Long-Term Risk of Coronary Events after AKI

Abstract

The incidence rate of AKI in hospitalized patients is increasing. However, relatively little attention has been paid to the association of AKI with long-term risk of adverse coronary events. Our study investigated hospitalized patients who recovered from de novo dialysis-requiring AKI between 1999 and 2008 using patient data collected from inpatient claims from Taiwan National Health Insurance. We used Cox regression with time-varying covariates to adjust for subsequent CKD and ESRD after discharge. Results were further validated by analysis of a prospectively constructed database. Among 17,106 acute dialysis patients who were discharged, 4869 patients recovered from dialysis-requiring AKI (AKI recovery group) and were matched with 4869 patients without AKI (non-AKI group). The incidence rates of coronary events were 19.8 and 10.3 per 1000 person-years in the AKI recovery and non-AKI groups, respectively. AKI recovery associated with higher risk of coronary events (hazard ratio [HR], 1.67; 95% confidence interval [95% CI], 1.36 to 2.04) and all-cause mortality (HR, 1.67; 95% CI, 1.57 to 1.79) independent of the effects of subsequent progression to CKD and ESRD. The risk levels of de novo coronary events after hospital discharge were similar in patients with diabetes alone and patients with AKI alone (P=0.23). Our results reveal that AKI with recovery associated with higher long-term risks of coronary events and death in this cohort, suggesting that AKI may identify patients with high risk of future coronary events. Enhanced postdischarge follow-up of renal function of patients who have recovered from temporary dialysis may be warranted.

The incidence rate of AKI in hospitalized patients is increasing¹ and the number of deaths associated with dialysis-requiring AKI has more than doubled.² In hospitalized patients, AKI results in increased in-hospital and posthospitalization resource use.³

Currently, the incidence rate of dialysis-requiring AKI is higher than the rate of ESRD, and its annual growth rate is as high as 10% in the United States.⁴ Along with the advances in critical care medicine and dialysis technologies, the probability of being discharged alive after temporary AKI has been rising among hospitalized patients.⁵ It has been noticed that patients ever suffering AKI have a greater risk for ESRD and higher long-term mortality than individuals with no such episode.^4,6,7 It is unclear whether patients ever having AKI have higher risk for other long-term adverse events.

Diabetes mellitus (DM) is well recognized as a contributing risk factor for coronary events, and advanced CKD has been recently acknowledged as another risk factor.⁸ Rapidly declining kidney function could associate with higher risk for cardiovascular disease (CVD) among patients with or without CKD.⁹ Nevertheless, relatively little effort has been devoted to establishing a clear association between AKI and the long-term incidence of adverse coronary events. In contrast to the currently accepted criteria that include diabetes as a CVD risk factor,¹⁰ no study has ascertained the contributing role of AKI in subsequent CVD incidence.

We hypothesized that hospitalized patients surviving temporary dialysis would have higher probability of developing coronary events in the long term and higher all-cause mortality than their counterpart patients with no such experience. We used data from a large population-based cohort to examine whether dialysis-requiring AKI would have a long-term effect similar to the effect from diabetes in terms of risk for coronary events.

Results

Patient Characteristics

Among 28,497 hospitalized patients with de novo dialysis-requiring AKI (AKI dialysis), 4192 (14.7%) patients died in the index hospitalization. Another 7199 (25.3%) patients died within 30 days after discharge (most of them were discharged under a do not resuscitate status). As a result, around 40% of the patients with de novo dialysis-requiring AKI died during their index hospitalization or within 1 month after discharge. There were 11,985 patients without recovery from dialysis and 5121 patients who survived 30 days after hospital discharge and recovered from dialysis.

After excluding 252 patients without counterparts from the propensity score matching, 4869 patients with AKI recovery and 4869 non-AKI matches (men=56.8%, mean age=63.6±16.7 years in both groups) were selected (Figure 1, Table 1). The propensity model for predicting the status of AKI recovery had a high discrimination power (estimated area under the curve of receiver operating characteristics=0.941) and fitted well with the observed binary data (adjusted generalized R²=0.355) (Supplemental Table 1).

Figure 1.

Flow diagram of selecting study subjects. Creation of the Acute Kidney Injury and Dialysis Study cohort. pt, patient.

Table 1.

Patient characteristics

Items	AKI Recovery Group (n=4869)	Non-AKI Group (n=4869)	P Value
Men	2768 (56.8%)	2768 (56.8%)	1.00
Age (yr)	63.6±16.7	63.6±16.7	0.99
Premorbid risk
Charlson score	3.02±2.02	3.02±2.09	0.57
Congestive heart failure	564 (11.6%)	525 (10.8%)	0.22
Peripheral vascular disease	79 (1.6%)	59 (1.2%)	0.10
Dementia	129 (2.6%)	120 (2.5%)	0.61
Chronic obstructive pulmonary disease	592 (12.2%)	591 (12.1%)	1.00
Rheumatologic disease	75 (1.5%)	113 (2.3%)	0.006
Peptic ulcer	682 (14%)	683 (14%)	1.00
Hemiplegia	70 (1.4%)	64 (1.3%)	0.66
Tumor	351 (7.2%)	364 (7.5%)	0.64
DM	1608 (33%)	1592 (32.7%)	0.75
Moderate or severe liver disease	349 (7.2%)	398 (8.2%)	0.07
CKD	656 (13.5%)	656 (13.5%)	1.00
Advanced CKDa	119 (2.4%)	107 (2.2%)	0.09
Index hospital comorbidity
Cardiovascular	350 (7.2%)	387 (7.9%)	0.17
Respiratory	920 (18.9%)	897 (18.4%)	0.57
Hepatic	101 (2.1%)	122 (2.5%)	0.18
Neurologic	85 (1.7%)	102 (2.1%)	0.24
Hematologic	75 (1.5%)	97 (2%)	0.11
Metabolic	130 (2.7%)	100 (2.1%)	0.05
Operative categories
Cardiothoracic	137 (2.8%)	139 (2.9%)	0.95
Upper gastrointestinal	40 (0.8%)	84 (1.7%)	<0.001
Lower gastrointestinal	108 (2.2%)	117 (2.4%)	0.59
Hepatobiliary	65 (1.3%)	78 (1.6%)	0.31
ICU admission during index hospitalization	2995 (61.5%)	3018 (62%)	0.65

ICU, intensive care unit.

^aPatients with creatinine more than 6 mg/dl with concomitant erythropoiesis-stimulating agents prescription.

Both groups were similar (Table 1). The Charlson score before admission was 3.0±2.1. The proportions of patients with DM, CKD, and advanced CKD were 33%, 13.5%, and 2.3%, respectively. Two confounders had more salient differences. The control group had a higher proportion with rheumatic disease (P=0.006) and was more likely to undertake upper gastrointestinal operations during index hospitalization (P<0.001).

Outcomes

Coronary Events.

Compared with the non-AKI group, the AKI recovery group had larger incidence rates of coronary events, ESRD, and death (Table 2) (all P values<0.001). After a mean follow-up of 3.31±2.71 years, the unadjusted rate of coronary events during follow-up was higher in the AKI recovery group than the non-AKI group (19.8 and 10.3 per 1,000 person-years, respectively). For each individual component within coronary events, the AKI recovery group had a higher incidence rate, particularly in nonfatal myocardial infarction (MI).

Table 2.

Long-term outcomes

Events	AKI Recovery Group (n=4869)	Non-AKI Group (n=4869)	P Value
Coronary events	299 (6.1%)	177 (3.6%)	<0.001
CABG			0.002
Single vessel	3 (0.06%)	3 (0.06%)
Two vessels	4 (0.08%)	1 (0.02%)
Three vessels	20 (0.4%)	6 (0.1%)
PTCA			0.09
Single vessel	69 (1.4%)	53 (1.1%)
Two vessels	24 (0.5%)	15 (0.3%)
Three vessels	2 (0.04%)	2 (0.04%)
Stents	64 (1.3%)	41 (0.8%)	0.03
Nonfatal MI	246 (5.1%)	140 (2.9%)	<0.001
Mortality	2793 (57.4%)	1685 (34.6%)	<0.001
ESRD	1034 (21.2%)	84 (1.7%)	<0.001

CABG, coronary artery bypass graft; PTCA, percutaneous transluminal coronary angioplasty.

For coronary events, the final Cox proportional hazards model with time-varying covariates had good validity (C index=0.74) (Table 3). In contrast to the non-AKI group, AKI recovery patients had a higher long-term risk for coronary events (hazard ratio [HR], 1.67; 95% confidence interval [95% CI], 1.36 to 2.04; P<0.001) independent of effects from age (HR per year, 1.03; 95% CI, 1.03 to 1.04; P<0.001), sex (HR for men, 1.33; 95% CI, 1.10 to 1.60; P=0.003), congestive heart failure (HR, 1.77; 95% CI, 1.39 to 2.25; P<0.001), DM (HR, 1.98; 95% CI, 1.63 to 2.41; P<0.001), and time-varying conditions of CKD (HR, 1.47; 95% CI, 1.11 to 1.96; P<0.001) and ESRD (HR, 2.09; 95% CI, 1.62 to 2.71; P<0.001).

Table 3.

Factors associated with long-term risk of coronary events

Risk Factors	HR	Lower 95% CI	Upper 95% CI	P Value
Age (yr)	1.03	1.03	1.04	<0.001
Men	1.33	1.10	1.60	0.003
Premorbid risk
Congestive heart failure	1.77	1.39	2.25	<0.001
DM	1.98	1.63	2.41	<0.001
AKI recovery versus non-AKI	1.67	1.36	2.04	<0.001
Time-varying factorsa
CKD	1.47	1.11	1.96	0.008
ESRD	2.09	1.62	2.71	<0.001

The final model had a good validity (C index=0.74). 95% CI, 95% confidence interval.

^aPatients with creatinine more than 6 mg/dl with concomitant erythropoiesis-stimulating agents prescription.

For sensitivity analysis, we excluded the diagnostic angiography procedures from coronary events and conducted estimation again. This alternative analysis also found that AKI recovery was significantly associated with higher long-term risk of coronary events (HR, 1.76; P<0.001). We also estimated another alternative model that did not adjust for time-varying CKD and ESRD. This model obtained a larger point estimate for the HR of AKI recovery (2.04; P<0.001).

All-Cause Mortality and Death Risk after Temporary AKI.

After a mean follow-up period of 3.4±2.7 years, the incidence rates of all-cause mortality were 178.6 per 1000 person-years in the AKI recovery group and 96.4 per 1000 person-years in the non-AKI group. The survival model had a good discrimination capability (C index=0.73). The disparity in all-cause mortality between the two groups was statistically significant (HR, 1.67; P<0.001) independent of risks from CKD (HR, 1.77; P<0.001) and ESRD (HR, 2.43; P<0.001) that were subsequently developed after index hospitalization.

We further conducted Cox regression analysis with time-varying covariates to examine the impact of temporary dialysis on long-term mortality after subsequent coronary events. The analysis again identified a significant association (HR, 1.42; P=0.03). This model also showed good validity, with a C index=0.73.

Probability of Freedom from Coronary Events in the Long Term under Different Scenarios of Morbid Conditions in Regard to DM, AKI, CKD, and ESRD.

Our simulation results (Figure 2, Supplemental Table 2, Table 4) indicate that a hospitalized patient without DM, AKI, CKD, and ESRD for the whole period of 10 years after discharge would have a 7% probability (100%−92.68%) of eventually encountering coronary events. In contrast, a hospitalized patient with AKI recovery and no DM would eventually have a probability of ∼12%–17% if there would be no ESRD developed in 10 years and a probability of ∼21%–28% if there would be ESRD. A non-AKI hospitalized patient with DM would eventually have a probability of ∼14%–20% if no ESRD developed in 10 years and a probability of ∼24%–32% if there would be ESRD. Hospitalized patients with both DM and AKI recovery had substantially higher long-term risk of coronary events. Such patients eventually have a probability of ∼22%–31% if there would be no ESRD developed in 10 years and a probability of ∼37%–48% if there would be ESRD.

Figure 2.

Future 10-year probability of freedom from coronary events based on different scenarios of morbid conditions. After obtaining the Cox regression equation, we estimated the hazard function along with time. Based on this hazard function, the probability of freedom from coronary events under different scenarios of morbid conditions with regard to DM, AKI, CKD, and ESRD was conducted in simulation to depict 10-year survival curves.

Table 4.

Scenarios of morbid conditions with regard to DM, AKI, CKD, and ESRD to depict 10-year survival curves of the probability of freedom from coronary events

	Disease/Condition and Time Span
	DM (Whole Duration)	Inpatient AKI (Whole Duration)	Subsequent CKD (1–4 Years after Discharge)	Subsequent ESRD (Years after Discharge)
				1–4	5–10
Reference	−	−	−	−	−
Scenario 1a	−	+	−	−	−
Scenario 1c	−	+	+	−	−
Scenario 1e	−	+	+	−	+
Scenario 2a	+	+	−	−	−
Scenario 2c	+	+	+	−	−
Scenario 2e	+	+	+	−	+
Scenario 3a	+	−	−	−	−
Scenario 3c	+	−	+	−	−
Scenario 3e	+	−	+	−	+

Comparison of Long-Term Risk of Coronary Events among Patient Groups Stratified by Status of DM and AKI.

With adjustment for propensity score, age, sex, premorbid risk, comorbidities, and subsequent time-varying CKD and ESRD after hospital discharge, the long-term risk level of de novo coronary events in patients with AKI alone (i.e., without DM) was statistically indifferent (Figure 3A) (P=0.23) from the risk level in the reference group (patients with DM alone; i.e., without AKI). Relative to the risk level of coronary events in patients with DM alone, coexistence of AKI recovery with DM greatly expanded risk (HR, 1.74; P<0.001).

Figure 3.

Coronary events stratified by DM and AKI. (A) Adjusted HRs for long-term risk of coronary events based on comparison among patient groups stratified by status of DM and AKI that selected patients having DM alone as the reference group and adjusted for propensity score, age, sex, premorbid risk, comorbidities, and subsequent time-varying CKD and ESRD after hospital discharge. 95% CI, 95% confidence interval. *P<0.005; ^#P<0.001. (B) Conditional effect plot showing the relationships of age, DM, and AKI with the incidence probability of coronary events during follow-up with adjustment for propensity score, sex, premorbid risk, and comorbidities based on the fitted multiple logistic regression model.

With adjustment for propensity score, sex, premorbid risk, and comorbidities, comparison of the relationship of age with the incidence probability of coronary events among patient groups stratified by status of DM and AKI (Figure 3B) again shows that coexistence of DM and AKI recovery significantly enlarged the risk. The size of additional harm was larger among older patients in terms of the increased probability of occurring coronary events. Like DM and AKI, older age was itself also a risk factor of coronary events.

Contrast Study on Long-Term Risk of Coronary Events among Patient Groups Validated by a Prospectively Collected Database.

From participants in the National Taiwan University Study Group on Acute Renal Failure (NSARF) 2002–2010, we identified 544 patients with the status of AKI recovery and 11,750 patients without acute dialysis who survived hospital discharge. We used the latter group as controls without AKI. The baseline CKD ratios were 15.3% and 1.7% among the AKI recovery and non-AKI groups, respectively. We used these patients’ data (with a median [interquartile range] of follow-up of 3.52 years [1.06–5.70 years]) to re-examine long-term risk of coronary events among patient groups stratified by status of DM and AKI and found consistent results. In adjusted comparison with a non-AKI and non-DM status that controlled for baseline comorbidities, including CKD status, AKI alone (HR, 3.25; P=0.002) and DM alone (HR, 2.78; P<0.001) were associated with higher long-term risk of coronary events.

Comparison of Long-Term Risk of Coronary Events between the AKI Recovery and Non-AKI Groups under a Framework of Subgroup Analysis.

To investigate consistency in long-term risk of AKI recovery among different patient groups, we conducted subgroup analysis with respect to baseline comorbidity that further adjusted for age, sex, and propensity score. We found that inpatients complicated by AKI recovery were consistently associated with higher long-term risk of coronary events across broad varieties of patient groups (Figure 4). The HRs were near 2 for all the patient groups.

Figure 4.

Risk of coronary events associated with AKI by participant characteristics. Adjusted HRs for long-term risk of coronary events based on comparison between the AKI dialysis recovery and non-AKI groups and subgroup analysis with respect to baseline comorbidity that further adjusted for age, sex, and propensity score. 95% CI, 95% confidence interval; COPD, chronic obstructive pulmonary disease; ICU, intensive care unit.

Unmatched Cases

We failed to find satisfactory counterparts from non-AKI patients for 252 (4.9%) patients with temporary dialysis by propensity score matching. These patients tended to have worse health than those patients with matches in the non-AKI group. Their incidence rates of coronary events and all-cause mortality were 42.3 and 302.2 per 1000 person-years, respectively.

Discussion

Our study is the first to find that patients with temporary dialysis have higher long-term CVD risk and subsequent mortality than those patients with neither AKI nor dialysis, and the hazard of its impact on long-term CVD risk is close to the hazard from DM (Figures 2 and 3, Table 3, Supplemental Table 2). We found similar results in analysis using Taiwan National Health Insurance (NHI) data (retrospectively collected) and contrast analysis using NSARF data (prospectively collected). These findings shed light on the importance of adequate care for patients who have recovered from dialysis-requiring AKI.

As highlighted by the collaborative campaign (World Kidney Day 2013) promoted by the International Society of Nephrology and the International Federation of Kidney Foundations, AKI is now a growing global health alert.^11,12 Threats from the growing AKI incidence rate are reflected in an expanding body of vulnerable populations with higher risks from ESRD and all-cause mortality as well as increasing posthospitalization resource use.¹¹ Additional evidence includes higher risk of long-term coronary events and mortality among AKI patients ever experiencing coronary angiography with contrast injury.¹³

Previous research has indicated that presence of underlying CKD partially accounts for the outcomes of AKI but not completely.¹⁴ Consistent with a previous report,⁶ our findings also highlight severe AKI as a cause of long-term ESRD (Table 2). This study further advances knowledge on threats from dialysis-requiring AKI, showing that even temporary dialysis leads to higher long-term risk of coronary events and mortality, independent of subsequent progression to CKD and ESRD. Interestingly, CKD and ESRD that were developed subsequently after discharge could also significantly aggravate long-term risk of coronary events (Figure 2, Supplemental Table 2, Table 4).

AKI Contributes to Cardiac Dysfunction

Our findings tend to suggest that AKI may trigger a cascade of perturbations that is never completely resolved. Several mechanisms have been proposed to explain cardiac dysfunction after AKI, such as fluid overload contributing to pulmonary edema, endothelial dysfunction, hypercoagulation, and myocardial depression activity presented during ultrafiltration.^15,16

Cytokine-mediated cardiac dysfunction secondary to AKI has been shown in animal models. AKI induces cardiac cell apoptosis and production of IL-1 and TNF, a product found in the renal ischemia reperfusion model of rats.¹⁷ Renal ischemia reperfusion, rather than bilateral nephrectomy, results in myocardial apoptosis. Endothelium dysfunction related to elevated homocysteine has also been observed after kidney ischemic reperfusion injury.¹⁸ Unregulated adhesion of receptor intercellular adhesion molecule-1, myeloperoxidase activity, and presence of apoptosis in the heart may result in inflammation after kidney ischemia.¹⁹ Because an episode of AKI can contribute to a faster subsequent rate of decline in kidney function, it may help identify patients at heightened long-term risks for MI and death.⁹

Although AKI is gradually being recognized as a contributing factor to late stage CKD, it is possible that endothelial phenotypic transition may have already taken place during AKI,²⁰ paving the way for progressive endothelial deterioration. Recently, fibroblast growth factor-23,²¹ a novel regulator of mineral metabolism, was shown to be markedly elevated in AKI and associated with cardiovascular events²² and poor long-term outcomes.²³ Likewise, the expression of Klotho, a cofactor that promotes the development of human artery calcification, has been seen to be downregulated during an episode of AKI.²⁴ Deteriorated microcirculatory dysfunction, in terms of elevated angiopoietin-2, has been identified during AKI.²⁵ Renal ischemia results in distant organ effects, and the alterations observed in the heart may have a significant impact on morbidity and mortality.

The role of AKI as a cause of mortality is evidenced by the fact that patients experiencing temporary dialysis and recovery from AKI have higher risk of coronary events than those patients without experience of severe AKI.²⁶ A comprehensive treatment strategy during hospitalization or even after discharge for AKI patients is necessary to reduce the risk of ensuing coronary events and subsequent mortality.

AKI Harmful Effect on Coronary Events Is Independent of and Comparable with the Harmful Effect from DM

Our findings indicate that the hazard of AKI harmful effect on long-term CVD risk is close to the hazard from DM after adjustment for perioperative and baseline confounding factors. As reported in the US National Cholesterol Education Program Adult Treatment Panel III, diabetes is a coronary heart disease risk equivalent that contributes to a 10-year risk of coronary death or MI.²⁷ Our results show that AKI harmful effects on coronary events are independent of and comparable with the harmful effects from DM. This finding highlights the potential value of preventing CVD by targeting AKI patients in addition to those patients with DM, and even dialysis-requiring AKI with subsequent recovery should be deemed as a risk category for CVD.

Treating and Preventing AKI Should Have High Priority

Our findings have important implications for postdischarge care among patients successfully treated with acute temporary dialysis as well as public health policy. Follow-up care after discharge among patients experiencing dialysis-requiring AKI and subsequent recovery are necessary to enhance secondary prevention. Taking into account the very high amount of expenditures for treating AKI and coronary events and the significant harm on life from the two conditions, AKI prevention should also receive high priority in the health public agenda.

Study Strengths and Limitations

Baseline GFR was not available from the International Classification of Diseases (ICD) codes, which added difficulty in identifying the severity of CKD. A meta-analysis suggested that CKD codes are insensitive but reasonably specific.²⁸ In Taiwan, the rate of misdiagnosing CKD may have been modest since 2004, because a CKD prevention program has made nearly all patients with CKD diagnosis receive specialist consultation.^29–31 In light of this information, our validation shows fair reliability for CKD identification in this study. Identifying AKI by the dialysis procedure also introduced limitations. Our validation shows that identifying AKI based on NHI data had a 98.5% positive predictive value and a 74.0% negative predictive value. Rates of dialysis withdrawal, preferences for withdrawal, and engagement in advance care planning also vary by age and race or ethnicity.³² Because the administrative data have varying sensitivity for diagnosis, we cannot completely exclude misclassification at baseline and residual confounding effects, although we have rigorously controlled for key characteristics.

This study has some strengths. The NHI and the NSARF databases have large sample sizes and long follow-up times. Reliability of ICD-9-CM codes for AKI and CKD was validated.^29–31 Investigation of AKI recovery effect was conducted across broad varieties of patient groups. Effects from time-varying covariates were inspected. The study used a propensity score method to reduce imbalance in key characteristics between AKI recovery and non-AKI groups and failed to find counterparts for only 1.5% of study patients. AKI recovery patients without matches tended to have higher rates of coronary events and mortality, suggesting that exclusion might underestimate the size of AKI harm, but it is unlikely that this information biased the study.

AKI requiring temporary dialysis increases long-term risk from coronary events and is related to all-cause mortality and death risk after coronary events independent of effects from baseline and subsequent CKD or ESRD. The harmful effect of AKI with recovery on coronary events is independent of and comparable with the effect from DM.

Concise Methods

Data Sources

The NHI is a nationwide insurance program that covers outpatient visits, hospital admissions, prescriptions, intervention procedures, and disease profiles for over 99% of the population in Taiwan (23.12 million people in 2009). The NHI database is one of the largest and most comprehensive databases in the world, and it has been offering research data to various studies on prescription use, diagnoses, and hospitalizations.³³

The National Health Research Institutes used original data from the NHI database to construct a longitudinal database of patients who were ever admitted between 1999 and 2008. This cohort included 2,619,534 hospitalized patients, representing 10% of all NHI enrollees. This sampling fraction (a 3.4:1 ratio) was based on a regulation that limits the maximal amount of NHI data that can be extracted for research purposes.

Research Variables

The demographic and clinical characteristics of our participants at their index hospitalization were recorded. The parameters included age, sex, year of admission, hospital characteristics, comorbid conditions, Charlson comorbidity index, organ dysfunction developed during the index hospitalization, categories of major operations, and intensive care unit admission. Preadmission comorbidity was determined as a specific entity recorded in at least one hospital admission or at least three outpatient department visits during the year before the index hospitalization, which was a relatively strict criterion that was well validated with good predicting power.^34,35

Identification of Cases and Controls

The study group consisted of patients ages≥18 years old who had a first-time diagnosis of dialysis-requiring AKI (identified by the procedure codes). The dialysis certificate data in Taiwan are highly reliable, because they are used for insurance payment.³³

To identify the patients’ baseline conditions, we used the 1-year period immediately before the index hospitalization to identify preadmission comorbidities, AKI, and dialysis. Patients with preadmission AKI or ESRD, MI, or cardiac revascularization were excluded as well as those patients who underwent kidney transplantation, vascular accesses creation, or peritoneal dialysis catheter implantation. Figure 1 depicts our procedures for selecting study subjects. Our study only enrolled patients who survived at least 30 days after discharge from the index hospitalization and had no rehospitalization or redialysis during the 30-day period.

For comparison, our study adopted a 1:1 ratio to construct a control group, in which each patient did not have AKI, dialysis, and coronary events before and during an index hospitalization (non-AKI group). The control group was matched with the exposure group (AKI recovery group) on the basis of age, sex, same calendar year of index hospitalization, and comorbidities before and during index hospitalization. We used the propensity score method for matching. To avoid interference effect, we excluded patients with at least one hospital stay (meeting the criteria of selection for the exposure group) from the pool of patients who could offer study cases in the control group.

Outcome Measures

The primary outcome was incidence of coronary events that included nonfatal MI, coronary artery bypass graft, and coronary angiography (Supplemental Table 2). The ICD-9 code of MI at hospitalization had high accuracy, which was indicated by previous research.^36,37 The records of coronary artery bypass graft and angiography were reliable, because they were constructed on the basis of NHI procedure codes that were tied to the NHI reimbursement system with auditing. The secondary outcome was all-cause mortality after hospital discharge.

Data Used for Validation and Contrast: NSARF Data of 2002–2010

We used NSARF,^38–42 a multicenter, prospectively constructed database, to validate CKD diagnosis recorded in NHI data. CKD was defined as an estimated GFR≤45 ml/min per 1.73 m².⁴¹ We also used the database to investigate how baseline kidney function was associated with probability of developing coronary events in different patient groups that were categorized on the basis of DM and AKI status. AKI was classified according to the risk, injury, failure, loss, and end stage criteria.⁴² We included NSARF 2002–2010 enrollees with complete information on serum creatinine (measured by using a standardized protocol) and followed them until September of 2012.

Statistical Analyses

Continuous variables are described as mean ± SD; discrete variables are presented as counts or percentages. We used R software, version 2.8.1 (Free Software Foundation, Inc., Boston, MA) and Stata software, version 12 (StataCorp, College Station, TX). A two-sided P value<0.05 was considered statistically significant.

To reduce bias in assessing the detrimental effects of AKI, we used a propensity score method to construct a comparable non-AKI group with balanced characteristics. Variables reflecting comorbid conditions reported in Table 1 were added to a nonparsimonious multivariable logistic regression model (Supplemental Table 1) to predict the probability of being in the AKI recovery group. The predicted probability derived from the estimated equation was used as the propensity score for each individual.

Taking into account the higher risk of developing CKD/ESRD after AKI incidence and the strong correlations of CKD/ESRD with coronary events and mortality,⁴³ we used a Cox proportional hazards model with time-varying covariates to account for the impacts of ESRD and CKD developed after discharge on the risks of coronary events and mortality. Tracking incidence of coronary events was censored at death or the end of 2008.

After obtaining the Cox regression equation, we estimated the hazard function along with the time for a reference patient who was non-AKI in the index hospitalization, did not have DM, AKI, CKD, or ESRD for 10 years after discharge, and had the mean values of all the other covariates for our study patients. Based on this hazard function, we further conducted simulation to depict 10-year survival curves of the probability of freedom from coronary events under different scenarios of morbid conditions in regard to DM, AKI, CKD, and ESRD. Specifically, we stratified patients by status of DM and AKI at the baseline time and the time points of developing CKD and ESRD after discharge (Supplemental Table 2).

We also conducted additional analyses as follows: adjusted comparison of long-term risk for coronary events among patient groups stratified by status of DM and AKI; adjusted comparison of the relationship of age with the incidence probability of coronary events among patient groups stratified by status of DM and AKI; and adjusted comparison of long-term risk for coronary events between the AKI dialysis recovery and non-AKI groups under a framework of subgroup analysis in regard to comorbidities.

Ethical Considerations

Because patients were anonymous in our study, no informed consent was required. Because the identification numbers of all individuals in the databases were encrypted to protect the privacy of the individuals, this study was exempt from a full ethical review by the institutional review board of National Taiwan University Hospital (201212021RINC).

Disclosures

None.