Outcomes, risk factors and health burden of contrast-induced acute kidney injury: an observational study of one million hospitalizations with image-guided cardiovascular procedures

Background Despite the use of low-osmolar contrast media that have significantly reduced the occurrence of severe adverse reactions, contrast-induced (CI) acute kidney injury (AKI) remains the third cause of AKI in hospitals. We sought to estimate the frequency of CI-AKI among hospitalized patients undergoing image-guided cardiovascular procedures, to quantify the effect of risk factors on the development of this complication and to assess relative organizational and economic burden in healthcare. Methods A retrospective cross-sectional population-based study using the extensive French hospital discharge database (PMSI) was carried out. Hospitalizations with image-guided cardiovascular procedures using a contrast media were identified in adults over a 2-year period (2012–2013). Suspected CI-AKI was defined as the presence, during hospitalization, of a diagnostic code of AKI (International Classification of Diseases, 10th revision [ICD-10] codes: N141, 142, N144, N990, N17, N19 or R392) or a code of renal replacement therapy procedure (Classification Commune des Actes Médicaux [CCAM] codes: JVJB001, JVJF002-005 and JVJF008) as creatinine criteria were not available. Results During 1,047,329 hospitalizations studied, 32,308 suspected CI-AKI were observed, yielding a frequency of 3.1 %. By multivariate analysis, factors that significantly increased the risk of suspected CI-AKI included cardiogenic shock (odds ratio [OR] = 20.5, 95 % confidence interval [95 % CI] [18.7; 22.5]), acute heart failure (OR = 2.5, 95 % CI [2.4; 2.6]) and chronic kidney disease (OR = 2.3, 95 % CI [2.2; 2.3]. Renal replacement therapy was initiated during 6,335 (0.6 %) hospitalizations. The mean length of stay and cost of hospitalizations associated with suspected CI-AKI was higher than in hospitalizations without suspected CI-AKI (20.5 vs 4.7 days, p < 0.00001 and €15,765 vs €3,352, p < 0.0001, respectively). Conclusions This is the first large-scale population-based study to estimate frequency and health burden of suspected CI-AKI occurring after image-guided cardiovascular procedures, and the first available data in a French population. We showed that this iatrogenic complication remains of high concern despite prevention efforts and contrast media product improvement. From our results, suspected CI-AKI is associated with particularly high mortality, significantly extends hospitalizations, and leads to additional costs reaching a total of €200M per year.


Background
The use of iodinated contrast media (ICM) improves the visualization of blood vessels during image-guided procedures [1]. However, it may lead to iatrogenic renal injury, causing contrast-induced (CI) acute kidney injury (AKI).
AKI, formerly known as acute renal failure, is common in hospitalized patients and is strongly associated with morbidity and mortality. It is a complex syndrome that may arise in response to various aetiologies, such as the use of nephrotoxic drugs.
CI-AKI is the third most common cause of AKI in hospitals, after functional and medicine-related causes [2]. Depending on the definition, the population and the procedure studied, CI-AKI frequencies reported in the literature range from 1.5 to 15 % [3][4][5][6][7]. In most patients involved, serum creatinine returns to normal values within 2 weeks [8]. However, a subset of this population will need renal replacement therapy with associated increased hospitalization times and costs, and increased mortality [9,10].
The frequency of CI-AKI was expected to decrease with the use of low-osmolar ICM. In order to provide insights on the current state of this complication to health professionals and authorities, we conducted a cross-sectional retrospective population-based study using hospital data. The objective was to estimate the frequency of CI-AKI among hospitalized patients who underwent image-guided cardiovascular procedures, to quantify the effect of risk factors on CI-AKI occurrence and to assess the relative organizational and economic burdens.

Data source
Subjects were identified from the extensive French hospital discharge database (Programme de Médicalisation des Systèmes d'Information [PMSI]), which covers all French public and private hospitals involved in medicine, surgery and obstetrics. In 2004, French hospitals adopted a prospective payment system. Since then, the PMSI database has become the basis of hospital funding by third-party payers. A standard discharge summary report is generated for each hospitalization, and includes information on patient characteristics (e.g. sex, age, residence code), the main diagnosis that led to hospital admission, examinations carried out during hospitalization, comorbidities and possible complications. Biological data are not recorded. Diagnoses are coded using the International Classification of Diseases, 10th revision (ICD-10); procedures using the French classification of procedures in medicine (Classification Commune des Actes Médicaux [CCAM]). The discharge summary is then linked to a diagnosis-related group, used for the classification of hospitalizations, which are associated with national fees. Permission to extract and use the PMSI data was obtained from the National Commission for Data Protection and Liberties.

Study population
The hospitalizations of adult patients (≥18 years) who had undergone image-guided cardiovascular interventions requiring intra-arterial or intra-venous ICM injections (Appendix 1) were extracted from 2012-2013 PMSI hospital databases. A committee of experts approved the selection of procedures.
Each hospitalization was considered as unique even though a patient could have had several different hospitalizations between January 1, 2012 and December 31, 2013.
When renal replacement therapy was used during the hospitalization of interest, the suspected CI-AKI was classified as requiring renal replacement therapy; otherwise, it was considered as without renal replacement therapy.
End-stage kidney disease patients were identified by searching for a relevant diagnosis (ICD-10 code: N185) or a procedure of renal replacement therapy for chronic kidney disease (ICD-10 code: Z992; CCAM code: JVJF004, JVJF008 or JVJB001) within the 2 years prior to the hospitalization of interest. These hospitalizations were excluded from the analysis of suspected CI-AKI requiring renal replacement therapy as the renal replacement therapy would most likely have been related to the underlying disease.

Collected data
All data were collected from the PMSI database. Known risk factors for developing CI-AKI (diabetes, chronic kidney disease, acute heart failure, onset of cardiogenic shock during the hospitalization) were recorded. The presence of diabetes (ICD-10 codes: E10-E14 or N083) or chronic kidney disease (ICD-10 code: N18) was searched for within the 2 years prior to the hospitalization, while acute heart failure (ICD-10 code: I50) and cardiogenic shock (ICD-10 code R570) were searched for only during the hospitalization.
In addition, sociodemographic characteristics (age and sex) and type of patient management (admission through the emergency department, hospitalization in an intensive care or resuscitation unit and hospital status) were collected.

Statistical analysis
Quantitative data are expressed as means with standard deviations (SDs) or 95 % confidence intervals (95 % CIs); categorical data as numbers and percentages. Quantitative variables were compared using the Student's t-test; categorical variables using the Chi-squared test.
Multivariable analysis using a mixed model with fixed effects (PROC GLIMMIX), taking into account the variability between hospitals, was performed to study suspected CI-AKI risk factors that required renal replacement therapy or not: gender, age, chronic kidney disease, diabetes, acute heart failure, cardiogenic shock, admission by the emergency unit and hospital status. Co-variables that were considered significant (p < 0.05) in suspected CI-AKI onset during bivariate analysis were included in the multivariable analysis.
Statistical analysis was performed using SAS 9.3® software (SAS Institute Inc. Cary, NC, USA). The significance level was fixed at 0.05 for all analyses.

Economic analysis
The French Health Insurance perspective was used for the economic evaluation of hospitalizations. Each hospitalization is associated to a diagnosis-related group that is linked to a fixed tariff, which varies between public and private hospitals. Extra fees are added for hospitalizations with a visit to the intensive care unit, resuscitation unit, etc. All costs are expressed in 2015 euros. Costs incurred prior to this year were reevaluated using the health service index (4011 E) (National Institute of Statistics and Economic Studies; Institut National de la Statistique et des Etudes Economiques [INSEE]).

Hospitalization characteristics
During 2012-2013, a total of 1,047,329 hospitalizations involving an image-guided cardiovascular procedure requiring ICM administration was recorded in France. Hospitalizations most often included a coronary angiography procedure (45.4 %) and intraluminal dilation of one or several coronary vessels with endoprosthesis insertion (26.3 %). Procedures were not mutually exclusive. The most frequently reported reasons for hospitalization were cardiovascular events (Table 1).
Among all of the hospitalizations, the mean age was 66.1 years, and 68.0 % were male. Concerning risk factors, 25.8 % of hospitalizations involved patients with diabetes, 9.2 % with chronic kidney disease, 3.4 % with acute heart failure and cardiogenic shock occurred in 2,202 (0.2 %) of hospitalizations, among which 539 led to intra-aortic balloon pump use.  Fig. 2). All co-variables were significant in the bivariate analysis.
Proportion of stays leading to patient's death was 1.9 %, with a higher for suspected CI-AKI hospitalizations (21.3 % vs 1.3 for hospitalizations without suspected CI-AKI, p < 0.0001). Proportion of stays leading to patient's death associated with suspected CI-AKI ranged from 8.7 to 49.3 % in hospitalizations involving patients who presented with risk factors (Fig. 3).

Requirement for renal replacement therapy
Among the total population, 0.6 % (n = 6,335) required renal replacement therapy. This was higher among hospitalizations of patients with risk factors (Fig. 1).
Multivariable analysis of risks factors for suspected CI-AKI requiring renal replacement therapy indicated that the occurrence of cardiogenic shock during a hospitalization also gave the highest OR for suspected CI-AKI requiring renal replacement therapy ( Fig. 4). All co-variables were significant in the bivariate analysis.
Hospitalizations during which suspected CI-AKI necessitated renal replacement therapy had a proportion of stays leading to patient's death higher than hospitalizations with suspected CI-AKI and no renal replacement therapy (52.3 vs 13.7 %; p < 0.0001). Proportion of stays leading to patient's during hospitalizations that involved patients with risk factors with suspected CI-AKI requiring renal replacement therapy ranged from 28.4 to 60.9 % (Fig. 3).

Additional costs associated with suspected CI-AKI Total population
Suspected CI-AKI onset during a hospitalization led to a mean additional length of stay of 15 Fig. 6 and Table 2). Overall, CI-AKI led to additional costs reaching a total of €401M over the 2-year period ( Table 2).

Specific populations
Extra lengths of stay associated with suspected CI-AKI onset ranged from 9.5 to 16.4 days and from 20.2 to 32.4 days, depending on comorbidities, for hospitalizations of patients with suspected CI-AKI and suspected CI-AKI requiring renal replacement therapy, respectively (Fig. 5). Additional costs for these hospitalizations varied, respectively, from €6,530 to €11,437 and from €19,830 to €27,875, depending on comorbidities (Fig. 6).
Suspected CI-AKI onset led to an additional length of stay of 10.6 days and was associated with an extra cost of €13,572 for hospitalizations including patients with cardiogenic shock.

Discussion
To the best of our knowledge, this is the largest study to estimate the frequency and health burden of CI-AKI occurring after image-guided cardiovascular interventions. Furthermore, this is the first available data on the general French population. Indeed, the few studies that have reported this complication in France focused on units treating severe cases, such as intensive care units [16,17].
Using a national claims database to obtain extensive real-life data, we identified more than 1 million  hospitalizations involving an image-guided cardiovascular procedure requiring ICM administration over a 2year period. This data sample allowed us to obtain robust results in the absence of important data such as creatinine results.
The frequency of suspected CI-AKI was 3.1 % in our study for the 1,047,329 hospitalizations including an image-guide procedure using ICM, which represents a substantial proportion. The reported frequencies of suspected CI-AKI vary widely in the literature, ranging from 1.5 to 15 % [3][4][5][6][7], depending on the patient population and baseline risk factors. Furthermore, as with any clinical event, the frequency also varies depending on the criteria by which it is defined. Differences in the definitions used make it difficult to compare results of clinical studies (Table 3). CI-AKI is classically defined in the recent literature as a rise in serum creatinine occurring within the first 24 h after contrast exposure and peaking up to 5 days afterward. In most instances, the rise in serum creatinine is expressed either in absolute terms (0.5-1.0 mg/dL; 44.2-88.4 μmol/L) or as a proportional rise in serum creatinine of 25 or 50 % above the baseline value.
Our results come from a large and extensive database including all types of patients, with an extensive selection of image-guided cardiovascular procedures using ICM, thus allowing assessment of patients at all levels of severity.
The frequency of suspected CI-AKI was higher in hospitalizations involving patients with comorbidities  Lengths of stay associated with contrast-induced acute kidney injury, with and without renal replacement therapy. CI-AKI = contrast-induced acute kidney injury ( Fig. 1) and reached 45.4 % in hospitalizations with cardiogenic shock onset. This latter event was the greatest risk factor in our multivariable analysis with an OR of 20.5. Patients with chronic kidney disease or acute heart failure, and those who were admitted via emergency departments, had around twice the risk of developing CI-AKI.
CI-AKI was associated with extremely high inhospital mortality of 21.3 %. The prognosis was even worse if the suspected CI-AKI required renal replacement therapy, with an in-hospital mortality reaching 52.3 %. Even though prior studies have reported mortality of 7 % for suspected CI-AKI without renal replacement therapy and 35 % for suspected CI-AKI requiring renal replacement therapy [9], but figures similar to ours have been reported in large retrospective cohort studies [10,13,21].
Surprisingly, mortality was lower during hospitalizations of patients with comorbidities than during those without (Fig. 3) despite a higher suspected CI-AKI frequency in these populations. One possible explanation for this lower mortality is the close monitoring of patients with important risk factors, leading to rapid detection and management of suspected CI-AKI. Furthermore, as it was not possible to study all of the confounding risks reported by Mehran et al., such as hypotension, contrast volume and haematocrit [22], the mortality results should be interpreted cautiously. Also, we could not analyse whether patients received non-steroidal anti-inflammatory drugs or were hyperglycaemic or volume depleted.
CI-AKI events were accompanied by an extreme increase in mean length of hospitalization (+15.8 days) and cost (+€12,413). The burden of this complication was higher when the suspected CI-AKI required renal replacement therapy, which increased the length of stay by 28.4 days and the cost by €28,627 compared with no CI-AKI. An economic analysis of the direct costs associated with CI-AKI has been conducted recently in the United States [23] this study calculated the mean extra cost of CI-AKI onset to be $10,345 for the hospitalization in question and $11,812 for 1 year. The  Fig. 6 Costs of hospitalizations associated with contrast-induced acute kidney injury, with and without renal replacement therapy. CI-AKI = contrast-induced acute kidney injury incidence data and results were calculated using a systematic search of the literature, and combined with unit costs, also obtained from the literature, using a decisionbased analytical model. In our study, over the 2-year study period, it was estimated that onset of suspected CI-AKI during 1,047,329 hospitalizations led to additional overall expenditure for French Health Insurance of €401M.
Our study had several limitations, the majority of which are intrinsically linked to the PMSI database. This database is used in hospitals for budgetary reasons, and was not developed for epidemiological studies. Consequently, the data do not include results of clinical or biological testing. The absence of creatinine data represents the biggest limitation of our study. Therefore, to identify cases of suspected CI-AKI we selected, with the assistance of a Department of Medical Information doctor, the codes used in practice when this complication appears. Although the coding used to be poor in the early 2000's, the quality has been widely improved in the past few years, showing high sensitivity and positive predictive value [24]. Furthermore, a selection of image-guided procedures usually requiring ICM administration was approved by a committee of experts. As the administration is not mandatory for all procedures selected and could not been verified, we conducted a secondary analysis on a restricted selection of image-guided cardiac procedures with mandatory use of ICM (DDQH and DGQH; Appendix 1). The results of this secondary analysis were similar to the main analysis and allowed us to validate the robustness of our results.
The PMSI database only allows observation of suspected CI-AKI with onset during hospitalization. However, CI-AKI onset can be up to 72 h after ICM administration and patients who are not considered at risk are often not hospitalized for such a long duration after the procedure. It is therefore possible that posthospital events have not been taken into account. Nevertheless, as the mean length of hospitalization in this study was 4.7 days for patients without CI-AKI, we have assumed that the great majority of patients would still have been in hospital at the time of suspected CI-AKI onset, if it occurred.
Moreover, the PMSI database does not allow us to find the exact chronology of onset events during the hospitalization. Therefore, we have assumed that if, during a hospitalization that included a procedure requiring ICM administration, the patient presented with AKI, then the two events were related. However, we cannot state with certainty that all cases of AKI occurring in patients having ICM were attributed to CI-AKI. All cases of AKI, in the wider sense, have been taken into account and considered as suspected CI-AKI in this study. Nevertheless, some could be associated with cardiogenic shock or underlying conditions. This is, however, a limitation for most CI-AKI analyses as the accountability of ICM is often difficult to establish, even in clinical practice, as renal injury is multifactorial [25]. Furthermore, we observed cardiogenic shock in only 0.2 % of all hospitalizations, therefore, even if these events induced an AKI unrelated to ICM, it would not significantly impact our results. However, we note that we could not extract hypotension or use of potentially nephrotoxic agents (e.g. antibiotics or non-steroidal anti-inflammatory drugs) from the database, which are important potential causes of AKI.

Conclusion
This is the largest population-based study that has estimated the frequency and health burden of suspected CI-AKI occurring after image-guided cardiovascular procedures. We have shown that this iatrogenic complication remains of high concern despite prevention efforts and contrast media product improvement. In this study, suspected CI-AKI was associated with particularly high mortality, significantly extended hospital length of stay, and additional costs reaching a total of €200M per year. Even though results should be interpreted cautiously due to limitations regarding absence of creatinine data and the impossibility to identify all risk factors of CI-AKI, we believe they are strengthened by the extremely large sample used. Besides, they are informative on the economic burden of CI-AKI when data are scarce.

Key points
CI-AKI occurred in 3.1 % of image-guided cardiovascular procedures. CI-AKI was associated with increased mortality. CI-AKI was associated with significant hospital costs due to increased length of stay.