Inpatient Pediatric CKD Health Care Utilization and Mortality in the United States

Abstract

Rationale & Objective:

The impact of chronic kidney disease (CKD) on inpatient health care utilization is unknown. This study aimed to describe the prevalence of pediatric CKD among children hospitalized in the United States and examine the association of CKD on hospital outcomes.

Study Design:

Cross-sectional national survey of pediatric discharges

Setting and Participants:

Hospital discharges of children (ages >28 days to 19 years) with a chronic medical diagnosis included in the Health Cost and Utilization Project Kids Inpatient Database for the years 2006, 2009, 2012, and 2016.

Predictor:

Presence of primary or co-existing CKD as identified by diagnosis codes

Outcomes:

Length of Stay (LOS), cost, and mortality

Analytic Approach:

Multivariable analysis using Poisson, Gamma, and logistic regressions were performed for LOS, cost, and mortality, respectively.

Results:

A chronic medical condition was present in 6,524,745 estimated discharges over the study period and CKD was present among 3.9% of discharges (96.1% without CKD). Those with CKD had a longer LOS with a median of 2.8 days (IQR: 1.4–6.0) to 1.8 days (IQR: 1.0–4.4) for those without a CKD diagnosis (p<0.001). Median cost was higher in the CKD group compared to the group without CKD, at $8,755(IQR: $4,563–18,345) per hospitalization and $5,016 (IQR: $2,860–10,109), respectively (p<0.001). Presence of CKD was associated with a longer LOS (29.9%, [CI 27.2–32.6%]), higher cost (61.3% [CI 57.4, 65.4%]), and higher risk of mortality (OR 1.51, CI 1.40, 1.63).

Limitations:

Lack of access to and adjustment for confounders including patient readmission and laboratory data

Conclusions:

Pediatric CKD was associated with longer LOS, higher costs, and a higher risk of mortality compared to hospitalizations with other chronic illnesses. Further studies are needed to better understand the health care needs and delivery of care to hospitalized children with CKD.

Plain Language Summary:

Children with CKD have high hospital health care utilization

Children with chronic kidney disease often require hospitalization for various reasons. However, outcomes of this high risk population are unknown. We used data collected from US hospitals to study the potential impact of CKD on hospitalization-related outcomes. We found that children with chronic kidney disease had longer hospital stays, incurred higher health care expenses, and were at higher risk of death than children hospitalized for other chronic illnesses. Our study suggests that these associations are related to the higher degree of medical complexity among children hospitalized with CKD. Further invstigation is needed to better understand the health care needs and delivery of care to hospitalized children with CKD.

INTRODUCTION

Chronic kidney disease (CKD), defined as long term abnormalities of kidney structure or function, remains a growing health problem.^{1, 2} Children with CKD can progress to end-stage kidney disease (ESKD) requiring dialysis or transplant. Even CKD patients without ESKD have multiple other chronic ailments including hypertension, cardiovascular disease, difficulties in growth, electrolyte anomalies, and metabolic bone disease.³ Additionally, these children are also at risk for acute deteriorations in health secondary to infection, dehydration, and medication associated side effects.^{4, 5} Outcomes of children with CKD and what resources they require when they are hospitalized remain unclear. The pediatric CKD population in the United States is poorly characterized.² This is particularly true for mild and moderate stages of pediatric CKD.⁶ Previous studies have noted the substantial financial stressors within families who have a child with CKD.^{7, 8} While annual Medicare expenditures exceed $50 billion for adult patients with CKD in the United States, similar statistics regarding national healthcare costs for pediatric CKD are unknown.⁹ A few studies have described health care utilization in disease-specific subsets of children with kidney disease, but these are not generalizable to the full population of children with CKD.^10–12 Lack of basic epidemiological information about this population makes policy and care decisions difficult. No study, to our knowledge, has evaluated the inpatient pediatric CKD population on a national scale.

Due to the gaps in our knowledge and the need to assure national capacity in support of the needs of chronically ill children with kidney disease, we undertook this study of inpatient health care utilization of children and adolescents with CKD. Our study goals included the generation of a description of pediatric CKD associated discharges in comparison to pediatric discharges with other chronic conditions, and improve our understanding of the utilization of health care resources and outcomes of pediatric CKD associated hospitalizations in terms of length of stay (LOS), cost, and mortality, compared to pediatric hospitalizations associated with other chronic illnesses.

METHODS

Data Source

De-identified data were obtained from the Agency for Healthcare Research and Quality’s Healthcare Cost and Utilization Project Kids’ Inpatient Database (HCUP-KID). HCUP-KID is a comprehensive, all-payer database of inpatient discharges for children in the United States.¹³ Data were compiled from over 12 months every three years. The data included were generated by taking a sample of 10% of uncomplicated live births and 80% of all other discharges of patients less than 20 years of age from non-rehabilitation hospitals in participating states. Each discharge record includes patient and hospital demographic data, LOS, total charges, mortality, International Classification of Diseases, Ninth or Tenth Revision (ICD-9 or ICD-10) diagnosis, and procedure codes. Cost-to-charge ratios are also provided.¹³ This study was deemed exempt by the University of Michigan Institutional review board, and as such, no consent procedures were implemented.

Sample and Variables of Interest

For our study, 2006, 2009, 2012, and 2016 survey years of HCUP-KID were examined, which included sampled data for only the cohort years indicated. Discharges from hospitalizations with age less than or equal to 28 days and discharges without a chronic illness diagnosis as defined by HCUP were excluded (Figure S1). Discharges with CKD as primary (first diagnosis position in data set) or secondary diagnosis (any other diagnosis position in data set) were identified among this sample. Demographic variables included age, sex, race/ethnicity, primary insurance type (private, public, and other), and median household income based on patient ZIP code. Hospital descriptors included geographic region (Northeast, Midwest, South, West), teaching status, and urban/rural status. Outcome variables of interest were LOS in days, cost, and mortality. Costs were calculated using cost-to-charge ratios provided by HCUP-KID based on hospital-specific all-payer inpatient costs obtained from hospital accounting reports collected by the Centers for Medicare and Medicaid Services. All costs were converted into 2016 US dollars using the Consumer Price Index for All Urban Consumers. We anticipated CKD patients would have an excess of co-occurent Compelx Chronic Conditions (CCC) relative to non-CKD patients. Number of CCC was included in a sensitivity analysis to evaluate if complexity of hospitalization explained the relationships between CKD and outcomes or if there was a residual impact of CKD beyond what was explained by disease complexity. CCCs are a classification system using diagnosis codes that allow for evaluation of the number of organ systems or other disease domains that a particular patient has a diagnosis in.^{14, 15} Importantly for this study, it does not allow for multiple diagnoses from a single organ system to be counted multiple times, limited the degree to multiple diagnoses related to a single overarching disease are counted, which could inflate the perceived complexity of certain discharges.

Definition of CKD

An established pediatric CKD ICD-9 code set was used to identify CKD diagnoses in cohort years 2006 through 2012 (Table S1).^{1, 16} The Centers for Medicare and Medicaid Services crosswalk followed by manual review was used to generate a corresponding pediatric CKD ICD-10 code set for the cohort year 2016 (Table S2). Codes utilized include formal CKD diagnoses (ICD-9 585, ICD-10 N18) and diagnoses codes for glomerular disease, congenital and other anomalies of the kidney and urinary tract, structural anomalies, tubular disorders, and end-stage renal disease, including kidney transplantation. CKD is clinically staged between stage 1 through stage 5, with higher stages equating to worse kidney function.¹ Discharges with a CKD stage diagnosis code were categorized by stage; those with a CKD stage not otherwise specified code were classified as CKD Stage NOS; and those with no CKD stage or NOS code were classified as CKD Other Diagnosis.

Statistical Analysis

Descriptive characteristics were shown using frequencies and percentages for categorical variables. Both medians with interquartile ranges (IQRs) and means with 95% confidence intervals were shown for continuous variables. Due to the complex survey data, sample weights, domain, cluster, and strata were used to generate nationally representative estimates, and Taylor series linearization variance procedures were used in all analyses.¹³ Multivariable analysis using Poisson and Gamma regressions were performed for LOS and cost, respectively.^{17, 18} Results of these models were reported as percentage differences. Logistic regression was performed for mortality (results reported as odds ratios). Trends of LOS, cost, and mortality were evaluated over the four cohort years using interaction terms between CKD and time. All multivariable regressions were adjusted for year, age, sex, race, median household income, insurance type, hospital region, and hospital location/teaching status. There were no missing data for the variables of interest. Sensitivity analyses were performed by stratifying CKD into those with CKD as a primary diagnosis vs. secondary diagnosis (i.e., primary CKD vs. secondary CKD vs. no CKD). A separate analysis stratified CKD by stage (1 vs. 2 vs. 3 vs. 4 vs. 5 vs. ESKD vs. stage NOS vs. other non-staged CKD diagnosis). Sensitivity analyses were performed excluding ESKD patients. Lastly, to evaluate if and how patient complexity attenuates any association of CKD with LOS, cost, and mortality, additional analyses controlling for number of CCCs were performed. Analyses were performed in SAS 9.4 (TS1M6) (SAS Institute, Cary, NC) and R 3.6.2 using survey packages.

RESULTS

Nationally, there were an estimated cumulative 6,524,745 discharges in the survey cohort years of 2006, 2009, 2012, and 2016, with 256,200 discharges of patients with CKD. Notably, only 49,945 discharges identified with CKD had stage of CKD. Demographic characteristics are detailed in Table 1.

Table 1.

Demographic Characteristics, by CKD discharge diagnosis status

Demographic Characteristics	CKD	No CKD	p-value
Total Discharges
Unweighted*	172357	4255608
Weighted†	256200 (3.9)	6268545 (96.1)
Age‡	9.9 [9.7–10.0]	11.0 [10.9–11.2]	<0.0001
Sex (Male)†	131639 (51.5)	2973480 (47.6)	<0.0001
Race/Ethnicity†			<0.0001
White	108261 (49.1)	2759264 (51.6)
Black	39716 (18.0)	1101057 (20.6)
Hispanic	52459 (23.8)	1042432 (19.5)
Asian/Pacific Islander	6817 (3.1)	125790 (2.4)
Native American	1812 (0.8)	47715 (0.9)
Other	11292 (5.1)	268896 (5.0)
Median household income†§			<0.0001
1st quartile	74124 (29.6)	1975130 (32.2)
2nd quartile	62767 (25.0)	1529443 (25.0)
3rd quartile	61500 (24.5)	1407107 (23.0)
4th quartile	52227 (20.8)	1214164 (19.8)
Insurance†			<0.0001
Private	102237 (39.9)	2562783 (41.0)
Public	132474 (51.8)	3118698 (49.8)
Other	21220 (8.3)	575391 (9.2)
Hospital Region†			<0.0001
Northeast	40188 (15.7)	1166250 (18.6)
Midwest	60287 (23.5)	1475376 (23.5)
South	93235 (36.4)	2382840 (38.0)
West	62490 (24.4)	1244078 (19.8)
Hospital Location/Teaching Status†			<0.0001
Rural	6898 (2.8)	444799 (7.3)
Urban nonteaching	25986 (10.6)	1271155 (20.9)
Urban teaching	213255 (86.6)	4361801 (71.8)
Year†			0.3251
2006	63313 (24.7)	1612878 (25.7)
2009	67890 (26.5)	1705428 (27.2)
2012	67478 (26.3)	1612260 (25.7)
2016	57520 (22.5)	1337979 (21.3)
Number of CCCs†			<0.001
0	85223 (33.3)	4527530 (72.2)
1	82390 (32.2)	1260954 (20.1)
2	54086 (21.1)	341674 (5.5)
3	22981 (9.0)	106702(1.7)
4	8532 (3.3)	26761 (0.4)
5	2439 (1.0)	4368 (0.1)
≥6	550 (0.2)	555 (0.1)

^*Unweighted frequency;

^†Weighted frequency (%);

^‡Weighted mean (95% confidence interval);

^§Calculated using median household income by patient ZIP code.

Abbreviations: CCC, Complex Chronic Conditions; CKD, chronic kidney disease; US, United States.

Length of Stay

Discharges with CKD accounted for 427,845 (95%CI: 391,411-464,279) hospital days per cohort year. Children discharged with CKD had a significantly longer median LOS of 2.8 days (IQR: 1.4–6.0) compared to 1.8 days (IQR: 1.0–4.4) in non-CKD-related discharges (Table 2). Mean LOS was 6.7 days (95%CI: 6.5, 6.8) in CKD-related discharges compared to 4.9 days (95%CI: 4.8, 5.0) in non-CKD-related discharges. Overall, discharges with CKD had a 29.9% longer LOS in multivariable analysis (Table 3). Over the four cohort years, the LOS for both CKD and non-CKD discharges remained stable (Figure S2).

Table 2:

Discharge Characteristics, by CKD discharge diagnosis status

Discharge Information	CKD	No CKD	p-value
Length of Stay (days) ‡
Weighted Mean (95% CI)	6.7 [6.5–6.8]	4.9 [4.8 −5.0]	<0.0001
Median (IQR)	2.8 (1.4–6.0)	1.8 (1.0–4.4)	<0.0001
Charges (2016 $US) ‡
Weighted Mean (95% CI)	65199 [62008–68390]	37740 [36325–39155]	<0.0001
Median (IQR)	26243(13,492–56150)	15579 (8460–32115)	<0.0001
Cost (2016 $US) ‡
Weighted Mean (95% CI)	21353 [20394–22311]	12129 [11649–12609]	<0.0001
Median (IQR)	8755 [4563–18345]	5016 [2860–10109]	<0.0001
Mortality (# in-hospital deaths)†	2358 (0.9)	33282 (0.5)	<0.0001

^*Weighted mean (95% confidence interval);

^†Weighted frequency (%); Abbreviations: IQR, interquartile range, CKD, chronic kidney disease; US, United States.

Table 3.

Regression Analyses of Length of Stay, Cost, and Mortality, in pediatric discharges (CKD vs No CKD diagnosis)^*

	Length of Stay	Cost	Mortality
	% Difference [95% CI]		OR [95% CI]
CKD‡ Status
No CKD	Reference	Reference	Reference
CKD	29.9 [27.2, 32.6]	61.3 [57.4, 65.4]⁋	1.51 [1.40, 1.63]
Time (per year)	−0.2 [−0.7, 0.3]	2.0 [1.0, 3.1] ⁋	0.98 [0.97, 0.99]
CKD*Time (per year)	---§	1.0 [1.0, 2.0] ⁋	---§
CKD vs. no CKD in 2006	---§	51.0 [44.5, 57.7]	---§
CKD vs. no CKD in 2009	---§	57.1 [52.7, 61.6]	---§
CKD vs. no CKD in 2012	---§	63.5 [59.3, 67.8]	---§
CKD vs. no CKD in 2016	---§	72.5 [64.9, 80.4]	---§
Age	−0.5 [−0.7, −0.3]	−1.7 [−1.9, −1.4]	0.97 [0.97, 0.98]
Sex
Male	Reference	Reference	Reference
Female	−7.5 [−8.6, −6.3]	−11.4 [−12.3, - 10.4]	0.71 [0.68, 0.74]
Race/Ethnicity
White	Reference	Reference	Reference
Black	−2.8 [−4.9, −0.6]	−6.2 [−8.6, −3.8]	0.95 [0.90, 1.00]
Hispanic	−3.0 [−5.4, −0.5]	5.9 [1.9, 10.1]	1.02 [0.96, 1.08]
Asian/Pacific Islander	9.0 [6.6, 11.5]	17.5 [13.2, 22]	1.36 [1.24, 1.49]
Native American	11.2 [2.1, 21.2]	−3.8 [−10.5, 3.4]	0.98 [0.81, 1.18]
Other	11.0 [7.6, 14.5]	14.2 [8.6, 20.0]	1.51 [1.37, 1.66]
Median Income Quartile†
1st	Reference	Reference	Reference
2nd	0.9 [−0.5, 2.2]	3.8 [1.8, 5.9]	0.97 [0.93, 1.02]
3rd	−0.0 [−1.7, 1.6]	6.5 [3.9, 9.1]	0.87 [0.83, 0.92]
4th	−0.2 [−2.3, 1.9]	12 [8.6, 15.5]	0.82 [0.77, 0.87]
Insurance
Private	Reference	Reference	Reference
Public	11.9 [9.7, 14.2]	−6.6 [−8.7, −4.4]	0.99 [0.95, 1.03]
Other	5.9 [2.6, 9.3]	1.1 [−3.6, 5.9]	1.51 [1.42, 1.61]
Hospital Region
Northeast	Reference	Reference	Reference
Midwest	−6.9 [−11.2, −2.4]	4.4 [−6.9, 17.0]	1.34 [1.19, 1.51]
South	−1.7 [−6.6, 3.5]	0.2 [−10.0, 11.5]	1.65 [1.48, 1.84]
West	−0.3 [−5.1, 4.7]	44.2 [29.5, 60.6]	1.72 [1.54, 1.93]
Hospital Location/Teaching Status
Rural	Reference	Reference	Reference
Urban nonteaching	25.4 [15.7, 35.9]	22.3 [9.1, 37.2]	2.41 [1.73, 3.36]
Urban teaching	63.1 [50.8, 76.4]	137.7 [113.4, 164.9]	5.34 [3.88, 7.34]

^*Adjusted for age, sex, race/ethnicity, median household income by zip code, insurance type, hospital region, and hospital type/teaching status.

^†Calculated using median household income by patient ZIP code.

^‡Chronic kidney disease

^§Interaction was non-significant and removed from this model

^⁋We caution the reader against interpreting these coefficients for the main effect of CKD and the interaction by themselves since they are entangled with time. They are shown here for completeness, but the examples below of the impact of CKD vs. no CKD in each survey year are included to illustrate the magnitude of the interaction.

Costs

The median cost for discharges with CKD diagnoses was $8,755 (IQR: $4,563–18,345) compared to $5,016 (IQR: $2,860–10,109) in those without CKD (Table 2). The mean cost for all discharges with CKD was $1.33 billion (95%CI: 1.19–1.46 billion) per cohort year. Costs trended up over the subsequent cohort years, with CKD discharges rising significantly faster than the other chronic disease discharges (Figure S3). On average, discharges with CKD had 61.3% higher costs compared to those without CKD (Table 3). However, the significant interaction revealed the magnitude of this difference increased over time from 51.0% in 2006 to 72.5% in 2016.

Mortality

The proportion of in-hospital mortality in CKD-related discharges was 0.9% almost double that of other chronic disease discharges at 0.5% (Table 2). Discharges with a primary diagnosis of CKD accounted for 93% in-hospital mortality among the CKD group in the adjusted analysis. Discharges with CKD had 51% higher odds of death compared to discharges without CKD (Table 3). Those with a median family income in top two quartiles had lower mortality compared to the lowest income quartile. Over the four cohort years, mortality remained stable in CKD and non-CKD discharges (Figure S4).

Sensitivity Analyses

Discharges where CKD was the primary diagnosis had a higher LOS, cost, and mortality than both discharges with secondary CKD diagnosis and non-CKD discharges (Table S3). However, longer LOS was attributable to the discharges with a primary CKD diagnosis but not a secondary CKD diagnosis (Table S4). Discharges with both primary (67.4%, 95%CI: 62.8, 72.1%) and secondary (44.6%, 95%CI: 39.2, 50.3%) diagnoses of CKD had higher cost compared with discharges with no CKD diagnosis. In this sensitivity analysis, the odds of mortality remained high in discharges with a primary discharge diagnosis of CKD, but was lower than discharges without CKD when the CKD diagnosis was secondary (0.36, 95%CI: 0.29–0.46).

CKD stage was available for a small portion of the overall discharges with CKD (Figure 1). The difference in LOS between CKD and non-CKD discharges increased in a relatively stepwise fashion with advancing CKD stage ranging between 11–52.8% higher in CKD patients. Cost difference was lowest in CKD Stage 1 (36.1%, 95%CI: 11.9–65.6%) and highest in CKD Stage 5 or ESKD (133.2% 95%CI: 120.2–145.5%). CKD Stage 3 and higher, NOS and other CKD discharges had a higher odds of mortality compared with non-CKD discharges.

Figure 1.

Sensitivity Analyses* of Length of Stay, Cost, and Mortality by CKD† Stage

*Adjusted for survey year, age, race, hospital region, income by ZIP code, hospital characteristics, and insurance type; †Chronic kidney disease; ‡Not otherwise specified

To better understand the nature of individual-level medical complexity on the effects noted in the primary analysis, additional models for LOS, cost, and mortality were created with the addition of the number of CCCs as a variable. With the inclusion of this variable, there was no residual impact of CKD on LOS (−1.9, 95%CI 4.5–0.7%) a smaller difference in cost (16.4%, 95%CI 12.8–20.2%) and an inverse association between CKD and mortality (OR=0.72, 95%CI=0.67–0.78) (Table S5). When adjusting for number of CCC and restricting the analysis to the 1,259,290 observations with at least one CCC, mortality findings remained similar but there were also significant inverse relationships with CKD and both LOS (−12.4, 95%CI −14.3 to −10.4%) and cost (−13.3, 95%CI −15.1 to −11.4%) (Table S6).

DISCUSSION

This study represents a national evaluation of pediatric hospitalizations with CKD in the US. Discharges with CKD accounted for over 60,000 discharges per cohort year, $1.3 billion in cost per cohort year, and more than 400,000 hospital days per cohort year. Compared to this study, previously reported data on kidney disease-related admissions among chronically ill children have shown shorter LOS and charges.¹⁹ This is likely due to the more comprehensive inclusion of kidney disease-related diagnoses in this study’s CKD case definition compared to previous work. Importantly, this study also evaluates CKD discharges relative to other chronic illnesses. Pediatric discharges with CKD in the primary or a secondary discharge diagnosis location had a 30% longer LOS, 61% higher cost, and a 51% increased odds of mortality compared to discharges without CKD. Discharges with CKD as the primary diagnosis appear to be driving the majority of these associations. Discharges with CKD as a secondary diagnosis were associated with 45% higher cost, similar LOS, and lower mortality than the no CKD chronic illness group. Lastly, after adjusting for the number of CCCs as a measure of patient complexity, there was no residual positive impact of CKD on any of the three outcomes. This finding suggests that the higher LOS, cost, and mortality seen in the primary analysis were driven by a higher degree of medical complexity among discharges with CKD compared to the other chronic conditions included in this analysis. A similar finding was recently reported in internal medicine practice.²⁰

Our results add to the existing knowledge of hospital utilization, and outcomes of children with chronic illnesses (Table 4).^21–23 Discharges with CKD have one of the higher point estimates of LOS compared with other chronic illnesses.^21–25 Similarly, CKD discharges had higher point estimates for cost and charges compared to other chronic illnesses with the exeption of similar estimates with immune thrombocytopenia and heart failure.^{24, 25} The costly provision of care for hospitalized CKD patients with end-stage kidney disease, including dialysis, transplantation, and associated complications may be comparable to hospitalized heart failure patients, including procedures such as ventricular assist devices, extracorporeal membrane oxygenation, and heart transplantation. In-hospital mortality for children with chronic illnesses has been reported infrequently, however CKD discharges had higher mortality than other reported chronic conditions with the exception of heart failure.

Table 4.

Comparative LOS, Charges, Cost, and in-hospital mortality of selected pediatric chronic diseases

	Length of Stay (Days)	Charges	Cost	Mortality
Chronic Kidney Disease				0.9%
Weighted Mean	6.7	$65,199	$21,353
Median	2.8	$26,243	$8,755
Lupus Nephritis10
Weighted Mean	5.5	$43,100
Median
Nephrotic Syndrome11,12				0.5%
Weighted Mean	5	$26,491
Median
Sickle Cell Disease21
Weighted Mean
Median	3	$10,691
Asthma22				0.01–0.06%
Weighted Mean
Median	1.5	$8,410
Inflammatory Bowel Disease23
Weighted Mean	6
Median		$10,176–11,836
Immune Thrombocytopenia24				0.3%
Weighted Mean	3.8
Median			$7,477–9,328
Heart Failure25				6.5–7.2%
Weighted Mean	14–18	$63,461–72,082
Median

In children with specific kidney diseases, the reported mean LOS for children with nephrotic syndrome and lupus nephritis were lower than CKD at 5 and 5.5 days, respectively (Table 4).^{10, 11} Charges of pediatric CKD discharges were more than double compared to discharges with nephrotic syndrome and about 55% higher than discharges with lupus nephritis.^{10, 11} Within subgroups of kidney disease, the reported 0.5% mortality for nephrotic syndrome associated discharges was lower than the all CKD associated discharges (0.9%) in our study.¹²

When primary and secondary CKD diagnoses were stratified, discharges with a secondary diagnosis of CKD had lower mortality compared to the non-CKD chronic disease discharges. This inverse association with mortality is of unclear etiology. This analysis was limited to diagnosed CKD, and unidentified pediatric CKD was not included. It is possible that children with CKD not known to the medical team have different outcomes. Presumably, children diagnosed with CKD before hospitalization are more likely to be seen by a pediatric nephrologist and receive appropriate surveillance and interventions, which may result in better outcomes. The degree to which undiagnosed CKD exists in the pediatric population and outcomes related to early identification is unknown. Additionally, pediatric hospitalizations with coexisting CKD cared for by non-nephrology services may be more inclined toward early nephrology consultation and closer monitoring of lab values. Inpatient consultation patterns in pediatric nephrology are unknown.

This study provides evidence that the level of medical complexity may be the driver of the higher LOS, cost, and mortality identified in pediatric CKD discharges. A recent study from the UK reported that adult kidney disease patients have a greater degree of medical complexity than patients seen by any other specialty.²⁰ While the etiologies of disease in pediatric and adult kidney disease are different, the causes of CKD in children include genetic disorders, congenital anomalies that may be part of a multi-organ system syndrome, and systemic inflammatory disorders. These diverse and complex primary etiologies, as well as CKD related to prematurity and repeated acute kidney injury in other pediatric chronic disease, make this finding of higher complexity is plausible in children. Further investigation into the details of medical complexity, the specific additional complex chronic conditions that are most responsible for higher health care utilization, and poorer outcomes is needed.

In multivariable analysis, median income in the two highest quartiles had lower mortality compared to the lowest income quartile. This inverse association between median income and mortality has been reported in one other study of hospitalized children.²⁶ The study found variability in this association based on clinical service, with services such as respiratory, hematology and oncology, and infectious disease having no association base on income while services such as cardiology, neonatology, and gastroenterology demonstrating this inverse association. While it is likely that hospitalizations represented in our study included inter-disciplinary teams, this association has not previously been reported specifically in kidney disease-related hospitalizations.

The strengths of this study are the national representative and large sample, as well as the inclusion of both community and academic hospitals. We acknowledge limitations to our study. The ICD-9 to ICD-10 transition in 2015 may make trends, including 2016 more difficult to interpret. Although we present longitudinal data, not all years between 2006–2016 are represented. The study unit for this investigation was discharges rather than individuals. Consequently, patients could be included more than once without opportunity for adjustment. We did not have access to treatment level data, including the need for intensive care, medication administration, laboratory data, or physician services providing care (e.g., general pediatrics, medical subspecialty, surgical subspecialty). As diagnosis codes were utilized for CKD identification, we were not able to evaluate discharges with unrecognized CKD. Due to this, CKD, particularly lower stages, may have been under-reported in hospital record discharge diagnosis codes, which may underestimate admissions complicated by CKD.²⁷ Finally, as this data include discharge diagnoses only, there was not reliable way to assess dialysis provision.

In four cohort years between 2006–2016, 250,000 discharges in the US were due to or complicated by pediatric CKD. Pediatric hospitalizations caused or complicated by CKD were associated with longer LOS, higher hospitalization-associated cost, and increased odds of mortality compared to hospitalizations without diagnosed CKD. These outcomes seem to be due to the higher complexity of CKD discharges compared to discharges with other chronic illnesses. Investigation is needed to identify modifiable patient characteristics and health care delivery with the aim of developing and testing interventions to reduce the adverse health outcomes of pediatric CKD in the US. It is likely that the costs reported represent only a small portion of pediatric CKD expenditures as much of CKD care is performed on an outpatient basis. Further investigation into direct medical and individual costs to families of children with CKD are needed fully grasp the economic burden that pediatric CKD has on families and the health care system at large.