Acute kidney injury in major gynaecological surgery: an observational study


  • AJ Vaught and T Ozrazgat-Baslanti have contributed equally to the manuscript.



To assess the prevalence, outcomes and cost associated with acute kidney injury (AKI) defined by consensus risk, injury, failure, loss, and end-stage kidney (RIFLE) criteria after gynaecologic surgery.


Retrospective single-centre cohort study.


Academic medical centre.


Two thousand three hundred and forty-one adult women undergoing major inpatient gynaecologic surgery between January 2000 and November 2010.


AKI was defined by RIFLE criteria as an increase in serum creatinine greater than or equal to 50% from the reference creatinine. We used multivariable regression analyses to determine the association between perioperative factors, AKI, mortality and cost.

Main outcome measures

AKI, combined major adverse events (hospital mortality, sepsis or mechanical ventilation), 90-day mortality and hospital cost.


Overall prevalence of AKI was 13%. The prevalence of AKI was associated with the primary diagnosis. Of women with benign tumour surgeries, 5% (43/801) experienced AKI compared with 18% (211/1159) of women with malignant disease (< 0.001). Only 1.3% of the whole cohort had evidence of urologic mechanical injury. In a multivariable logistic regression analysis, AKI patients had nine times the odds of a major adverse event compared to patients without AKI (adjusted odds ratio 8.95, 95% confidence interval 5.27–15.22). We have identified several readily available perioperative factors that can be used to identify patients at high risk for AKI after in-hospital gynaecologic surgery.


AKI is a common complication after major inpatient gynaecologic surgery associated with an increase in resource utilisation and hospital cost, morbidity and mortality.


Acute kidney injury (AKI) is a serious complication among hospitalised patients associated with increased morbidity and mortality. With the introduction of the consensus RIFLE (risk, injury, failure, loss, and end-stage kidney) criteria, the adverse effects of less severe AKI, characterised by changes in serum creatinine (sCr) level reflecting acute decline in glomerular filtration rate, have increasingly been recognised.[1, 2] Whereas the RIFLE definition is based on at least a 50% change in sCr relative to the reference sCr[3], the recent KDIGO (kidney disease: improving global outcomes) clinical practice guideline further expanded RIFLE to include sCr changes as small as 0.3 mg/dl.[4] Furthermore, recommendations in this guideline provide a series of stage-based management clinical steps to be considered for all women with AKI or at high risk of developing it.

Among surgical patients, an association between small sCr changes and short-and long-term mortality has emerged in the literature.[5-10] Although the prevalence and risk factors for AKI have been increasingly studied in general surgical patients,[7, 10] studies describing the prevalence and outcomes of AKI defined by RIFLE criteria among gynaecologic surgical patients are lacking. Although an increasing number of gynaecologic procedures are being performed on an outpatient basis, in-hospital gynaecologic procedures constitute almost 12% of all surgical procedures.[11, 12] In the absence of data based on consensus AKI definitions in such a large surgical population, the health-care burden of AKI after gynaecological surgeries is difficult to determine and insight into the clinical risk stratification is unavailable for practising surgeons.

In a large single-centre cohort of women undergoing major inpatient gynaecologic surgery, we assessed the prevalence, outcomes, risk factors and cost for AKI defined using consensus RIFLE criteria.


Data source and patient population

Using the University of Florida (UF) Integrated Data Repository we assembled a single-centre retrospective cohort by integrating multiple clinical databases. We included women 18 years and older admitted to the hospital for longer than 24 hours following an in-hospital gynaecologic procedure between 1 January 2000 and 30 November 2010. We excluded women with chronic kidney disease prior to admission[13] and those with any obstetric procedure (= 214). We used the combination of the operating physician's specialty and the first two diagnostic and procedure International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes to identify gynaecologic procedures and to group them according to the primary diagnosis and anatomic location. The study was approved by the University of Florida Institutional Review Board and the Privacy Office.

Definition of AKI

We applied two AKI definitions using sCr changes only: RIFLE (AKI) and the American College of Surgeons National Surgical Quality Improvement Program (NSQIP-AKI). RIFLE defines AKI using at least a 50% sCr change from a reference sCr (corresponding to at least 25% decline in glomerular filtration rate)[3] and NSQIP defines AKI as a rise in sCr greater than 2 mg/dl from the preoperative value or as the acute requirement for renal replacement therapy.[14] We defined reference sCr either as the minimum sCr within 6 months of the admission (used for the main results), or as the mean and minimum sCr within 7 days of the admission (used for sensitivity analyses).[15] Women with AKI were stratified according to the worst RIFLE stage reached during the hospitalisation. RIFLE Risk corresponds to a 50% increase in sCr or 25–50% decline in glomerular filtration rate, RIFLE Injury to a doubling in sCr or 51–75% decline in glomerular filtration rate, and RIFLE Failure to a tripling in sCr or >75% decline in glomerular filtration rate compared with preoperative renal function. Complete renal recovery existed if the discharge sCr returned to a level less than 50% above reference sCr. Partial renal recovery existed if there was a persistent increase in sCr more than 50% above reference sCr but no need for renal replacement therapy. Less than 5% of the sCr values were missing in the overall cohort.

Outcomes and covariates

The main outcomes included AKI, combined major adverse events (hospital mortality, severe sepsis or mechanical ventilation), 90-day mortality and hospital cost. Patient survival status was determined using hospital discharges and the Social Security Death Index. We defined severe sepsis by adding ICD-9-CM codes for acute organ dysfunction to the sepsis diagnosis.[16, 17] The exact dates were used to calculate the duration of mechanical ventilation, intensive care unit (ICU) and hospital stay. Cost of hospitalisation was estimated using the ratio of cost-to-charge for urban hospitals in the South Atlantic division.[18] We converted all costs to 2012 US dollars using the Consumer Price Index to adjust for inflation over the years. Postoperative complications were defined as previously described.[19, 20] We identified urologic injuries related to gynaecologic procedure using validated ICD-9-CM codes for bladder and ureteral injuries and manual review of medical records.[21]

The presence of underlying comorbidities and gynaecologic tumours was identified by ICD-9-CM codes using previously validated criteria[22, 23] and by calculating the Charlson–Deyo comorbidity index.[24] Emergent surgery was defined as either non-elective surgery or emergent admission using clinical data. Using residency zip code, we linked to US Census data[25] to calculate residing neighbourhood characteristics and distance from hospital.[26]

Statistical analysis

The analytical plan followed the STROBE recommendations.[27] The Pearson chi-squared test or Fisher's exact test was used to test independence between categorical variables, while Student's t-test, analysis of variance and the Kruskal–Wallis test were used for comparison of continuous variables as appropriate. All significance tests were two-sided with α < 0.05 considered statistically significant unless Bonferroni correction was used to adjust for multiple comparisons. In those cases α < 0.0083 was considered statistically significant. Statistical analyses were performed with SAS v.9.3 (SAS, Cary, NC, USA).

For all multivariable analyses we selected explanatory variables based on their significance in a prior univariate analysis and reported association in the literature. We used separate logistic models to determine the association between (1) perioperative factors (age, gender, ethnicity, primary diagnosis, emergent surgery status, weekend admission, socio-economic status and comorbid conditions on admission), and occurrence of any stage of AKI and (2) AKI and combined major adverse eventswhile adjusting for other postoperative complications in addition to perioperative factors listed above. Adjusted odds ratios (OR) with 95% confidence intervals (95% CI) were reported for logistic regression models. We used the area under the receiver operating characteristics curve values (AUC) and Hosmer–Lemeshow goodness-of-fit test to assess model fit and discrimination. We performed sensitivity analyses by restricting analyses only to women within each primary diagnosis group and by comparing the effect of different definitions for reference sCr on the model fit.

We constructed risk-adjusted generalised log-linear models[28] for hospital cost with logarithmic transformation due to the skewness of the distribution. Each model was adjusted for perioperative risk factors and other postoperative complications. Adjusted incremental cost with 95% CI was calculated using non-parametric transformation of the regression coefficients to US dollars. Standard errors were calculated using the smearing estimate to adjust for bias due to the transformation of costs back to the original scale.[29]


Prevalence of acute kidney injury

Among the 2341 adult women undergoing major gynaecologic surgery, the most common admission diagnoses were malignant (= 1159) and benign (= 801) neoplastic disease, accounting for 84% of the cohort. The most common primary procedures were abdominal hysterectomies among uterine/fallopian surgeries (= 1349), salpingo-oophorectomies among ovarian surgeries (= 345), vulvectomy (= 41) and repair of colorectovaginal fistulas and rectoceles (= 39) among vulva/perineum/vaginal surgeries.

The overall prevalence of AKI was 13% (295/2341). The prevalence of AKI was strongly associated with the primary diagnosis: only 5% (43/801) of women with benign tumour surgeries experienced AKI compared with 18% (211/1159) among women with malignant disease (< 0.001) (Supporting Information Table S1). Mild to moderate AKI (stages Risk and Injury) constituted the majority of all AKI, with only 15% (45/295) of AKI patients developing RIFLE Failure stage and 3% (8/295) requiring renal replacement therapy. AKI was significantly more prevalent when defined using RIFLE criteria than using NSQIP criteria, which captured only 7% of the women with RIFLE AKI. Only 14% of all AKI patients had a diagnosis of AKI listed in their discharge summaries (Table 1).

Table 1. Clinical characteristics of the patients stratified by severity of acute kidney injury
VariablesNo AKI (= 2046)All AKI (= 295)AKI Stage RIFLE Risk (= 185)AKI Stage RIFLE Injury (= 65)AKI Stage RIFLE Failure (= 45)
  1. AKI, Acute kidney injury; ICD-9-CM, International Classification of Diseases, Ninth Revision, Clinical Modification; NSQIP, National Surgical Quality Improvement Program; SD, Standard Deviation; RIFLE, risk, injury, failure, loss, and end-stage kidney.

  2. P < 0.0083 when compared with *no AKI group, **RIFLE Risk group and ***RIFLE Injury group (using Bonferroni correction for the Pearson chi-squared test or Fisher's exact test for categorical variables and Kruskal–Wallis test for continuous variables).

Age (years), mean (SD) 53 (15)60 (15)*60 (15)*59 (14)*61 (14)*
Ethnicity, n (%)
Caucasian1475 (72)226 (77)141 (76)54 (83)31 (69)
African-American380 (18)44 (15)26 (14)7 (11)11 (24)
Hispanic55 (3)9 (3)8 (4)0 (0)1 (2)
Other or missing136 (7)16 (5)10 (6)4 (6)2 (4)
Rural area residency, n (%) 800 (39)97 (33)*61 (33)20 (31)16 (36)
Distance from residing neighbourhood to hospital (km), median (25th,75th) 33 (15, 71)36 (21, 79)36 (19, 77)38 (27, 102)29 (9, 67)
Population living in poverty in residing neighborhood, % (SD) 16 (8)16 (8)15 (8)14 (7)18 (9)
Insurance, n (%)
Medicare621 (30)143 (48)*87 (47)*29 (45)27 (60)*
Medicaid394 (19)50 (17)30 (16)14 (22)6 (13)
Private878 (43)84 (28)*57 (31)*16 (25)11 (24)
Uninsured153 (7)18 (6)11 (6)6 (9)1 (2)
Emergent surgery, n (%)244 (12)108 (37)*60 (32)*25 (38)*23 (51)*
Weekend admission, n (%)60 (3)29 (10)*15 (8)*7 (11)*7 (16)*
Weekend discharge, n (%)695 (34)54 (18)*37 (20)*13 (20)*4 (9)*
Charlson's Comorbidity Index, Median (25th, 75th) 1 (0, 3)5 (2, 8)*5 (2, 8)*8 (2, 8)*3 (2, 8)*
Comorbid conditions on admission, n (%)
Diabetes mellitus306 (15)61 (21)*38 (21)14 (22)9 (20)
Chronic obstructive pulmonary disease246 (12)54 (18)*37 (20)*9 (20)8 (18)
Cardiovascular disease104 (5)38 (13)*24 (13)*5 (8)9 (20)*
Cerebrovascular disease24 (1)10 (3)*7 (4)2 (1)1 (2)
Metastatic tumour307 (15)144 (49)*91 (49)*37 (57)*16 (36)*
Admission serum creatinine (mg/dl), median (25th, 75th)0.7 (0.6, 0.8)0.8 (0.6, 1.1)0.7 (0.6, 1.0)*0.85 (0.6, 1.1)*1.3 (0.75, 2.1)*,**,***
AKI onset in the first 72 hours, n (%) 265 (68)85 (46)49 (75)**35 (78)**
AKI by NSQIP definition, n (%) 22 (7)0 (0)0 (0)22 (49) **,***
ICD-9-CM AKI diagnostic code in discharge summary, n (%) 40 (14)11 (6)12 (18)**17 (38)**

Most AKI episodes developed within 72 hours of admission, whereas only 9% of AKI patients had onset of AKI after the 7th day of admission. AKI was more likely to occur among women undergoing emergent surgery and those admitted on the weekend. The majority of emergent surgeries were related to benign and malignant uterine and ovarian neoplasms, abscess of the vulva, menorrhagia, torsion of the ovary, and infections. Only 1.3% of the whole cohort had evidence of urologic mechanical injury. Although women with AKI have a higher prevalence of urologic mechanical injury compared with women without AKI (5.4%, 16/295 versus 0.7%, 14/2046), this subgroup of women represented only a small proportion of AKI.

Perioperative factors associated with acute kidney injury

We constructed a multivariable logistic regression model to determine perioperative factors independently associated with the occurrence of any stage AKI (Table 2). The model demonstrated good discrimination and model fit. Increasing age, emergent surgery and weekend admission as well as the presence of congestive heart failure and chronic pulmonary disease on admission were associated with AKI. Women with benign gynaecologic tumours had lower odds for AKI, whereas women with any type of cancer, and especially those with metastatic cancer, had significantly higher odds. Diabetes mellitus was a risk factor for AKI only among women with surgeries for benign tumours and non-neoplastic disease (OR 2.36, 95% CI 1.32–4.21) but not among women with surgery for malignant tumours.

Table 2. Perioperative factors associated with acute kidney injury
VariablesAdjusted odds ratio (95% confidence interval)
  1. Adjusted odds ratios were derived using the multivariable logistic regression that included all listed variables in the model simultaneously. The modeled outcome was any stage of acute kidney injury. The cross-validation estimate of the area under receiver operator curve (95% confidence interval) was 0.79 (0.77, 0.82) by bootstrap resampling, and the Hosmer–Lemeshow test indicated good fit (= 0.64).

  2. a

    P-value <0.05.

Age, per 1-year increase 1.015 (1.005, 1.03)a
African-American ethnicity (vs. others) 1.15 (0.78, 1.69)
Primary diagnosis (vs. benign tumour)
Malignant tumour1.80 (1.17, 2.76)a
Non-neoplastic disease of female organs1.57 (0.94, 2.63)
Emergent surgery (vs. elective) 4.10 (2.84, 5.67)a
Weekend admission (vs. weekday admission) 1.84 (1.08, 3.13)a
Metastatic cancer (Yes vs. No) 4.01 (2.84, 5.67)a
Congestive heart failure (Yes vs. No) 2.25 (1.23, 4.11)a
Chronic pulmonary disease (Yes vs. No) 1.80 (1.24, 2.61)a
Diabetes mellitus (Yes vs. No) 1.40 (0.98, 1.99)

Adverse hospital outcomes and acute kidney injury

Although the overall hospital mortality and 90-day mortality were only 0.9 and 1.7%, respectively, in women with AKI, hospital mortality increased to 7% and 90-day mortality to 10%. Of the 22 women who died in hospital, 91% had AKI during their hospitalisation. The primary causes of hospital death were neoplastic disease (19/22), chronic liver disease (1/22), immune disorders (1/22) and sepsis (1/22) (Table 3).

Table 3. Postoperative complications and hospital outcomes stratified by severity of acute kidney injury
 VariablesNo AKI (n = 2046)All AKI (n = 295)AKI stage RIFLE Risk (= 185)AKI stage RIFLE Injury (= 65)AKI stage RIFLE Failure (= 45)
  1. AKI, acute kidney injury; CI, confidence interval; RIFLE, risk, injury, failure, loss, and end-stage kidney.

  2. < 0.0083 when compared with *no AKI group, RIFLE Risk group and *RIFLE Injury group (using Bonferroni correction for Pearson chi-squared or Fisher's exact test for categorical variables and Kruskal–Wallis test for continuous variables).

Hospital outcomes
Major adverse event (mortality, sepsis, or mechanical ventilation), n (%)29 (1.4)72 (24.4)*27 (15)*21 (32)*,**24 (53)*,**
Hospital mortality, n (%)2 (0.1)20 (7)*1 (0.5)*9 (14)*,**10 (22)*,**
90-day mortality, n (%)10 (0.5)29 (10)*7 (4)*11 (17)*,**11 (24)*,**
Days in hospital, median (25th, 75th)4 (3, 5)10 (7, 17)*10 (7, 13)*10 (7, 26)*,**19 (10, 43)*,**,***
Discharge to home, n (%)1997 (98)232 (79)*162 (88)*48 (74)*,**22 (49)*,**
Resource utilization
Intensive care unit admission, n (%)73 (3.6)110 (37)*50 (27)*32 (49)*,**28 (62)*,**
Days in intensive care unit, median (25th, 75th)2 (1, 4)4 (2, 9)*2 (1, 4)5 (2, 9)*,**13 (5, 28)*,**
Mechanical ventilation, n (%)25 (1)61 (21)*24 (13)*16 (25)*21 (47)*,**
Days on mechanical ventilation, median (25th, 75th)1 (1, 2)5 (2, 16)*2 (2, 4)*7 (2, 19)*,**15 (5, 27)*,**
Postoperative complications
Mechanical wound complications, n (%)30 (1)23 (8)*9 (5)*4 (6)*10 (22)*,**
Surgical infections, n (%)56 (3)18 (6)*7 (4)8 (12)*6 (13)*,**
Pulmonary complications, n (%)50 (2)44 (15)*20 (11)*10 (15)*14 (31)*
Cardiovascular complications, n (%)30 (1)17 (6)*6 (3)*7 (10)*4 (9)*
Venous thromboembolism, n (%)15 (0.7)14 (5)*8 (4)*2 (3)*4 (9)*
Gastrointestinal complications, n (%)62 (3)36 (12)*23 (12)*6 (9)*7 (16)*
Number of postoperative complications, n (%)
None1757 (86)172 (58)*120 (65)*34 (52)*18 (40)*
1257 (13)77 (26)*46 (25)*21 (32)*10 (22)*,**
≥232 (1)46 (16)*19 (10)*10 (15)*17 (38)*,***

Increasing severity of AKI was associated with increasing rates of most of the postoperative complications, including surgical infections, and pulmonary and cardiovascular complications. Close to half of the women with AKI (120/295, 42%) had at least one other postoperative complication, whereas the majority of women without AKI had no other postoperative complications (1757/2046, 86%). Renal recovery from AKI was promising, with full recovery noted 89% of the time and only 11% of all AKI patients having partial or no renal recovery.

We constructed two multivariable logistic regression models using a composite of major adverse events (occurrence of hospital death, sepsis or mechanical ventilation) as an outcome and using AKI stages separately or together while adjusting for perioperative factors and other complications. Both models had good discrimination, showing that AKI had a significant association with major adverse events proportional to the severity of AKI (AUC 0.88, 95% CI 0.84–0.92) (Table 4). Women with AKI had a nine times higher adjusted odds of major adverse event compared with women without AKI. The adjusted odds increased with the severity of AKI, with RIFLE-F having an odds ratio of 23.56 (10.46–53.06). The number of other postoperative complications and emergent surgery also showed an increase in odds of major adverse events.

Table 4. The risk-adjusted association between acute kidney injury and major adverse events
 VariablesAdjusted odds ratio (95% confidence interval)
  1. RIFLE, risk, injury, failure, loss, and end-stage kidney.

  2. Major adverse events were defined as any occurrence of death, sepsis or mechanical ventilation during hospitalisation.

  3. a

    Adjusted odds ratios were derived using the multivariable logistic regression that included all listed variables in the model simultaneously. We used two separate multivariable logistic regression models to calculate odds ratios for all stages of acute kidney injury together and separately. The cross-validation estimate of the area under receiver operator curve (95% confidence interval) was 0.88 (0.84, 0.92) by bootstrap resampling, and the Hosmer–Lemeshow test indicated good fit (P = 0.49).

  4. b

    P-value <0.05.

  5. c

    Number of complications sums all postoperative complications from Table 3.

Acute kidney injury (vs. no acute kidney injury)
Acute kidney injury, all stagesa8.95 (5.27, 15.22)b
Acute kidney injury, stage RIFLE-Risk5.39 (2.96, 9.80)b
Acute kidney injury, stage RIFLE-Injury13.26 (6.42, 27.41)b
Acute kidney injury, stage RIFLE-Failure23.56 (10.46, 53.06)b
Age, per 1-year increase1.01 (0.99, 1.04)
African-American ethnicity (vs. others)1.67 (0.91, 3.08)
Charlson Comorbidity Index score, per unit increase1.05 (0.96, 1.12)
Emergent surgery (vs. elective)2.90 (1.74, 4.82)b
Weekend admission (vs. weekday admission)1.16 (0.50, 2.71)
Primary diagnosis (vs. benign tumour)
Malignant tumour1.23 (0.56, 2.75)
Non-neoplastic diseases of female organs1.64 (0.74, 3.63)
Insurance type (vs. Private)
Medicare1.00 (0.49, 2.04)
Medicaid1.04 (0.51, 2.10)
Uninsured1.37 (0.54, 3.50)
Number of postoperative complicationsc(vs No complications)
One3.53 (2.07, 6.00)b
≥Two10.20 (5.23, 19.89)b

The strong association between AKI and major adverse events remained in sensitivity analyses where we limited analyses to women with malignant tumour surgeries only (OR 5.60, 95% CI 3.11–11.54), benign tumour surgeries (OR 33.03, 95% CI 7.85–139.02) or non-neoplastic disease surgeries (OR 12.42, 95% CI 3.67–42.07). The sensitivity analyses examining the effect of different methods of assigning the RsCr demonstrated no significant change in model fit for each outcome (data not shown).

Utilisation of hospital resources

Women with AKI stayed almost a week longer in the hospital compared with women without AKI, and were less likely to be discharged home (Table 3). When compared with women with no AKI, women with AKI were ten times more likely to be admitted to ICU, and also to have a significantly prolonged ICU stay. Among women admitted to ICU, those with AKI had a significant increase in resource utilisation, including prolonged mechanical ventilation [17/109 (16%) of women with AKI versus 0/72 (0%) of those without AKI, < 0.05] and increased use of invasive monitoring-arterial line placement [29/109 (27%) versus 6/72 (8%), respectively, < 0.05], central venous line placement [74/109 (68%) versus 13/72 (8%), respectively, < 0.05], and tracheostomy [13/109 (12%) versus 0/72 (0%), respectively). The majority of ICU women with AKI (95/109, 87%) were ordered at least one unit of packed red blood cells compared with 47% (34/72) of those without AKI.

In a multivariable regression analysis of hospital cost, when gynaecologic surgical patients had no other adverse events besides AKI, risk-adjusted average hospital cost doubled ($18 900 with AKI versus $8700 without AKI). However, the adjusted incremental cost for major adverse event was four times higher if patients had concomitant AKI ($54 700 versus $13 200, respectively). We found a similar relationship for almost all postoperative complications, as the presence of AKI significantly increased the incremental cost of each complication (Table 5).

Table 5. Risk-adjusted incremental cost of postoperative complications stratified by acute kidney injury occurrence
Additional complicationsNo acute kidney injury (n = 2046)Acute kidney injury (= 295)
  1. CI, confidence interval.

  2. Adjusted average cost was calculated using non-parametric transformation of the regression coefficients generated using generalized log-linear models adjusted for age, ethnicity, Charlson Comorbidity Index, emergent surgery status, admission day, procedure type, primary insurance, and each postoperative complication.

  3. a

    P < 0.001 calculated using two independent samples t-test for comparison with no AKI group.

Patient with no other postoperative complication, n (%)1738 (85.0)156 (52.9)
Adjusted average cost ($1000) (95% CI)8.7 (8.6, 8.9)18.9 (17.3, 20.4)a
Major adverse event, n (%)29 (1.4)72 (24.4)
Adjusted average cost ($1000) (95% CI)13.2 (11.5, 14.9)54.7 (47.2, 62.2)a
Postoperative pulmonary complications, n (%)50 (2.4)44 (14.9)
Adjusted average cost ($1000) (95% CI)10.9 (9.8, 12.0)45.9 (38.3, 53.5)a
Postoperative procedural complications, n (%)91 (4.5)42 (14.2)
Adjusted average cost ($1000) (95% CI)10.8 (10.1, 11.6)37.9 (31.4, 44.4)a
Postoperative gastrointestinal complications, n (%)62 (3.0)36 (12.2)
Adjusted average cost ($1000) (95% CI)12.1 (11.0, 13.1)33.2 (27.1, 39.3)a
Postoperative surgical infections or wound complications, n (%)79 (3.9)35 (11.9)
Adjusted average cost ($1000) (95% CI)11.6 (10.7, 12.5)46.2 (37.7, 54.8)a
Postoperative cardiovascular complications, n (%)30 (1.5)17 (5.8)
Adjusted average cost ($1000) (95% CI)11.1 (9.7, 12.5)49.0 (35.9, 62.1)a
Postoperative venous thromboembolism, n (%)10 (0.5)11 (3.7)
Adjusted average cost ($1000) (95% CI)13.0 (10.0, 15.9)37.4 (25.0, 49.8)a


Main findings

In a large single-centre cohort of women undergoing major in-hospital gynaecologic surgery we have demonstrated that acute kidney injury defined by the RIFLE criteria is a common complication associated with significant increases in mortality, hospital cost and resource utilisation, even after adjustment for other postoperative complications. The gynaecologic surgeries for malignant cancer were associated with the highest prevalence of AKI. Patients with AKI had nine times higher odds of a major adverse event regardless of whether the primary diagnosis was a malignant or a benign tumour or non-neoplastic disease. Consistent with other reports, AKI occurred early, within 72 hours of admission for most patients, emphasising the importance of early detection and intervention in the management of this disease process.[30] The use of hospital resources increased among AKI patients even after excluding those not requiring ICU admission: AKI patients had a two-fold increase in ICU length of stay, and prolonged MV and invasive monitoring were utilised more in ICU patients with AKI than in those without AKI. Not surprisingly, the incremental cost of any other postoperative complication was quadrupled in the presence of AKI. Whereas the average cost for laparoscopic and vaginal hysterectomy is $7000–$8521,[31] the risk-adjusted cost in our cohort doubled from $8700 to $18 900 for those patients who developed AKI only, and quadrupled to $54 700 when AKI was further complicated by additional postoperative complications. The three-fold increase in median hospital length of stay of 11 days among women with the least severe AKI (RIFLE-R) compared with the reported average length of stay of between 2.6 and 3.9 days for women undergoing abdominal and laparoscopic hysterectomies[32] is reflective of the strong association between higher cost and resource utilisation once a complication develops, and reinforces the importance of early risk assessment and diagnosis. Any effort to intervene in the course of AKI prior to a severe and often irreversible stage will require preventive techniques that can be deployed after early risk stratification by frontline clinical providers. The KDIGO clinical practice guideline for AKI was intended specifically for such practitioners and offers a variety of kidney-sparing strategies including haemodynamic and renal monitoring, adequate volume status and avoidance of radiographic and nephrotoxic medications among women at-risk for AKI.[4] Our study demonstrates that several readily available perioperative factors can be used to identify such patients at high risk for AKI after gynaecologic surgery, allowing primary surgical providers to apply preventive strategies in a timely manner.

Strengths and limitations, and interpretation

Similar to other surgical populations, the American College of Surgeons National Surgical Quality Improvement Program criteria for AKI underestimated the prevalence of AKI at only 1% in our cohort, thus failing to identify 90% of the AKI patients, mostly those with mild to moderate stages.[5] The lack of reports of AKI in the gynaecologic population may reflect this misconception of AKI as a rare surgical complication. In contrast to reports of AKI prevalence of <2% when utilising the American College of Surgeons National Surgical Quality Improvement Program criteria,[33] the prevalence of AKI defined by RIFLE in the surgical population is much higher, ranging from 25 to 50%, depending on the type of procedure.[5-8, 10, 34] Not only is AKI one of the most prevalent postoperative complications but it is associated with other adverse events, an increase in resource utilisation, and both short- and long-term mortality.[6-8, 35] This is true not only for the most severe AKI that requires renal replacement therapy but also for the whole spectrum of AKI severity, including only small changes in sCr.[5] With a prevalence of 13% after major gynaecologic surgery, AKI was more common than previously reported postoperative ileus (3%),[36] infections and pelvic abscesses (3–10%),[37] vaginal dehiscence (0.29–4.9%),[38] ureteral injury (0.3–4.8%),[39, 40] and bowel injury (0.3%).[41]

Perioperative AKI is a complex syndrome associated with preoperative patient susceptibilities and intraoperative exposures to nephrotoxins, hypotension, bleeding and surgical stress. In our cohort, AKI patients were more likely to require invasive haemodynamic monitoring and use of vasopressors and blood products, all of which may reflect unstable perioperative haaemodynamic status.[42] Urinary output is a common end-point during resuscitation and its decrease often precedes use of invasive monitoring for assessment of volume status.[8, 34] The increased risk of blood loss and hypotension during the large debulking procedures for gynaecologic cancer and exposure to nephrotoxic agents may contribute to the high odds for AKI among women with cancer.[43] The increased risk for AKI in congestive heart failure and chronic pulmonary disease may be associated with decreased renal perfusion due to low cardiac output or neurohormonal response to hypercapnoea.[44] Mechanical urologic injuries did not appear to be a significant mechanism of AKI, as only 1.28% of the women had any evidence of the injury, consistent with previous reports.[39, 40] Although we have used a validated approach for identifying urologic injuries with ICD-9-CM codes, we acknowledge that underestimation could occur. However, the cumulative urologic injury rate reported with the use of universal intra-operative cystoscopy of 4.3% still cannot account for the reported prevalence of AKI in our study.[40]

Our study is a retrospective cohort analysis and we cannot exclude bias due to unmeasured factors, although we attempted to control for selection bias with multivariable statistical methods. Although we included only major inpatient surgeries for patients with considerable comorbidity burden that may not reflect the patient profile seen in the outpatient setting, in-hospital gynaecologic procedures still constitute almost 12% of all surgical procedures.[11] The major strength of our study was the ability to use longitudinal changes in sCr to define AKI using consensus RIFLE criteria rather than relying on ICD-9-CM diagnostic codes. While we assessed comorbidities and complications using previously validated criteria, this approach relies on accurate coding and thus the potential for underassessment of risk exists.


In conclusion, AKI is a common complication after major inpatient gynaecologic surgery associated with a marked increase in major adverse events. The presence of AKI significantly increases hospital costs associated with other postoperative complications. We have identified several risk factors for AKI that can allow early risk stratification in the preoperative setting in order to improve the delivery of recommended kidney-sparing interventions.[4]

Disclosure of interests

The authors have no conflicts of interest to declare.

Contribution to authorship

The authors were involved as follows: AB – conception, design, data acquisition and supervision, data analysis and interpretation, and drafting and revising the article; AJV – conception, data interpretation, drafting and revising the article; TOB – data acquisition, data analysis and interpretation, and drafting and revising the article; AJ –data analysis and interpretation, and drafting the article; LM – revising the article; CEH – data interpretation, and drafting and revising the article. All authors approved the final version of the article to be published.

Details of ethics approval

The study was approved by the UF Privacy Office and the UF Institutional Review Board (#5-2009) on 05/15/2009.


AB is supported by Award Number K23GM087709 from the National Institutes of Health – National Institute of General Medical Sciences. AB has received grant funding from Astute Medical, Inc. The content is solely the responsibility of the authors and does not represent the official views of the National Institutes of Health. Funding agency had no involvement in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. AJ was supported by the University of Florida Medical Student Summer Research Fellowship.


The authors thank Gigi Lipori, Christine Bono, and Yue Du for assistance with data retrieval. The preliminary report from this research was presented as an oral presentation and awarded at the 38th annual Gulf Atlantic Anesthesia Residents’ Research Conference hosted by the University of Puerto Rico Department of Anesthesiology in Puerto Rico from 27 to 29 April 2012.