista MC, Pereira Junior VG, Dos Santos BFC, et al. Outcomes from a cohort of patients with acute kidney injury subjected to continuous venovenous hemodiafiltration: The role of negative fluid balance. PloS One . 2017;12(4):e0175897. doi: http://doi.org/10.1371/journal.pone.0175897. [ Links ]

74. Shum HP, Chan KC, Tam CW, Yan WW, Chan TM. Impact of renal replacement therapy on survival in patients with KDIGO Stage 3 acute kidney injury: A propensity score matched analysis. Nephrology. 2018;23:1081-9. doi: http://doi.org/10.1111/nep.13164. [ Links ]

75. McDonald JS, McDonald RJ, Williamson EE, Kallmes DF, Kashani K. Post-contrast acute kidney injury in intensive care unit patients: a propensity score-adjusted study. Intens Care Med . 2017;43(6):774-84. doi: http://doi.org/10.1007/s00134-017-4699-y. [ Links ]

76. Kashani K, Thongprayoon C, Cheungpasitporn W, Iacovella GM, Akhoundi A, Albright RC, Jr. Association between mortality and replacement solution bicarbonate concentration in continuous renal replacement therapy: A propensity-matched cohort study. PloS One . 2017;12(9):e0185064. doi: https://doi.org/10.1371/journal.pone.0185064. [ Links ]

77. Kovacheva VP, Aglio LS, Boland TA, Mendu ML, Gibbons FK, Christopher KB. Acute kidney injury after craniotomy is associated with increased mortality: A cohort study. Neurosurgery. 2016;79(3):389-96. doi: http://doi.org/10.1227/neu.0000000000001153. [ Links ]

78. Legouis D, Galichon P, Bataille A, Chevret S, Provenchere S, Boutten A, et al. Rapid occurrence of chronic kidney disease in patients experiencing reversible acute kidney injury after cardiac surgery. Anesthesiology . 2017;126(1):39-46. doi: http://doi.org/10.1097/aln.0000000000001400. [ Links ]

79. Marik PE, Khangoora V, Rivera R, Hooper MH, Catravas J. Hydrocortisone, vitamin c, and thiamine for the treatment of severe sepsis and septic shock: A retrospective before-after study. Chest. 2017;151(6):1229-38. doi: http://doi.org/10.1016/j.chest.2016.11.036. [ Links ]

80. Yu SC, Kuo JR, Shiue YL, Yu ZX, Ho CH, Wu CC, et al. One-year mortality of patients with chronic kidney disease after spinal cord injury: A 14-year population-based study. World Neurosurg. 2017;105:462-9. doi: http://doi.or-g/10.1016/j.wneu.2017.06.009. [ Links ]

81. Huang ST, Ke TY, Chuang YW, Lin CL, Kao CH. Renal complications and subsequent mortality in acute critically ill patients without pre-existing renal disease. CMAJ. 2018;190(36):E1070-E80. doi: http://doi.org/10.1503/cmaj.171382. [ Links ]

82. Wang AY, Trongtrakul K, Bellomo R, Li Q Cass A, Gallagher M, et al. HMG-CoA reductase inhibitors (statins) and acute kidney injury: A secondary analysis of renal study outcomes. Nephrology (Carlton). 2019;24(9):912-8. doi: http://doi.org/10.1111/nep.13597. [ Links ]

83. Praca APA, Nassar AP, Jr., Caruso P Outcomes of cancer patients discharged from ICU after a decision to forgo life-sustaining therapies. Crit Care Med . 2019;47(6):e454-e60. doi: http://doi.org/10.1097/CCM.0000000000003729. [ Links ]

84. McCoy IE, Montez-Rath ME, Chertow GM, Chang TI. Estimated effects of early diuretic use in critical illness. Crit Care Explor. 2019;1(7):e0021. doi: http://doi.org/10.1097/CCE.0000000000000021. [ Links ]

85. Zeeshan M, Hamidi M, Feinstein AJ, Gries L, Jehan F, Sakran J, et al. Four-factor prothrombin complex concentrate is associated with improved survival in trauma-related hemorrhage: A nationwide propensity-matched analysis. J Trauma Acute Care Surg . 2019;87(2):274-81. doi: http://doi.org/10.1097/TA.0000000000002262. [ Links ]

86. Gaudard P Saour M, Morquin D, David H, Eliet J, Villiet M, et al. Acute kidney injury during daptomycin versus vancomycin treatment in cardiovascular critically ill patients: a propensity score matched analysis. BMC Infect Dis. 2019;19(1):438. doi: http://doi.org/10.1186/s12879-019-4077-1. [ Links ]

87. Pitta RD, Gasparetto J, De Moraes TP Telles JP, Tuon FF. Antimicrobial therapy with aminoglycoside or meropenem in the intensive care unit for hospital associated infections and risk factors for acute kidney injury. Eur J Clin Microbiol Infect Dis. 2020 39(4)723-8. doi: http://doi.org/10.1007/s10096-019-03779-6. [ Links ]

88. Schubert SA, Hawkins RB, Mehaffey JH, Fonner CE, Rich JB, Speir AM, et al. Preoperative beta-blocker use correlates with worse outcomes in patients undergoing aortic valve replacement. J Thorac Cardiovasc Surg. 2019;158(6):1589-97-e3. doi: http://doi.org/10.1016/j.jtcvs.2018.12.108. [ Links ]

89. Jiang W, Shen B, Wang Y, Xu J, Luo Z, Ding X, et al. Potentially modifiable predictors for renal replacement therapy in patients with cardiac surgery associated-acute kidney injury: A Propensity Score-matched case-control study. Braz J Cardiovasc Surg. 2019 34(1)33-40. doi: http://doi.org/10.21470/1678-9741-2018-0206. [ Links ]

90. Shen Y, Zhang W, Shen Y. Early diuretic use and mortality in critically ill patients with vasopressor support: a propensity score-matching analysis. Crit Care (London, England) . 2019;23(1):9. doi: http://doi.org/10.1186/s13054-019-2309-9. [ Links ]

91. Miyamoto Y, Iwagami M, Aso S, Yasunaga H, Matsui H, Fushimi K, et al. Association between intravenous contrast media exposure and non-recovery from dialysis-requiring septic acute kidney injury: a nationwide observational study. Intens Care Med . 2019;45(11):1570-9. doi: http://doi.org/10.1007/s00134-019-05755-2. [ Links ]

92. Chou YH, Huang TM, Wu VC, Chen WS, Wang CH, Chou NK, et al. Associations between preoperative continuation of renin-angiotensin system inhibitor and cardiac surgery-associated acute kidney injury: a propensity score-matching analysis. J Nephrol. 2019;32(6):957-66. doi: http://doi.org/10.1007/s40620-019-00657-4. [ Links ]

93. Goto Y, Koyama K, Katayama S, Tonai K, Shi-ma J, Koinuma T, et al. Influence of contrast media on renal function and outcomes in patients with sepsis-associated acute kidney injury: a propensity-matched cohort study. Crit Care (London, England) . 2019;23(1):249. doi: http://doi.org/10.1186/s13054-019-2517-3. [ Links ]

94. Williams LS, Walker GR, Loewenherz JW, Gi-del LT. Association of contrast and acute kidney injury in the critically Ill: A Propensity-Matched Study. Chest. 2020;157(4):866-76. doi: http://doi.org/10.1016/j.chest.2019.10.005. [ Links ]

95. Abudayyeh A, Song J, Abdelrahim M, Dah-bour I, Page VD, Zhou S, et al. Renal replacement therapy in patients with stage IV Cancer admitted to the intensive care unit with acute kidney injury at a comprehensive cancer center was not associated with survival. Am J Hosp Palliat Care. 2020 37(9)707-15. doi: http://doi.org/10.1177/1049909120902115. [ Links ]

96. Zhao GJ, Xu C, Ying JC, Lu WB, Hong GL, Li MF, et al. Association between furosemide administration and outcomes in critically ill patients with acute kidney injury. Crit Care (London, England) . 2020;24(1):75. doi: http://doi.org/10.1186/s13054-020-2798-6. [ Links ]

97. Austin PC. Propensity-score matching in the cardiovascular surgery literature from 2004 to 2006: a systematic review and suggestions for improvement. J Thor Cardiovasc Surg. 2007;134(5):1128-35. doi: http://doi.or-g/10.1016/j.jtcvs.2007.07.021. [ Links ]

98. Rosenbaum PR RD. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983 70(1):4. [ Links ]

99. Rosenbaum PR RD. Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat. 1985;39(1):3. [ Links ]

100. Ospina-Tascón GA, Hernández G, Álvarez I, Calderón-Tapia LE, Manzano-Núnez R, Sánchez-Ortiz AI, et al. Effects of very early start of norepinephrine in patients with septic shock: a propensity score-based analysis. Crit Care (London, England) . 2020;24(1):52. doi: http://doi.org/10.1186/s13054-020-2756-3. [ Links ]

101. Austin PC. Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies. Pharm Stat. 2011;10(2):150-61. doi: http://doi.org/10.1002/pst.433. [ Links ]

102. Steiner PM, Cook TD, Shadish WR, Clark MH. The importance of covariate selection in controlling for selection bias in observational studies. Psychol Methods. 2010;15(3):250-67. doi: http://doi.org/10.1037/a0018719. [ Links ]

103. Hullsiek KH, Louis TA. Propensity score modeling strategies for the causal analysis of observational data. Biostatistics. 2002;3(2):179-93. doi: http://doi.org/10.1093/biostatistics/3.2.179. [ Links ]

104. Petersen ML, Porter KE, Gruber S, Wang Y, van der Laan MJ. Diagnosing and responding to violations in the positivity assumption. Stat Methods Med Res. 2012;21(1):31-54. doi: http://doi.org/10.1177/0962280210386207. [ Links ]

105. VanderWeele TJ, Hernan MA. Causal Inference Under Multiple Versions of Treatment. J Causal Inference. 2013;1(1):1-20. doi: http://doi.org/10.1515/jci-2012-0002. [ Links ]

106. Rubin DB. On principles for modeling propensity scores in medical research. Pharma-coepidemiol Drug Saf. 2004;13(12):855-7. doi: http://doi.org/10.1002/pds.968. [ Links ]

107. VanderWeele TJ, Hernan MA, Tchet-gen Tchetgen EJ, Robins JM. Re: Causality and causal inference in epidemiology: the need for a pluralistic approach. Int J Epidemiol. 2016;45(6):2199-200. doi: http://doi.org/10.1093/ije/dyw162. [ Links ]

108. Ali MS, Groenwold RH, Belitser SV, Pestman WR, Hoes AW, Roes KC, et al. Reporting of covariate selection and balance assessment in propensity score analysis is suboptimal: a systematic review. J Clin Epidemiol.2015;68(2):112-21. doi: http://doi.org/10.1016/j.jclinepi.2014.08.011. [ Links ]

109. Brookhart MA, Schneeweiss S, Rothman KJ, Glynn RJ, Avorn J, Sturmer T. Variable selection for propensity score models. Am J Epidemiol. 2006;163(12):1149-56. doi: http://doi.org/10.1093/aje/kwj149. [ Links ]

Assistance with the study None declared.

Financial support and sponsorship None declared.

Conflict of interest The authors have no conflict of interest to disclose.

Presentaciones None declared.

How to cite this article: Oliveros Rodríguez H, Buitrago G, Castellanos Saavedra P. Use of matching methods in observational studies with critical patients and renal outcomes. Scoping review. Colombian Journal of Anesthesiology. 2021;49:e944.

CONTENIDO COMPLEMENTARIO

COMPLEMENTAR CONTENT 1. Study characteristics assessed. 

Year Author Journal Country Study type Exposure Variables Matching algorithm Closeness Balance assessment Analytical method used Outcomes
2002 Ravindra L. The American Journal of Medicine United States Cohort Timing of nephrology consults in acute renal failure (early vs. late). Age, sex, urinary output, liver failure, hematological insufficiency, heart rate, and serum creatinine and BUN Covariable in one model Not reported ROC values Logistic regression Mortality
2002 Mehta RL JAMA United States Cohort Association of diuretics with adverse or favorable outcomes in critically ill patients with acute renal failure Age, nephrotoxic etiology of acute renal failure, BUN, acute respiratory failure and congestive heart failure. Covariable in one model Not reported ROC values Logistic regression Mortality
2004 Uchino, S. Crit Care Med Australia Cohort Determination of the impact of diuretics on mortality rate in critically ill patients with acute renal failure Patient age, simplified acute physiology score II, creatinine on ICU admissions, need for renal replacement therapy, time between ICU admission and inclusion in the study, central venous pressure, Glasgow coma score, vasopressor use, urine volume during 6 hours prior to inclusion, platelet count, creatinine, arterial pH, septic etiology, low cardiac output etiology and other etiologies Covariable in one model Not reported C statistic Logistic regression Mortality
2006 Cho, K. C. J Am Soc Nephrol United States Cohort Survival by dialysis modality (continuous vs. intermittent) in critically ill patients with acute renal injury Age, liver failure, sepsis, thrombocytopenia, blood urea nitrogen and serum creatinine Not reported Not reported ROC values Logistic regression and Cox regression Survival
2006 Liu, K. D. Clin J Am Soc Nephrol United States Cohort Association between dialysis initiation (early vs. late) and mortality. Age, liver failure, sepsis, thrombocytopenia, and serum creatinine and stratification by site and modality of initial dialysis Not reported Not reported ROC values Logistic regression and Cox regression Mortality
2007 S. Uchino The International Journal of Artificial Organs United States Cohort Renal replacement therapy (continuous vs. intermittent) to assess impact on survival and renal recovery ATrterial pH, furosemide dose, creatinine, gastrointestinal medical admission, immunomodulation, use of pulmonary artery catheter, hepatorenal syndrome, mechanical ventilation, mean arterial pressure and Glasgow coma score Stratification Not reported P Value Logistic regression Survival
2008 Schortgen, F. Intensive Care Med France Cohort The risk of renal adverse events in patients with shock resuscitated with hypo-oncotic colloids, artificial hyperoncotic colloids, hyperoncotic albumin or crystalloids Demographic characteristics, admission category (i.e., planned or unplanned medical, surgical, with the type of surgery), potentially life-threatening underlying disease, immunodeficiency, risk factors for acute renal failure and use of nephrotoxic drugs. Simplified acute physiology score (SAPS II) and the number of organ dysfunction Covariable in one model Not reported P Value Logistic regression Renal failure
2009 Bagshaw, S. M. Intensive Care Med Canada Cohort In patients with septic shock, describe the association between time of onset of hypotension and effective antimicrobial therapy and onset and severity of early acute renal failure Age, sex, comorbidity, source of infection, surgical status, need for mechanical ventilation, need for vasopressors, APACHE II score, hospital setting and study year Covariable in one model Not reported ROC values Logistic regression and Cox regression Renal failure
2009 Benedetto, U. The Annals of thoracic surgery Italy Cohort Association between on-pump coronary artery bypass grafting (CABG) vs. mini cardiopulmonary bypass systems (mini-CPB) and reduced incidence of acute renal failure Age, sex, body mass index (BMI), preoperative GFR, diabetes mellitus, chronic obstructive pulmonary disease requiring treatment, hypertension, peripheral vascular disease, NYHA class III/IV, left ventricular ejection fraction, prior percutaneous coronary intervention, myocardial infarction within 30 days, emergency surgery, preoperative medications including blockers, angiotensin converting enzyme inhibitors or angiotensin receptor blockers, aspirin and clopidogrel within 5 days after surgery; primary surgeon, number of grafts per patient and CPB time; postoperative variables including red blood cell transfusions and low cardiac output incidence Coincidence matching Coincidence matching Nearest neighbor C statistic Logistic regression Renal failure
2010 Billings Journal of Cardiothoracic and Vascular Anesthesia United States Cohort Pre- and postoperative use of statins and incidence of acute renal failure in patients taken to cardiac surgery Preoperative treatment with statins, baseline creatinine, age, sex, history of diabetes,history of hypertensive disease, BMI, baseline left ventricular ejection fraction,preoperative diuretic, preoperative use of ACEi, valve surgery, use of CBP, time on CBP, aortic cross-clamping time, intraoperative inotropic support, intraoperative coronary catheterization, postoperative surgical re-exploration, postoperative diuretic, and extubation time Stratification y Covariable in one model Not reported P Value Logistic regression Renal failure
2010 Chamchad, D. J Cardiothorac Vasc Anesth United States Cohort Immediate extubation in the operating room after cardiac surgery associated with intensive care and hospital length Age, sex, BMI, ethnicity, preoperative serum creatinine, NYHA, reoperation, CABG, valve procedure, DM, hypertension, COPD, use ofBB, angina, reintubation, bleeding, postoperative renal failure, postoperative atrial fibrillation, time on CPB, priority of the surgery, off-pump operation, surgical priority Coincidence matching Greedy matching Standardized measurements Logistic regression and Cox regression Estancia
2010 Engoren, M. Anesthesiology United States Cohort Red blood cell transfusion vs. no transfusion in patients with acute pulmonary injury for the development of acute renal failure Not described Not reported Not reported Not reported Logistic regression Renal failure
2011 Chou, Y. H. Crit Care Taiwán Cohort Impact of early or late initiation of renal replacement therapy in patients with acute renal failure of septic origin Demographics, comorbidities, septic AKI development after surgery (or not), and indications for TSR. Biochemical data such as complete red blood cell count, blood urea nitrogen (BUN), serum creatinine (sCr), serum glutamate-oxaloacetate transaminase (GOT), total serum bilirubin, serum albumin and serum potassium (sK+) were documented on admission to the ICU and initiation of RRT. Clinical parameters included heart rate, systolic and diastolic blood pressure, central venous pressure (CVP), partial oxygen pressure in arterial blood gases and inspired oxygen fraction. Severity scores included Glasgow Coma Scale (GCS), acute physiology score and chronic health evaluation II (APACHE II), sequence organ failure assessment (SOFA) and simplified acute physiology score III (SOFA III). The use of mechanical ventilation was documented and inotropic equivalent dose was calculated. Coincidence matching Not reported P Value Logistic regression and Cox regression Mortality
2011 Kawar, E. Jaapa United States Cohort Comparison of clinical outcomes between pages admitted to an intensive care unit managed by residents vs. medical assistant Age, sex, admission diagnoses, patient origin before admission to the ICU and insurance status, tachycardia (heart rate 150 beats per minute), arrhythmia, hypotension, stroke, coma, intracranial mass, gastrointestinal bleeding and need for mechanical ventilation, cardiopulmonary resuscitation (CPR) within 24 hours before admission, liver cirrhosis, serum creatinine or metastatic cancer, readmission to the ICU. Coincidence matching Not reported P Value Logistic regression and Cox regression Length of stay and mortality
2011 Le Manach, Y. Anesthesiology France Cohort Chronic statin treatment compared to no statin treatment in terms of the main postoperative adverse events in patients undergoing vascular surgery Age, sex, admission diagnoses, patient origin before admission to the ICU and insurance status, tachycardia (heart rate 150 beats per minute), arrhythmia, hypotension, stroke, coma, intracranial mass, gastrointestinal bleeding and need for mechanical ventilation, cardiopulmonary resuscitation (CPR) within 24 hours before admission, liver cirrhosis, high serum creatinine levels (< 2 mg/dL) or metastatic cancer. Physician and staff names and whether the patients was being readmitted to the ICU Covariable in one model Not reported C statistic Logistic regression Mortality
2011 Lee, E. H. J Cardiothorac Vasc Anesth Korea Cohort Immediate postoperative hypoalbuminemia value as a marker of outcome after off-pump coronary artery bypass grafting (OPCABG) Sex, age, body mass index (BMI), diabetes mellitus, hypertension, chronic pulmonary disease, renal failure, prior stroke, preoperative ejection fraction, preoperative hematocrit, preoperative albumin, European System for Cardiac Operative Risk (EuroSCORE) and EuroSCORE logistic additive Coincidence matching Not reported P Value Logistic regression Renal failure
2011 Shahin, J. Crit Care Canada Cohort Relationship between inotrope use and hospital mortality and renal dysfunction in patients undergoing cardiac surgery Parsonnet score and low left ventricular ejection fraction, age, sex, bypass time, procedure type, CABG only or other procedures and the perioperative use of intra-aortic balloon pump Stratification Greedy matching P Value Logistic regression Mortality
2012 Attaran European Journal of Cardio-Thoracic Surgery UK Cohort Radiofrequency ablation for atrial fibrillation in patients undergoing cardiac surgery in relation to a better outcome and postoperative survival rate Age, BMI, female gender,class IV angina, recent acute MI, current smoker, DM, hypercholesterolemia, hypertension, respiratory disease, cerebrovascular disease, renal dysfunction, three-vessel disease, priority surgery, rheumatic valve disease, euroSCORE, LVEF < 30% Coincidence matching Not reported P Value Logistic regression Mortality
2012 Azuma, N. Eur J Vasc Endovasc Surg Japan Cohort Determination of factors that affect ischemic wound healing and the role of angiosome in bypass surgery Diabetes, end-stage renal disease, heel ulcer, Rutherford classification, serum albumin and CRP Coincidence matching Not reported P Value Logistic regression Survival
2012 Clec'h, C. Crit Care France Cohort Association of renal replacement therapy with in-hospital mortality in ICU patients with acute renal failure Increase in creatinine reflected by the RIFLE score, oliguria reflecte in 24-hour urine output and SAPS II score, blood urea nitrogen, serum creatinine and potassium (before RRT initiation), fluid accumulation (reflected by the difference in patient weight when reaching maximum RIFLE class as compared to weight on admission to the ICU) and SAPS II score Coincidence matching Not reported P Value Logistic regression Mortality
2012 Lee, E. H. Intensive Care Med Korea Cohort Association between preoperative low levels of serum albumin and acute renal injury after offpump coronary artery bypass Age, body mass index, baseline renal function, left ventricular ejection fraction, liver disease, diabetes mellitus, occlusive peripheral vascular disease and chronic obstructive pulmonary disease Coincidence matching Greedy matching C statistic Logistic regression Renal failure
2012 Vaara, S. T. Acta Anaesthesiol Scand Finland Cohort Volume of annual cases in an intensive care unit of patients treated with renal replacement therapy due to acute renal injury and its association with inhospital mortality Gender, age, BMI, elective surgery, intra-aortic balloon, cardiopulmonary resuscitation, comorbidities, systemic organ failure, among others Covariable in one model Not reported P Value Logistic regression Mortality
2012 Wu, V. C. PLoS One Taiwan Cohort Impact of diuretic use and dose on mortality in critically ill patients with acute renal injury 30 days after dialysis Gender, age, BMI, elective surgery, intra-aortic balloon, cardiopulmonary resuscitation, comorbidities, systemic organ failure, among others Covariable in one model Not reported P Value Logistic regression and Cox regression Mortality
2013 Apel, M. Crit Care Germany Cohort End-stage renal disease is independently associated with a higher risk of death following major surgical procedures Age, sex, comorbidities, SAPS II, surgery type and SOFA subitems on admission to the ICU Coincidence matching Greedy matching P Value Logistic regression Mortality
2013 Bayer, O. Critical Care Medicine Germany Cohort Comparison of perioperative fluid management in cardiac surgery patients using colloids vs. crystalloids with dialysis, cardiovascular stabilization, fluid balance and mortality Age, hypertension, DM, cirrhosis, NYHA IV, acute MI, pulmonary hypertension, bypass time, serum creatinine, use of noradrenaline or adrenaline, NSAIDs, diuretics, ACEi, aminoglycosides, glycopeptides, antifungals, iodinated contrast media, septic shock, SAPSII score Stratification Not reported P Value Logistic regression Mortality
2013 Ehrmann Critical Care Medicine France Cohort Use vs. non use of iodinated contrast medium as related to an increase in the incidence of acute renal injury in ICU patients . Age, hypertension, creatinine levels, chronic renal failure, diabetes, respiratory failure, coma, trauma, fluid use 12 h before inclusion, use of mechanical ventilation, shock, lactate, Sequential Organ Failure Assessment (SOFA) and sepsis. Coincidence matching Not reported P Value Logistic regression Renal failure
2013 Leite, T. T. Crit Care Brazil Cohort Determination of the association between early vs. late initiation of dyalisis after AKIN stage 3 classification and in-hospital mortality in critically ill patients Demographic, clinical and laboratory data on ICU admission and on the day of renal replacement therapy initiation IPW (weighting) Not reported Standardized measurements + significance Logistic regression Mortality
2013 Verônica Torres Costa e Silva PLOS ONE Brazil Cohort Time of early vs. late nephrology consult and association with prognosis in patients with acute renal failure in the ICU Late consult: Diuresis, serum creatinine levels, surgical AKI and mechanical ventilation Nephrology consult: SAPS 3, CSA score, diuresis, serum creatinine levels, potassium and bicarbonate levels Covariable in one model Not reported ROC values Logistic regression Mortality and Renal failure
2014 Brewer, R. Ann Thorac Surg United States Cohort Comparison between on-pump and off-pump coronary revascularization surgery Age, gender, race, mortality risk, body surface area, smoker, preoperative creatinine, hypertension, DM, dyslipidemia, dialysis, chronic pulmonary diseases, peripheral vascular disease, cerebrovascular disease, prior stroke, prior acute MI, heart failure, type of surgery, number of compromised vessels, VEF <40%, left trunk stenosis >50%, preoperative intra-aortic balloon Coincidence matching Not reported P Value Logistic regression Mortality
2014 Elmiste-kawy, E Ann Thorac Surg Canada Cohort Incidence and associated risk factors in mild renal failure and determination of clinical impact following cardiovascular surgery Acute myocardial infarction, age, angina class, atrial fibrillation, cardiopulmonary bypass, CARE score, central nervous system complications, left ventricular functional class, coronary artery disease, endocarditis, sex, hypertension, intraoperative blood transfusion, hospital length of stay, length ofICU stay, time on ventilation, surgical priority, peripheral vascular disease, preoperative creatinine, preoperative hemoglobin, preoperative intra-aortic balloon pump, pulmonary edema, repeat cardiac surgery, shock, smoking Coincidence matching Nearest neighbor P Value Logistic regression Mortality
2014 Iwagami, M. Crit Care Med Japan Cohort Effect of postoperative polymyxin B hemoperfusion on mortality in patients with abdominal septic shock Age, sex, surgery-related findings; use of noradrenaline, dopamine and dobutamine; previously defined organ dysfunctions (6), hospital volumes and hospital type (teaching or non-teaching hospital) Coincidence matching Nearest neighbor C statistic Logistic regression Mortality
2014 Raghunathan Critical Care Medicine United States Cohort Association between the use of crystalloids and in-hospital mortality in critically ill adult patients with sepsis Not described Coincidence matching Greedy matching P Value Logistic regression Mortality
2014 Linder, A. Am J Respir Crit Care Med Canada Cohort At least one episode of mild acute renal failure (stage 1) vs. absence of renal failure associated with long-term survival after recovery from critical disease Sex, age, diabetes, MAP (lower baseline), noradrenaline treatment, lactate level on day 1, chronic heart failure, COPD, chronic liver failure, chronic steroid treatment, malignancy and HIV Coincidence matching Caliper P Value Regresión Survival
2014 Oh, H. J. Crit Care Korea Cohort Benefit of a team that specializes in acute renal injury patients undergoing continuous renal replacement therapy Age, gender, mean arterial pressure, APACHE II, organ failure, age-adjusted Charlson comorbidity index, RIFLE scale, contributing factors, use of anticoagulation, use of diuretics, laboratory tests (hemoglobin, leukocytes, creatinine, bicarbonate, BUN, total cholesterol, albumin, CRP, arterial pH and total bilirubin) Coincidence matching Greedy matching P Value Logistic regression and Cox regression Survival
2014 Picard, W Antimicrob Agents Chemother France Cohort Determine the nephrotoxicity risk associated with aminoglycoside treatment Age, gender, nosocomial sepsis, sepsis origin, sepsis severity, organ failure, rhabdomyolysis, renal graft of single functional kidney, DM, cirrhosis, pre-existing renal failure, nephrotoxic treatments, ACE inhibitors/ARBs, use of diuretics, high osmolarity contrast medium, use of hydoxyethyl starches, NSAIDs, antibiotics Covariable in one model Not reported P Value Cox regression Renal failure
2014 Vaara, S. T. Crit Care Med Finland Cohort Excess mortality attributable to acute renal injury Excess mortality attributable to acute renal injury Coincidence matching Not reported Standardized measurements + significance Logistic regression Mortality
2014 Wald, R. Crit Care Med Canada Cohort Intermittent vs. continuous dialysis associated with the risk of chronic dialysis Age, sex, hospitalization year, number of admissions and emergency room visits, number of visits to family physician, nephrologist, internist, cardiologist, geriatrician and psychiatrist; coronary artery, carotid or peripheral artery revascularization, heart valve surgery, coronary angiography, cardiac stress test, echocardiography, 12-lead individual electrocardiography or outpatient 24-hour electrocardiography, pacemaker insertion or carotid Doppler ultrasound within the 5 years prior to hospitalization and ARI diagnosis, major infection, alcoholism, chronic renal disease, liver dysfunction, heart failure, myocardial infarction, diabetes mellitus, cancer, cerebrovascular disease, myocardial infarction or angina and peripheral artery disease within 5 years prior to the date of hospital admission, Charlson comorbidity index, sepsis, non-ruptured aortic aneurysm repair, heart surgery, mechanical ventilation and initiation of dialysis in a teaching hospital Coincidence matching Nearest neighbor Standardized measurements Logistic regression and Cox regression Renal failure
2015 Chiang, Y. J Cardiothorac Vasc Anesth United States Cohort Effect of pulmonary artery catheterization on clinical outcomes after cardiac surgery in high risk patients Demographic characteristics, preoperative characteristics (hypertension, hypercholesterolemia, diabetes mellitus, obesity, prior stroke, coronary heart disease, endocarditis, chronic obstructive pulmonary disease, pulmonary hypertension, peripheral vascular disease, chronic renal failure, dialysis, liver disease, substance abuse, previous CABG, prior valve surgery), elective surgery, type of cardiac surgery, CABG, single valve CABG surgery, double or triple CABG surgery Coincidence matching Not reported P Value Logistic regression Mortality
2015 Ding, W. Cardiology China Cohort Intra-aortic balloon pump in relation to early clinical outcome in high risk patients taken to elective off-pump coronary artery bypass grafting (OPCABG) Age, sex, body mass index, smoker, DM, hypertension, dyslipidemia, renal dysfunction, prior stroke, prior acute MI, prior cardiac surgery, number of compromised coronary arteries, SYNTAX, BNO, euroSOCRE, LVEF, distal anastomosis Coincidence matching Not reported P Value Logistic regression Mortality
2015 Horkan, C. M. Crit Care Med United States Cohort Acute renal injury in patients who survived critical care and association of post-discharge outcomes (readmission beyond 30 days after discharge, postdischarge mortality and progression to endstage renal disease) Age, sex, race, Deyo-Charlson index, type of patient and risk classification, injury, failure, loss of renal function or end-stage renal disease Covariable in one model Caliper P Value Logistic regression and Cox regression Mortality and Renal failure
2015 Leite, T. T. Clin J Am Soc Nephrol Brazil Cohort Renal outcomes in critically ill patients receiving propofol or midazolam Sex, age, non-renal admission, simplified acute physiology score I and sequential organ failure evaluation, main comorbidities, type of ICU admission (medical, surgical or coronary), sepsis diagnosis, admission, Glasgow Coma Scale score, renal function on admission to the ICU (first serum creatinine measurement available), need for vasoactive drugs during the first 48 hours in the ICU and mean arterial pressure median in the first 48 hours in the ICU Coincidence matching Nearest neighbor Standardized measurements + significance Logistic regression Renal failure
2015 Alexandre Liborio Nephrology Dialysis Transplantation United States Cohort Propofol vs. midazolam. Sex, age, non-renal admission, simplified acute physiology score I and sequential organ failure evaluation, main comorbidities, type of ICU admission (medical, surgical or coronary), sepsis diagnosis, admission, Glasgow Coma Scale score, renal function on admission to the ICU (first serum creatinine measurement available), need for vasoactive drugs during the first 48 hours in the ICU and mean arterial pressure median in the first 48 hours in the ICU Coincidence matching Nearest neighbor + caliper Standardized measurements Logistic regression Renal failure
2015 Liborio, A. B. Clin J Am Soc Nephrol Brazil Cohort Complications of renal failure in critically ill patients associated with mortality and dialysis rates Age, SAPS-I/SOFA scores on admission, maximum BUN, cumulative fluid balance, minimum bicarbonate, maximum serum potassium and main comorbidities, including heart failure Coincidence matching Nearest neighbor P Value Logistic regression Mortality and Renal failure
2015 Sang, B. H. Crit Care Med Korea Cohort Influence of postoperative albumin level on the prevalence of acute renal injury after live donor liver transplant Age, sex, BMI, diabetes, hypertension, mean arterial pressure, CHILD PUGH score, model for end-stage liver diseasescore, hemoglobin, platelets, albumin, sodium, total bilirubin, creatinine, prothrombin time and intraoperative data IPW (weighting) Not reported C statistic Logistic regression and Cox regression Renal failure
2015 Shaw, A. D. Crit Care United States Cohort Type of intravenous fluid administered to patients with systemic inflammatory response syndrome and association with outcome Age, sex, race, admission source, admission type, urban, comorbidities Coincidence matching Greedy matching P Value Logistic regression Mortality
2015 Shen, T. C. Medicine (Baltimore) Japan Cohort Risk of septicemia in patients with endstage renal disease with or without renal transplant Age, sex, index date, comorbidities (hepatitis B, hepatitis C, Charlson comorbidity index score) and medications (steroids and immunosuppressors [cyclosporine, azathioprine, mycophenolate mofetil and tacrolimus]). Coincidence matching Not reported P Value Logistic regression and Cox regression Mortality
2016 Allen, C. J. Mil Med United States Cohort Risk of acute renal failure or death in patients with penetrating trauma receiving a single bolus of hydroxyethyl starch during initial fluid resuscitation Age and sex, injury characteristics (injury severity score, initial systolic blood pressure, heart rate, Glasgow scale), excess base, hematocrit, need for blood transfusion, ICU admission and surgery Coincidence matching Nearest neighbor P Value Logistic regression Mortality and Renal failure
2016 Bonnet, V. Medicine (Baltimore) France Cohort Comparison of postoperative complications of mitral valve repair taking into account age under and over 80 years and comorbidities Age over and under 80 years, sex, BMI, comorbidities, NYHA >=2, LVEF <45%, pulmonary hypertension, type of surgery, surgical time Coincidence matching Not reported C statistic Logistic regression Mortality
2016 Carmona, p Interact Cardiovasc Thorac Surg Spain Cohort On or offpump coronary revascularization surgery comparing intraoperative and 30-day postoperative outcomes Age, sex, EuroSCORE logistic, smoking, obesity BMI >30, hypertension under treatment, dislipidemia under treatment, diabetes mellitus under with insulin treatment, past medical history of stroke (stroke or transient ischemic attack), chronic pulmonary disease: pulmonary disease under treatment with bronchodilators or steroids, clinical symptoms of stable or unstable angina, past history of percutaneous coronary intervention, history of acute coronary syndrome within the previous 6 months. Surgery, intra-aortic balloon pump implantation before surgery. Chronic renal disease: defined as preoperative creatinine >2 mg/dL or need for renal replacement therapy, peripheral artery disease with intermittent clinical symptoms. Lower limb claudication, more than 50% carotid artery occlusion, amputation of any limb due to arterial disease, prior aortic or lower limb artery surgery, priority of the surgery: if the case was intervened within the first 24 hours after indication of surgery or later Coincidence matching Not reported ROC values Logistic regression Mortality
2016 Chao, p. W. Crit Care Med Taiwan Cohort Long-term results in critically ill septic patients who received cardiopulmonary resuscitation vs. those who did not receive resuscitation Age, year of onset, month of onset, sex, urbanization level, hospital level, Charlson comorbidity idex score, infection site, use of inotropic agents, dialysis, ventilation, days in the ICU, comorbidities, concomitant medications Stratification Nearest neighbor Standardized measurements Logistic regression and Cox regression Survival
2016 Charat Thongprayoon BMC Nephrology United States Cohort Effect of several measurement methods to determine baseline creatinine levels on the accuracy of acute renal injury diagnosis in critically ill patients Age, race, diabetes mellitus, hypertension, coronary heart disease, stroke, peripheral vascular disease, congestive heart failure and APACHE III on admission to the ICU Not reported Not reported C statistic Logistic regression Renal failure
2016 Cipolle, M. D. J Trauma Acute Care Surg United States Cohort Differences in results between trauma patients who received care in the trauma hospital program and patients with similar medical complexity trauma who did not receive care in the trauma program Age, injury severity score and pre-existing diabetes mellitus conditions requiring insulin therapy, hypertension, congestive heart failure and stroke Coincidence matching Not reported P Value Logistic regression Mortality
2016 Clark, E. Intensive Care Med Canada Cohort Clinical characteristics and in-hospital mortality in patients with end-stage chronic renal failure o chronic dialysis with septic shock as compared to patients with septic shock not on chronic dialysis Age, sex, comorbidities (cancer, immune compromise, ICC, coronary artery disease, elective or emergency surgery, COPD, dibetes) and APACHE II score Coincidence matching Nearest neighbor Standardized measurements and C statistic Logistic regression and Cox regression Mortality
2016 Hammond, D. A. Pharmacotherapy United States Cohort Comparison of the incidence of acute renal failure development in critically ill patients who received concomittant vancomycin with piperacillin-tazobactam or cefepime Age, sex, BMI, initial creatinine, creatinine at the time of antibiotic initiation, SOFA, APACHE IV, septic shock, number of concomitant antibiotics, initial vancomycin levels, number of nephrotoxic medications, length of stay, admission unit, infection-associated length of stay, comorbidities IPW (weighting) Not reported P Value Logistic regression Renal failure
2016 Henriikka Mildh Mildh et al. Ann. Intensive Care Finland Cohort Three-year mortality in survivors of 30 days of critical care with acute renal injury Age, comorbidities and ICU admission and treatment characteristics Coincidence matching Caliper Standardized measurements Logistic regression and Cox regression Mortality
2016 J. Latour-Péreza Med Intensiva Spain Cohort Assessment of the use and effectiveness of the routine invasive strategy in patients with non-ST elevation acute coronary syndrome with renal dysfunction Age, sex, mode of access, transport, delays and access to the center, cardiac catheterization availability, coronary risk factors, prior cardiovascular disease, prior treatments (pharmacological and coronary) and baseline clinical status (blood pressure, heart rate, initial EKG, initial Killip, TIMI score, GRACE and CRUSADE). Stratification Not reported Standardized measurements Not reported Mortality
2016 Joung, K. W. Medicine (Baltimore) Korea Cohort Comparison of postoperative acute renal injury in ileal conduit and urinary shunting of the neobladder following radical cystectomy Sex, years, height, weight, body mass index, history of hypertension, diabetes mellitus, coronary heart disease or cerebrovascular disease, cancer stage and grade, application of neoadjuvant chemotherapy, angiotensin converting enzyme inhibitor, angiotensin receptor II blocker or diuretic medications, albumin, hematocrit, creatinine, uric acid and placement of ureteral stent Coincidence matching Not reported P Value Logistic regression Renal failure
2016 Jun, I. G. Liver Transpl Korea Cohort Incidence of acute renal failure in patients with live donor liver transplant with and without ABO compatibility ABO compatibility, sex, age, body weight, height, body mass index (BMI), ABO type and end-stage liver disease score model (MELD); comorbidities such as diabetes mellitus, hypertension and coronary artery disease and preoperative laboratory values such as platelet count, prothrombin time, normalized international proportion and hemoglobin, glucose, serum albumin, serum sodium, total serum bilirubin, sCr, serum aspartate aminotransferase, serum alanine aminotransferase, and the highest intraoperative IA titre. Stratification Nearest neighbor Standardized measurements Logistic regression and Cox regression Survival
2016 Karamanos, E. World J Surg United States Cohort Impact of diabetes mellitus on outcomes of patients undergoing emergency laparotomy due to small bowel adhesive obstruction Age, gender, BMI, comorbidities, wound classification, Charlson index, inflammatory status at the time of admission. Not reported Not reported C statistic Logistic regression Renal failure
2016 Kerry C. Cho Clinical Science Articles United States Cohort Age, liver failure, sepsis, thrombocytopenia, blood urea nitrogen and serum creatinine Covariable in one model Not reported P Value Logistic regression and Cox regression Survival
2016 Lai, W. H. Scand J Trauma Resusc Emerg Med Taiwan Cross-sectional Incidence and clinical presentation of post-traumatic acute renal failure in hospitalized adult patients and its association with shock Sex, years, comorbidity, Glasgow scale, head/neck, chest, abdomen or limb injury according to severity on the abbreviated injury scale for each body region and injury severity score (ISS). Coincidence matching Not reported P Value Logistic regression Mortality and Renal failure
2016 Lin, C. Y. Medicine (Baltimore) Taiwàn Cohort Correlation of preoperative renal failure with mortality and morbidity after aortic valve replacement vs. patients without renal failure. Preoperative demographics, associated comorbidities, surgical procedures and hospital recovery were compared between groups. In-hospital mortality was defined as death occurring during hospitalization. Emergency operations included hemodynamically unstable patients requiring a high dose of inotropic agents, endotracheal intubation or mechanical circulatory support, including intra-aortic balloon pump (IABP) or extracorporeal membrane oxygenation (ECMO) before the surgery. Coincidence matching Nearest neighbor P Value Logistic regression Mortality
2016 Mazzeffi, M. J Anesth United States Cohort Preoperative use of aspirin associated with cardiac surgery patients with massive transfusion. Type of surgery, urgency of the surgery, year of the surgery, cardiopulmonary bypass time, use of antifibrinolytics, age, sex, weight, height, preoperative platelet count, preoperative hematocrit, preoperative INR, baseline hemodialysis, baseline creatinine, chronic pulmonary disease, hypertension, diabetes, dislipidemia, peripheral vascular disease, cerebrovascular disease, infectious endocarditis, prior myocardial infarction, prior cardiac intervention, congestive heart failure within the 2 weeks following surgery, use of betablockers, use of statins and use of warfarin Coincidence matching Greedy matching P Value Logistic regression Mortality
2016 Moise, p. A. J Anesth Clin Ther United States Cohort Comparing the results of early treatment with daptomycin vs. vancomycin for methicillin-resistant Staphylococcus aureus bacteremia Variables identified for matching were: age (<60 vs. >60), disease severity (ICU status) and type of infection (intravascular, extravascular or catheter). Bacteremia clearing was used to validate the hierarchy for the type of infection. Coincidence matching Not reported P Value Logistic regression Mortality
2016 Oscar Peñuelas Journal of Intensive Care Medicine Spain Cross-sectional Assessment of predictors for ICU-acquired paresis and associated short-term outcomes Age, SAPS II, chronic obstructive pulmonary disease, sepsis (leading to mechanical ventilation or as a complication during mechanical ventilation), use of neuromuscular blockade, cardiovascular failure, renal failure, hematological insufficiency and days on mechanical ventilation before weaning Coincidence matching Nearest neighbor y Caliper Standardized measurements y ROC values Logistic regression Mortality
2016 Park, J. Y. Crit Care South Korea Cohort Results of early vs. late initiation of renal replacement therapy Age, sex, CCI, systolic arterial pressure, prothrombin time and total bilirubin level Coincidence matching Nearest neighbor y Caliper Standardized measurements +significance Logistic regression and Cox regression Survival
2016 Ruan, S. Y. Crit Care Taiwan Cohort Risk of acute renal failure associated with inhaled nitric oxide therapy in patients with acute respiratory distress syndrome Age, sex, year of entry to the cohort, height, body weight, ARDS etiology, simplified acute physiology score II (SAPS II) 15, lung injury score 16, comorbidities, ventilator settings, vital signs, arterial blood gases, complete blood count, creatinine and bilirubin levels, urine output, radiological pattern, and use of vasopressors on the first ARDS day. Creatinine clearance was estimated using the Cockcroft-Gault. Stratification Caliper P Value Logistic regression and Cox regression Renal failure
2016 Thongprayoon, C. J Am Soc Nephrol United States Cohort Assessing the risk of acute renal failure in patients taken to transcatheter aortic valve replacement (TAVR) vs. surgical aortic valve replacement (SAVR) STS risk score, eGFR, sex, body mass index, type of surgery (elective vs. urgent/emergent), presence of anemia, history of diabetes mellitus, hypertension, congestive heart failure, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, NYHA (classes 1 and 2 vs. 3 and 4), preoperative use of angiotensin converting enzyme inhibitor or angiotensin receptor blocker, prior percutaneous cardiac intervention, prior coronary bypass grafting, prior aortic valve surgery, cardiac rhythm during surgery (normal vs other) and aortic valve regurgitation (none or trivial vs, mild, moderate or severe) Coincidence matching Caliper Standardized measurements Logistic regression Renal failure
2017 Bentzer, p. Scand J Trauma Resusc Emerg Med Sweden Cohort Effect of dextran-70 on outcomes in patients with severe sepsis Age, sex, chronic obstructive pulmonary disease (COPD), renal failure, diabetes, outcome variables (except mortality) and fluid administration data. Coincidence matching Greedy matching Standardized measurements Logistic regression Mortality and Renal failure
2017 Chan, M. J. Medicine (Baltimore) Taiwàn Cross-sectional Effects of 3 different surgical methods for coronary revascularization in renal injury Age, sex, preoperative creatinine level, cardiac ejection fraction, emergent/ urgent intervention, use of intra-aortic balloon pump (IABP), preoperative ventilator use, diabetes mellitus, recent myocardial infarction, shock and chronic obstructive pulmonary disease Coincidence matching Not reported P Value Logistic regression Renal failure
2017 Dos Santos, T. O. C. PLoS One Brazil Cross-sectional Mortality-related factors in patients admitted to the ICU with acute renal failure requiring dialysis who underwent continuous veno-venous hemodiafiltration (CVVHDF) Age, comorbidities, severity of the disease that caused AKI, metabolic disorders, organ dysfunction and weight gain during the resuscitation phase. Other factors that appear to influence prognosis in these patients are those related with the therapy itself, including type of dialysis modality, time of initiation of the procedure, administered dialysis dose, metabolic control and volume status obtained over the course of treatment Coincidence matching Not reported P Value Logistic regression and Cox regression Survival
2017 Dou, L. Nephron United States Cohort Obstructive sleep apnea associated with the risk of acute renal failure in the intensive care unit Age, BMI, history of CHF, CKD, CPD, hypertension, diabetes, high risk surgery, exposure to nephrotoxins, mechanical ventilation and APACHE III score Coincidence matching Caliper Standardized measurements Logistic regression Renal failure
2017 Hoi-Ping SHUM Nephrology China Cross-sectional Impact ofrenal replacement therapy on 90-day mortality in critically ill patients with KDIGO stage 3 renal failure Age, baseline GFR, worse creatinine/potassium/ bilirubin/platelet count level, pH within the first 24 hours of admission to the ICU and APACHE IV score Coincidence matching Nearest neighbor y Caliper C statistic Logistic regression and Cox regression Mortality
2017 Jennifer S. McDonald Intensive Care Med United States Cohort Association between intravenous iodinated contrast material administration and later development of post-contrast AKI, emergent dialysis and short-term mortality Age, serum creatinine before CT (sCr), changes in sCr prior to exploration and SOFA score, comorbidities, use of nephrotoxic drugs, IV fluids used, stability prior to CT Stratification Not reported P Value Logistic regression Mortality and Renal failure
2017 Kianoush Kashan PLOS ONE United States Cohort Effect of bicarbonate concentration in the replacement solution on patients in the ICU that required CRRT for AKI. pH, bicarbonate, interaction between pH and bicarbonate, presence of sepsis, use of mechanical ventilation, vasopressor administration, intra-aortic balloon pump (IABP) and Charlson comorbidity index Coincidence matching Caliper Standardized measurements Logistic regression and Cox regression Mortality
2017 Kovacheva, V. p. Neurosurgery United States Cohort Determine ifAKI in patients after craniectomy is associated with higher 30-day mortality Age, sex, race (white vs. non-white), Deyo-Charlson index, sepsis and chronic renal disease covariables, severity of the disease, intracranial bleeding, subarachnoid hemorrhage, acute respiratory failure and trauma Coincidence matching Caliper P Value Logistic regression y Cox Mortality
2017 Latour-Perez, J Med Intensiva Spain Cohort Assess the use an effectiveness of the routine invasive strategy in patients with non-ST elevation acute coronary syndrome with renal dysfunction Age, sex, clinics, mode of access, transportation and delay in accessing the center, availability of cardiac catheterization, coronary risk factors, prior cardiovascular disease, previous treatments (pharmacological and coronary) and baseline clinical status (blood pressure, heart rate, initial EKG, initial Killip, TIMI 24, GRACE and CRUSADE scores Stratification Not reported Standardized measurements + significance Logistic regression Mortality
2017 Legouis, D. Anesthesiology France Cohort Impact of acute renal injury in patients who experience reversible acute renal injury after cardiac surgery Age, body mass index, preoperative eGFR, clamping and bypass time, European logistic system for assessing the cardiac surgery risk, left ventricular ejection fraction and type of surgery Coincidence matching Nearest neighbor y Caliper Standardized measurements Logistic regression and Cox regression Renal failure
2017 Marik, p. E. Chest United States Cohort Hydrocortisone, vitamin C and thiamine for the treatment of severe sepsis and septic shock Age, sex, diagnosis on admission, comorbidities, need for mechanical ventilation, use of vasopressors, daily urine output, fluid balance after 24 and 72 hours, length of stay in the ICU and laboratory values, Immunocompromised patients if they received cytotoxic therapy or if diagnoses with acquired immune deficiency syndrome. Serum creatinine levels, leukocyte counts (WB), platelet count, total bilirubin, PCT and lactate. Acute renal injury (AKI), APACHE II and APACHE IV scores on admission, including foreseen in-hospital mortality in APACHE IV Not reported Not reported P Value Logistic regression Survival
2017 Shou-Chun Yu, World Neurosurgery Tai wan Cohort One-year mortality rate in patients with spinal cord injury with CKDandTRDand comparison in patients with spinal cord injury without CKD Age, sex, comorbidities, length of stay in the intensive care unit and total length of stay Stratification Not reported P Value Cox regression Mortality
218 Etienne Gayat Intensive Care Med France Cohort Association between the use of ACEi and ARA II in patients with acute renal failure in the ICU with mortality within 1 year Age, Charlson score, CKD, diabetes mellitus, chronic heart failure, ACEi/ARB administration before ICU admission, vasopressor administration while in the ICU, RRT during ICU stay, systolic blood pressure at discharge from the ICU Coincidence matching Nearest neighbor Standardized measurements Cox regression Mortality
2018 Buyun Wu Scientific Reports China Cohort Impact of low AKI recognition (more than 3 days after onset) on short-term prognosis Sex, age, setting in which AKI occurred, eGFR on admission, oliguria, history of malignant neoplasm, APACHE II score, SOFA score, risk factors for AKI (hypovolemia, heart failure, sepsis, surgery), AKI stage, blood urea nitrogen and whether RRT was used IPW (weighting) y Stratification Caliper Standardized measurements Logistic regression Mortality
2018 Shih-Ting Huang_ nuevo CMAJ Tai wan Cohort Renal outcomes in patients with no preexisting renal disease who were admitted due to critical acute disease Age (5-year periods), sex and index year with control patients, comorbidities Coincidence matching Not reported Standardized measurements Logistic regression and Cox regression Mortality and Renal failure
2019 AMANDA Y WANG Nephrology Australia Cohort HMG-CoA reductase inhibitors (statins) and acute renal injury Age, sex, sepsis, mechanical ventilation and APACHE III scores at the time of randomization Coincidence matching Not reported P Value Cox regression Renal failure
2019 Ana p. A ccmjournal Brazil Cohort Assessment of inhospital mortality and mortality risk factors in patients with cancer discharged from the ICU after the decision to refuse vital support therapies SAPS 3 and SOFA on admission to the ICU, setting prior to admission to the ICU, infection on admission to de ICU, use of non-invasive mechanical ventilation, acute renal injury and delirium during while in the ICU Coincidence matching Caliper P Value Logistic regression Mortality
2019 Ian E. McCoy, Crit Care Expl United States Cohort Estimation of the effects of diuretic use in the first 24 hours in the ICU Age, sex, race, comorbidities, type of admission, ICU type, serum creatinine on admission, mechanical ventilation and each of the SOFA score items on admission, out of six ICU items Coincidence matching Greedy matching Standardized measurements Logistic regression Renal failure
2019 Muhammad Zeeshan American Association for the Surgery of Trauma United States Cohort Four-factor prothrombin complex plus fresh frozen plasma vs. fresh frozen plasma alone in patients with trauma-related bleeding Age, sex, race, body mass index, vital signs (systolic blood pressure, heart rate, Glasgow scale), time to therapy initiation, injury mechanisms, injury severity score, abbreviated injury scale (AIS), Head-AIS, chest-AIS,abdominal-AIS, spine-AIS, limb-AIS, comorbidities, use of anti-platelet agents before injury, and trauma center level Stratification Nearest neighbor P Value Logistic regression Mortality and Renal failure
2019 Philippe Gaudard BMC Infectious Diseases France Cohort Acute renal injury during treatment with daptomycin vs. vacomycin in critically ill cardiovascular patients Severity score such as simplified acute physiology (SAPS II) on admission to the ICU and sequential organ failure assessment (SOFA) upon initiation of treatment, history of chronic renal failure (creatinine clearance under 50 mL/min), creatinine and AKI before initiation of treatment (defined as baseline), sepsis-related or heart-failure related circulatory shock (defined as systolic blood pressure undere 90 mmHg and resistant to fluid challenge, with evidence of end-organ hypoperfusion such as oliguria, altered mentation or high plasma lactate level), need for cardiopulmonary bypass for surgery (within 48 h of antibiotic initiation), use of other nephrotoxic agents (iodine contrast, aminoglycosides, cyclosporin), bacteremia documented by CPG Coincidence matching Not reported P Value Logistic regression Renal failure
2019 Raphael Donadio Pitta European Journal of Clinical Microbiology & Infectious Diseases Brazil Cross-sectional Comparison of the incidence of AKI in critically ill patients treated with aminoglycosides or meropenem in the intensive care setting Age, sex and use of vancomycin, polymyxin, diuretic and amphotericin Coincidence matching Nearest neighbor y Caliper P Value Logistic regression Renal failure
2019 Sarah A. Schubert The Journal of Thoracic and Cardiovascular Surgery United States Cohort Determine the influence of perioperative administration of betablockers before aortic valve replacement Rates of hypertension, anemia, coronary artery disease, preoperative atrial fibrillation or heart failure, among others Stratification Greedy matching Standardized measurements Logistic regression Mortality
2019 Wuhua Jiang Braz J Cardiovasc Surg China Cohort Discover potentially modifiable perioperative predictors for renal replacement therapy (RRT) in patients with cardiac surgery-associated acute renal injury Age, comorbidities, history of contrast media exposure, preoperative cardiac function status (NYHA classification), baseline eGFR (CKD-EPI), creatinine levels surgery-related factors (proceduresk urgency, time on cardiopulmonary bypass, red blood cell transfusion) and operative characteristics of central venous pressure and red blood cell transfusion. If >1 cardiac surgery procedures were performed during the same admission, only the data for the first surgery were considered Coincidence matching Not reported P Value Logistic regression Renal failure
2019 Yanfei Shen Critical Care BMC China Cohort Relationship between the use of loop diuretics and in-hospital mortality in critically ill patients with vasopressor support Age, weight, diabetes mellitus, hypertension, heart disease, AKI, SOFA score on admission to the ICU, white count, fluid volume intake and proportions of various vasopressors Stratification Nearest neighbor y Caliper Standardized measurements Logistic regression Mortality and Renal failure
2019 Yoshihisa Miyamoto Intensive Care Med Japan Cohort Examine the association between the use of intravenous contrast and no recovery from acute renal injury requiring dialysis (AKI-D) and hospital in-hospital mortality in patients with sepsis Age, sex, CCI, congestive heart failure, diabetic nephropathy, CKD, consciousness level with JCS, discharge year, source of infection, emergency surgery due to infectious source, hispital volume and treatments performed within the next 2 days after discharge Coincidence matching Nearest neighbor Standardized measurements and C statistic Logistic regression Mortality and Renal failure
2019 Yu Hsiang Chou Journal of Nephrology Taiwan Cohort Preoperative continuation of renin-angiotensin system inhibitor and cardiac surgery-associated acute kidney injury Age, sex, CKD, DM, decompensated heart failure, coronary heart disease, occlusive peripheral artery disease, hypertension, hyperlipidemia, chronic obstructive pulmonary disease, cirrhosis, hyperuricemia, metastatic cancer, smoking, hemoglobin, baseline creatinine, albumin, urinary protein, surgery type, antihypertensives and statins Coincidence matching Caliper P Value Logistic regression and Cox regression Renal failure
2019 Yuya Goto Goto et al. Critical Care Japan Cohort Assess the effect of contrast media on renal function and mortality in patients with sepsis who already had AKI Age, sex, infection site, history of chronic kidney disease, diabetes mellitus, septic shock, immune compromise, use of aminoglycosides or vancomycin, laboratory data (hemoglobin, Cr, total bilirubin, platelets, lactate), partial oxygen pressure/ inspired oxygen fraction ratio, and Glasgow coma scale on admission to the ICU Coincidence matching Not reported P Value Logistic regression Mortality and Renal failure
2019 Lisa-Mae Chest United States Cohort Association between contrast and acute renal injury in critically ill patients with normal renal function Age, gender, comorbidities, use of contrast, race, obesity, primary diagnosis, APACHE Stratification Nearest neighbor y Caliper Standardized measurements Logistic regression Renal failure
2020 Ala Abudayyeh American Journal of Hospice United States Cohort Determine the impact of dialysis on survival of patients admitted to the ICU with stage IV cancer and renal failure Age, sex, race/ethnicity and vital status (alive vs. dead). Clinical data included type of malignant neoplasm, peripheral vascular disease and/or dementia diagnosis, and consults for dialysis and palliative care. Dementia, peripheral vascular diseases and reduced albumin, length of stay in the ICU, SOFA score and disposition. First albumin level obtained in the first 3 days in the ICU Coincidence matching Nearest neighbor Standardized measurements Logistic regression and Cox regression Renal failure
2020 Guangju Zhao Zhao et al. Critical Care BMC United States Cohort Administración de furosemida y los resultados en pacientes críticos con lesión renal aguda. Edad, género, etnia, tipo de admisión, comorbilidades, puntuación simplificada de fisiología aguda II (SAPSII), puntuación de evaluación de falla orgànica secuencial (SOFA), presión arterial media (MAP), nivel de creatinina en suero, uso de vasopresores e inotrópicos, entrada diaria de líquidos, fluido equilibrio, tipos de líquidos, cirugía cardiaca, TSR y ventilación mecànica. Coincidence matching Nearest neighbor y Caliper Standardized measurements Logistic regression and Cox regression Renal failure
2020 Gustavo A.Ospina-Tascón Critical care Colombia Cohort Effects of very early initiation of noradrenaline in patients with septic shock Age, APACHE II, comorbidities and patient origin (emergency room, general ward or intensive care unit) were documented. Heart rate and blood pressure were also documented and at 2, 4, 6, 8 and 24 h after vasopressor administration. Multiple organ dysfunction was assessed using the sequential organ failure assessment score (SOFA). Days off ventilation, renal replacement therapy requirement and days without RRT were also calculated. Finally, ICU and hospital length of stay were recorded together with ICU, in-hospital and 28-day mortality Coincidence matching Nearest neighbor P Value Cox regression Mortality and Renal failure

SOURCE. Authors.

SOURCE: Authors.

COMPLEMENTAR CONTENT 2. Evolution of publications using PS, adjusted for every 10000 publications. 

COMPLEMENTAR CONTENT 3. Number ofpublications with propensity score by journal. 

Journal n (%)
Critical Care 13(13.1)
Critical Care Medicine 11(11.1)
Intensive Care Medicine 7(7.1)
Medicine (Baltimore) 5(5.1)
Plos One 4(4.0)
Clinical Journal of the American Society ofNephrology 3(3.0)
The Annals of Thoracic Surgery 3(3.0)
Journal of Cardiothoracic and Vascular Anesthesia 3(3.0)
Anesthesiology 3(3.0)
Journal of the American Society of Nephrology 2(2.0)
Medicina Intensiva 2(2.0)
Scandinavian Journal of Trauma Resuscitation and Emergency Medicine 2(2.0)
Nephrology 2(2.0)
Chest 2(2.0)
The American journal of medicine 1(1.0)
JAMA 1(1.0)
The International Journal of Artificial Organs 1(1.0)
JAAPA 1(1.0)
European Journal of Cardio-Thoracic Surgery 1(1.0)
European Journal of Vascular and Endovascular Surgery 1(1.0)
Acta Anaesthesiologica Scandinavica 1(1.0)
American Journal ofRespiratory and Critical Care Medicine 1(1.0)
Antimicrobial Agents and Chemotherapy 1(1.0)
Cardiology 1(1.0)
Military Medicine 1(1.0)
Interactive Cardiovascular and Thoracic Surgery 1(1.0)
BMC Nephrology 1(1.0)
The Journal of Trauma and Acute Care Surgery 1(1.0)
Pharmacotherapy 1(1.0)
Annals of Intensive Care 1(1.0)
Liver Transplantation 1(1.0)
World Journal of Surgery 1(1.0)
Clinical Science Articles 1(1.0)
Journal ofAnesthesia 1(1.0)
Clinical Therapeutics 1(1.0)
Journal of Intensive Care Medicine 1(1.0)
Nephron 1(1.0)
Neurosurgery 1(1.0)
World Neurosurgery 1(1.0)
Canadian Medical Association Journal 1(1.0)
critical care explorations 1(1.0)
American Association for the Surgery of Trauma 1(1.0)
BMC Infectious Diseases 1(1.0)
European Journal ofClinical Microbiology & Infectious Diseases 1(1.0)
The Journal of Thoracic and Cardiovascular Surgery 1(1.0)
Brazilian Journal of Cardiovascular Surgery 1(1.0)
Journal ofNephrology 1(1.0)
American Journal of Hospice 1(1.0)
Scientific Reports 1(1.0)
Nephrology Dialysis Transplantation 1(1.0)

SOURCE. Authors.

COMPLEMENTAR CONTENT 4. Number ofpublications with propensity score by country. 

Country n (%)
United States 35(35.3)
Taiwàn 10(10.1)
France 9(9.1)
Brasil 7(7.1)
Canada 6(6.1)
Korea 6(6.1)
Japan 5(5.1)
China 5(5.1)
Spain 4(4.0)
Finland 3(3.0)
Australia 2(2.0)
Germany 2(2.0)
Italy 1(1.0)
UK 1(1.0)
South korea 1(1.0)
Sweden 1(1.0)
Colombia 1(1.0)

SOURCE. Authors.

COMPLEMENTAR CONTENT 5. Exposures in critically ill patients with renal outcomes. 

Exposure n (%)
Intervención farmacológica
Antibiotic 7(7.1)
IV fluids 6(6.1)
Diuretics 5(5.1)
Statins 3(3.0)
Midazolam. propofol 2(2.0)
ACEi/ARAII 2(2.0)
Inotropes 1(1.0)
Inhaled nitric oxide 1(1.0)
Hydrocortisone. vitamin C and thiamine 1(1.0)
Prothrombin + fresh frozen plasma vs. fresh plasma 1(1.0)
Betablockers 1(1.0)
Diuretics + vasopressors 1(1.0)
Vasopressors 1(1.0)
Aspirin 1(1.0)
Risk factor
History of acute renal failure 10(10.1)
End-stage renal disease 2(2.0)
Factors affecting healing. angiosome 1(1.0)
Renal transplant 1(1.0)
Age under or over 80 years 1(1.0)
Diabetes mellitus 1(1.0)
Obstructive sleep apnea 1(1.0)
Critical disease in patient with no past history of acute renal injury 1(1.0)
Post-traumatic acute renal failure 1(1.0)
Dialysis therapy
Early/late 5(5.1)
Continuous/intermittent 4(4.0)
Renal replacement therapy 4(4.0)
Chronic dialysis 1(1.0)
Continuous veno-venous hemodiafiltration 1(1.0)
Bicarbonate in replacement solution 1(1.0)
Surgical and other invasive procedures
Iodinated contrast medium 5(5.1)
Coronary revascularization surgery (bypass) 4(4.0)
Invasive techniques in patients with NSTEMI and renal dysfunction 2(2.0)
Immediate postoperative extubation 1(1.0)
Red blood cell transfusion 1(1.0)
Radiofrequency ablation for atrial fibrillation 1(1.0)
Pulmonary artery catheterization 1(1.0)
Intra-aortic balloon pump 1(1.0)
Radical cystectomy 1(1.0)
Liver transplant (ABO compatibility) 1(1.0)
Transcatheter/surgical aortic valve replacement 1(1.0)
Cardiopulmonary resuscitation 1(1.0)
Administrative processes and biomarkers
Postoperative hypoalbuminemia 3(3.0)
Care by residents/medical assistance/ trauma team 2(2.0)
Nephrology consult (early/late ) 1(1.0)
Specialized equipment for RRT 1(1.0)
Volume of patients on dialysis in the ICU 1(1.0)
Paresis in ICU 1(1.0)
Refusal of vital support therapies in the ICU 1(1.0)
Creatinine measurement methods 1(1.0)

SOURCE: Authors.

COMPLEMENTAR CONTENT 6. Fòrmule 1. The difference in standardized measurements is defined as follows:

d=100xX⃛Tratados-X⃛ControlSTratados2+SControl22

SOURCE: Austin (6).

Where X⃛Tratados and X⃛Control represents variable averages between treated and not treated, while STratados2 and SControl2 correspondsto standard deviations of the covariables between treated and not treated subjects.

source. Austin (6).

Received: April 24, 2020; Accepted: August 10, 2020; other: October 05, 2020

*E-mail:h.oliverosr@javeriana.edu.co

Correspondence: Cra. 7 No. 40-62, PhD Program in Clinical Epidemiology, Department of Clinical Epidemiology and Biostatistics, Pontificia Universidad Javeriana, Bogotá, Colombia.

Creative Commons License This is an open-access article distributed under the terms of the Creative Commons Attribution License

 
Cra. 15a No. 120-74
Teléfono: (57-1) 6196077 Ext. 4111