Abstract
Background:
The transition from hospital to home is a vulnerable period for all patients, especially for those who have limited English proficiency (LEP).
Methods:
We retrospectively studied adults discharged home from a hospital in 2018–2019 to determine the association of LEP with (1) reach of a care transitions outreach program phone call (automated call within three days after discharge or a subsequent manual phone call) and (2) post-discharge issues reported on the phone calls. All results were adjusted for measured confounders; we described associations using predicted probabilities and average marginal effects.
Results:
A total of 13,860 patients were included, and 11% had LEP. After adjustment, the program reached most patients regardless of LEP status; automated calls were more likely to reach EP patients (81.1% vs. 75.6%; p < 0.01), and when the automated call was unsuccessful, manual calls were more likely to reach LEP patients (47.8% vs. 28.3%, p < 0.001). After adjustment, patients with LEP reported more difficulty with all measured issues: understanding discharge instructions (11.3% vs. 6.5%), obtaining prescriptions (8.3% vs. 5.5%), medications concerns (12.9% vs 10.6%), follow-up questions (16.1% vs. 13.3%), new or worsening symptoms (15.1% vs. 11.9%), and any other clinical issues (16.6% vs. 13.0%); p < 0.05 for all comparisons.
Conclusion:
While reach was high for the care transitions program, among patients with LEP, important disparities exist in patient-reported post-discharge issues. These results indicate the need for better discharge processes that focus on communication quality and health equity.
Patients with limited English proficiency (LEP) commonly face barriers in communicating with their clinicians and understanding their treatment plans.1 These barriers are particularly problematic in the hospital, where patients with LEP are more likely to suffer adverse events, have higher mortality rates, and have higher readmission rates than English proficient (EP) patients.2,3 While prior studies have documented in-hospital adverse events, the experience of patients with LEP during and after hospital discharge is not well understood.
The transition from hospital to home is a vulnerable period for all patients,4,5 especially for those who have LEP.6 To identify and address issues that arise during this transition, hospitals have implemented programs to contact all recently discharged patients routinely. These programs have been shown to improve patient satisfaction and increase post-discharge medication adherence.7 To reach discharged patients efficiently, care transition programs have increasingly relied on technological solutions such as automated phone calls and text messages. Whether such tools effectively reach those with LEP is not known; however, evidence from prior technology implementation suggests that new technologies tend to widen healthcare disparities.8,9
Additionally, little is known about the issues and barriers that patients with LEP face after hospital discharge. Although one study found that patients with LEP have poor comprehension of hospital discharge instructions,10 other dimensions of the care transition such as worsening symptoms, difficulty obtaining prescriptions, or issues taking medications have not been described.
To address these two knowledge gaps in the care of patients with LEP, we examined data from a care transitions outreach program at a large academic medical center. Our first goal was to evaluate whether the program’s technology and processes reached patients with LEP and EP (English proficiency) equally. Our second goal was to measure the prevalence of post-discharge, patient-reported issues, and their association with English proficiency.
METHODS
Design, Setting, Subjects
We performed a retrospective cohort study of patients aged 18 years or older discharged home from an academic medical center between May 1, 2018, and April 30, 2019. We included all patients who were discharged home from participating clinical services and were not part of a bundled payments program that coordinated post-discharge care separately. We excluded patients who were discharged with home hospice (1.2%), lacked a listed telephone number (0.4%), were admitted for less than one day (0.5%), were readmitted to the same hospital within 72 hours (1.9%), or were missing exposure covariate or outcome data (2.8%). Additionally, we excluded 3.7% of discharges because medical records could not be matched to automated calling program data (Appendix 1). For patients with multiple admissions during the study period, we examined the first discharge. At the study site, clinicians typically write discharge instructions in English. To provide high quality care and to adhere to Joint Commission and Federal standards, physicians and nurses use professional interpreters to review discharge instructions with patients and families whose preferred language is not English. Clinicians are also able to provide patient educational materials created by the EMR vendor and available in multiple languages.
The Care Transitions Outreach Program
The study site implemented a hospital-wide care transitions outreach program in March 2017. The program’s goal was universal contact with all patients discharged from the hospital to identify and address care transition problems. In their discharge instructions, patients were notified that they would receive an automated call within 72 hours of discharge. A third party (CipherHealth Voice, New York, NY, USA) delivered a scripted, automated call to ask patients if they had questions or concerns in six post-discharge domains. The program called all patients regardless of their reported language preference. Respondents could choose to hear and respond to the questions in English, Spanish, or Cantonese, the three most common languages spoken by patients in the health system. Appendix 2 provides the complete text of the script.
The automated calling program attempted to call patients up to five times per day for two consecutive days. If a patient failed to answer the automated call and met any of the specified criteria—age > 85 years, discharged home with home services, limited English proficiency, or part of the study site’s Medicare accountable care organization—a centralized care transition nurse would review the patient’s chart and call the patient manually if they had not already been contacted by a clinician (physician, nurse practitioner, physician’s assistant) or had another health care encounter. If a patient did answer the automated call initially and identified an issue, the care transition nurse would call the patient manually to follow up. In either situation, the nurse would use a professional language interpreter if they were not fluent in the patient’s preferred language.
Measures: Limited English Proficiency
There is no gold standard for defining limited English proficiency. We classified patients as having limited English proficiency (LEP) if in the electronic medical record (EMR), their preferred language for health care was a language other than English and if they self-identified as needing an interpreter. This definition was validated through chart review in a prior study.3 In that study, compared to the chart reviewers’ assessment of English proficiency, the definition had a positive predictive value of 100%; thus, any misclassification in this study would err on the side of categorizing patients with LEP as EP. Preferred language and need for an interpreter were obtained on registration and directly reflected the patients’ stated responses.
Measures: Sociodemographics and Clinical Comorbidities
We obtained demographic and clinical data using structured elements from the EMR. Using administrative data, we determined patients’ discharge diagnoses with International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis codes; we used these codes to calculate the Elixhauser score for each patient. The Elixhauser score is comprised of 30 comorbidities (claims-defined) and was developed to predict hospital length of stay, cost, and mortality.11
Outcomes: Reach Rate
We classified “patient reach” into one of three categories. If a patient was able to answer the automated call and at least one clinical question, we classified them as having “answered an automated call.” If a patient did not answer the automated call, we determined if they were eligible for a manual follow up call (see The Care Transitions Outreach Program). Among those eligible for a manual phone call, if the nurses successfully reached a patient by manual call, we classified them as having been “reached by a manual call.”
Outcomes: Patient-reported, Post-discharge Issues
The care transitions outreach program asked patients about six post-discharge issues: questions about their discharge instructions, difficulty getting their prescriptions, medication concerns, questions about follow-up care, new or worsening symptoms, and any other clinical issues not already addressed (specific questions listed in Appendix 2).
In sensitivity analyses, we further characterized the severity of issues in two ways. First, following an outreach call, care transition nurses noted the urgency of reported issues using templated documentation and categorized these from least to most urgent as follows: the nurse counseled the patient on the phone, the nurse notified a clinician, the nurse requested a non-urgent action by a clinician, or the nurse needed urgent action by the clinician within 24 hours. Second, when an issue was identified, we determined how often nurses needed to involve other health care professionals such as nonclinical health staff (e.g., social workers, case managers), clinicians, or pharmacists to resolve the issue.
Analysis
To compare baseline variables, we used the chi-square test for categorical variables. For continuous variables, we used the t-test for normally distributed measures and the Wilcoxon rank sum test for nonnormal measures. We determined the association between LEP and program reach by fitting separate models for each stage of outreach (successful reach by automated call, eligibility for manual call, and successful reach by manual call). Similarly, we determined the association between LEP and each post-discharge issue by fitting a separate model for each patient-reported issue. For each model, patients were included if they responded to the question corresponding to the post-discharge issue (see Appendix 2). All models were fit using generalized estimating equation with a log link and Poisson distribution12 adjusted for the following confounders: age, race, ethnicity, marital status, insurance, discharging service, Elixhauser score, discharge with home services, and length of stay. We report the regression results as the predicted population rates and average marginal effects (AME) along with the 95% confidence intervals.13 We conduced post hoc, exploratory analyses to test for heterogeneity within patients with LEP by preferred language. Because within patients with LEP race and ethnicity are strongly correlated with preferred language, race and ethnicity were not used in the multivariable model testing for heterogeneity by language. Also, p values in post hoc analyses were adjusted for multiple comparisons using the False Discovery Rate correction.14 We completed analyses in SAS 9.4 (Cary, NC) and STATA 15.1 (College Station, TX). The UCSF Committee on Human Research approved this study. The code used to generate the cohort and perform the analyses can be found online.15
RESULTS
Participant Characteristics
During the one-year study period, 13,860 unique patients were discharged home from inpatient hospitalizations, and 1,566 (11%) had LEP (Table 1). Patients with LEP were more likely to be excluded for any reason (13.7% vs. 11.9%). While patients with LEP were more likely to be excluded because of hospice discharge and inability to match data sources, they were less likely to be excluded because of missing variables (Appendix 1). The most common languages spoken by patients with LEP were Spanish (42%) and Cantonese (23%). Compared with English speakers, patients with LEP were more likely to be older, Hispanic/Latinx, and Asian. Patients with LEP were more likely to be insured through Medicaid or Medicare. Finally, patients with LEP were more likely to be discharged home with services.
Table 1.
Patient Characteristics of Discharged Patients by English Proficiency
| Limited English proficiency (n = 1566) | English proficient (n = 12294) | p value | |
|---|---|---|---|
|
| |||
| Age in years, median (IQR) | 65 (52, 77) | 57 (42, 67) | < 0.001 |
| Male (%) | 789 (50) | 6467 (53) | 0.098 |
| Language (%)* | < 0.001 | ||
| English | 0 (0) | 12095 (98) | |
| Spanish | 650 (42) | 68 (1) | |
| Cantonese | 353 (23) | 57 (0) | |
| Russian | 125 (8) | 14 (0) | |
| Mandarin | 116 (7) | 6 (0) | |
| Vietnamese | 50 (3) | 6 (0) | |
| All others | 272 (17) | 48 (0) | |
| Race (%) | < 0.001 | ||
| Asian | 682 (44) | 1444 (12) | |
| Black or African American | 9 (1) | 1145 (9) | |
| Other | 640 (41) | 2017 (16) | |
| White or Caucasian | 235 (15) | 7688 (63) | |
| Hispanic ethnicity (%) | 635 (41) | 1657 (13) | < 0.001 |
| Marital status (%) | < 0.001 | ||
| Divorced/Separated | 105 (7) | 979 (8) | |
| Married/Partnered | 994 (63) | 6593 (54) | |
| Single | 253 (16) | 4136 (34) | |
| Widowed | 214 (14) | 586 (5) | |
| Insurance status (%) | < 0.001 | ||
| Commercial | 203 (13) | 5133 (42) | |
| Medicaid | 576 (37) | 2535 (21) | |
| Medicare | 763 (49) | 4504 (37) | |
| Other | 24 (2) | 122 (1) | |
| Clinical service (%) | < 0.001 | ||
| General Surgery | 175 (11) | 1535 (12) | |
| Hospital Medicine | 455 (29) | 2597 (21) | |
| Neurosurgery | 126 (8) | 1628 (13) | |
| Other | 810 (52) | 6534 (53) | |
| Length of stay in days (median [IQR]) | 3.2 (2.0, 5.8) | 3.1 (1.5, 5.4) | < 0.001 |
| ICU stay (%) | 196 (13) | 1644 (13) | 0.368 |
| Elixhauser score (mean [SD])† | 2.0 (1.8) | 2.3 (1.8) | < 0.001 |
| Discharge home with services (%) | 372 (24) | 1970 (16) | < 0.001 |
IQR, interquartile range.
Of the 1,566 patients with LEP, the most common reported preferred languages were Spanish (42%), Cantonese (23%), Russian (8%), and Mandarin (7%). All other languages each made up less than 4% of discharges for patients with LEP. All the patients with LEP self-identified as needing an interpreter.
We classified patients as having limited English proficiency (LEP) if in the electronic medical record, their preferred language for healthcare was a language other than English and if they self-identified as needing an interpreter. In this specific definition, some patients were classified as English proficient whose preferred language was not English because they did not self-identify as needing an interpreter when offered.
The Elixhauser score is comprised of 30 comorbidities (claims-defined) and was developed to predict hospital length of stay, cost, and mortality.11
Limited English Proficiency and Patient Outreach
The care transition program reached a majority of discharged patients, although there were some differences by English proficiency (Figure 1; unadjusted results in Appendix 3). Fewer patients with LEP answered automated calls compared with patients with EP (adjusted, 75.6% vs. 81.1%; average marginal effect [AME] −5.5%, confidence interval [CI] −3.1% to −7.9%). When patients did not answer the automated call, patients with LEP were more likely to meet criteria for a manual call (adjusted, 35.3% vs. 12.7%; AME 22.5%, CI 16.4% to 28.6%). When the care transition nurses called manually, they were more likely to reach patients with LEP than those with EP (47.8% vs. 28.3%; AME 19.5%, CI 6.4% to 32.6%).
Figure 1:
Predicted probabilities from the adjusted model are presented. Separate models are fit at each stage (i.e., reached by automated call, met criteria for manual call, reached by manual call). All models were adjusted for the following confounders: age, race, ethnicity, marital status, insurance, discharging service, Elixhauser score, discharge disposition, length of stay. For all differences, p < 0.01. Unadjusted results can be found in Appendix 3.
All patients in the study were eligible for an automated call. If a patient failed to answer the automated call and met any of the following specific criteria—age > 85 years, discharged home with home services, limited English proficiency, or part of the study site’s Medicare accountable care organization—a centralized care transition nurse would review the patient’s chart and call the patient manually if they had not already been contacted by a clinician (physician, nurse practitioner, physician’s assistant) or had another health care encounter.
In post-hoc exploratory analyses, care transition program reach varied by language among patients with LEP (Appendix 4). Among patients with LEP, Spanish-speaking patients were more likely to be reached by an automated call when compared to patients who preferred Cantonese (adjusted 85.3% vs. 69.9%; AME 15.4%, CI 9.7% to 21.1%) and more likely to be reached when compared to patients who preferred a language other than Spanish or Cantonese (adjusted 85.3% vs. 73.1%; AME 12.3%, CI 7.5% to 17.1%). When patients with LEP were not reached by automated calls, Spanish-speaking patients were more likely to be eligible for manual calls compared to patients who preferred Cantonese (adjusted 47.9% vs. 29.6%; AME 18.3%, CI 6.1% to 30.6%). When manual calls were attempted, there were no observed differences by preferred language among patients with LEP.
When patients with LEP were reached by automated calls, 76% of Spanish-speaking patients answered in Spanish, 48% of Cantonese-speaking patients answered in Cantonese, and 89% of non-Spanish, non-Cantonese-speaking patients answered questions in English (Appendix 5). Data are not available on whether the patient themselves or a caregiver answered the automated call.
Limited English Proficiency and Patient Outcomes
Patients with LEP were more likely to report a problem with all measured post-discharge issues (Figure 2; unadjusted results in Appendix 3). A greater number of patients with LEP had questions regarding information in their discharge instructions compared to patients with EP (adjusted, 11.3% vs. 6.5%; AME 4.8%, CI 2.7% to 6.9%). More patients with LEP needed help to get their prescriptions filled (adjusted, 8.3% vs. 5.5%; AME 2.9%, CI 0.6% to 5.1%) and had concerns about their medications (adjusted 12.9% vs 10.6%; AME 2.3%, CI 0.0% to 4.6%). After discharge, patients with LEP were more likely to have questions about follow-up care (adjusted, 16.1% vs. 13.3%; AME 2.8%, CI 0.3% to 5.3%). They were also more likely to have new or worsening symptoms (adjusted 15.1% vs. 11.9%; AME 3.2%, CI 0.7% to 5.8%), and other clinical questions for the nurses (adjusted 16.6% vs. 13.0%; AME 3.6%, CI 1.1% to 6.1%). Although patients with LEP faced more post-discharge issues, there was no significant difference in issue severity, as defined by the need for assistance after the phone call (Appendix 6). Furthermore, when patients reported an issue and the care transition nurse escalated the concern, nurses involved nonclinical health staff, clinicians, and pharmacists at similar rates (Appendix 7).
Figure 2:
Predicted probabilities from the adjusted model are presented. Separate models were fit for each issue and included patients who answered the question: discharge instruction (n = 10,458), getting prescriptions (n = 7,849), medication concerns (n = 10,693), follow-up questions (n = 10,554), new or worsening symptoms (n = 11,224), any other clinical issues (n = 10,170). Models were adjusted for the following confounders: age, race, ethnicity, marital status, insurance, discharging service, Elixhauser score, discharge disposition, length of stay.
* The difference between groups is statistically significant with p < 0.05. Discharge instruction difference 4.8% (95% confidence interval [CI], 2.7% to 6.9%), p < 0.001; getting prescriptions difference 2.9% (95% CI, 0.6% to 5.1%), p = 0.012; medications concerns difference 2.3% (95% CI, 0.0% to 4.6%), p = 0.0495; follow-up questions difference 2.8% (95% CI, 0.3% to 5.3%), p = 0.027; new or worsening symptoms difference 3.2% (95% CI, 0.7% to 5.8%), p = 0.012; any other clinical issues difference 3.6% (95% CI, 1.1% to 6.1%), p = 0.004. Unadjusted results can be found in Appendix 3.
LEP, limited English proficiency; EP, English proficient.
In post-hoc exploratory analyses, there was little observed variation in post discharge issues by language among patients with LEP (Appendix 8). There was no observed heterogeneity in needing help to get prescriptions filled, concerns about medications, question about follow up care, or new or worsening symptoms. Among patients with LEP, patients who preferred Spanish were more likely to report issues with discharge instructions when compared to patients who preferred Cantonese (19.4% vs. 8.7%; AME 10.8% CI 5.7% to 15.9%) and when compared to those who preferred a language other than Spanish or Cantonese (19.4% vs. 6.1%; AME 13.3%, CI 9.1% to 17.6%). Additionally, among patients with LEP, patients who preferred Spanish were more likely to have other clinical questions for the nurses when compared to those who preferred a language other than Spanish or Cantonese (24.6% vs. 10.7%; AME 14.0%, CI 9.1% to 18.8%).
DISCUSSION
Overall, the care transitions program reached a substantial proportion of patients discharged from our adult hospital, including both English and non-English-speaking patients. The program’s robust efforts to reach patients with LEP, using multi-lingual automated scripts and prioritizing patients with LEP for manual outreach, enabled the program to reach a majority of patients with LEP. While automated phone calls reached fewer patients with LEP, the program prioritized patients with LEP for manual calls and reached proportionally more patients with LEP when automated calls were unsuccessful. Within the group of patients classified as LEP, the outreach program was most successful at reaching Spanish-speaking patients.
Additionally, we found that limited English proficiency was strongly associated with higher rates of patient-reported, post-discharge issues. These associations were clinically significant and robust, persisting after accounting for confounders. Despite these disparities in reported post-discharge issues for patients with LEP, there were no differences in the urgency of reported issues or the need for assistance from other professionals.
Interestingly, more than half of Cantonese-speaking patients replied to automated calls in English. While we lack the data to interrogate this further, possibilities include the involvement of English-proficient caregivers and the use of English-proficient caregivers’ phone numbers as patient’s contact number. Elucidating the reasons for this finding would require additional study and would be important for any implementation project using automated calls to improve the care of Cantonese-speaking patients.
The findings from this study confirm and expand on existing literature related to the communication barriers faced by hospitalized patients with LEP, which can manifest throughout the hospital course from admission to discharge. When patients with LEP are hospitalized, they receive adequate information for informed consent less than a third of the time, a statistic that is improved by easy access to professional interpreters.16 When critically ill, patients with LEP and their families receive less information and emotional support from their health care providers.17 During the hospital discharge process, few have a professional interpreter at the bedside to assist with discharge instructions,6 so after hospital discharge, it may not be surprising that patients with LEP report all post-discharge issues at higher rates. This study finds discharge disparities in multiple domains, from understanding instructions to follow up appointments to symptoms. Structural barriers, during both the hospital course and the care transition period, may explain the disparities patients with LEP face in outcomes such as readmission and mortality.18,19
Clinicians and health systems can take steps to improve the discharge experience for patients with LEP. At a minimum, clinicians must use a professional interpreter via telephone, video, or in person to communicate discharge instructions effectively to patients with LEP, standards that have been long articulated by The Joint Commission and Federal Statutes.20 Further, these issues go beyond the use of interpreters and require an understanding that communication is often limited by co-occurring low health literacy in patients with LEP.21,22
The results from this study highlight the central role of communication. While this study showed higher rates of reported issues in patients with LEP, there was no difference in the rates of the most severe issues requiring escalation, suggesting that nurse counseling was sufficient to resolve many of the excess issues reported by patients with LEP. Additionally, written translation technology can play an increasingly important role as technology improves. While the current implementation is imperfect, future iterations of artificial intelligence–driven translation services may provide real-time, scalable text translation for written discharge instructions.23 It will be important to ensure that these translations are accurate to avoid adverse events post-discharge.
One emerging program, Meds-to-Beds, relies on pharmacists to deliver medications and medication education to patients in the hospital before discharge, utilizing professional interpreters when needed, and has been shown to reduce readmission rates.24 Utilizing a program such as Meds-to-Beds along with appropriate written translations for patients to take home with them could lead to less miscommunication at the pharmacy when patients with LEP attempt to obtain their medications, a task often performed without an interpreter’s help. Finally, post-discharge population health programs are commonly used to identify issues that may otherwise go unaddressed. These programs rely on automated phone calls and text messages to scale outreach. To promote equity, population health programs should support non-English languages and examine their data to ensure they are reaching patients with LEP, including examining for heterogeneity by individual languages. These methods of improving care for patients with LEP are established but require health system leadership and buy-in, including financial resources.
Addressing disparities for patients with LEP during care transitions will require health systems to assess the underlying causes of the disparities, including communication barriers, access to professional medical interpreters, and systemic bias. However, there are already several interventions available that can improve outcomes for patients with LEP. The first step in understanding barriers is to have robust data collection, including the appropriate classification of patients by English proficiency.25,26 Next, hospital systems must provide professional interpreters to all patients with LEP as mandated by the Civil Rights Act of 1964.27,28 Professional interpreters improve patients’ communication and understanding of their medical care.16,29 However, the provision of interpreter services does not guarantee adequate use; studies have documented that even when interpreters are available, they may not be routinely used.30 To assess interpreter use and support quality improvement programs to improve use, the electronic medical record must have a way for clinicians to easily document professional interpreter use during language-discordant encounters.25 This critical intervention is only effective if it is utilized.
Limitations
This study has several limitations inherent to the data available and the nature of the care transitions outreach program. First, this study was conducted at an urban academic medical center with a substantive and linguistically diverse LEP population, and the findings may not generalize to other settings. Second, while the study results suggest communication may explain the higher rates of reported issues by patients with LEP, we were not able to measure professional interpreter use at the time of discharge and thus we are unable to assess if interpreter use would mitigate the observed disparities. Third, there is the potential for unmeasured confounding to influence our observed relationship between LEP and post-discharge issues, although we accounted for demographic variables, insurance status, patient complexity, and clinical service. Fourth, patients with LEP were more likely to meet exclusion criteria—the difference is statistically significant, but the absolute difference is small. While the risk of bias appears low, due to the nature of exclusions (e.g., missing data, failure to merge) we cannot estimate the degree of potential bias. Fifth, the sample size of patients with LEP did not have sufficient power to test for heterogeneity in reach and outcomes by patients’ preferred language or language used when replying to automated calls. While such results may have important implications for external implementation, they are exploratory and being underpowered may falsely find no difference. Finally, our criteria for defining LEP favored specificity over sensitivity; as such, there may be a small group of patients in the EP cohort who required language assistance during their hospital stay. This tradeoff biases the measured disparities toward the null; that is, the true disparity may be more substantial than what is measured in this study.
CONCLUSION
In this study, we demonstrated that a care transitions outreach program that supports non-English languages can reach a majority of patients with LEP. We also showed the importance of such an outreach program for patients with LEP as they experienced substantially more post-discharge issues than English speakers across many domains. Future work should focus on implementing and refining known solutions that improve the quality of hospital discharges and care transitions for patients with limited English proficiency.
Acknowledgments
Sponsor’s Role:
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Funding:
Dr. Karliner's time was supported by NIH/NIA K24 AG067003. This study was funded by the UCSF Division of Hospital Medicine and UCSF Health Office of Population Health. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Appendix for “Disparities After Discharge. The Association of Limited English Proficiency and Post-Discharge Patient Reported Issues”
Appendix-1: Identification of Study Cohort
We examined language status among those excluded and found that patients with LEP were more likely to be excluded after removing 30 discharges where language status could not be ascertained (13.7% LEP vs. 11.9% EP, p<0.05). While these differences are statistically significant, the absolute difference is small in magnitude. Due to the nature of the exclusion (e.g., missing data, failure to merge), we cannot estimate the degree of potential bias on the primary results.
When examining the specific reasons for exclusion, we found patients with LEP were more likely to be excluded due to hospice discharge (2.5% vs 1.0%, p<0.001) and because we could not find a match between the hospital database and calling program database (7.0% vs 5.2%, p<0.001). Patients with LEP were less likely to be excluded because of missing exposure, covariate, or outcome data (1.6% vs. 3.0%, p<0.001). There were no significant differences in exclusion because of missing a phone number, LOS <1d, or 3-day readmission.
Appendix-2: The Automated Script
| Please note, responses in red indicate that an alert will be sent to your follow-up staff. |
|---|
|
Step 1:
Language Hi, this is Registered Nurse Name of Nurse from [hospital name redacted for review] with an important follow-up call about your health, or the health of your family member. • To continue in English, press 1 → Ok,thanks. • To continue in Spanish, press 2 → Ok,thanks. • Or to continue in Cantonese, press 3 → Ok,thanks. |
|
Step 2: Who
Answered: • If you or your family member were recently discharged from [hospital name redacted for review] Medical Center, please press 1 → Ok, thanks. Caregivers and parents, please complete this call on behalf of the patient. • If you would like us to call you back later, press 2 → Ok, we’ll call you back at a later time. • Or if we have reached the wrong number, press 3 → Ok, we’re sorry to have bothered you. Have a nice day. |
|
Step 3:
Symptoms We want to know if you are doing well or if you need any help, so please answer all eight questions. If at any time you experience a medical emergency, please call 9-1-1 or go to the nearest emergency room. Ok, let’s get started! • Are you having any new symptoms, or symptoms that are getting worse? Please press 1 for yes → Ok, we will call you back as soon as we can. If it’s an emergency, call 9-1-1. • Or press 2 for no → Ok, thanks. |
|
Step 4:
Prescriptions Taking your medication correctly is important for your wellness and recovery. • If you were prescribed new medicines when you left the hospital, press 1 → Proceed to Step 5. • If you were not prescribed new medication when you left the hospital, press 2 → Ok, thanks. |
|
Step 5: Prescriptions II
(Only asked if they pressed 1 to Step
4) • Do you need help getting your new medicines? Please press 1 for yes → Ok, we will call you back. • Or press 2 for no → Ok, that’s great to hear. |
|
Step 6:
Medications • Do you have any medication questions or concerns? Please press 1 for yes → Ok, we’ll call you back. • Or press 2 for no → Ok, thanks. |
|
Step 7: Follow-Up
Help Many patients are told to schedule or attend home visits or follow-up appointments after going home from the hospital. • Do you have questions about your follow-up plan or need the phone number to schedule your appointment or home visit? Please press 1 for yes → Ok, we’ll call you back. • Or press 2 for no → Ok, thanks for letting us know. |
|
Step 8: Discharge
Instructions When you left the hospital, you were given instructions on how to care for yourself at home. • Do you have any questions about those instructions? Please press 1 for yes → Ok, we’ll call you back. • Press 2 for no → OK, great. |
|
Step 9:
Satisfaction We strive to provide the best experience for our patients and families. Were you satisfied with your stay at [hospital name redacted for review] Medical Center? • If you were completely satisfied, press 1 → Great, we’re happy that you were very satisfied with your stay. • If you were somewhat satisfied, press 2 → OK, thanks for letting us know, we are sorry that we did not meet your expectations. • If you were not satisfied, press 3 → We’re sorry to hear that we did not meet your expectations. If you would like to speak with our Patient Relations department regarding your experience, please call [redacted for review]. |
|
Step 10: Other Clinical
Issues One last question. Would you like the opportunity to discuss another clinical issue about your hospital stay with a [hospital name redacted for review] nurse? • Please press 1 for yes → Ok, we’ll call you back. • Press 2 for no → Ok, thanks. |
|
Step 11:
Goodbye Thank you for answering our questions. Have a nice day. Goodbye. |
Appendix-3: Unadjusted versions of main manuscript results
3.1: Post-discharge program reach by English proficiency, unadjusted
Unadjusted results that correspond to Figure 1 in the main manuscript. All differences are significant p < 0.01
3.2. Patient-reported, post-discharge issues by English proficiency
Adjusted and unadjusted results that correspond to Figure 1 in the main manuscript
| Post-discharge issue | Unadjusted | Adjusted |
|---|---|---|
|
| ||
| Discharge instructions | ||
| EP | 6.4% | 6.5% |
| LEP | 13.5% | 11.3% |
| Difference | 7.2% (95% CI, 5.1 to 9.2%) | 4.8% (95% CI, 2.7 to 6.9%) |
| Getting prescriptions | ||
| EP | 5.3% | 5.5% |
| LEP | 10.6% | 8.3% |
| Difference | 5.3% (95% CI, 3.1 to 7.6%) | 2.9% (95% CI, 0.6 to 5.1%) |
| Medication concerns | ||
| EP | 10.5% | 10.6% |
| LEP | 13.9% | 12.9% |
| Difference | 3.4% (95% CI, 1.3 to 5.5%) | 2.3% (95% CI, 0.0 to 4.6%) |
| Follow up questions | ||
| EP | 13.1% | 13.3% |
| LEP | 17.9% | 16.1% |
| Difference | 4.8% (95% CI, 2.5 to 7.2%) | 2.8% (95% CI, 0.3 to 5.3%) |
| New or worsening symptoms | ||
| EP | 11.9% | 11.9% |
| LEP | 14.8% | 15.1% |
| Difference | 2.8% (95% CI, 0.8 to 4.9%) | 3.2% (95% CI, 0.7 to 5.8%) |
| Any other clinical issues | ||
| EP | 12.6% | 13.0% |
| LEP | 20.4% | 16.6% |
| Difference | 7.8% (95% CI, 5.3 to 10.3%) | 3.6% (95% CI, 1.1 to 6.1%) |
Appendix-4: Variation in program reach by language among patients with LEP
In post-hoc exploratory analyses we examined heterogeneity in program reach by language within patients with LEP. These results are adjusted for adjusted for the following confounders: age, marital status, insurance, discharging service, Elixhauser score, discharge with home services, and length of stay. Because race and ethnicity are highly correlated with language within patients with LEP, we did not adjust for them. Given the exploratory analyses, we calculated p values for each measure of difference and adjusted for multiple testing using the False Discovery Rate correction for p value.
| Reach | Spanish | Cantonese | Non-Spanish/non-Cantonese | Difference between Cantonese and Spanish | Difference between non-Spanish/non-Cantonese and Spanish | Difference between non-Spanish/non-Cantonese and Cantonese |
|---|---|---|---|---|---|---|
|
| ||||||
| Reached by automated call | 85.3% | 69.9% | 73.1% | 15.4% (CI 9.7% to 21.1%, p<0.001) | 12.3% (CI 7.5% to 17.1%, p<0.001) | 3.1% (CI −3.0% to 9.3%, p=0.47) |
|
| ||||||
| Eligible for manual call | 47.9% | 29.6% | 38.4% | 18.3% (CI 6.1% to 30.6%, p=0.01) | 9.5% (CI −2.7% to 21.6%, p=0.29) | 8.9% (CI 0.6% to 17.1%, p=0.09) |
|
| ||||||
| Reached by manual call | 36.9% | 39.6% | 46.3% | −2.7% (CI −21.8% to 16.4%, p=0.83) | −9.4% (CI −26.0% to 7.3%, p=0.44) | 6.7% (CI −10.3% to 23.6%, p=0.55) |
Appendix-5: Automated phone call response language
Patients were able to listen to and respond to the automated phone call prompts in English, Spanish, or Cantonese. The table below outlines response language by the patients’ preferred language as noted in the EMR.
| EMR preferred language | Automated call response language (row percent) | |||
|---|---|---|---|---|
| Cantonese | English | Spanish | Total | |
| Spanish | 0 (0%) | 133 (24%) | 414 (76%) | 547 |
| Cantonese | 116 (48%) | 126 (52%) | 0 (0%) | 242 |
| Other | 41 (10%) | 360 (89%) | 3 (1%) | 404 |
| Total | 157 | 619 | 417 | 1193 |
Appendix-6: Problem urgency as documented
When a problem was identified care transition nurses would document the problem in a note in the EMR with a structured header that denoted the urgency of the problem as they understood it. These categories include: RN spoke with the patient, clinician notified, clinician action requested, clinician action needed, not reached. These categories are hierarchical and mutually exclusive.
6.1. Problem urgency by English proficiency, unadjusted
| Problem urgency by RN note | Limited English proficiency (n=555) | English proficient (n=3340) |
|---|---|---|
|
| ||
| RN spoke with patient | 11.5% | 10.1% |
| Clinician notified | 59.1% | 55.2% |
| Clinician action requested | 11.4% | 15.4% |
| Clinician action needed | 1.6% | 1.6% |
| Not reached | 16.4% | 17.6% |
Raw column percentages presented. P value 0.093 (Chi-square)
6.2. Problem urgency by English proficiency, adjusted
| Problem urgency by RN note | Limited English proficiency (predicted probability) | English proficient (predicted probability) |
|---|---|---|
|
| ||
| RN spoke with patient | 10.3% | 10.3% |
| Clinician notified | 55.8% | 55.8% |
| Clinician action requested | 14.8% | 14.8% |
| Clinician action needed | 1.6% | 1.6% |
| Not reached | 17.4% | 17.4% |
Predicted probabilities from the adjusted model are presented. Because of the hierarchical nature of the outcome, we used an ordered logit model. We adjusted for the following confounders: age, race, ethnicity, marital status, insurance, discharging service, Elixhauser score, discharge disposition, length of stay. Group p value =0.99.
Appendix-7: Issue Escalation and Involvement of Other Professionals
| Unadjusted | Adjusted | |
|---|---|---|
|
| ||
| Escalation to nonclinical health staff | ||
| EP | 2.5% | 2.4% |
| LEP | 1.6% | 1.9% |
| Difference (LEP-EP) | −0.9% (−2.0 to 0.3%) | −0.6% (−2.1 to 0.9%) |
| Escalation to clinician | ||
| EP | 5.4% | 5.1% |
| LEP | 2.9% | 4.2% |
| Difference (LEP-EP) | −2.5% (−4.1 to −0.9%) | −1.0% (−3.4 to 1.4%) |
| Escalation to pharmacist | ||
| EP | 7.7% | 7.8% |
| LEP | 6.3% | 6.0% |
| Difference (LEP-EP) | −1.4% (−3.6 to 0.8%) | −1.8% (−4.2 to 0.6%) |
LEP - limited English proficiency, EP - English proficient; Nonclinical health staff includes Health Care Navigator, Case Manager, Hospital Assistant, Medical Assistant, Social Worker
Predicted probabilities from the adjusted model are presented. Separate models were fit for each outcome and included patients who reported one or more issues (n=3895; LEP = 555, EP = 3340). All models were adjusted for the following confounders: age, race, ethnicity, marital status, insurance, discharging service, Elixhauser score, discharge disposition, length of stay.
In the unadjusted analyses, escalation to clinician was statistically significant. However, no differences were statistically significant following adjustment.
Appendix-8: Variation in issue rates by preferred language among patients with LEP
In post-hoc exploratory analyses we examined heterogeneity in issue rates by language within patients with LEP. These results are adjusted for the following confounders: age, marital status, insurance, discharging service, Elixhauser score, discharge with home services, and length of stay. Because race and ethnicity are highly correlated with language within patients with LEP, we did not adjust for them. Given the exploratory analyses, we calculated p values for each measure of difference and adjusted for multiple testing using the False Discovery Rate correction for p value.
| Issue | Comparison | Difference | P value adjusted for multiple comparisons |
|---|---|---|---|
|
| |||
| Discharge instruction | Cantonese vs. Other | 8.7% vs. 6.1% (AME 2.6%; CI −1.7% to 6.8%) | 0.42 |
| Cantonese vs. Spanish | 8.7% vs. 19.4% (AME −10.8%; CI −15.9% to −5.7%) | <0.001 | |
| Other vs. Spanish | 6.1% vs. 19.4% (AME −13.3%; CI −17.6% to −9.1%) | <0.001 | |
|
| |||
| Follow up help | Cantonese vs. Other | 21.1% vs. 17.0% (AME 4.1%; CI −2.1% to 10.2%) | 0.42 |
| Cantonese vs. Spanish | 21.1% vs. 14.1% (AME 7.0%; CI 1.1% to 12.9%) | 0.07 | |
| Other vs. Spanish | 17.0% vs. 14.1% (AME 2.9%; CI −1.7% to 7.5%) | 0.42 | |
|
| |||
| Medications | Cantonese vs. Other | 12.3% vs. 13.5% (AME −1.2%; CI −6.3% to 3.9%) | 0.72 |
| Cantonese vs. Spanish | 12.3% vs. 12.3% (AME 0.0%; CI −5.0% to 5.0%) | 1.00 | |
| Other vs. Spanish | 13.5% vs. 12.3% (AME 1.2%; CI −3.0% to 5.4%) | 0.7 | |
|
| |||
| Other clinical issues | Cantonese vs. Other | 17.5% vs. 10.7% (AME 6.9%; CI 1.3% to 12.5%) | 0.06 |
| Cantonese vs. Spanish | 17.5% vs. 24.6% (AME −7.1%; CI −13.2% to −1.0%) | 0.07 | |
| Other vs. Spanish | 10.7% vs. 24.6% (AME −14.0%; CI −18.8% to −9.1%) | <.0001 | |
|
| |||
| Prescriptions | Cantonese vs. Other | 10.8% vs. 9.3% (AME 1.5%; CI −4.3% to 7.2%) | 0.72 |
| Cantonese vs. Spanish | 10.8% vs. 7.3% (AME 3.5%; CI −1.9% to 8.9%) | 0.42 | |
| Other vs. Spanish | 9.3% vs. 7.3% (AME 2.0%; CI −2.1% to 6.1%) | 0.48 | |
|
| |||
| Symptoms | Cantonese vs. Other | % vs. 10.8% (AME −2.2%; CI −7.1% to 2.7%) | 0.50 |
| Cantonese vs. Spanish | 10.8% vs. 9.3% (AME −2.4%; CI −7.1% to 2.4%) | 0.48 | |
| Other vs. Spanish | 9.3% vs. 7.3% (AME −0.1%; CI −4.3% to 4.1%) | 0.99 | |
Footnotes
Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author). Kristin Gagliardi reports that she was employed by the third party (CipherHealth) prior to her employment at University of California, San Francisco and prior to any involvement with this study. All other authors report no conflicts of interest.
Competing interests: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author). LM, MG, LK, and SJS have nothing to disclose. KG reports that she was employed by the third party (CipherHealth) prior to her employment at UCSF and prior to any involvement with this study.
Ethical approval: The Human Research Protection Program Institutional Review Board at the University of California, San Francisco, approved this study (IRB# 19–27987).
Data sharing: Data used to complete this analysis used protected health elements. Researchers can contact the corresponding authors to request access to the study data. Code used to perform the analyses can be accessed online (https://github.com/sachinjshah).
References
- 1.Wilson E, Hm Chen A, Grumbach K, Wang F, Fernandez A. Effects of limited English proficiency and physician language on health care comprehension. J Gen Intern Med. 2005;20(9):800–806. doi: 10.1111/j.1525-1497.2005.0174.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Karliner LS, Perez-Stable EJ, Gregorich SE. Convenient Access to Professional Interpreters in the Hospital Decreases Readmission Rates and Estimated Hospital Expenditures for Patients With Limited English Proficiency. Med Care. 2017;55(3):8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Jacobs ZG, Prasad PA, Fang MC, Abe-Jones Y, Kangelaris KN. The Association between Limited English Proficiency and Sepsis Mortality. J Hosp Med. 2019;(2019–11-20 ONLINE FIRST). doi: 10.12788/jhm.3334 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Kripalani S, Jackson AT, Schnipper JL, Coleman EA. Promoting effective transitions of care at hospital discharge: A review of key issues for hospitalists. J Hosp Med. 2007;2(5):314–323. doi: 10.1002/jhm.228 [DOI] [PubMed] [Google Scholar]
- 5.Krumholz HM. Post-Hospital Syndrome — An Acquired, Transient Condition of Generalized Risk. N Engl J Med. 2013;368(2):100–102. doi: 10.1056/NEJMp1212324 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Semere W, Nápoles AM, Gregorich S, Livaudais-Toman J, Karliner L. Caregiving for Older Adults with Limited English Proficiency: Transitioning from Hospital to Home. J Gen Intern Med. 2019;34(9):1744–1750. doi: 10.1007/s11606-019-05119-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Braun E, Baidusi A, Alroy G, Azzam ZS. Telephone follow-up improves patients satisfaction following hospital discharge. Eur J Intern Med. 2009;20(2):221–225. doi: 10.1016/j.ejim.2008.07.021 [DOI] [PubMed] [Google Scholar]
- 8.Phelan JC, Link BG, Tehranifar P. Social Conditions as Fundamental Causes of Health Inequalities: Theory, Evidence, and Policy Implications. J Health Soc Behav. 2010;51(1_suppl):S28–S40. doi: 10.1177/0022146510383498 [DOI] [PubMed] [Google Scholar]
- 9.Chang VW, Lauderdale DS. Fundamental Cause Theory, Technological Innovation, and Health Disparities: The Case of Cholesterol in the Era of Statins. J Health Soc Behav. 2009;50(3):245–260. doi: 10.1177/002214650905000301 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Karliner LS, Auerbach A, Nápoles A, Schillinger D, Nickleach D, Pérez-Stable EJ. Language Barriers and Understanding of Hospital Discharge Instructions: Med Care. 2012;50(4):283–289. doi: 10.1097/MLR.0b013e318249c949 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity Measures for Use with Administrative Data: Med Care. 1998;36(1):8–27. doi: 10.1097/00005650-199801000-00004 [DOI] [PubMed] [Google Scholar]
- 12.Zou G. A Modified Poisson Regression Approach to Prospective Studies with Binary Data.pdf. Published online September 25, 2003. [DOI] [PubMed] [Google Scholar]
- 13.Norton EC, Dowd BE, Maciejewski ML. Marginal Effects—Quantifying the Effect of Changes in Risk Factors in Logistic Regression Models. JAMA. 2019;321(13):1304–1305. doi: 10.1001/jama.2019.1954 [DOI] [PubMed] [Google Scholar]
- 14.Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J R Stat Soc Ser B Methodol. 1995;57(1):289–300. [Google Scholar]
- 15.sachinjshah - Overview. GitHub. Accessed August 26, 2021. https://github.com/sachinjshah [Google Scholar]
- 16.Lee JS, Pérez-Stable EJ, Gregorich SE, et al. Increased Access to Professional Interpreters in the Hospital Improves Informed Consent for Patients with Limited English Proficiency. J Gen Intern Med. 2017;32(8):863–870. doi: 10.1007/s11606-017-3983-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Thornton JD, Pham K, Engelberg RA, Jackson JC, Curtis JR. Families with limited English proficiency receive less information and support in interpreted intensive care unit family conferences*: Crit Care Med. 2009;37(1):89–95. doi: 10.1097/CCM.0b013e3181926430 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Narula N, Curran T, Nagle D. Return To The Hospital In Limited English Proficiency Patients After Inpatient Colorectal Surgery. [Google Scholar]
- 19.Karliner LS, Kim SE, Meltzer DO, Auerbach AD. Influence of language barriers on outcomes of hospital care for general medicine inpatients. J Hosp Med. 2010;5(5):276–282. doi: 10.1002/jhm.658 [DOI] [PubMed] [Google Scholar]
- 20.Lee JS, Nápoles A, Mutha S, et al. Hospital discharge preparedness for patients with limited English proficiency: A mixed methods study of bedside interpreter-phones. Patient Educ Couns. 2018;101(1):25–32. doi: 10.1016/j.pec.2017.07.026 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Sudore RL, Landefeld CS, Pérez-Stable EJ, Bibbins-Domingo K, Williams BA, Schillinger D. Unraveling the relationship between literacy, language proficiency, and patient–physician communication. Patient Educ Couns. 2009;75(3):398–402. doi: 10.1016/j.pec.2009.02.019 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Sentell T, Braun KL. Low Health Literacy, Limited English Proficiency, and Health Status in Asians, Latinos, and Other Racial/Ethnic Groups in California. J Health Commun. 2012;17(sup3):82–99. doi: 10.1080/10810730.2012.712621 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Khoong EC, Steinbrook E, Brown C, Fernandez A. Assessing the Use of Google Translate for Spanish and Chinese Translations of Emergency Department Discharge Instructions. JAMA Intern Med. 2019;179(4):580. doi: 10.1001/jamainternmed.2018.7653 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Lash DB, Mack A, Jolliff J, Plunkett J, Joson JL. Meds-to-Beds: The impact of a bedside medication delivery program on 30-day readmissions. J Am Coll Clin Pharm. 2019;2(6):674–680. doi: 10.1002/jac5.1108 [DOI] [Google Scholar]
- 25.Taira BR, Kim K, Mody N. Hospital and Health System–Level Interventions to Improve Care for Limited English Proficiency Patients: A Systematic Review. Jt Comm J Qual Patient Saf. 2019;45(6):446–458. doi: 10.1016/j.jcjq.2019.02.005 [DOI] [PubMed] [Google Scholar]
- 26.Karliner LS, Napoles-Springer AM, Schillinger D, Bibbins-Domingo K, Pérez-Stable EJ. Identification of Limited English Proficient Patients in Clinical Care. J Gen Intern Med. 2008;23(10):1555–1560. doi: 10.1007/s11606-008-0693-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.LEP.GOV -- Commonly Asked Questions and Answers Regarding Limited English Proficient (LEP) Individuals. (Accessed October 15, 2020). https://www.lep.gov/faq/faqs-rights-lep-individuals/commonly-asked-questions-and-answers-regarding-limited-english
- 28.Commonly Asked Questions and Answers Regarding the Protection of Limited English Proficient (LEP) Individuals under Title VI of the Civil Rights Act of 1964 and Title VI Regulations. https://www.lep.gov/sites/lep/files/resources/042511_Q%26A_TitleVI_and_Regulations.pdf
- 29.Karliner LS, Jacobs EA, Chen AH, Mutha S. Do Professional Interpreters Improve Clinical Care for Patients with Limited English Proficiency? A Systematic Review of the Literature. Health Serv Res. 2007;42(2):727–754. doi: 10.1111/j.1475-6773.2006.00629.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Tang AS, Kruger JF, Quan J, Fernandez A. From Admission to Discharge: Patterns of Interpreter Use among Resident Physicians Caring for Hospitalized Patients with Limited English Proficiency. J Health Care Poor Underserved. 2014;25(4):1784–1798. doi: 10.1353/hpu.2014.0160 [DOI] [PubMed] [Google Scholar]