The Association of Acute Kidney Injury in the Critically Ill and Postdischarge Outcomes: A Cohort Study* Clare M. Horkan, MB, BCh1; Steven W. Purtle, MD2; Mallika L. Mendu, MD, MBA3; Takuhiro Moromizato, MD4; Fiona K. Gibbons, MD5; Kenneth B. Christopher, MD6

Objective: Hospital readmissions contribute significantly to the cost of inpatient care and are targeted as a marker for quality of care. Little is known about risk factors associated with hospital readmission in survivors of critical illness. We hypothesized that acute kidney injury in patients who survived critical care would be associated with increased risk of 30-day postdischarge hospital readmission, postdischarge mortality, and progression to end-stage renal disease. Design: Two center observational cohort study. Setting: Medical and surgical ICUs at the Brigham and Women’s Hospital and the Massachusetts General Hospital in Boston, Massachusetts. Patients: We studied 62,096 patients, 18 years old and older, who received critical care between 1997 and 2012 and survived hospitalization. Interventions: None Measurements and Main Results: All data was obtained from the Research Patient Data Registry at Partners HealthCare. The exposure of interest was acute kidney injury defined as meeting Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease Risk, Injury or Failure criteria occurring 3 days prior to 7 days after critical care initiation. The primary outcome was hospital readmission in the *See also p. 490. 1 Department of Medicine, Brigham and Women’s Hospital, Boston, MA. 2 Division of Pulmonary Sciences and Critical Care Medicine, University of Colorado, Boulder, CO. 3 Renal Division, Brigham and Women’s Hospital, Boston, MA. 4 Department of Medicine, Hokubu Prefectural Hospital, Okinawa, Japan. 5 Pulmonary Division, Massachusetts General Hospital, Boston, MA. 6 The Nathan E. Hellman Memorial Laboratory, Renal Division, Brigham and Women’s Hospital, Boston, MA. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/ccmjournal). The authors have disclosed that they do not have any potential conflicts of interest. For information regarding this article, E-mail: [email protected] Copyright © 2015 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins DOI: 10.1097/CCM.0000000000000706

354

www.ccmjournal.org

30 days following hospital discharge. The secondary outcome was mortality in the 30 days following hospital discharge. Adjusted odds ratios were estimated by multivariable logistic regression models with inclusion of covariate terms thought to plausibly interact with both acute kidney injury and readmission status. Adjustment included age, race (white vs nonwhite), gender, Deyo-Charlson Index, patient type (medical vs surgical) and sepsis. Additionally, long-term progression to End Stage Renal Disease in patients with acute kidney injury was analyzed with a risk-adjusted Cox proportional hazards regression model. The absolute risk of 30-day readmission was 12.3%, 19.0%, 21.2%, and 21.1% in patients with No Acute Kidney Injury, Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease class Risk, Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease class Injury, and Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease class Failure, respectively. In patients who received critical care and survived hospitalization, acute kidney injury was a robust predictor of hospital readmission and post-discharge mortality and remained so following multivariable adjustment. The odds of 30-day post-discharge hospital readmission in patients with Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease class Risk, Injury, or Failure fully adjusted were 1.44 (95% CI, 1.25–1.66), 1.98 (95% CI, 1.66–2.36), and 1.55 (95% CI, 1.26–1.91) respectively, relative to patients without acute kidney injury. Further, the odds of 30-day post-discharge mortality in patients with Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease class Risk, Injury, or Failure fully adjusted per our primary analysis were 1.39 (95% CI, 1.28–1.51), 1.46 (95% CI, 1.30–1.64), and 1.42 (95% CI, 1.26–1.61) respectively, relative to patients without acute kidney injury. The addition of the propensity score to the multivariable model did not change the point estimates significantly. Finally, taking into account age, gender, race, Deyo-Charlson Index, and patient type, we observed a relationship between acute kidney injury and development of end-stage renal disease. Patients with Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease class Risk, Injury, Failure experienced a significantly higher risk of end-stage renal disease during follow-up than patients without acute kidney injury (hazard ratio, 2.03; 95% CI, 1.56–2.65; hazard ratio, 3.99; 95% CI, 3.04–5.23; hazard ratio, 10.40; 95% CI, 8.54–12.69, respectively). February 2015 • Volume 43 • Number 2

Clinical Investigations Conclusions: Patients who suffer acute kidney injury are among a highrisk group of ICU survivors for adverse outcomes. In patients treated with critical care who survive hospitalization, acute kidney injury is a robust predictor of subsequent unplanned hospital readmission. In critical illness survivors, acute kidney injury is also associated with the odds of 30-day postdischarge mortality and the risk of subsequent end-stage renal disease. (Crit Care Med 2015; 43:354–364) Key Words: acute kidney injury; critical care; hospital readmission; outcomes; risk, injury, failure, loss of kidney function, and endstage kidney disease

MATERIALS AND METHODS

A

Data Sources Data on critically ill patients were collected prospectively in a central computerized registry called the Research Patient Data Registry (RPDR) (29) that serves as a central clinical data warehouse for all inpatient and outpatient records at Partners HealthCare sites including BWH and MGH. The RDPR has been used for other clinical research studies and mortality and coding data from the RPDR has been validated in at least one prior study (30). Approval for the study was granted by the Partners Human Research Committee Institutional Review Board. Between 1997 and 2012, there were 76,597 unique patients, 18 years old or older, assigned the CPT code 99291 (critical care, first 30–74 min) (30) who had baseline creatinine measured prior to or at hospital admission and had a diagnosisrelated group (DRG) assigned following hospitalization. A total of 14,501 patients were excluded from the analysis: 491 patients with ESRD prior to hospital admission per the United States Renal Data System (USRDS) records (31), 10,297 patients who died in the hospital, and 3,713 patients who had readmissions determined to be planned. 62,096 patients constituted the study cohort. In the cohort, we also identified 923 patients who died within 30 days of discharge, who were not admitted to the hospital, and who were excluded in the primary analysis.

cute kidney injury (AKI) is a sequela of intensive care and is associated with high morbidity and mortality (1, 2). The prevalence of AKI ranges from 35% to 70% of critically ill patients and approximately 10% of AKI cases will need renal replacement therapy (3). AKI episodes may hasten progression of renal disease (4–8). Mortality in patients with AKI is associated with the severity of renal injury (9, 10). AKI in the critically ill is commonly observed as part of the multiple organ dysfunction syndrome, which is associated with adverse outcomes (11). Hospital readmissions contribute significantly to the cost of inpatient care and are targeted as a marker for quality of care (12). Approximately 37% of Medicare spending and 31% of total health care expenditures can be attributed to the costs of hospitalization (13, 14). Hospital readmissions are a significant driver of healthcare costs. Hospital readmission within 30 days of discharge occurs in 18% of Medicare beneficiaries and results in an annual cost of $15 billion (15). In a study of adult patients 65 years old or older, covariates associated with readmission included older age, male gender, African-American race, discharge to long-term care, admission to the medical service and Medicare-only insurance (16). Hospital readmissions are used as a marker for quality of care, yet little is known regarding risk factors associated with readmission in critically illness survivors (17–19). AKI in patients with heart failure or following cardiac surgery is shown to be associated with readmission (20, 21). With heightening societal and political interest in cost-effective healthcare delivery, AKI at the time of critical care may be a marker for critical illness survivors who are at high risk for subsequent adverse events. Although quality of life and longterm survival have been explored in critically ill patients with AKI (9, 11, 22–26), 30-day postdischarge hospital readmission in these patients is not known. Given the heightened mortality in critically ill patients with AKI (27), we sought to determine whether critically ill patients who develop AKI have a increased risk of 30-day postdischarge hospital readmission, increased 30-day mortality following hospital discharge and accentuated progression to end-stage renal disease (ESRD). We hypothesized that AKI in patients who survived critical care would be associated with increased risk of 30-day postdischarge hospital readmission and of postdischarge mortality.

Critical Care Medicine

Source Population We abstracted laboratory and administrative data from the electronic medical records of two teaching hospitals in Boston, MA: Brigham and Women’s Hospital (BWH), with 793 beds and Massachusetts General Hospital (MGH) with 902 beds. Each hospital has approximately 45,000 hospital admissions per year. The 2010 30-day readmission rates for MGH in surgical and medical patients were 13.9% and 17.3%, respectively; for BWH, 30-day readmission rates in surgical and medical patients were 15.7% and 18.8%, respectively (28).

Exposure of Interest and Comorbidities The primary exposure was AKI. AKI was defined as RIFLE class Risk, Injury, or Failure (32) occurring between 3 days prior to critical care initiation and 7 days after critical care initiation (33). We applied the serum creatinine criteria only to determine the maximum RIFLE class (33, 34). We classified patients according to the maximum RIFLE class (class Risk, class Injury, or class Failure) (3) defined as a fold change in serum creatinine from preadmission serum creatinine (34). RIFLE class was defined as Risk (fold change ≥ 1.5), Injury (fold change ≥ 2.0), or Failure (fold change ≥ 3.0) (3). Patients with baseline serum creatinine of greater than 4.0 mg/dL who had an absolute change in serum creatinine greater than 0.5 mg/dL were considered to have RIFLE class Failure (32). Preadmission creatinine was obtained in all patients from 7 to 365 days prior to hospital admission with the creatinine closet to hospital admission recorded. If baseline creatinine prior to hospital admission was not available, we used the first creatinine drawn at hospital admission. www.ccmjournal.org

355

Horkan et al

Table 1.

Patient Characteristics of the Study Population Risk, Injury, Failure, Loss of Kidney Function, and End-Stage Kidney Disease Class

Characteristic

No Acute Kidney Injury

Risk

Injury

53,907

4,376

2,049

56.9 (18.2)

61.1 (15.7)

60.2 (15.7)

 Female

21,547 (40)

1,932 (44)

905 (44)

 Male

32,360 (60)

2,444 (56)

1,144 (56)

 White

42,088 (78)

3,464 (79)

1,610 (79)

 Non-white

11,819 (22)

912 (21)

439 (21)

 Medical

25,976 (48)

1,879 (43)

934 (46)

 Surgical

27,931 (52)

2,497 (57)

1,115 (54)

 0–1

24,639 (46)

1,213 (28)

423 (21)

 2–3

18,006 (33)

1,509 (34)

696 (34)

 4–6

9,447 (18)

1,304 (30)

707 (35)

  ≥7

1,815 (3)

350 (8)

223 (11)

Sepsis, n (%)

4,238 (8)

845 (19)

615 (30)

 Stage 0–2

40,350 (77)

3,102 (74)

 Stage 3

10,046 (19)

895 (21)

483 (24)

 Stage 4

1,735 (3)

196 (4)

116 (6)

 Stage 5

592 (1)

19 (1)

10 (1)

 0

21,552 (40)

805 (18)

173 (8)

 1

18,933 (35)

1,390 (32)

456 (22)

 2

9,036 (17)

1,175 (27)

615 (30)

 3

3,203 (6)

658 (15)

482 (24)

  ≥4

1,183 (2)

348 (8)

323 (16)

30-d postdischarge hospital readmission rate, %b

12.3

19.0

21.2

30-d postdischarge mortality rate, %

3.0

5.5

8.0

10.2 (12.5)

16.5 (18.5)

No. of cases Age, mean (sd) Sex, n (%)

Race, n (%)

Patient type, n (%)

Deyo–Charlson Index, n (%)

Chronic kidney disease stage, n (%) 1,364 (69)

Acute organ failure, n (%)

Length of stay, mean (sd)

20.3 (22.1)

p values determined by chi-square unless designated by (a) then p value determined by Kruskal–Wallis. b Patients who died within 30 days of hospital discharge who were not readmitted were excluded in the 30-day postdischarge hospital readmission rate determination.

Critical care initiation was defined as the first day of CPT code 99291 (critical care, first 30–74 min) assignment, an approach validated in our administrative database (30). Patient type was defined as medical or surgical and incorporates the DRG methodology, devised by Centers for Medicare 356

www.ccmjournal.org

& Medicaid Services (35). We used the Deyo–Charlson Index to assess the burden of chronic illness (36). We used the validated International Classification of Diseases, 9th Edition (ICD9) coding algorithms developed by Quan et al (37) to derive a comorbidity score for each patient. February 2015 • Volume 43 • Number 2

Clinical Investigations

Failure

Total

1,764

62,096

56.6 (15.7)

57.3 (17.9)

p

< 0.001a < 0.001

688 (39)

25,072 (40)

1,076 (61)

37,024 (60) < 0.001

1,264 (72)

48,426 (78)

500 (28)

13,670 (22) < 0.001

956 (54)

29,745 (48)

808 (46)

32,351 (52) < 0.001

230 (13)

26, 505 (43)

580 (33)

20,791 (33)

716 (41)

12,174 (20)

238 (13)

2,626 (4)

541 (31)

6, 239 (10)

< 0.001 < 0.001

895 (52)

45,711 (75)

211 (12)

11,635 (19)

170 (10)

2,217 (4)

446 (26)

1,067 (2) < 0.001

136 (8)

22,666 (37)

359 (20)

21,138 (34)

474 (27)

11,300 (18)

376 (21)

4,719 (8)

419 (24)

2,273 (4)

21.1

13.3

< 0.001a

6.4

3.4

< 0.001a

20.2 (21.8)

11.2 (14.0)

< 0.001a

Sepsis was defined by the presence of any of the following ICD9-CM codes: 038.0–038.9, 020.0, 790.7, 117.9, 112.5, or 112.81, 3 days prior to critical care initiation to 7 days after critical care initiation (38). Number of organs with failure was adapted from Martin et al (38) and defined by a combination of ICD-9-CM Critical Care Medicine

and CPT codes relating to acute organ dysfunction assigned from 3 days prior to critical care initiation to 30 days after critical care initiation (39, 40). Patients with ESRD prior to hospital admission were identified by submitting the administrative data to the USRDS (31). Chronic kidney disease stage was determined by the www.ccmjournal.org

357

Horkan et al

Modification of Diet in Renal Disease equation from the baseline creatinine, age, gender, and race of cohort patients(41). Endpoints The primary outcome was 30-day postdischarge hospital readmission (42) determined from RPDR hospital admission data. To determine patients with unplanned hospital readmissions, the first or initial instance of critical care was identified. The first hospitalization associated with critical care was identified as the index critical care exposure, and a 30-day unplanned readmission was defined as a subsequent or unscheduled admission to BWH or MGH within 30 days of discharge following the hospitalization associated with the index critical care exposure (43). To evaluate unplanned readmissions, we excluded readmissions with DRG codes that are commonly associated with planned readmissions: DRG 001, 075, 105, 109, 110, 113, 120, 209, 263, 315, 336, 410, 462, 478, 515, 517, 518, 527, and 533 (44) in addition to DRGs for transplantation: 103, 302, 480, 481 and 495; procedures related to pregnancy: 364, 370, 371, 372, 374, 381; dental procedures: 185, 187; and psychiatric issues: 425, 426, 428, 430, 433, 434, 435, 521, and 523. If a patient had one or more admissions (for any reason) within 30 days after discharge from the hospital admission with the index critical care exposure, only one was counted as a readmission. Information on vital status for the study cohort was obtained from the Social Security Administration Death Master File, which we have validated for in-hospital and out of hospital mortality in our administrative database (30). Hundred percent of the cohort had vital status present at 365 days following critical care initiation. The censoring date was February 25, 2012. Progression to ESRD was determined by submitting the administrative data to the USRDS (31) to extract the first ESRD service date from the Medical Evidence Report (form CMS-2728). Power Calculations and Statistical Analysis We assumed that 30-day postdischarge hospital readmission rate would increase 5% in patients with RIFLE Risk, Injury, or Failure class when compared with those without AKI. With an α error level of 5% and a power of 80%, the minimum sample size thus required for our primary end point 30-day postdischarge hospital readmission is 1,890 total patients. Categorical covariates were described by frequency distribution, and compared across RIFLE groups using contingency tables and chi-square testing. Continuous covariates were examined graphically and in terms of summary statistics and compared across exposure groups using one-way ANOVA. The outcomes considered were 30-day postdischarge hospital readmission (42) and 30-day mortality following hospital discharge. Unadjusted associations between RIFLE groups and outcomes were estimated by Mantel Haenszel methods and by bivariable logistic regression analysis. Adjusted odds ratios were estimated by multivariable logistic regression models with inclusion of covariate terms thought to interact with both AKI and postdischarge hospital readmission plausibly. For the primary model (postdischarge hospital readmission), 358

www.ccmjournal.org

specification of each continuous covariate (as a linear vs categorical term) was adjudicated by the empiric association with the primary outcome using Akaike’s Information Criterion; overall model fit was assessed using the Hosmer–Lemeshow (HL) test. Analyses based on fully adjusted models were performed to evaluate the AKI -30-day postdischarge hospital readmission association, and p for interaction was determined to explore for any evidence of effect modification. We individually tested for effect modification by gender, race, and chronic kidney disease by adding an interaction term to the multivariate models. Unadjusted event rates were calculated with the use of the Kaplan–Meier methods and compared with the use of the log-rank test. Sensitivity analyses were performed for patients with and without advanced chronic kidney disease. Long-term progression to ESRD in patients with AKI was analyzed with a risk-adjusted Cox proportional hazards regression model. To evaluate the subsequent risk of the need for chronic hemodialysis in our cohort additionally, we analyzed cohort data with the primary exposure as AKI and the outcome of interest as ESRD per the USRDS (31) at 1 year following discharge. To reduce potential bias from the nonrandomized assignment of the diagnosis of AKI, we constructed propensity scores for the allocation of AKI (45, 46) and used these in the primary and secondary analyses. Using logistic regression, propensity scores were calculated for each cohort subject to estimate the probability for the presence or absence of a diagnosis of AKI. We based covariate selection for the propensity score development on a previous study of AKI in our administrative database (33) including age, sex, race (white vs non-white), patient type (surgical vs medical), Deyo–Charlson Index, and sepsis covariates. The propensity score (as a continuous variable) was entered into the final multivariable logistic regression and Cox models. In addition, two smaller cohorts were obtained where a treated subject (with AKI) was matched to a control subject (without AKI) on the basis of the propensity score. We used Mahalanobis metric matching within calipers defined by the propensity score to match the smaller cohorts (47) using the publicly available matching algorithm “psmatch2” (48). All values of p presented are two-tailed; values less than 0.05 were considered nominally significant. All analyses are performed using STATA 12.0MP (College Station, TX).

RESULTS Table 1 shows characteristics of the study population. Most patients were men (60%) and white (78%). A total of 48% had medically related DRGs. The mean age at hospital admission was 57 years (sd = 18) years. Patient characteristics of the study cohort were stratified according to RIFLE class (Table 1). Factors that significantly differed between stratified groups included age, sex, race, Deyo–Charlson Index, acute organ failure, chronic kidney disease, and sepsis. Table 2 indicates that RIFLE class, age, sex, race (white vs non-white), patient type (surgical vs medical), Deyo–Charlson Index, and sepsis are significant predictors of 30day postdischarge hospital readmission. Thirty-day postdischarge hospital readmission rate was 13%, and the 30-day postdischarge mortality rate was 3%. The median follow-up time for ESRD was 5.9 years. There were 604 patients subsequently diagnosed with February 2015 • Volume 43 • Number 2

Clinical Investigations

Table 2. Multivariable-Adjusted Associations Between Covariates and 30-Day Postdischarge Hospital Readmission (n = 61,173) Characteristic

Odds Ratio

95% CI

p

Age (per 1 yr)

1.00

1.00–1.00

< 0.001

 Female

1.00

Reference

 Male

0.91

0.87–0.96

 White

1.00

Reference

 Non-white

0.91

0.86–0.97

 Medical

1.00

Reference

 Surgical

1.17

1.11–1.22

 0–1

1.00

Reference

 2–3

1.42

1.34–1.50

< 0.001

 4–6

1.77

1.66–1.90

< 0.001

  ≥7

2.33

2.09–2.59

< 0.001

 Sepsis

1.61

1.50–1.72

< 0.001

Sex < 0.001

Race 0.003

Patient type < 0.001

Deyo–Charlson Index

Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease class  No acute kidney injury

1.00

Reference

 Risk

1.39

1.28–1.51

< 0.001

 Injury

1.46

1.30–1.64

< 0.001

 Failure

1.42

1.26–1.61

< 0.001

Adjusted odds ratios were estimated by a multivariable logistic regression model with inclusion of covariate terms thought to associate with acute kidney injury and 30-day postdischarge hospital readmission plausibly. Estimates for each variable are adjusted for all other variables in the table. Patients who died within 30 days of hospital discharge who were not readmitted were excluded.

ESRD with 101,643 person-years of follow-up in critical illness survivors, yielding an ESRD rate of 5.9 per 1,000 person-years. Primary Outcome AKI was a strong predictor of 30-day postdischarge hospital readmission (Fig. 1 and Table 3). The odds of 30-day postdischarge hospital readmission in patients with RIFLE class Risk, Injury, and Failure were 1.7-, 1.9-, and 1.9-fold that of patients without AKI. AKI remained a significant predictor of the odds of 30-day postdischarge hospital readmission after adjustment for age, sex, race, Deyo–Charlson Index, patient type, and sepsis. The adjusted odds of 30-day postdischarge hospital readmission in patients with RIFLE class Risk, Injury, and Failure were 1.4-, 1.5-, and 1.4-fold that of patients without AKI (Table 3). The adjusted model showed good calibration (HL chi-square 8.98; p = 0.11). The results did not materially differ with additional adjustment for propensity score (Table 3) or hospital site covariate (data not shown). We repeated the primary analyses by using a selected propensity score–matched cohort (n = 15,844). Thirty-day postdischarge Critical Care Medicine

hospital readmission rates in the matched cohort were 20.0% in patients with AKI and 15.7% in patients without AKI. There were no significant differences between the groups (AKI: n = 7,922; without AKI: n = 7,922) with respect to all of the variables listed in Table 2 (all p > 0.20, not shown). Again, RIFLE class was a significant predictor 30-day postdischarge hospital readmission. The odds of 30-day postdischarge hospital readmission in patients with RIFLE class Risk, Injury, or Failure in the matched cohort fully adjusted per our primary analysis were 1.31 (95% CI, 1.18–1.44), 1.41 (95% CI–1.24, 1.59), and 1.35 (95% CI–1.18, 1.55), respectively, relative to patients without AKI. Secondary Outcomes AKI was a strong predictor of 30-day postdischarge mortality (Table 4). The odds of 30-day postdischarge mortality in patients with RIFLE class Risk, Injury, and Failure were 1.9-, 2.9-, and 2.2-fold that of patients without AKI. AKI remained a significant predictor of the odds of 30-day postdischarge hospital readmission after adjustment for age, sex, race, Deyo–Charlson Index, and patient type. The adjusted odds www.ccmjournal.org

359

Horkan et al

AKI was associated with subsequent development of ESRD (Fig. 2). The hazard ratio of ESRD developing following hospital discharge adjusted for age, sex, race, patient type, sepsis, Deyo–Charlson Index, and in patients with RIFLE class Risk, Injury, or Failure were 1.97 (95% CI, 1.54–2.53), 3.46 (95% CI, 2.65–4.52), and 9.11 (95% CI, 7.53–11.01), respectively, relative to patients without AKI. With additional adjustment for chronic kidney disease stage, the adjusted hazard ratios of ESRD developing with RIFLE class Risk, Injury, or Failure were 2.85 (95% CI, 2.22–3.66), 4.67 (95% CI, Figure 1. Time-to-event curves for hospital readmission. Note: Unadjusted hospital readmission rates were 3.59–6.09), and 3.92 (95% CI, calculated with the use of the Kaplan–Meier methods and compared with the use of the log-rank test. 3.17–4.85), respectively, relaCategorization of acute kidney injury by Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease (RIFLE) class is per the primary analysis. The global comparison log-rank p value is less than 0.001. tive to patients without AKI. Finally, following adjustment of 30-day postdischarge hospital readmission in patients with for age, sex, race, patient type, sepsis, Deyo–Charlson Index, RIFLE class Risk, Injury, and Failure were 1.4-, 2.0-, and 1.6and chronic kidney disease stage, the odds of ESRD developing fold that of patients without AKI (Table 4). The results did at 1 year in patients with RIFLE class Risk, Injury, or Failure not materially differ with additional adjustment for propenwere 3.20 (95% CI, 2.06–4.97), 5.96 (95% CI, 3.72–9.54), and sity score (Table 4), hospital site covariate, or length of stay 6.45 (95% CI, 4.60–9.04), respectively, relative to patients (data not shown). without AKI. Table 3. Unadjusted and Adjusted Associations Between Acute Kidney Injury and 30-Day Postdischarge Hospital Readmission (n = 61,173) Risk, Injury, Failure, Loss of Kidney Function, and End-Stage Kidney Disease Class Characteristic

No Acute Kidney Injury

Risk

Injury

Failure

1.68 (1.55–1.82)

1.93 (1.73–2.16)

1.92 (1.30–2.16)

< 0.001

< 0.001

< 0.001

1.39 (1.28–1.51)

1.46 (1.30–1.64)

1.42 (1.26–1.61)

< 0.001

< 0.001

< 0.001

1.39 (1.28–1.51)

1.47 (1.31–1.64)

1.43 (1.26–1.62)

< 0.001

< 0.001

< 0.001

Unadjusted  OR (95% CI)

1.00 (Referent)

  p Covariate adjusted  OR (95% CI)

1.00 (Referent)

  p Propensity score and covariate adjusted

a

 OR (95% CI)

1.00 (Referent)

  p

OR = odds ratio. a Estimates adjusted for age, sex, race (white vs non-white), patient type (surgical vs medical), Deyo–Charlson Index, sepsis, and propensity score. Propensity scores were calculated utilizing logistic regression to estimate the probability for the presence or absence of a diagnosis of acute kidney injury. Unadjusted associations between Risk, Injury, Failure, Loss of Kidney Function, and End-Stage Kidney Disease (RIFLE) class groups and 30-day postdischarge hospital readmission were estimated by bivariable logistic regression models. Patients who died within 30 days of hospital discharge who were not readmitted were excluded. Adjusted odd ratios were estimated by multivariable logistic regression models with inclusion of covariate terms thought to associate with both RIFLE class groups and 30-day postdischarge hospital readmission plausibly. Estimates adjusted for age, sex, race (white vs non-white), patient type (surgical vs medical), Deyo–Charlson Index, and sepsis.

360

www.ccmjournal.org

February 2015 • Volume 43 • Number 2

Clinical Investigations

Table 4. Unadjusted and Adjusted Associations Between Acute Kidney Injury and 30-Day Postdischarge Mortality (n = 62,096) Risk, Injury, Failure, Loss of Kidney Function, and End-Stage Kidney Disease Class Characteristic

No Acute Kidney Injury

Risk

Injury

Failure

Unadjusted  OR (95% CI)

1.00 (Referent)

  p

1.89 (1.64–2.17)

2.85 (2.41–3.36)

2.24 (1.84–2.73)

< 0.001

< 0.001

< 0.001

1.44 (1.25–1.66)

1.98 (1.66–2.36)

1.55 (1.26–1.91)

< 0.001

< 0.001

< 0.001

1.44 (1.24–1.66)

1.98 (1.66–2.37)

1.55 (1.26–1.91)

< 0.001

< 0.001

< 0.001

Covariate adjusted  OR (95% CI)

1.00 (Referent)

  p Propensity score and covariate adjusted

a

 OR (95% CI)   p

1.00 (Referent)

OR = odds ratio. a Estimates adjusted for age, sex, race (white vs non-white), patient type (surgical vs medical), Deyo–Charlson Index, sepsis, and propensity score. Propensity scores were calculated using logistic regression to estimate the probability for the presence or absence of a diagnosis of acute kidney injury. Unadjusted associations between Risk, Injury, Failure, Loss of Kidney Function, and End-Stage Kidney Disease (RIFLE) class groups and 30-day postdischarge mortality were estimated by bivariable logistic regression models. Patients who died within 30 days of hospital discharge who were not readmitted were included. Adjusted odd ratios were estimated by multivariable logistic regression models with inclusion of covariate terms thought to associate with both RIFLE class groups and 30-day postdischarge mortality plausibly. Estimates adjusted for age, sex, race (white vs non-white), patient type (surgical vs medical), Deyo–Charlson Index, and sepsis.

Subanalyses When patients with chronic kidney disease stage 5 were excluded (n = 59,563), the odds of 30-day postdischarge hospital readmission in patients with RIFLE class Risk, Injury, or Failure fully adjusted per our primary analysis were 1.35 (95% CI, 1.24–1.47), 1.41 (95% CI, 1.26–1.59), and 1.41 (95% CI,

1.26–1.59) respectively, relative to patients without AKI. In addition, adjustment for chronic kidney disease stage in place of Deyo–Charlson Index did not materially alter the point estimates; the odds of 30-day postdischarge hospital readmission in patients with RIFLE class Risk, Injury, or Failure adjusted for age, sex, race, patient type, sepsis, and chronic kidney disease stage were 1.46 (95% CI, 1.35–1.60), 1.63 (95% CI, 1.45–1.82), and 1.58 (95% CI, 1.39–1.80), respectively, relative to patients without AKI. In the cohort (n = 61,173), we performed a sensitivity analysis where the exposure RIFLE class Risk, Injury, or Failure was determined from the maximum creatinine obtained throughout the entire hospitalization following critical care admission. Inclusion of all creatinine measurements from ICU admission to hospital discharge did not materially alter the AKI-readmission association; the odds of 30-day postdischarge hospital readmission in patients with RIFLE Figure 2. Time-to-event curves for end-stage renal disease progression. Note: Unadjusted end-stage renal disease prevalence was calculated with the use of the Kaplan–Meier methods and compared with the use of class Risk, Injury, or Failure the log-rank test. Categorization of acute kidney injury by Risk, Injury, Failure, Loss of kidney function, and Enddetermined throughout the stage kidney disease (RIFLE) class is per the primary analysis. The global comparison log-rank p value is less hospitalization following than 0.001. Critical Care Medicine

www.ccmjournal.org

361

Horkan et al

critical care admission adjusted for age, sex, race, patient type, Deyo–Charlson index, and sepsis were 1.37 (95% CI, 1.26– 1.49), 1.39 (95% CI, 1.24–1.55), and 1.41 (95% CI, 1.28–1.58) respectively, relative to patients without AKI. Effect Modification and Confounding Gender, race, chronic kidney disease, or hospital did not emerge as an effect modifier of the association between AKI and 30-day postdischarge hospital readmission (p-interaction: 30-day postdischarge hospital readmission > 0.10 for all variables). Analysis following stratification for age, chronic kidney disease stage, and Deyo–Charlson Index shows that the AKIreadmission association is preserved across strata (Supplemental Table 1, Supplemental Digital Content 1, http://links. lww.com/CCM/B124). Furthermore, individually running the adjusted model with and without terms for chronic kidney disease or baseline glomerular filtration rate, the 30-day postdischarge hospital readmission estimates in each case are similar (data not shown). This indicates that the AKI 30-day postdischarge hospital readmission relationship is not materially confounded by chronic kidney disease stage or baseline glomerular filtration rate.

DISCUSSION In this study, we investigated whether AKI in the critically ill was associated with in-hospital survivors with the risk of postdischarge adverse outcomes. Our data demonstrate that AKI is associated with a significant increase in the odds of 30-day postdischarge hospital readmission, 30-day postdischarge mortality, and progression to ESRD in the following year. However, as our study is observational and not interventional, a causal relationship between AKI status and postdischarge outcomes cannot be inferred from these data alone. Risk factors for adverse events in ICU survivors include high ICU bed occupancy, time of day at ICU discharge, severity of illness scores, organ failure indices, and discharge facility (49-52). It is well described that ICU survivors suffer significant long-term morbidity and mortality (53). From our data, it appears that a substantial minority of patients who survive to hospital discharge are readmitted to the hospital within 30 days and this is accentuated in those with AKI. Such a difference is clinically significant in light of the high priority to reduce readmissions among Medicare beneficiaries (54). Studies have identified organizational issues thought to be important in reducing unplanned readmissions (55–57), including discharge planning effectiveness, premature discharge prevention, quality of inpatient care improvement, and care coordination improvement inside the hospital and following discharge (58). The 30-day hospital readmission metric has not been rigorously tested in critical illness ­survivors. Approximately 15% of patients who survive an ICU stay will die during the first 6 months following hospital discharge (59). The risk factors for posthospital death in critical illness survivors are not well described. Whereas previous studies have explored predictive models for mortality in the critically ill, 362

www.ccmjournal.org

there are few known prognostic factors for patients who survive critical illness. What our study illustrates is that postdischarge outcomes in critical illness survivors appear to be associated with comorbidity and with organ failure. The present study may have limitations. The impact of AKI on readmission, mortality, and subsequent ESRD cannot be studied outside of the observational study framework and as such, this is the only study design available for studying this association. We are not able to follow readmissions to hospitals outside of MGH and BWH so our estimates are less than the published prevalence of readmission (28). We cannot exclude the possibility that other unmeasured variables influence readmission independently of AKI, which may have biased estimates. Reliance on ICD-9 codes to determine the Deyo–Charlson Index or sepsis measures the treated prevalence rather than the true prevalence which is likely higher (60). Despite adjustment for multiple potential confounders, there might be residual confounding of unmeasured variables leading to observed differences in outcomes. Furthermore, we have not included measures of functional status, health literacy, social support, and medication adherence, which are recognized as predictors of unplanned readmission(61). Our finding that AKI is a significant predictor of postdischarge outcomes does not include Acute Physiology and Chronic Health Evaluation II scores, which are strong predictors of critical illness outcome (62). It is conceivable that inclusion of a physiologic score in the analysis may materially alter the AKI -readmission association. With the addition of age and gender data, the Deyo–Charlson index is reported to be an alternative (albeit inferior) method of risk adjustment without physiologic data (63). However, despite multivariable adjustment, the absence of physiologic data is a potential limitation of our study. The present study has several strengths. The RIFLE criteria has been evaluated in several clinical studies of critically ill patients with AKI (3, 64–67). The RIFLE criteria are clinically relevant for the diagnosis of AKI, correlate with patient outcomes (68, 69), and have been validated in the ICU (3, 66). In our administrative dataset, we have previously shown that the Deyo–Charlson Index provides sufficient adjustment for comorbid conditions with RIFLE Injury/Failure as an outcome (33). We have previously validated the use of CPT code 99291 to identify patients in the RDPR administrative dataset who are admitted to an ICU (30). The 30-day time frame for hospital readmission is commonly used in outcomes research (70–73), has been demonstrated to be the statistically optimal choice for identifying readmission rates (74, 75), and is most relevant for targeting interventions and for public reporting (76). Our study compliments observations regarding the risk of the development of ESRD related to the severity of AKI (5).

CONCLUSIONS In aggregate, these data demonstrate that AKI is associated with the odds of 30-day hospital readmission, postdischarge mortality, and development of ESRD in critically illness survivors. With the importance of reducing the cost of healthcare February 2015 • Volume 43 • Number 2

Clinical Investigations

delivery, exposures that are predictive of postdischarge outcomes may be useful for targeted interventions. Accurate estimates of prognosis may enlighten the postdischarge management of critically illness survivors who suffered an episode of AKI. If our hypothesis is correct, those critical illness survivors who suffer AKI might benefit from a more thorough follow-up schedule, enhanced patient education (77), and longitudinal care focusing on the prevention of ESRD.

ACKNOWLEDGMENTS This article is dedicated to the memory of our dear friend and colleague Nathan Edward Hellman, MD, PhD.

REFERENCES

1. Dennen P, Douglas IS, Anderson R: Acute kidney injury in the intensive care unit: An update and primer for the intensivist. Crit Care Med 2010; 38:261–275 2. Barrantes F, Tian J, Vazquez R, et al: Acute kidney injury criteria predict outcomes of critically ill patients. Crit Care Med 2008; 36:1397–1403 3. Hoste EA, Clermont G, Kersten A, et al: RIFLE criteria for acute kidney injury are associated with hospital mortality in critically ill patients: A cohort analysis. Crit Care 2006; 10:R73 4. Ishani A, Xue JL, Himmelfarb J, et al: Acute kidney injury increases risk of ESRD among elderly. J Am Soc Nephrol 2009; 20:223–228 5. Chawla LS, Amdur RL, Amodeo S, et al: The severity of acute kidney injury predicts progression to chronic kidney disease. Kidney Int 2011; 79:1361–1369 6. Amdur RL, Chawla LS, Amodeo S, et al: Outcomes following diagnosis of acute renal failure in U.S. veterans: Focus on acute tubular necrosis. Kidney Int 2009; 76:1089–1097 7. Wald R, Quinn RR, Luo J, et al; University of Toronto Acute Kidney Injury Research Group: Chronic dialysis and death among survivors of acute kidney injury requiring dialysis. JAMA 2009; 302:1179–1185 8. Lo LJ, Go AS, Chertow GM, et al: Dialysis-requiring acute renal failure increases the risk of progressive chronic kidney disease. Kidney Int 2009; 76:893–899 9. Maynard SE, Whittle J, Chelluri L, et al: Quality of life and dialysis decisions in critically ill patients with acute renal failure. Intensive Care Med 2003; 29:1589–1593 10. Garzotto F, Piccinni P, Cruz D, et al; NEFROINT Investigation Group: RIFLE-based data collection/management system applied to a prospective cohort multicenter Italian study on the epidemiology of acute kidney injury in the intensive care unit. Blood Purif 2011; 31: 159–171 11. Korkeila M, Ruokonen E, Takala J: Costs of care, long-term prognosis and quality of life in patients requiring renal replacement therapy during intensive care. Intensive Care Med 2000; 26:1824–1831 12. Rumball-Smith J, Hider P: The validity of readmission rate as a marker of the quality of hospital care, and a recommendation for its definition. N Z Med J 2009; 122:63–70 13. Catlin A, Cowan C, Hartman M, et al; National Health Expenditure Accounts Team: National health spending in 2006: A year of change for prescription drugs. Health Aff (Millwood) 2008; 27:14–29 14. Medicare Payment Advisory Commission: Healthcare Spending and the Medicare Program: A Data Book. Washington, DC, Medicare Payment Advisory Commission, 2006 15. Medicare Payment Advisory Commission: Report to the Congress: Promoting Greater Efficiency in Medicare. Washington, DC, Medicare Payment Advisory Commission, 2007 16. Silverstein MD, Qin H, Mercer SQ, et al: Risk factors for 30-day hospital readmission in patients ≥65 years of age. Proc (Bayl Univ Med Cent) 2008; 21:363–372 17. Morris DS, Rohrbach J, Rogers M, et al: The surgical revolving door: Risk factors for hospital readmission. J Surg Res 2011; 170:297–301

Critical Care Medicine

18. Rosenberg AL, Watts C: Patients readmitted to ICUs*: A systematic review of risk factors and outcomes. Chest 2000; 118:492–502 19. Lone NI, Seretny M, Wild SH, et al: Surviving intensive care: A systematic review of healthcare resource use after hospital discharge*. Crit Care Med 2013; 41:1832–1843 20. Thakar CV, Parikh PJ, Liu Y: Acute kidney injury (AKI) and risk of readmissions in patients with heart failure. Am J Cardiol 2012; 109:1482–1486 21. Brown JR, Parikh CR, Ross CS, et al; Northern New England Cardiovascular Disease Study Group: Impact of perioperative acute kidney injury as a severity index for thirty-day readmission after cardiac surgery. Ann Thorac Surg 2014; 97:111–117 22. Noble JS, Simpson K, Allison ME: Long-term quality of life and hospital mortality in patients treated with intermittent or continuous hemodialysis for acute renal and respiratory failure. Ren Fail 2006; 28:323–330 23. Morgera S, Kraft AK, Siebert G, et al: Long-term outcomes in acute renal failure patients treated with continuous renal replacement therapies. Am J Kidney Dis 2002; 40:275–279 24. Gopal I, Bhonagiri S, Ronco C, et al: Out of hospital outcome and quality of life in survivors of combined acute multiple organ and renal failure treated with continuous venovenous hemofiltration/hemodiafiltration. Intensive Care Med 1997; 23:766–772 25. Ahlström A, Tallgren M, Peltonen S, et al: Survival and quality of life of patients requiring acute renal replacement therapy. Intensive Care Med 2005; 31:1222–1228 26. Hamel MB, Phillips RS, Davis RB, et al: Outcomes and cost-effectiveness of initiating dialysis and continuing aggressive care in seriously ill hospitalized adults. SUPPORT Investigators. Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments. Ann Intern Med 1997; 127:195–202 27. Gammelager H, Christiansen CF, Johansen MB, et al: One-year mortality among Danish intensive care patients with acute kidney injury: A cohort study. Crit Care 2012; 16:R124 28. Dartmouth Atlas Project, PerryUndem Research & Communications: The Revolving Door: A Report on U.S. Hospital Readmissions. Princeton, NJ, Robert Wood Johnson Foundation, 2013. 29. Murphy SN, Chueh HC: A security architecture for query tools used to access large biomedical databases. Proc AMIA Symp 2002:552–556 30. Zager S, Mendu ML, Chang D, et al: Neighborhood poverty rate and mortality in patients receiving critical care in the academic medical center setting. Chest 2011; 139:1368–1379 31. Meduri GU, Golden E, Freire AX, et al: Methylprednisolone infusion in early severe ARDS: Results of a randomized controlled trial. Chest 2007; 131:954–963 32. Bellomo R, Ronco C, Kellum JA, et al; Acute Dialysis Quality Initiative workgroup: Acute renal failure - definition, outcome measures, animal models, fluid therapy and information technology needs: The Second International Consensus Conference of the Acute Dialysis Quality Initiative (ADQI) Group. Crit Care 2004; 8:R204–R212 33. Braun AB, Litonjua AA, Moromizato T, et al: Association of low serum 25-hydroxyvitamin D levels and acute kidney injury in the critically ill. Crit Care Med 2012; 40:3170–3179 34. Bagshaw SM, Uchino S, Cruz D, et al; Beginning and Ending Supportive Therapy for the Kidney (BEST Kidney) Investigators: A comparison of observed versus estimated baseline creatinine for determination of RIFLE class in patients with acute kidney injury. Nephrol Dial Transplant 2009; 24:2739–2744 35. Rapoport J, Gehlbach S, Lemeshow S, et al: Resource utilization among intensive care patients. Managed care vs traditional insurance. Arch Intern Med 1992; 152:2207–2212 36. Charlson ME, Pompei P, Ales KL, et al: A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Dis 1987; 40:373–383 37. Quan H, Sundararajan V, Halfon P, et al: Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005; 43:1130–1139 www.ccmjournal.org

363

Horkan et al 38. Martin GS, Mannino DM, Eaton S, et al: The epidemiology of sepsis in the United States from 1979 through 2000. N Engl J Med 2003; 348:1546–1554 39. Braun A, Chang D, Mahadevappa K, et al: Association of low serum 25-hydroxyvitamin D levels and mortality in the critically ill. Crit Care Med 2011; 39:671–677 40. Beier K, Eppanapally S, Bazick HS, et al: Elevation of blood urea ­nitrogen is predictive of long-term mortality in critically ill patients ­independent of “normal” creatinine. Crit Care Med 2011; 39:305–313 41. Levey AS, Bosch JP, Lewis JB, et al: A more accurate method to estimate glomerular filtration rate from serum creatinine: A new prediction equation. Modification of Diet in Renal Disease Study Group. Ann Intern Med 1999; 130:461–470 42. Koehler BE, Richter KM, Youngblood L, et al: Reduction of 30-day postdischarge hospital readmission or emergency department (ED) visit rates in high-risk elderly medical patients through delivery of a targeted care bundle. J Hosp Med 2009; 4:211–218 43. Landrum L, Weinrich S: Readmission data for outcomes measurement: Identifying and strengthening the empirical base. Qual Manag Health Care 2006; 15:83–95 44. Jencks SF, Williams MV, Coleman EA: Rehospitalizations among patients in the Medicare fee-for-service program. N Engl J Med 2009; 360:1418–1428 45. Uchino S, Doig GS, Bellomo R, et al; Beginning and Ending Supportive Therapy for the Kidney (B.E.S.T. Kidney) Investigators: Diuretics and mortality in acute renal failure. Crit Care Med 2004; 32:1669–1677 46. Bagshaw SM, Lapinsky S, Dial S, et al; Cooperative Antimicrobial Therapy of Septic Shock (CATSS) Database Research Group: Acute kidney injury in septic shock: Clinical outcomes and impact of duration of hypotension prior to initiation of antimicrobial therapy. Intensive Care Med 2009; 35:871–881 47. D’Agostino RB Jr: Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med 1998; 17:2265–2281 48. Leuven E, Sianesi B: PSMATCH2: Stata module to perform full Mahalanobis and propensity score matching, common support graphing, and covariate imbalance testing. 2003 [cited July 18, 2013]. Available at http://ideas.repec.org/c/boc/bocode/s432001.html 49. Frost SA, Alexandrou E, Bogdanovski T, et al: Severity of illness and risk of readmission to intensive care: A meta-analysis. Resuscitation 2009; 80:505–510 50. Lone NI, Walsh TS: Impact of intensive care unit organ failures on mortality during the five years after a critical illness. Am J Respir Crit Care Med 2012; 186:640–647 51. Chrusch CA, Olafson KP, McMillan PM, et al: High occupancy increases the risk of early death or readmission after transfer from intensive care. Crit Care Med 2009; 37:2753–2758 52. Beck DH, McQuillan P, Smith GB: Waiting for the break of dawn? The effects of discharge time, discharge TISS scores and discharge facility on hospital mortality after intensive care. Intensive Care Med 2002; 28:1287–1293 53. Desai SV, Law TJ, Needham DM: Long-term complications of critical care. Crit Care Med 2011; 39:371–379 54. Stefan MS, Pekow PS, Nsa W, et al: Hospital performance measures and 30-day readmission rates. J Gen Intern Med 2013; 28: 377–385 55. Coleman EA, Parry C, Chalmers S, et al: The care transitions intervention: Results of a randomized controlled trial. Arch Intern Med 2006; 166:1822–1828

364

www.ccmjournal.org

56. Naylor MD, Brooten DA, Campbell RL, et al: Transitional care of older adults hospitalized with heart failure: A randomized, controlled trial. J Am Geriatr Soc 2004; 52:675–684 57. Jack BW, Chetty VK, Anthony D, et al: A reengineered hospital discharge program to decrease rehospitalization: A randomized trial. Ann Intern Med 2009; 150:178–187 58. Kanel K, Elster S, Colleen V. PRHI Readmission Briefs Brief 1: Overview of Six Target Chronic Diseases. 2010 [cited July 15, 2013]. Available at http://www.chqpr.org/downloads/PRHI_ ReadmissionBrief_ChronicDisease_June2010.pdf 59. Wunsch H, Guerra C, Barnato AE, et al: Three-year outcomes for Medicare beneficiaries who survive intensive care. JAMA 2010; 303:849–856 60. Linde-Zwirble WT, Angus DC: Severe sepsis epidemiology: Sampling, selection, and society. Crit Care 2004; 8:222–226 61. Coleman EA, Min SJ, Chomiak A, et al: Posthospital care transitions: Patterns, complications, and risk identification. Health Serv Res 2004; 39:1449–1465 62. Knaus WA, Draper EA, Wagner DP, et al: APACHE II: A severity of disease classification system. Crit Care Med 1985; 13:818–829 63. Quach S, Hennessy DA, Faris P, et al: A comparison between the APACHE II and Charlson Index Score for predicting hospital mortality in critically ill patients. BMC Health Serv Res 2009; 9:129 64. Uchino S, Bellomo R, Goldsmith D, et al: An assessment of the RIFLE criteria for acute renal failure in hospitalized patients. Crit Care Med 2006; 34:1913–1917 65. Abosaif NY, Tolba YA, Heap M, et al: The outcome of acute renal failure in the intensive care unit according to RIFLE: Model application, sensitivity, and predictability. Am J Kidney Dis 2005; 46:1038–1048 66. Ahlström A, Kuitunen A, Peltonen S, et al: Comparison of 2 acute renal failure severity scores to general scoring systems in the critically ill. Am J Kidney Dis 2006; 48:262–268 67. Bagshaw SM, George C, Dinu I, et al: A multi-centre evaluation of the RIFLE criteria for early acute kidney injury in critically ill patients. Nephrol Dial Transplant 2008; 23:1203–1210 68. Ricci Z, Cruz D, Ronco C: The RIFLE criteria and mortality in acute kidney injury: A systematic review. Kidney Int 2008; 73:538–546 69. Kellum JA: Acute kidney injury. Crit Care Med 2008; 36:S141–S145 70. Oddone EZ, Weinberger M, Horner M, et al: Classifying general medicine readmissions. Are they preventable? Veterans Affairs Cooperative Studies in Health Services Group on Primary Care and Hospital Readmissions. J Gen Intern Med 1996; 11:597–607 71. Marcantonio ER, McKean S, Goldfinger M, et al: Factors associated with unplanned hospital readmission among patients 65 years of age and older in a Medicare managed care plan. Am J Med 1999; 107:13–17 72. Frankl SE, Breeling JL, Goldman L: Preventability of emergent hospital readmission. Am J Med 1991; 90:667–674 73. Jiang HJ, Andrews R, Stryer D, et al: Racial/ethnic disparities in potentially preventable readmissions: The case of diabetes. Am J Public Health 2005; 95:1561–1567 74. Halfon P, Eggli Y, van Melle G, et al: Measuring potentially avoidable hospital readmissions. J Clin Epidemiol 2002; 55:573–587 75. Heggestad T, Lilleeng SE: Measuring readmissions: Focus on the time factor. Int J Qual Health Care 2003; 15:147–154 76. Donzé J, Aujesky D, Williams D, et al: Potentially avoidable 30-day hospital readmissions in medical patients: Derivation and validation of a prediction model. JAMA Intern Med 2013; 173:632–638 77. Bradley EH, Curry L, Horwitz LI, et al: Contemporary evidence about hospital strategies for reducing 30-day readmissions: A national study. J Am Coll Cardiol 2012; 60:607–614

February 2015 • Volume 43 • Number 2