Trends in Resource Utilization by Children with Neurological Impairment in the United States Inpatient Health Care System: A Repeat Cross-Sectional Study

Jay Berry and colleagues report findings from an analysis of hospitalization data in the US, examining the proportion of inpatient resources attributable to care for children with neurological impairment.


Introduction
Neurological impairment (NI) comprises a heterogeneous group of static and progressive health conditions that involve the central and peripheral nervous systems and result in functional and/or intellectual impairment. This group includes children with epilepsy, infants who are born prematurely and have hypoxicischemic injury to the brain, and children with genetic and metabolic disorders that affect the nervous system.
Although the impact that children with NI have on the pediatric health care system remains largely unknown, emerging evidence describes epidemiologic and care delivery phenomena that suggest this impact may be substantial. The rising prevalence and improved survival of extremely premature infants was associated with an increased risk of cerebral palsy during the 1990s [1,2]. Children with spina bifida, muscular dystrophy, and other NIrelated diagnoses are also surviving longer [3,4], in part because of improved co-morbid condition management of gastroesophageal reflux disease, oromotor dysfunction, and chronic lung disease [5,6].
In the US, many children with NI receive uncoordinated, crisisdriven care that is believed to contribute to excessive health care utilization and cost [7][8][9][10][11]. They have a disproportionately high use of emergency care health services, and they experience frequent hospitalizations, with subsequently high readmission rates [12,13]. When hospitalized, they tend to stay longer and have more expensive admissions than other children [14]. It is estimated that a large proportion of lifetime health-related costs for a children with severe NI is attributable to inpatient care [15].
This evidence and our clinical experience suggest that children with NI may account for an increasing proportion of hospital resources. However, this assumption has not been tested at a national level for the US. We undertook this study to (1) evaluate national trends in hospitalizations for children with NI over time within different types of hospitals and (2) describe the patient characteristics and reasons why children with NI are hospitalized.

Ethics Statement
This study was approved by the Children's Hospital Boston Institutional Review Board with a waiver for informed consent. The investigation was conducted according to the principles expressed in the Declaration of Helsinki.

Study Design and Setting
This is a retrospective, repeat cross-sectional analysis of the Healthcare Cost and Utilization Project Kids' Inpatient Database (KID) for the years 1997, 2000, 2003, and 2006. The KID is a multi-state database of US hospitalizations for children aged 0-18 y [16,17]. From 1997 to 2006, the KID sample increased from 1.9 million hospital discharges from 2,521 hospitals in 22 states to 3.1 million hospital discharges from 3,739 hospitals in 38 states [16,17]. The dataset includes a weight variable for each observation that was used to produce national estimates of inpatient resource utilization for children with specific diagnoses while accounting for differences in the KID hospital sampling frame and oversampling of patient sub-populations that could have occurred over time.

Study Population
We identified hospitalizations for children aged 0-18 y with NI as determined by International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnostic codes. Up to 15 separate ICD-9-CM diagnostic codes are available for each hospitalization in the KID [17]. We assembled an initial list of 606 NI ICD-9-CM codes from previous studies and the tabular index of ICD-9-CM coding [18][19][20].
Pediatric neurologist review of NI ICD-9-CM codes. Two board-certified pediatric neurologists (A. P. and J. L. B.) from different institutions reviewed the initial NI code list [21]. They considered candidate NI ICD-9-CM codes using the following definition: ''diagnosis is consistent with NI (static or progressive) and typically results in either functional and/or intellectual impairment'' ( Figure 1) [19,20]. The reviewers independently classified each code as ''yes,'' (diagnosis consistent with NI), ''no'' (diagnosis not consistent with NI), or ''maybe'' (diagnosis may be consistent with NI). Code consensus was defined as classification agreement between the reviewers. The reviewers then discussed the codes for which there was classification disagreement (n = 276 [46%]) until consensus was reached (n = 265 [96%]) or was not reached (n = 11 [4%]). A third pediatric neurologist (F. Filloux) classified the remaining non-consensus codes.
The NI restricted codes were then compared to a list describing coding changes over time assembled by the National Center for Health Statistics and the Centers for Medicare and Medicaid Services. Eleven NI ICD-9-CM codes changed from 1997 to 2006. We incorporated these code changes into our analysis. NI-related hospitalizations were identified from the presence of one or more NI ICD-9-CM codes in any of the discharge diagnoses. The NI ICD-9-CM codes are presented in Table S1.

Main Outcomes
We evaluated the number of hospitalizations, total number of days spent in the hospital, and total charges for admissions associated with the NI inclusive and NI restricted admissions for each year.
Hospital type. We analyzed the measures of NI inpatient resource utilization within all hospitals, children's hospitals, and non-children's hospitals [16,17]. ''Children's hospital'' designation was given to hospitals based on specifications from the National Association of Children's Hospitals and Related Institutions. Children's hospitals included freestanding children's hospitals, children's hospitals within adult general hospitals, and children's specialty hospitals.
Reasons for hospitalization. We evaluated patient demographic and clinical characteristics that may correlate with trends in hospital resource use for children with NI over time. Patient demographic variables included patient age (,1 y, 1-4 y, 5-12 y, and 13-18 y) and insurance type (public, private, and other). Race/ethnicity was described but not analyzed because of the extent of missing data (17%-28% for the years under study).
We also analyzed each NI hospitalization to determine if an NIrelated surgical procedure occurred. These procedures were classified based on ICD-9-CM procedure codes from previous literature on surgical procedures that these children frequently encounter [6,14,20,22,23]. We specifically assessed hospitalizations where a gastrostomy (codes beginning with 43.1 or 44.32) and/or fundoplication (codes 44.66 or 44.67) occurred.
We analyzed the principal reason for each hospitalization in children with NI based on the ICD-9-CM Major Diagnostic Category (MDC). MDCs represent 25 mutually exclusive diagnosis domains primarily organized by organ system (nervous system, respiratory system, circulatory system, etc.). One MDC is assigned for each hospitalization from the patient's principal ICD-9-CM diagnosis code. The ''newborns and other neonates'' MDC includes newborns cared for in a neonatal intensive care unit or in a healthy newborn unit of a maternity hospital.

Statistical Analyses
We evaluated national inpatient resource utilization trends for children with and without NI using all of the years of KID data with a Mantel-Haenszel chi-square test. We used SAS version 9.1.2 (SAS Institute) survey procedures to account for the changing KID sampling frame through a weight variable calculated from the demographic characteristics of the hospitals contributing to the database (ownership/control, bed size, teaching status, children's hospital status, rural/urban location, and state/region) and their hospitalized patients. As the sampling frame expanded over time, the discharge weight decreased so that the increasing proportion of actual records did not falsely inflate the national estimates [24]. All patient-level data were clustered within each hospital and stratified by the demographic characteristics of each hospital to generate national estimates and associated variances of the proportion of inpatient resource utilization attributable to children with NI. Significance threshold was defined as p#0.05. We adjusted total hospital charges for inflation to 2006 US dollars [25,26].

Results
There were 25,747,016 hospitalizations for children aged 0-18 y identified from 1997, 2000, 2003, and 2006 combined. Of these, 1,338,590 (5.2%) hospitalizations were associated with children who had NI inclusive diagnoses, and 1,032,829 (4.0%) hospitalizations were associated with children who had NI restricted diagnoses (i.e., the NI codes classified as ''maybemaybe'' were excluded). NI inclusive results are presented throughout the Results, with NI restricted results displayed in the table and figures. The most prevalent diagnoses among all hospitalized children with NI were epilepsy (52.2% [n = 538,978]) and cerebral palsy (15.9% [n = 164,665]) ( Figure 2). The NI diagnoses varied by age: cerebral palsy was diagnosed in 1.1% (n = 4,708) of hospitalized infants with NI, compared with 17.4% (n = 102,590) of children aged 13-18 y.
Throughout the study period the proportion of hospitalized children aged 13-18 y that had NI increased from 7.3% to 9.9% (p,0.001) (

Children's Hospitals
Inpatient resource utilization. Throughout the study period, children with NI accounted for an increasing proportion of hospitalizations (11.7% in 1997 to 13.5% in 2006, p,0.001) within children's hospitals (Figure 3). They experienced a greater increase in admissions (

Discussion
The main findings from this study suggest that inpatient resource utilization for children with NI in the US has changed from 1997 to 2006. The changes were largely attributable to children with NI utilizing children's hospitals more over time. Within children's hospitals, they accounted for both increasing trends and a substantial proportion of resources over the study period, including nearly one-third of all hospital charges. Within all hospitals, infants and children with NI experienced a greater increase in neonatal and neurological admissions and a greater decrease in respiratory admissions than children without NI. The majority of admissions for children with NI occurred for nonnervous-system problems, suggesting that co-morbid conditions affecting other organ systems were a major contributor to inpatient utilization.
Our study contributes to existing literature by describing national inpatient utilization for children with NI who had underlying primary neurological diagnoses (e.g., cerebral palsy) or other diagnoses that are characteristically associated with comorbid nervous system impairment (e.g., Down Syndrome). As expected, studies restricted to a cohort of children with primary neuromuscular diagnoses report less inpatient utilization (e.g., 2% of all pediatric admissions and 7% of all pediatric hospital charges) than the present study [27]. Our findings are consistent with adult studies reporting that NI associated with intellectual disabilities and dementia ranks among the top determinants of health care cost [28,29]. Although the underlying diagnoses and reasons for NI may be different across the age spectrum, patients with NI may account for a substantial share of both pediatric and adult health care resources. There is a rising proportion of children's hospital use by children with NI, as the total number of hospitalizations for children without NI did not increase within children's hospitals over time. Most US child neurologists work within children's hospitals, and there has been an emergence of multi-disciplinary care coordination clinics that provide comprehensive care for children with NI at children's hospitals [30,31]. Children with NI may be utilizing children's hospitals more often to seek care from providers who are more familiar with and comfortable managing their health problems. Further study is necessary to evaluate whether clinician NI knowledge and care delivery differ between children's and non-children's hospitals.
US physicians have demonstrated deficiencies in working knowledge of the most common forms of NI, and negative provider attitudes toward this population are associated with substandard acute care practices [32,33]. Parents of adolescents with NI report this as a contributing factor that complicates the transfer of their child's health care from pediatric to adult providers [34]. In the present study, we observed a trend of rising hospitalizations in adolescent children with NI that was greater than that for children without NI. Further investigation is necessary to determine whether hindered care transition is contributing to this trend.
The rise in the number of admissions associated with gastrostomy with and without fundoplication operations in children with and without NI over time may represent a change in approach to more interventions to bypass oromotor dysfunction, mitigate aspiration from oral feeds, and improve gastroesophageal reflux management in these children. The appropriateness and effectiveness of these operations in children with NI are currently under investigation. It is possible that the decrease in the absolute number of respiratory admissions for children with NI could be related to the rise in gastrostomy and/or fundoplication operations: future studies should investigate whether these operations help prevent pneumonia admissions due to reflux and aspiration [35]. We were unable to evaluate this directly in the present study, as the KID does not track individual patients before and after operations.
In the present study, the absolute number of respiratory admissions decreased for children both with and without NI. Prior small area studies report that children with intellectual disability are at higher risk for respiratory hospitalizations that may be ambulatory care sensitive than children without intellectual disability [36]. Our study suggests that, nationally, children with NI may have received an equivalent or greater benefit from efforts to improve respiratory heath in children, including better respiratory preventive measures and acute and chronic respiratory care treatments [37,38]. It is possible that some children with NI could have gained exposure to these efforts through improved care coordination within medical homes or other structured clinical programs [31].
The increasing trend of neonatal admissions for infants with NI is supported by prior studies that report a rising incidence of NI associated with increasing preterm infant deliveries and preterm infant survival [1]. However, some recent studies show that the neonatal prevalence of certain NI-related diagnoses, such as cerebral palsy, is now decreasing [39]. This may be related to perinatal care improvements that minimize the risk of brain injury, as well as to changing practices associated with prenatal congenital malformation screening and elective termination [40]. Using inpatient administrative billing codes to detect NI among neonates limits the interpretation of the trends observed in the present study. Some infants may not be diagnosed with NI until well after their neonatal hospitalization.
Our study has several other limitations. We were unable to determine whether the trends could have been a function of the changing population prevalence of children with NI at large in the US. US census data indicate that the population of all children aged 0-18 y increased by 5.6% from 1997 to 2006. However, to our knowledge, true population prevalence trends of all children with NI remain unknown.
The large shift in the inclusion of hospitals from different states in the KID sampling frame could have affected the study findings. In a post hoc analysis, we restricted the cohort to patients admitted to hospitals that were present in both the 1997 and 2006 KID samples. Similar results were observed. The KID contains hospital discharge rather than individual patient data, which prevented us from being able to determine whether the same sized population of children with NI was being hospitalized more frequently over time. This will be important to evaluate in future health resource utilization and epidemiologic studies of children with NI.
Continuous years of data may be preferable for studying hospital utilization trends when compared with four discrete time points in a 10-y period. The hospitals within the KID used ICD-9-CM coding for all the years of the study. It will be important to evaluate the identification of NI diseases using ICD-10 and other coding systems, as many countries are currently using or implementing them. Errors in ICD-9-CM coding exist, and billing reimbursement may influence which codes are entered for a patient during hospitalization. In some cases, NI may be regarded as a subjective medical condition that cannot be correlated with a specific diagnosis code. As such, the hospital resource use described in the present study may underrepresent children with NI who are hospitalized with that situation. The KID contains hospital charge data, but not cost data. Charge data represent the amount that hospitals billed for services but does not reflect how much hospital services actually cost or the specific amounts that hospitals received in payment.
Despite these limitations, this study demonstrates that children with NI account for a substantial proportion of inpatient resources utilized in the US. Their impact is growing within children's hospitals. Use of large administrative databases like the KID, by itself or in conjunction with other databases, may provide unique perspectives to answer the clinical questions raised by the study findings, including the following. (1) Does care quality differ for children with NI depending on the type of hospital they use? (2) Does inpatient resource utilization for children with NI vary depending on their access to medical homes and preventive care? And (3) How does substandard transition of health care services from pediatric to adult health care providers affect hospital use for adolescents and young adults with NI?
As a higher proportion of hospital resources are accounted for by children with NI over time, children's hospitals, in particular, will need to ensure that (1) adequate clinical and coordinated expertise is focused on the needs of these children, (2) NI clinical assessment and care management training is developed for trainees and junior graduates in pediatric postgraduate educational programs, (3) partnerships between families of children with NI and hospitals are developed and implemented, and (4) care treatment strategies of both nervous and non-nervous-system problems are rigorously evaluated for these children. These system-based efforts have the potential to promote continued quality of care for children with NI. We must ensure that the current health care system is staffed, educated, and equipped to serve, with efficiency and quality, this growing segment of vulnerable children.

Supporting Information
Table S1 Neurological impairment ICD-9-CM codes used in this study. (DOC)