BACKGROUND: The relationship between limited English proficiency (LEP) and worse pediatric health outcomes is well documented.
OBJECTIVES: To determine the relationship between LEP status and pediatric hospital readmissions.
METHODS: We performed a retrospective cohort analysis of children ≤ 18 years old admitted to a tertiary children’s hospital from 2008 to 2014. The main exposure was LEP status. Independent variables included sex, age, race/ethnicity, insurance, median household income, surgical/medical status, severity of illness (SOI), the presence of a complex chronic condition, and length of stay. Primary outcome measures were 7- and 30-day readmission.
RESULTS: From 67 473 encounters, 7- and 30-day readmission rates were 3.9% and 8.2%, respectively. LEP patients were more likely to be younger, poorer, and Hispanic; have lower SOI; and government-subsidized insurance. Adjusted odds for 7- or 30-day readmission for LEP versus English-proficient (EP) patients were 1.00 (P = .99) and 0.97 (P = .60), respectively. Hispanic ethnicity (adjusted odds ratio [aOR]: 1.26 [P = .002] and 1.14 [P = .02]), greater SOI (aOR: 1.04 [P < .001] and 1.05 [P < .001]), and the presence of a complex chronic condition (aOR: 2.31 [P < .001] and 3.03 [P < .001]) were associated with increased odds of 7- and 30-day readmission, respectively. White LEP patients had increased odds of 7- and 30-day readmission compared with white EP patients (aOR: 1.46 [P = .006] and 1.32 [P = .007]) and the poorest LEP patients had increased odds of 7- and 30-day readmission compared with the poorest EP patients (aOR: 1.77 [P = .04] and 2.00 [P < .001]).
CONCLUSIONS: This is the first large study evaluating the relationship between LEP and pediatric hospital readmission. There was no increased risk of readmission in LEP patients compared with EP patients.
Pediatric hospital readmissions pose a significant burden to patients, their families, and the health care system. Readmissions within 30 days account for 2% to 6% of the >2 million hospital admissions for children each year, and >20% of hospitalized pediatric patients will experience at least 1 readmission in 1 year.1–3 These admissions incur indirect costs and opportunity loss on individual patients and families, in addition to their psychosocial and physical impacts.4 Readmissions also constitute a financial burden on the national health care system, with an estimated total cost of >$1 billion annually for children readmitted within 1 year.2
Due to the negative effects of hospitalization on children, their families, and the health care system as a whole, reducing hospital readmissions has become a focus for quality-improvement efforts. In 2012, the Centers for Medicare and Medicaid Services began decreasing Medicare payments to hospitals with excess readmissions.5 More recently, there has been an increasing focus on pediatric readmission rates, because it is 1 of the first measures that will be developed by the Pediatric Quality Measures Project (established by the federal Children’s Health Insurance Program Reauthorization Act),6 with several state Medicaid programs already penalizing hospitals with excessive readmissions.7–10
Research to date has shown that an increase in pediatric hospital readmission rates is associated with complex chronic health conditions,2,11,12 longer hospital length of stay (LOS),11–13 and socioeconomic factors, such as Medicaid insurance use and black race.2,11 Despite this growing literature investigating the relationship between socioeconomic factors and pediatric hospital readmission rates, there are no studies evaluating limited English proficiency (LEP) status of the caregiver and patient as a potential risk factor for readmission, even though LEP status has been shown to be a risk factor for other poor health outcomes, including increased rates of serious adverse medical events,14 worse postoperative pain management,15 unnecessary admissions from and longer LOS in the emergency department,16,17 longer hospitalizations,18 and poorer access to health care.19 Even in the adult literature there are limited data on LEP and hospital readmission, with 1 study showing no difference in readmission on the basis of LEP status20 and another study showing a difference among specific ethnicities.21
This investigation seeks to address an existing knowledge gap in the current literature. If LEP status is found to be a risk factor for pediatric hospital readmissions, future interventions designed to improve care coordination for LEP families at the time of discharge may reduce readmission rates. On the basis of the known risk factors associated with pediatric hospital readmission and the association between LEP status and health outcomes, we aimed to determine whether pediatric patients with LEP in a tertiary hospital setting have increased hospital readmission rates compared with English-proficient (EP) patients.
Study Design and Data Source
We conducted a retrospective database analysis of all patients aged ≤18 years admitted to Lucile Packard Children’s Hospital (LPCH) at Stanford University Medical Center from January 1, 2008, to May 2, 2014. LPCH is a free-standing children’s hospital that provides tertiary care in a university-based setting.
LPCH has in-person interpretation for Spanish, Cantonese, Mandarin, Vietnamese, Korean, and American Sign Language. Through a vendor, there is access to phone interpretation for >200 languages. In addition, there is document translation that requires advance notification.
We extracted all data by using the Stanford Translational Research Integrated Database Environment (STRIDE). STRIDE consists of 3 integrated components: a clinical data warehouse based on the HL7 Reference Information Model, which contains clinical information on >1.6 million pediatric and adult patients cared for at Stanford University Medical Center since 1995; an application development framework for building research data management applications on the STRIDE platform; and a biospecimen data management system. Oversight for STRIDE is provided by Stanford’s Department of Biomedical Informatics.22 The institutional review board at Stanford University approved the study, and a waiver of consent was granted.
We defined LEP status for a patient as (1) the patient or guardian selecting a non-English language as the primary language on admission, (2) documentation of in-person interpreter use during the hospitalization, and/or (3) having “language barrier” selected as a barrier to education during the initial nursing assessment. Patients who did not satisfy any of these criteria were defined as EP. We excluded patients if they were admitted to labor and delivery, left against medical advice, transferred to another acute care hospital, died during the admission, or had missing data for the outcome measures or covariates.
We evaluated readmissions within 7 and 30 days. Thirty-day readmission is the metric used by the Centers for Medicaid and Medicare Service. We included the 7-day readmission as well because a shorter interval of time between readmissions may be more indicative of issues related to the index admission and transition of care, whereas longer intervals may reflect issues with chronic outpatient management.23 Elective admissions (eg, chemotherapy, scheduled infusions, and elective surgeries) could not be considered a readmission. Admissions were labeled as elective or urgent/emergent in the electronic medical record. Administrators in patient placement and the admitting office and inpatient and outpatient schedulers had the ability to input the data into electronic medical records. Index admissions could only be associated with 1 readmission within the designated time intervals. An encounter could be considered a readmission from a previous encounter and an index admission for a subsequent encounter.
We determined a priori those covariates most likely to be associated with our outcome measure. These include age, sex, race/ethnicity, insurance status, median income quintile, medical versus surgical status, All Patient Refined Diagnosis Related Groups (APR-DRG) severity of illness (SOI) scoring, presence of a complex chronic condition (CCC), and hospital LOS. We categorized age as <1 year, 4 to 9 years, 5 to 12 years, and 13 to 18 years.
We used data from the US Census Bureau’s American Community Survey 5-year estimates24,25 encompassing the years 2008–2012 to estimate median household income by zip code tabulation area (ZCTA). Administrative data from STRIDE include zip code and not ZCTA; therefore, zip codes were converted to ZCTAs by using data provided by the Department of Health and Human Services.26 We created income quintiles on the basis of 2011 income data from the US Census Bureau’s Current Population Survey.27
We determined if an admission was surgical or medical on the basis of 2015 Procedure Classes software developed as part of the Healthcare Utilization Project, a federal-state-industry partnership and sponsored by the Agency for Healthcare Research and Quality.28 The Procedure Classes program assigns all International Classification of Diseases, Ninth Revision, procedure codes into 1 of 4 categories: minor diagnostic, minor therapeutic, major diagnostic, and major therapeutic. Minor procedures are non–operating room procedures, whereas major procedures are operating room–based procedures. Patients with no procedure codes or only minor procedures were designated “medical,” whereas patients with major procedure(s) were considered surgical.
To control for disease severity, we used APR-DRG versions 25 through 31 (3M Health Information Systems) SOI levels. APR-DRGs include mutually exclusive groupings of clinically related diagnosis and procedural codes from the International Classification of Diseases, Ninth Revision, Clinical Modification.29 We then used publicly available service intensity weights created by the New York State Department of Health based on costs from 3 years of cases (2009–2011)30,31 to transform the categorical APR-DRG SOI levels into a continuous variable (from 0.1287 to 29.228) that can be compared across APR-DRGs.
To define CCCs, we used an open-source set of International Classification of Diseases, Ninth Revision, diagnosis and procedure codes related to childhood complex chronic health conditions that are strongly associated with mortality, morbidity, functional limitations, high health resource utilization, and use of a complex care clinical program.32,33 This tool has been used in previous studies to predict readmission rates in children with chronic conditions.11,34
For all statistical analyses we used Stata/SE 14.0 (StataCorp, College Station, TX) and SAS Enterprise 6.1 (SAS Institute, Cary, NC). We defined primary outcome measures as 7-day and 30-day readmission rates. We summarized patient characteristics by using frequency and percentages for categorical variables. We performed univariate logistic regressions on each variable that was determined a priori to be a potential confounder.
To assess the association between LEP and the primary outcome of 7-day and 30-day readmission rates, we performed multivariable logistic regression, controlling for patient characteristics as listed above in our a priori–determined covariates. In our model, we adjusted for patients (medical record numbers) having multiple encounters by using the clustered by robust SE method. The results are presented as adjusted odds ratios (aORs). P values <.05 were considered significant.
We hypothesized that the effect of LEP on pediatric hospital readmissions may have strata-specific variation, as seen in previous studies.21,35,36 We tested for interactions between our outcome variables and age, income quintile, race/ethnicity, weighted APR-DRG SOI score, CCC, LOS, and surgical/medical admission by including interaction terms one at a time in separate adjusted models. Variables to test for interaction were determined a priori. We then examined and presented stratum-specific odds ratios for all variables in which the interaction P value was <.10.
Over the study period there were 71 918 total admissions. We excluded 1904 of these admissions because the patient’s disposition was not to home, and another 1637 admissions due to missing data. We analyzed the remaining 67 473 admissions for 49 002 unique patients during the study period. A total of 17 754 (26.3%) admissions were associated with LEP families and 49 719 (73.7%) admissions were associated with EP families (Fig 1). Table 1 lists the demographic and clinical characteristics of the 2 cohorts in the study population. For all encounters, there were 2536 readmissions (3.8%) within 7 days and 5288 readmissions (7.8%) within 30 days. English was reported as the primary language in 81.5% of all encounters in the study population. The most common non-English language spoken was Spanish, which was reported in 16.1% of all encounters and in 61.0% of LEP encounters in the study population.
In univariate analyses, we found that age <1 year, Hispanic ethnicity, lower income quintile, having a CCC, and higher weighted APR-DRG SOI score were associated with an increased odds of readmission within 7 and 30 days (Table 2). The unadjusted odds ratios for 7- and 30-day readmissions for LEP patients were 1.17 (95% confidence interval [CI]: 1.05–1.29; P = .003) and 1.15 (95% CI: 1.06–1.26; P = .002), respectively, compared with EP patients. After controlling for potential confounders in our multivariable logistic regression, LEP was not associated with increased odds of readmission at 7 or 30 days (Table 3). In the final model, age, Hispanic ethnicity, government insurance, medical patients, and having a CCC were associated with increased odds of readmission at 7 and 30 days. For every 1-unit increase in weighted APR-DRG SOI score there was a 4% and 5% increase in odds of readmission at 7 and 30 days, respectively. There was an increased odds of 30-day readmission with increasing hospital LOS but not with 7-day readmission.
For effect modification, we found statistically significant interaction terms (P ≤ .10) at both 7 and 30 days for the following variables: race (.02, .03) and income (.03, <.001). Compared with white EP patients, white LEP patients had aORs of 1.46 (95% CI: 1.11–1.92) and 1.32 (95% CI: 1.08–1.61) for 7- and 30-day readmissions, respectively. LEP patients in the combined first and second income quintile (lowest quintiles) had aORs of 1.77 (95% CI: 1.03–3.05) and 2.00 (95% CI: 1.37–2.92) for 7- and 30-day readmissions, respectively, compared with EP patients in the combined first and second income quintile (Table 4).
To our knowledge, this is the first study to evaluate the relationship between LEP status and pediatric hospital readmission rates. Our analysis found no increased risk of readmission at 7 or 30 days on the basis of LEP status for all patients after adjusting for covariates. Our rates of readmission at 7 and 30 days were comparable to those of previous studies in pediatric patients.3 Similar to previous studies, after adjusting for potential confounders, we found increased odds of readmission for patients with a CCC, public insurance, and hospital LOS.2,311,12 The finding of a small increase in the odds of readmission for Hispanic patients, but not for black patients, differs from previous studies, which found a statistically significant odds of readmission for black patients but not for Hispanic patients.11
Although our primary analysis did not reveal an association between LEP and hospital readmission, our secondary effect modification analysis did reveal vulnerable populations among LEP patients. We found that LEP status had a larger effect on white patients compared with other races and ethnicities. This finding may be due to the fact that these patients are more likely to speak a non-English, non-Spanish language and that non-Spanish language interpretation may be more difficult to access in general. Although not tested, this clinically relevant disparity in access to interpretation may amplify poorer communication during discharge and transition of care. Our analyses also revealed that, among low-income patients, being an LEP patient was associated with an increase in hospital readmissions compared with low-income EP patients. This finding implies that LEP status affects readmission among the poorest patients more than among middle- and upper-class patients. To reduce pediatric hospital readmissions, it may be effective to focus our attention and resources to understanding the underlying causes of increased hospital readmissions and develop interventions targeted at low-income LEP patients and white LEP patients.
We acknowledge limitations of our study. Our institution serves a community with a higher percentage of Latinos, which may not be generalizable to other areas in the country. Being in a community with a large Spanish-speaking population may provide benefit to our patient population through better access to direct and translated communication, through the Spanish-speaking hospital staff, including, but not limited to, the admitting staff, bedside nurses, physicians and ancillary staff. Due to the large Spanish-speaking population, the hospital provides 24-hour, in-person Spanish interpreter services, which may not be available at other hospitals.
As a single-center retrospective investigation, we were unable to capture readmissions that may have occurred elsewhere.37 In addition, our definition of LEP was based on the subjective opinion of the patient/guardian, admitting personnel, and bedside nursing staff. Due to the subjective nature of this variable, patients who had LEP may have been labeled as EP, leading to a minimization of a true effect that LEP status had on odds of readmission.
In choosing to use APR-DRG SOI subclass as a measure of SOI, we needed to convert the data to a weighted score, because an SOI level is not equivalent across different APR-DRGs. To do so, we used publicly available weights developed on a different geographic population in New York State, which uses both pediatric and adult data and therefore may not accurately reflect the SOI of our patient population. Despite this situation, we believe that we were able to adequately capture the SOI of our patient population, and that SOI is important factor to consider in hospital readmissions. Last, there may be other recognized and unrecognized confounders that we were unable to extract from the data available to us, such as maternal educational level and other measures of socioeconomic status.
LEP was not associated with increased odds of readmission to a single-center tertiary pediatric hospital. Hispanic ethnicity and high resource utilization (weighted APR-DRG SOI level and CCCs) were independently associated with increase odds of readmission.
We thank Dr Suzan Carmichael, PhD, and Ms Peiyi (Peggy) Kan, MS, of the Department of Pediatrics Research and Statistical Unit at Stanford for assistance in statistical analysis. We also thank Dr Alan Schroeder, MD, for his invaluable advice and feedback during the production of this manuscript.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: All phases of this study were supported by the Jackson Vaughan Critical Care Research Award (Dr Ju) and the National Institutes of Health (K08 DK094868A) (Dr Park). Funded by the National Institutes of Health (NIH).
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Bardach NS,
- Vittinghoff E,
- Asteria-Peñaloza R,
- et al
- ↵Centers for Medicare and Medicaid Services. Readmission Reduction Program (HRRP). 2014. Available at: www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/AcuteInpatientPPS/Readmissions-Reduction-Program.html. Accessed March 9, 2016
- ↵CHIPRA Measures by CHIPRA Categories—Initial Core Set and PQMP COE Measure Assignments. 2012.
- ↵New York State Health Foundation. Reducing hospital readmissions in New York State: a simulation analysis of alternative payment incentives. 2011. Available at: http://nyshealthfoundation.org/uploads/resources/reducing-hospital-readmissions-payment-incentives-september-2011.pdf. Accessed March 9, 2016
- MassHealth. Payment for out-of-state acute hospital services, and in-state acute hospital services. 2014. Available at: www.mass.gov/eohhs/docs/masshealth/acutehosp/2014-notice-final-payment-acute-hospital-services.pdf. Accessed March 9, 2016
- Texas Medicaid and Healthcare Partnership. Potentially preventable events. 2014. Available at: www.tmhp.com/Pages/Medicaid/Hospital_PPR.aspx. Accessed March 9, 2016
- ↵Illinois Department of Healthcare and Family Services. Potentially preventable readmissions policy. 2014
- Feudtner C,
- Levin JE,
- Srivastava R,
- et al
- Neuman MI,
- Hall M,
- Gay JC,
- et al
- Cohen AL,
- Rivara F,
- Marcuse EK,
- McPhillips H,
- Davis R
- Jimenez N,
- Jackson DL,
- Zhou C,
- Ayala NC,
- Ebel BE
- López L,
- Rodriguez F,
- Huerta D,
- Soukup J,
- Hicks L
- ↵US Census Bureau. American FactFinder. Available at: http://factfinder.census.gov/faces/nav/jsf/pages/searchresults.xhtml?refresh=t. Accessed July 18, 2016
- ↵US Department of Commerce Economics and Statistics Administration. A Compass for Understanding and Using American Community Survey Data: What General Data Users Need to Know. Washington, DC: US Government Printing Office; 2008
- ↵Department of Health and Human Services. Zip code to ZCTA crosswalk. In: Uniform Data System Resources from the Health Resource and Services Administration. 2014
- ↵US Census Bureau. Historical income tables: household. Available at: www.census.gov/hhes/www/income/data/historical/household/. Accessed November 4, 2015
- ↵Healthcare Cost and Utilization Project. Procedure classes. 2015. Available at: https://www.hcup-us.ahrq.gov/toolssoftware/procedure/procedure.jsp. Accessed August 26, 2015
- ↵3M Health Information Systems. 3M APR DRG software. Available at: http://solutions.3m.com/wps/portal/3M/en_US/Health-Information-Systems/HIS/Products-and-Services/Products-List-A-Z/APR-DRG-Software/.
- ↵New York State Department of Health. APR-DRG service intensity weights and average length of stay. July 1, 2014.
- ↵New York State Department of Health. Acute hospital inpatient rebasing and service intensity weights (SIWs) effective July 1, 2014.
- Mendu ML,
- Zager S,
- Moromizato T,
- McKane CK,
- Gibbons FK,
- Christopher KB
- Wang L,
- Haberland C,
- Thurm C,
- Bhattacharya J,
- Park KT
- Davies SM,
- Saynina O,
- McDonald KM,
- Baker LC
- Copyright © 2017 by the American Academy of Pediatrics