Characteristics and Outcomes of AKI Treated with Dialysis during Pregnancy and the Postpartum Period

Abstract

Acute kidney injury (AKI) is a rare complication of pregnancy, but may be associated with significant morbidity and mortality in young and often otherwise healthy women. We conducted a retrospective population-based cohort study of all consecutive pregnancies over a 15-year period (1997–2011) in Ontario, Canada, and describe the incidence and outcomes of AKI treated with dialysis during pregnancy or within 12 weeks of delivery. Of 1,918,789 pregnancies, 188 were complicated by AKI treated with dialysis (incidence: 1 per 10,000 [95% confidence interval, 0.8 to 1.1]). Only 21 of 188 (11.2%) women had record of a preexisting medical condition; however, 130 (69.2%) women experienced a major pregnancy-related complication, including preeclampsia, thrombotic microangiopathy, heart failure, sepsis, or postpartum hemorrhage. Eight women died (4.3% versus 0.01% in the general population), and seven (3.9%) women remained dialysis dependent 4 months after delivery. Low birth weight (<2500 g), small for gestational age, or preterm birth (<37 weeks’ gestation) were more common in pregnancies in which dialysis was initiated (35.6% versus 14.0%; relative risk, 3.40; 95% confidence interval, 2.52 to 4.58). There were no stillbirths and fewer than five neonatal deaths (<2.7%) in affected pregnancies compared with 0.1% and 0.8%, respectively, in the general population. In conclusion, AKI treated with dialysis during pregnancy is rare and typically occurs in healthy women who acquire a major pregnancy-related medical condition such as preeclampsia. Many affected women and their babies have good short-term outcomes.

Pregnancy-related AKI is associated with significant morbidity and mortality in young and often otherwise healthy women.^1,2 Since the 1960s, the incidence of pregnancy-related AKI in developed countries has declined from 1 in 3000 to 1 in 18,000.¹ This trend has been closely linked to improved obstetrical care and dramatic reductions in first-trimester AKI attributable to septic abortions.² However, these trends are less relevant in contemporary populations. The changing risk profile of pregnant North American women includes lower parity, older age, higher body mass index, and greater comorbidities such as hypertension, diabetes mellitus, and chronic kidney disease, as well as increased use of reproductive technologies resulting in multiple gestations. These changes may have a unique impact on the incidence, etiology, and outcomes of pregnancy-related AKI.^3–7

Reports that describe the burden of pregnancy-related AKI in more contemporary developed cohorts are limited by small sample sizes, highly selected populations, and use of definitions for AKI that have not been validated in pregnancy.^8–11 Other reports fail to characterize AKI in the context of patient demographics, standards of health care delivery, and physician practice patterns.¹ We conducted the present study to describe the incidence and outcomes of the most severe cases of AKI in pregnancy, namely, those treated with dialysis. We analyze data from 1.9 million pregnancies in Ontario, Canada from 1997 to 2011, and describe the comorbid conditions associated with AKI treated with dialysis (AKI-D) during pregnancy and the postpartum period, and characterize maternal and perinatal outcomes after development of AKI-D.

Results

Pregnancies in Ontario, Canada

There were 1,918,789 deliveries among 1,204,797 unique mothers over the 15-year study period in Ontario, Canada (Supplemental Figure 1). The median number of pregnancies per mother during the study period was 1 (interquartile range [IQR] 1–2; range 1–11). The median maternal age at delivery was 30 years (IQR 26–34). The prevalence of at least one recorded preexisting medical condition, including hypertension, diabetes mellitus, or chronic kidney disease, was 5.3%. There were 31,258 (1.6%) multiple gestation pregnancies. Cesarean sections were performed in 488,901 (25.5%) pregnancies.

Acute Kidney Injury Treated with Dialysis

Of 1,918,789 pregnancies over the 15-year period, 188 were complicated by AKI-D, representing a cumulative incidence of 1 per 10,000 (95% confidence interval [95% CI], 0.8 to 1.1). These events occurred in 188 different women. There was a significant decline in the incidence over time, from 1.4 per 10,000 in 1997 to 0.8 per 10,000 in 2011 (P<0.001). Among the 188 mothers with AKI-D during pregnancy, 92 (48.9%) had at least one other pregnancy during the study period, 41 (44.6%) of which occurred after the index pregnancy complicated by AKI-D.

Dialysis was initiated within 12 weeks postpartum (median of 6 days [IQR 3–28] after delivery) in over half (n=101 [53.7%]) of affected pregnancies. The remainder of acute dialysis events occurred during pregnancy: 22 (11.7%) in the first trimester, 23 (12.2%) in the second trimester, and 42 (22.3%) in the third trimester (at a median of 31 days [IQR 13–85] before delivery). Once dialysis was initiated, the median follow-up time was 84 (IQR 72–187) days. Over this time, 139 women (73.9%) received at least one intermittent hemodialysis treatment, 55 (29.3%) received at least one continuous veno-venous hemodialysis treatment, and five or less (≤2.7%) received at least one peritoneal dialysis treatment. Sixty-two women (33.0%) initiated dialysis during an intensive care unit stay; 37 of these women were mechanically ventilated on the date of dialysis initiation.

Maternal Characteristics

Compared with the general population, women who developed AKI-D in pregnancy were older, had a lower neighborhood income, fewer prenatal visits, and were more likely to have preexisting hypertension or chronic kidney disease (Table 1). Although only 21 of the 188 women (11.2%) who developed AKI-D had record of a preexisting medical condition, 40 (21.3%) women had been seen by an internal medicine specialist and 11 (5.9%) had been seen by a nephrologist at least once in the 5 years prior to pregnancy (Table 1). Women with record of a preexisting medical condition were significantly more likely to develop AKI-D during their pregnancy (relative risk [RR] 2.24; 95% CI, 1.42 to 3.52).

Table 1.

Maternal characteristics among pregnancies with and without AKI-D

Characteristic	AKI-D n=188	No AKI-D n=1,918,601	P Valuea
Unique mothers, n	188	1,204,609
Age at delivery (years), median (IQR)	32 (25–35)	30 (25–34)	0.03
Age category at delivery (years), n (%)			0.01
<20	11 (5.9)	78,037 (4.1)
20–29	61 (32.4)	823,665 (42.9)
30–39	106 (56.4)	956,009 (49.8)
≥40	10 (5.3)	60,890 (3.2)
Neighborhood income quintile, n (%)			0.04
1–2 (lowest)	95 (50.5)	818,899 (42.7)
3 (middle)	26 (13.8)	385,375 (20.1)
4–5 (highest)	66 (35.1)	704,477 (36.7)
Rural residence, n (%)	18 (9.6)	202,373 (10.5)	0.66
Number of previous pregnanciesb			0.49
0	104 (55.3)	990,884 (51.6)
1	57 (30.3)	659,676 (34.4)
≥2	27 (14.4)	268,041 (14.0)
Time since previous pregnancy (years), median (IQR)	0 (0–3)	0 (0–3)	0.55
Recorded preexisting medical conditionsc n (%)
Hypertension	10 (5.3)	25,439 (1.3)	<0.001
Diabetes mellitus	≤5 (≤2.7)	14,935 (0.8)	0.70
Chronic kidney disease	8 (4.3)	4,568 (0.2)	<0.001
Systemic lupus erythematosis	≤5 (≤2.7)	627 (0.03)	<0.001
Thrombotic microangiopathy treated with plasma exchange	0 (0.0)	24 (0.001)	0.96
Any of the above preexisting medical conditions	21 (11.2)	102,026 (5.3)	<0.001
Pre-pregnancy physician consultationc n (%)
General practitioner/family physician	178 (94.7)	1,814,275 (94.6)	0.94
Internal medicine specialist	40 (21.3)	301,323 (15.7)	0.04
Nephrologist	11 (5.9)	9,080 (0.5)	<0.001
Number of prenatal physician visitsd median (IQR)	10 (6.5–12)	10 (8–12)	0.02
Number of ultrasonographic examinationsd median (IQR)	2 (0–4)	2 (0–4)	0.53
Use of assisted reproductive technologiese n (%)	6 (3.2)	37,471 (2.0)	0.22

To comply with privacy regulations for minimizing the chance of identification of a study participant, numbers of participants are suppressed in the case of five or fewer participants (reported as ≤5).

^aP values were derived from two-sample t-tests when comparing continuous variables and chi-squared tests when comparing categorical variables.

^bInformation on previous pregnancies was available since 1991 and included all pregnancies in Ontario resulting in livebirth or stillbirth after 20 weeks’ gestation.

^cInformation on preexisting medical conditions and consultations was ascertained from administrative database codes in the 5 years prior to pregnancy.

^dInformation on prenatal visits and ultrasonographic examinations was ascertained from administrative database codes during pregnancy.

^eUse of assisted reproductive technologies was ascertained from administrative database codes in the year prior to conception and during pregnancy and the postpartum period, and includes tuboplasty after previous sterilization, artificial insemination, ovarian hyperstimulation, in vitro fertilization, and other unspecified fertilization methods.

Medical complications during pregnancy are shown in Table 2. Diabetes mellitus, preeclampsia, thrombotic microangiopathy, heart failure, pyelonephritis, sepsis, placental abruption, and postpartum hemorrhage were significantly more prevalent among women with AKI-D. Women who experienced a major medical complication during pregnancy were nearly four times more likely to develop AKI-D (RR 3.73; 95% CI, 2.78 to 5.08). These medical complications were more frequently recorded among women who presented with AKI-D in the postpartum period (88 [87.1%] versus 10 [45.4%] in the first trimester, 8 [34.8%] in the second trimester, and 24 [57.1%] in the third trimester) and preeclampsia was the most frequent medical complication recorded during this time frame (Supplemental Table 1). Compared with the general population, women with AKI-D were also more likely to deliver by cesarean section (Table 2). There was no difference in the use of assisted reproductive technologies or multiple gestations between the two groups (Tables 1 and 2).

Table 2.

Complications among pregnancies with and without AKI-D

Complication, n (%)	AKI-D n=188	No AKI-D n=1,918,601	P Valuea
Multifetal gestation	≤5 (≤2.7)	31,255 (1.6)	0.97
Cesarean delivery	81 (43.1)	488,820 (25.5)	<0.001
Medical complicationsb
Hyperemesis gravidarum	11 (5.9)	77,107 (4.0)	0.20
Diabetes mellitusc	13 (6.9)	45,615 (2.4)	<0.001
Gestational hypertension	6 (3.2)	34,619 (1.8)	0.15
Preeclampsia	40 (21.3)	99,686 (5.2)	<0.001
Thrombotic microangiopathyd	25 (13.3)	8143 (0.4)	<0.001
Pregnancy-related liver diseasee	≤5 (≤2.7)	1475 (0.1)	<0.001
Heart failure	30 (16.0)	2665 (0.1)	<0.001
Pyelonephritis	6 (3.2)	13,585 (0.7)	<0.001
Sepsis	34 (18.1)	6252 (0.3)	<0.001
Amniotic fluid embolism	≤5 (≤2.7)	39 (0.0)	<0.001
Placenta previa	≤5 (≤2.7)	10,275 (0.5)	0.05
Placental abruption	9 (4.8)	46,242 (2.4)	0.03
Postpartum hemorrhage	33 (17.6)	63,144 (3.3)	<0.001
Any of the above medical complications	130 (69.2)	719,973 (37.5)	<0.001

To comply with privacy regulations for minimizing the chance of identification of a study participant, numbers of participants are suppressed in the case of five or fewer participants (reported as ≤5).

^aP values were derived from two-sample t-tests when comparing continuous variables and from chi-squared tests when comparing categorical variables.

^bMedical complications were ascertained from administrative database codes during pregnancy and the peripartum period.

^cDiabetes mellitus includes preexisting type 1 diabetes mellitus, preexisting type 2 diabetes mellitus, and gestational diabetes.

^dThrombotic microangiopathy includes HELLP syndrome and thrombotic thrombocytopenic purpura. Eight of the 25 patients in the AKI-D group were also treated with plasma exchange.

^ePregnancy-related liver disease includes unspecified liver and biliary diseases in pregnancy.

Maternal Mortality and Dialysis Dependence

Among the 188 pregnancies complicated by AKI-D, eight women (4.3%) died within 90 days of delivery, compared with 229 (0.01%) maternal deaths in the non–AKI-D group (P<0.001) (Table 3). All of these deaths occurred within 4 weeks of delivery and at a median of 2 (IQR 1–8) days from the initiation of dialysis. Listed causes of death included severe preeclampsia, pregnancy-related liver or biliary tract disorders, hemorrhage, and complications of obstetric surgery. Among the 180 pregnancies in which the mother survived beyond 90 days, seven (3.9%) women remained dependent on chronic dialysis after 90 days postpartum (Table 3).

Table 3.

Incidence of 90-day maternal mortality and chronic dialysis dependence among pregnancies with and without AKI-D

Maternal outcome, n (%)	AKI-D n=188	No AKI-D n=1,918,601	P Valuea
Maternal mortality	8 (4.3)	229 (0.01)	<0.001
Chronic dialysis dependenceb	7 (3.9)c

To comply with privacy regulations for minimizing the chance of identification of a study participant, numbers of participants are suppressed in the case of five or fewer participants (reported as ≤5).

^aP values represent the difference in unadjusted relative risk of the specified maternal outcome associated with the presence versus absence of AKI-D in pregnancy. This was estimated from log-binomial models using generalized estimating equations to account for within-mother clustering from additional pregnancies during follow-up. Unadjusted relative risks were not reported due to unstable estimates as a result of the low number of events in the group without AKI-D.

^bChronic dialysis dependence was not relevant in the No AKI-D group.

^cRefers to the 180 (95.7%) pregnancies complicated by AKI-D in which the mother survived beyond 90 days.

Perinatal Outcomes

Of 188 pregnancies affected by AKI-D, 67 infants (35.6%) experienced adverse outcomes (Table 4). Compared with the general population, infants from pregnancies affected by AKI-D were more likely to be of low birth weight (RR 4.7; 95% CI, 3.6 to 6.0), small for gestational age (RR 3.2; 95% CI, 1.9 to 5.3), or born preterm (RR 2.5; 95% CI, 2.0 to 3.1). There were no stillbirths and fewer than five neonatal deaths (<2.7%) in affected pregnancies compared with 0.1% and 0.8% in the general population, respectively.

Table 4.

Risk of adverse perinatal outcomes associated with AKI-D in pregnancy and the postpartum period

Perinatal outcome, n (%)	AKI-D n=188	No AKI-D n=1,918,601	Relative Riska (95% CI)
Low birth weight	50 (26.6)	102,958 (5.4)	4.66 (3.64–5.96)
Small for gestational age	15 (8.0)	46,080 (2.4)	3.16 (1.90–5.27)
Preterm birth	61 (32.5)	242,844 (12.7)	2.49 (2.03–3.06)
Stillbirth	0 (0.0)	1416 (0.1)	…
Perinatal mortality	≤5 (≤2.7)	7320 (0.4)	…
Neonatal death	≤5 (≤2.7)	15,858 (0.8)	…
Any of the above perinatal outcomes	67 (35.6)	268,582 (14.0)	3.40 (2.52–4.58)

To comply with privacy regulations for minimizing the chance of identification of a study participant, numbers of participants are suppressed in the case of five or fewer participants (reported as ≤5).

^aAll relative risks are unadjusted, estimated from log-binomial models using generalized estimating equations to account for within-mother clustering from additional pregnancies during follow-up. Ellipses indicate unstable estimates of relative risk due to the low number of events in the AKI-D group.

Discussion

In this study of 1.9 million pregnancies over a 15-year period, 1 in 10,000 pregnancies were affected by AKI-D. Most women who developed AKI-D had no record of a preexisting health condition; however, those with prepregnancy hypertension, diabetes, chronic kidney disease, or systemic lupus erythematosus were twice as likely to develop AKI-D in pregnancy compared with those without such a record. Those with a major complication of pregnancy, including preeclampsia, thrombotic microangiopathy, heart failure, pyelonephritis, sepsis, placental abruption, and postpartum hemorrhage were nearly four times more likely to develop AKI-D in pregnancy. Maternal mortality after pregnancy-related AKI-D was 4%, and 4% of survivors remained dialysis dependent after delivery. Adverse perinatal outcomes such as low birth weight and preterm birth occurred in 35% pregnancies affected by AKI-D; however, there were no stillbirths and fewer than five neonatal deaths (<3%).

To our knowledge, this is the first population-based study to characterize the incidence and outcomes following severe AKI in pregnancy in a developed nation without apparent selection bias or loss to follow-up, containing over three times the number of cases treated with dialysis as other reports.⁸ Previous population-based reports on pregnancy-related AKI refer almost exclusively to secular trends in infection-related first-trimester AKI coinciding with the legalization of abortion and improved obstetrical care in low-income countries.^1,12 Pregnancies in high-income regions are more frequently complicated by hypertensive disorders of pregnancy, gestational diabetes, postpartum hemorrhage, and postpartum fever of unknown origin, whereas genitourinary and obstetric infections appear to be on the decline.^6,13 Results from a recent Canadian study of AKI in pregnancy defined by hospital diagnosis codes suggest that hypertensive disorders of pregnancy may be driving increases in AKI as well.¹¹ Our data suggest that these complications, alone or in combination, may also occur in a setting necessitating acute dialysis. These findings are in keeping with known pathophysiologic mechanisms of AKI and two recent studies of severe cases of pregnancy-related AKI from Brazil and Egypt, defined by treatment with dialysis and AKI criteria within admission to the intensive care unit, respectively.^8,9

Preeclampsia was the most prevalent medical complication among women with AKI-D in our cohort and the most frequent medical complication recorded among women initiated on dialysis in the postpartum period. Preeclampsia is the leading cause of maternal morbidity and mortality worldwide, and a major contributor to premature births and intrauterine growth restriction.¹⁴ The frequency of preeclampsia ranges between 2% and 7% in healthy nulliparous women and increases substantially in women with multiple gestation, chronic hypertension, diabetes mellitus, and chronic kidney disease.^15,16 In addition to the direct effect of preeclampsia on renal function via endothelial injury, secondary effects of intravascular volume depletion, vasoconstriction, and activation of inflammatory and coagulation cascades in these settings can increase the risk of kidney failure in pregnancy.¹⁷ As the low prevalence of preexisting medical conditions in our study was likely underestimated, women with chronic hypertension, prepregnancy diabetes mellitus, and chronic kidney disease may drive a greater proportion of preeclampsia-related AKI-D events than realized herein.^6,7 In contrast to previous studies, our data suggest that although AKI events during pregnancy frequently occur in the context of acute catastrophic medical complications (one third of women with AKI-D were admitted to the intensive care unit at the time of dialysis initiation, half of whom were also mechanically ventilated), the impact may be short-lived. Maternal mortality was 4% and fewer than 4% of women remained dialysis dependent after 12 weeks postpartum, compared with previous studies indicating that maternal mortality associated with AKI-D in pregnancy may be as high as 30%, with up to 50% of survivors remaining dependent on chronic dialysis.^1,8 However, these studies represent single-center experiences prone to selection biases.

The lack of a consensus definition for AKI in pregnancy and variations in demographics and standard of care among populations has made estimation of the incidence of pregnancy-related AKI from epidemiologic studies difficult.^8,9,11,18 Widely accepted definitions for AKI in the general population using serum creatinine concentration have not been validated in pregnancy; hence, mild AKI in pregnancy may be missed given the normal physiologic drop in serum creatinine from hyperfiltration.^19–21 Our study avoided such measurement issues by defining the most severe cases of AKI based on receipt of acute dialysis, a universally recognized clinical indicator of high disease severity. However, our study has some limitations. First, we did not have access to clinical parameters at the time of dialysis initiation, including renal function, blood pressure, and basis for starting or stopping dialysis, so women with progressive chronic kidney disease initiating dialysis electively during pregnancy to optimize maternal and fetal wellbeing could have been misclassified as having AKI-D in this study. However, the low prevalence of preexisting chronic kidney disease and low proportion of women that remained dialysis dependent postpartum suggests that this was infrequent. Second, we conducted our analysis using hospital diagnostic and physician billing codes from administrative data. Our strategy for identifying many of the perinatal and maternal indicators with these codes may lack sensitivity and, therefore, underestimate their true prevalence. Further, the cause of AKI-D could not be determined for certain using these data. Third, our cohort was limited to pregnancies that resulted in a livebirth or stillbirth after 20 weeks’ gestation. This may have resulted in underestimation of first trimester AKI-D events. Assuming a 42-week gestation for the purposes of timing of AKI-D events prior to delivery may also have contributed to underestimation and misclassification of first-trimester events (favoring third-trimester events) for preterm deliveries. However, we would expect first-trimester AKI-D events to be infrequent in our population.⁶ Finally, we restricted our definition of AKI to those treated with dialysis, and these results may not apply to less severe presentations of AKI in pregnancy.

In conclusion, AKI treated with dialysis in pregnancy is rare and typically occurs in otherwise healthy women who acquire a major pregnancy-related medical condition. Fortunately, with ongoing improvements in obstetrical care, multidisciplinary approaches, and new insights into the diagnosis and management of associated conditions such as preeclampsia, maternal and perinatal mortality in this setting are largely avoidable.

Concise Methods

Design and Setting

We conducted a retrospective population-based cohort study using linked health care databases in Ontario, Canada. Ontario provides universal access to hospital and physician services for all permanent residents, with about 140,000 births per year over the previous 5 years (2009–2013).²² During our study period, the population of Ontario was approximately 75% Caucasian, 14% Asian, 4% Black, 2% Aboriginal, and 1% Latin American.²³ Study conduct and reporting follow guidelines (STROBE) for observational studies (Supplemental Table 2).²⁴ A research ethics board at Sunnybrook Health Sciences Centre in Toronto approved the prespecified protocol. However, numbers of participants were suppressed in the case of five or fewer participants (reported as ≤5) to comply with privacy regulations for minimizing the chance of identification of a study participant.

Data Sources

We identified our pregnant population by childbirth and ascertained maternal characteristics and outcome data from records in seven linked databases from the Institute of Clinical Evaluative Sciences (ICES). Vital statistics and location of residence were obtained from the Registered Persons Database, which contains demographic information for all Ontario residents ever issued a health card.²⁵ We used inpatient and outpatient diagnostic and procedural information to define maternal characteristics and outcome data. The Canadian Institute for Health Information Discharge Abstract Database (CIHI-DAD) collects demographic, diagnostic, and procedural variables for all acute care, chronic, and day surgery admissions. The Ontario Health Insurance Plan (OHIP) database contains fee-for-service claims for inpatient and outpatient physician services.^25,26 Each claim record includes information about the physician, service provided, and diagnostic information. Diabetic and hypertensive patients were identified using the Ontario Diabetes Database and the Ontario Hypertension Database. These databases are derived by ICES and use a validated algorithm to identify diabetic and hypertensive patients based on data from CIHI-DAD and OHIP.^27,28 Childbirths and perinatal outcomes were identified using the MOMBABY dataset. This is also an ICES-derived dataset that links the CIHI-DAD inpatient admission records of delivering mothers and their newborns.²⁵ Maternal cause of death was determined from the Ontario Registrar General Death database, which contains information on all deaths registered in Ontario.²⁶ We have used these databases in previous studies to research health outcomes and health services including those related to AKI-D.^29–32 All variables used in this study were complete, with minor exceptions detailed in Supplemental Table 3.

In Ontario, hospital admissions were coded using the International Classification of Diseases, Ninth Revision prior to April 2002 and the International Classification of Diseases, Tenth Revision thereafter. Billing claims are submitted using fee and diagnosis codes outlined in the OHIP Schedule of Benefits. Whenever possible, we defined pregnancy characteristics and maternal and perinatal outcomes using database codes that have been proven feasible and reliable in prior studies.^{26–28,33–37} These codes typically have high specificity and modest sensitivity for capturing the variable of interest. Detailed information on coding definitions and validity is provided in Supplemental Table 4.

Population

We established a cohort of all pregnancies resulting in childbirth between January 1, 1997, and December 31, 2011. Our data sources capture all hospital childbirths in the province, defined as delivery of an infant at a gestational age of ≥20 weeks.^38,39 For the purpose of defining a baseline period in which to assess comorbid conditions, we assigned an index date to each woman that was 42 weeks prior to delivery, which represents the longest term gestational duration. Prepregnancy characteristics were assessed in the 5 years prior to the index date and pregnancy outcomes were assessed between the index date and 12 weeks after delivery (or 17 weeks for dialysis dependence). We excluded pregnancies among mothers with a baseline history of end-stage renal disease (defined as chronic dialysis treatment or kidney transplantation).

Maternal Outcomes

We defined AKI in pregnancy (AKI-D) as receipt of at least one acute dialysis treatment during pregnancy or within 12 weeks postpartum (pregnancy complications such as preeclampsia may occur up to 12 weeks postpartum).⁴⁰ We documented all deaths up to 90 days after delivery and defined chronic dialysis dependence as the receipt of two or more chronic dialysis treatments between 90 and 120 days after the date of delivery.

Perinatal Outcomes

Perinatal outcomes included low birth weight (<2500 g irrespective of gestational age), small for gestational age (physician diagnosis code for small for gestational age or intrauterine growth restriction on hospital discharge abstraction), preterm birth (birth of a live infant before 37 weeks’ gestation), stillbirth (fetal death after 22 weeks’ gestation), perinatal mortality (stillbirth or death of a live-born infant within 7 days) and neonatal death (death of a live-born infant within 28 days).^41–44

Statistical Analysis

Pregnancy was the unit of analysis in this study. We determined the cumulative incidence of AKI-D in pregnancy between 1997 and 2011 and used the Cochran–Armitage trend test to determine if there was a significant change in the incidence over time. To compare maternal characteristics and complications among pregnancies with and without AKI-D, we used two-sample t tests for continuous variables and chi-squared tests for categorical variables. Maternal mortality rates were compared between women with and without AKI-D using generalized estimating equations to account for the within-mother clustering from additional pregnancies during follow-up. Among AKI-D affected pregnancies, we also determined the incidence of chronic dialysis dependence. We used generalized estimating equations and log-binomial models to estimate (1) the relative risk of developing AKI-D in women with and without a preexisting medical condition (listed in Table 1), (2) the relative risk of AKI-D in women who experienced a major medical complication of pregnancy (listed in Table 2), and (3) the relative risk of perinatal outcomes in pregnancies with and without AKI-D. The reference group in all analyses was the general population. Confidence intervals of 95% (95% CI) were calculated and we set the level of significance at P<0.05. All analyses were conducted using SAS, version 9.3 (SAS Institute Inc., Cary, NC).

Disclosures

All authors have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any organization for the submitted work; no financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years; no other relationships or activities that could appear to have influenced the submitted work.