Effects of the Limited English Proficiency of Parents on Hospital Length of Stay and Home Health Care Referral for Their Home Health Care–Eligible Children With Infections FREE

A total of 310 languages other than English are spoken in the United States, 149 of which are languages originating from outside the United States.¹ A language other than English is spoken in 14 million US households by more than 55 million (roughly 1 in 5) US residents, nearly half of whom describe themselves as having limited English proficiency (LEP) or speaking English less than very well.^2- 3 As recently as 2009, approximately 25 million individuals were identified as having LEP by the American Community Survey.² Spanish is the primary language of more than 34 million persons living in the United States who are 5 years of age or older, making the United States the second-largest Spanish-speaking community in the world, second only to Mexico.⁴ Medical communities across the nation are faced with communication challenges that accompany such a linguistically diverse population. In 2007, the American Medical Association released a guideline for health care providers on communicating with patients with LEP, highlighting the risks that result from inadequate communication during health care encounters, including misunderstanding of the care plan, delayed diagnosis, errors in prescribing medication, misuse of health services, and lack of confidence in the medical system.^5- 9

The relationship between language barriers and lower quality health care has been well documented.^5- 10 In 1998, Flores et al⁵ demonstrated that language is the most important barrier to health care access in the Latino population. In 2002, the same group chronicled multiple adverse health care consequences due to language differences between the patient and medical staff, including poor medical care, misdiagnosis, inappropriate medications, and unnecessary hospitalization.⁶ The consequences of miscommunication have been shown to contribute to poor adherence, inappropriate follow-up, and patient dissatisfaction.¹¹ In the emergency department, even when controlling for socioeconomic status, LEP has been associated with longer stays, increased resource use, and higher hospitalization rates.^7- 10,12 Furthermore, within a cultural or racial/ethnic group, English proficiency plays a larger role in health care equity than either culture or race/ethnicity alone.¹³

An important, indirect measure of inpatient quality of care is hospital length of stay (LOS), which is associated with cost of care, patient outcome, and hospital-associated complications.¹⁴ The importance of a patient's primary language to LOS in the pediatric emergency department was described by Goldman et al,⁷ who found that language other than English, among other factors, was associated with longer stays. Whether a similar relationship exists in the inpatient pediatric setting is unclear; however, the inpatient setting offers a unique opportunity to address LEP issues because the interaction between provider and family is more extensive, including history taking, clinical assessments, and discussion of management. Limited English proficiency could have an effect on all of these interactions, and it is conceivable that patients with LEP are at risk for longer hospital stays and other adverse health care consequences due to language differences between the patient and medical staff.

Inpatient hospital costs account for the largest portion of our nation's health care expenditures, and it has been well established that home health care is less expensive and often has better outcomes than inpatient treatments.^14- 17 Home health care can be a surrogate marker for health care utilization because patients receiving this service typically use fewer resources, including daily physician care and hospital support services.¹⁸

The aim of our study was to determine whether the LEP of parents is associated with length of hospital stay and referral of home health care services for pediatric inpatients. Multiple factors contribute to the decision of whether a patient will receive a home health care referral, such as provider perception of family ability to adequately manage home health care, family comfort level with patient care, insurance coverage of home health care services, and availability of home health care services in the community. The purpose of our study is to determine whether LEP is a contributing factor in LOS and home health care referral. Patients receiving prolonged intravenous antibiotic therapy are commonly eligible for home health care, providing a useful study population for evaluating both LOS and home health care referral. Identifying disparities in measurable health service utilization among these patients is an important first step to understanding the effect of language barriers on health care service delivery. If a group of patients or families are unable, on the basis of LEP, to benefit from the lower cost and equal or better health outcome that home health care can offer, identifying such an inequity would allow for targeted improvement of the health care system and health outcomes, alike.

Our study was conducted in an urban, tertiary care, free-standing Midwestern children's hospital serving as a regional pediatric referral center. It has a multistate referral area and serves a multiethnic population. Of the approximately 17 000 inpatient hospitalizations in 2008, 31% were for nonwhite patients, 49% either self-pay or are on Medicaid, and 3% are for families with LEP (Delynn Archer, information technologist, written communication, January 2009).

All children 0 to 18 years of age admitted to the hospital during the period from January 1, 2000, to December 31, 2008, for infections typically requiring prolonged antibiotic therapy were included in the LOS portion of our study. Cases of infection were selected in consultation with the hospital's infectious disease specialists on the basis of the likelihood of home health care eligibility. They included joint and/or bone infections (International Classification of Diseases, Ninth Revision [ICD-9] codes 730.20-730.29 for osteomyelitis and ICD-9 codes 711.00-711.09 for infected joints) and central nervous system infections (ICD-9 codes 320.9, 320.7, 320.0, and 036.0 for bacterial meningitis, ICD-9 code 324.0 for brain abscess, and ICD-9 code 376.01 for orbital cellulitis). Because home health care data were only available starting after 2003, a subset of data was analyzed for home health care referral (723 patients). Patients were excluded if they died while hospitalized.

Study groups were determined by the English proficiency of the patient's primary caregiver, identified through a combination of data on self-reported primary language spoken at home and need for an interpreter. By hospital protocol, a patient's primary caregiver is asked by the intake nurse or an admissions specialist to report on the primary language spoken at home and whether there is a need for an interpreter on admission to an inpatient service. Because the study hospital does not currently ask families their self-identified ability to speak English, for the purposes of our study, patients with primary caregivers who reported a primary language spoken at home other than English and the need for an interpreter were defined as having LEP. All other patients were deemed to be proficient in English. Hospital LOS was defined as the time between inpatient admission and subsequent hospital discharge, measured in hours. Home health care referral was defined as the documentation of an order for home health care in the discharge-planning portion of the medical chart.

Covariates included patient age in years, sex, race/ethnicity (white, African American, Latino, Asian, and other), insurance type (private, Medicaid, or self-pay), presence of a primary care provider (yes or no), and presence of comorbidities (asthma, malignancy, sepsis, diabetes, sickle cell disease, congenital cardiac disease, liver disease, and renal disease). These comorbidities were considered by the authors to be potentially related to both study outcomes (LOS and home health care referral). For the analysis, the comorbidity variable was dichotomized as presence or absence of a comorbidity.

Means, standard deviations, and proportions were used for descriptive analyses. Categorical variables were compared using the χ² test or the Fisher exact test, and continuous variables were compared using the t test.

Hospital LOS was compared between language proficiency groups using simple linear regression. Owing to a skewed distribution, LOS was transformed using the natural logarithm (ln) to achieve a normal distribution. The relationship of all study variables to language proficiency and ln(LOS) were analyzed by use of binary logistic regression and simple linear regression, respectively. Covariates were included in multiple linear regression models of the relationship between language proficiency and ln(LOS). Odds ratios (ORs) with 95% confidence intervals (CIs) expressed the relationship between the 2.

Because differences in ln(LOS) are not easily interpretable in the clinical setting, regression results were converted to relative LOS by exponentiating the unstandardized regression coefficient for each variable. This relative LOS represents the ratio of the geometric means of the compared groups, offering a measure of relative difference (eg, a relative LOS of 1.5 would indicate that 1 group stayed an average of 50% longer than the other).

Because home health care data were only available starting after 2003, a subset of data was analyzed for home health care referral (723 patients). Home health care referral was compared between language proficiency groups using logistic regression. Covariates were included in multiple logistic regression models of the relationship between language proficiency and home health care referral. Covariates were chosen to be included in the model using stepwise logistic regression. Odds ratios with 95% CIs were used to express the relationship between the two. Data were analyzed using SPSS version 16.0.1 (SPSS Inc, Chicago, Illinois). Our study was approved by the hospital's institutional review board.

On the basis of primary diagnoses, a total of 1282 subjects were eligible during the period of our study; however, only 1257 patients are included in the analysis because 25 deaths occurred during hospitalization. Demographic and clinical characteristics of the study population are detailed in Table 1. Of the 1257 patients, 39 (3.1%) had LEP and 1218 (96.9%) were proficient in English. Patients with LEP were more likely to be Latino and either uninsured or insured by Medicaid (P < .001).

The median LOS for all subjects was 4.1 days (interquartile range, 2.6-7.2 days), with a median LOS among the English-proficient and LEP groups of 4.0 and 6.1 days, respectively (P = .098). Adjusted LOS results are shown in Table 2. Limited English proficiency was associated with longer LOS (adjusted relative LOS, 1.6 [95% CI, 1.1-2.3]), indicating that patients with LEP stayed 60% longer on average than English-proficient patients. Absence of a primary care provider, presence of any comorbidity, receipt of home health care, and Medicaid insurance status were all associated with longer LOS.

Home health care referral results are shown in Table 3. Home health care referrals were made for 32.6% of English-proficient patients and 6.9% of patients with LEP. In the multivariate model, LEP was significantly associated with a decreased number of home health care referrals (OR, 0.2 [95% CI, 0.04-0.8]). Medicaid insurance status was also related to a decreased number of home health care referrals (OR, 0.6 [95% CI, 0.4-0.8]). In exploratory analyses, Latino race/ethnicity and LEP were strongly collinear, negating each other's effect in the adjusted regression models. Owing to the primary objectives of our study, LEP was chosen for the final model, and race/ethnicity was removed. Having any comorbidity more than doubled the chance of home health care referral (OR, 2.1 [95% CI, 1.4-3.1]). Interaction terms were run for LEP and insurance status, and their interactions were found to be insignificant in both models (P = .08 for the LOS model and P = .05 for home health care model).

Limited English proficiency was associated with longer LOS among pediatric inpatients admitted to the hospital with infections requiring prolonged antibiotic therapy. Specifically, in adjusted analysis, patients from families with LEP had a 60% longer relative LOS than those of English-proficient families. We also found a disparity in home health care referral based on LEP, with patients who had primary caregivers with LEP 80% less likely to receive a home health care referral than those with English-speaking primary caregivers. This disparity persisted even after adjustment for relevant covariates.

Although much of the literature addressing topics related to our study has to do with adult health care,^19- 21 research shows that health care disparities in pediatrics continue to grow.²² The findings from our study add to the known associations between language barriers and health care quality.^8,11,13 Many studies specifically addressing pediatric disparities have focused on race/ethnicity, insurance coverage, and socioeconomic status.^{18- 19,22- 31} Our study adds to the understanding of language barriers in pediatric care and is consistent with previous work identifying disparities in health and health care outcomes for children from families with LEP³² (ie, children who reside in households in which a language other than English is the primary language spoken).^9,33

Our study has important limitations. It was conducted at a single site in the Midwest. Other locations may have more extensive interpreter services available for parents, decreasing the influence of language as a barrier to home health care for families with LEP. Differences in treatment styles among infectious disease specialists or between regions or institutions may also affect home health care recommendations. Such variability is seen in the recent trend in some institutions to transition patients with diagnoses described in our study to oral antibiotics earlier than in the past.³⁴ Earlier transition to oral antibiotics could decrease the necessity for home health care by nullifying the need for antibiotic infusion; however, one would not expect a differential effect in this trend by English proficiency status. Despite the large sample size of patients with the defined infections, only 39 of 1257 patients (3.1%) had LEP. This is inconsistent with the overall percentage of admissions of patients who had LEP to the study hospital and raises the question of why our study population differed demographically from the larger hospital population. Because of the retrospective nature of our study, severity of illness could not be included. No standardized severity scores currently exist or are used for the infections in question. Although severity of illness is a factor in LOS and in home health care referral decisions, to our knowledge, there are neither studies nor empirical evidence associating LEP with severity of the infections reviewed in our study. Therefore, it is difficult to imagine that the difference in home health care referral in our study was driven by patients with LEP having different levels of severity of infection. Lastly, the gold standard for determining LEP is self-report of ability to speak English, not self-report of primary language spoken at home or need for interpreter services. It is reasonable to assume that the need for interpreter services is a proxy for LEP status. However, misclassification may occur in either direction. There is a possibility that primary caregivers identified as needing an interpreter did not have LEP but were bilingual. It is equally possible that caregivers identified as being proficient in English were not bilingual and had LEP.

The clinical importance of increased relative LOS (60% longer in this sample) depends on the reference LOS. Because the median LOS in our study population was 4.1 days, a 60% increase represents approximately 2.5 more days in the hospital. This could be significant from a financial, clinical, and social standpoint for hospitals, patients, and payors. Although formal cost calculations were not made, at the time of our study, the average daily charge for an inpatient bed at the study hospital was approximately $2500. Using the difference in median LOS, this implies an increase of $6250 for total-stay room charge for a patient having LEP compared with an English-proficient patient. This estimate does not account for other charges but shows the potential scale of cost difference based on language. Besides financial strains, families who stay in the hospital longer have to deal with other difficulties such as time off from work, child care for their offspring who are not ill, and transportation.

Patients who receive home health care typically use fewer resources, including daily physician care and hospital support services.¹⁵ Patients who receive home health care also tend to have superior outcomes.¹⁶ In our study, the lower likelihood of home health care for patients having LEP may represent a denied opportunity for equitable health and health care outcomes.

The language-based home health care referral gap described herein provides a specific opportunity for improvement in the provision of adequate and equitable health care for the children of our diverse, multilingual society. Assessing the fit between home health care services and the communities they serve may help in clarifying the causes of the home health care gap. Providers, who ultimately decide which patients to recommend for home health care, could be studied with regard to attitudes and knowledge about families with LEP and the services available to them. One possible solution for addressing the disparities in LOS and home health care referral would be to educate physicians, housestaff, and medical students about the home health care system and to foster closer work with medical interpreters. Education might also highlight the need for vigilance in assessing whether families with LEP are willing or able to use home health care as thoroughly as English-proficient families. Lastly, an LOS of greater-than-expected duration for orthopedic and brain infections could be seen as a quality improvement issue and should merit further investigation.

With the population of people with LEP in the United States growing exponentially,^1- 2,4 the medical community must ensure that all patients with LEP receive adequate interpreter services. Increasing the number and quality of trained medical interpreters and translators, improving the infrastructure for a multilingual approach to care, and further minimizing multitiered care based on language are important areas for advocacy. Future studies on interventions to address language-related disparities such as those described in our study are needed to improve the care of patients with LEP.