Inequities in Pain Assessment and Care of Hospitalized Children With Limited English Proficiency

This was a prospective, cross-sectional study consisting of in-person surveys and a subsequent electronic medical record (EMR) chart review. The study was conducted in the Southeastern United States, where 69.4% of the county’s population identifies as Hispanic or Latino and 74.9% of the county’s population speaks a language other than English at home according to the 2015–2019 US Census Bureau data.²⁹  The hospital is a 309-bed, tertiary care, freestanding children’s hospital with >10 000 inpatient admissions annually and multiple pediatric subspecialties, including a consulting pain service and palliative care service. Official hospital interpreter services are provided via a 24-hour telephone language line. The WIRB-Copernicus Group (now known as WCG) institutional review board approved this study.

The in-person survey was administered on 4 nonconsecutive days over a 1-month period from March to April 2021. The dates were known only to the research team, which was composed of pediatric nurse practitioners and physicians, all of whom were trained in a standard interview procedure. Each parent–child dyad was approached by a 2-person research team that was fluent in both English and Spanish. For non-English and non-Spanish–speaking patients, an interpreter was used via the hospital’s telephone language line. A standard short script described the study objectives, and patients were asked if they would like to participate voluntarily with the understanding that their responses had no direct effect on patient care. Verbal consent from parents and assent from patients was obtained to participate. The verbal survey answers were recorded on paper questionnaires and later compiled with the EMR data into Research Electronic Data Capture hosted by the hospital.^30,31 

Patients aged 0 to 21 years admitted to the medical and surgical wards for at least 12 hours as of the 8 am census on each survey day were eligible for participation. Patients in the PICU, NICU, and hematology/oncology wards were excluded because these patients have unique concerns with pain prevalence and analgesia. Patients were interviewed directly if they were aged ≥5 years and were developmentally capable of understanding the questions as judged by the research team and parent. This age limit was based on previous audits and research indicating the age at which children have the developmental readiness to self-report pain.^18,32,33  If it was not possible to interview the child, then a parent at bedside was interviewed as a proxy. If a child was alone in the room, but deemed able to complete the survey, a parent was contacted by phone for consent. Patients were excluded if they were not in their hospital room after 3 attempts, if a parent was unreachable by telephone for consent, or if the patient was unable to complete the survey and there was no parent at bedside to act as a proxy.

The survey tool consisted of a structured questionnaire based on previously published pain audits.^5,6  The 5-minute survey asked the child or parent proxy (hereafter referred to together as “patients”) about the child’s preferred language, what language their pain was assessed in during the previous 24 hours, the expectations for pain during this hospitalization (categorized as none, mild, moderate, or severe), and whether the child experienced any pain during the previous 24 hours. If patients answered “yes” to having pain in the last 24 hours, they were further asked about the worst intensity of the pain using a 0-10 Revised Faces Pain Scale (FPS-R), the specific cause of this worst pain, and the utility of medications and adjunctive therapies used for this worst pain. Regardless of the presence or cause of pain, patients were also asked if they had undergone a needle poke (including intravenous catheter insertion, phlebotomy, or intramuscular medication injection) in the previous 24 hours and, if applicable, to give the needle poke a numerical rating using the same 0-10 FPS-R. The FPS-R was used because it has been validated in children as young as 5 years of age from different ethnicities to accurately reflect acute pain intensity.^32–34 

An EMR chart review was then conducted for all patients who participated in the survey. Data collection was limited to the 24 hours preceding the survey administration. The EMR data included patient age, gender, race, ethnicity, insurance provider, reason for admission, surgery/procedure within the preceding 48 hours, length of stay in the hospital at the time of the survey, type and frequency of pain documentation, highest nurse-documented pain intensity, and type of pain medications prescribed in the last 24 hours. Race and ethnicity data were collected because disparities often exist in pain assessment and management for minority groups.^16,35–42 

The primary outcome of this study was a difference of ≥3 points on the pain scale between patient-reported and nurse-documented worst pain scores; this was chosen a priori as a clinically significant difference. Patient demographics and characteristics were summarized using descriptive statistics. Differences in numerical values across groups were analyzed by using student t tests for normative data and Mann-Whitney U tests for nonnormative data. Categorical variables were analyzed using Fisher’s exact tests. A multiple logistic regression model was estimated to predict a discrepant pain score on the basis of preferred language; additional covariates included in the model were children’s age, sex, and survey respondent. These covariates were prioritized given hypothesized associations with pain assessment and observed significant differences between patients with and without LEP in bivariate analyses. P values of ≤.05 were considered statistically significant. Analyses were conducted using GraphPad PRISM software (San Diego, CA).

A total of 199 patients were admitted to the medical and surgical wards on the 4 study days. A total of 161 parent–child dyads were approached for the study and 155 interviews were completed, representing a 96% response rate (Fig 1). The principal interviewee was the child in 50% of interviews. Patients ranged in age from 2 weeks to 21 years with a median age of 8 years (interquartile range 1–15) and 50% (n = 77) of patients were female. A total of 64% (n = 99) of patients spoke English as their preferred language, 35% (n = 54) spoke Spanish, and 1% (n = 2) spoke French or Creole. Patients who preferred a language other than English were considered to have LEP, and these patients were more likely to be younger, female, Hispanic, parental proxies, and have government insurance than English-proficient respondents. The most common reason for admission to the hospital was for an acute illness/infection (67%, n = 104), which was the same between the 2 exposure groups. Additional information about the sample is provided in Table 1.

As seen in Table 2, 60% (n = 93) of all patients surveyed reported experiencing pain in the previous 24 hours. Of these patients, 6% (n = 6) reported mild pain (1–3), 46% (n = 43) reported moderate pain (4 6), and 47% (n = 44) reported severe pain (7–10) on the FPS-R. The most frequent cause of worst pain was an acute illness/infection. The average worst pain score in the previous 24 hours on the 0–10 FPS-R was 6.6 (SD = 2.4). The average pain score for any child having a needle poke in the previous 24 hours was 5.0 (SD = 3.7) on the FPS-R scale. There were no differences in self-reported worst pain scores or needle poke pain scores in terms of preferred language.

Multiple pain tools using a 0–10 scale were used by nurses for documentation, including the Revised Face, Legs, Activity, Cry, Consolability Behavioral Pain Scale, the CRIES neonatal pain assessment tool for infants <2 months of age, the Numerical Rating Scale, and the FPS-R. Children frequently had multiple pain scales documented in the last 24 hours because nurses used the Revised Face, Legs, Activity, Cry, Consolability Behavioral Pain Scale for children who were asleep and either the Numerical Rating Scale or FPS-R for awake, developmentally appropriate children. A total of 34% (n = 32) of the patients who self-reported pain only had pain scores of 0 documented in the EMR during the last 24 hours. Every patient who reported English as a preferred language was asked about their pain only in English during the previous 24 hours (100%, n = 99). However, only 39% (n = 22) of patients with LEP had their pain assessed exclusively in their preferred language in the previous 24 hours, indicating that a majority (61%, n = 34) of these patients were asked about their pain at least once in their nonpreferred language.

In the analysis of our primary outcome, which was limited to those who self-reported pain on the survey (n = 93), 65% (n = 60) had a difference in pain scores of ≥3 (on the 10-point FPS-R scale) between their self-reported worst pain score and the highest nurse-documented pain score in the EMR. There was a statistically significant difference in the number of patients with discrepant pain scores between patients with and without LEP, with 55% (n = 33) of English-proficient patients and 82% (n = 27) of patients with LEP having a difference of ≥3 between their self-reported and nurse-documented worst pain score (P = .01). The unadjusted odds of patients with LEP having a discrepant pain score was 3.68 (95% confidence interval: 1.39–11.03) compared with patients without LEP. When adjusted for age, sex, and survey respondent, the adjusted odds ratio (aOR) was attenuated slightly but remained statistically significant (aOR = 3.23 [95% confidence interval: 1.13–10.31]).

On chart review, the most common pain medications prescribed were acetaminophen, followed by nonsteroidal antiinflammatory drugs, opioids, and adjunctive medications (eg, neuropathic analgesia and muscle relaxants). Of respondents who reported pain on the survey, 74% (n = 69) self-reported receiving analgesia at the time of their worst pain. There was a statistically significant difference in the number of patients receiving analgesia based on preferred language, with 82% (n = 49) of English-proficient patients and 60% (n = 20) of patients with LEP receiving pain medications at the time of their worst pain (P = .03). Although there was a relatively low overall utilization of opioids, more English-proficient patients (22%, n = 22) were prescribed opioids than patients with LEP (9%, n = 5) (P = .04).

Pain assessment is a crucial determinant of pain management, and adequate analgesia is an important component of patients’ well-being. Because pain is subjective, clinicians often rely on the gold standard of self-reported pain scores, which may be influenced by language discordance between patients and clinicians. Although patients with LEP often have worse health outcomes in terms of quality, satisfaction with care, and rates of adverse events, there is little data on how pain is assessed and managed in children with LEP. Therefore, we replicated previous pain audits in a large pediatric population with LEP and found that, although children belonging to minority language groups did not have any differences in self-reported pain measures, there were striking discrepancies in their pain process outcomes, including nursing pain assessments and analgesia.

Although numerical self-reported pain scores are often oversimplifications of a complex pain experience, they are necessary to provide effective pain therapies.^17–19  Our study had the highest number of child respondents (50%) when compared with similar audits, but previous studies with a heavy reliance on proxy-reported pain scores suggest that parental proxy reports are equal to or lower than a child’s self-reported score.^10,43,44  However, although parents may serve as reliable proxies for pediatric pain scores, nurse-reported pain scores have consistently been shown to underreport pain.^7,9,12,16  In a similar audit in 2017, Shomaker et al demonstrated that 51% of pediatric inpatients reported more intense pain on self-report than was documented by nursing.⁷  In our study, nurses frequently underreported pain, with 65% of patients self-reporting a worst pain score of ≥3 points higher than any nurse-documented score in the last 24 hours. This may be because of a lack of procedural pain documentation in the EMR or recall bias of self-reported pain. However, this finding is worrisome since nurse-documented pain is the trigger for pain interventions, and if nurses are not assessing pain accurately, then patients are likely not receiving appropriate analgesia.

Although there were no differences in patient-reported pain prevalence or worst pain scores between English-proficient patients and patients with LEP at our institution, there were significant discrepancies in the process outcomes between these groups. Our study found that patients with LEP had significantly greater odds (odds ratio = 3.68) of having discrepant self-reported and nurse-documented worst pain scores compared with English-proficient patients, and this significant discrepancy persisted after adjusting for age, sex, and survey respondent (aOR = 3.23). This may be secondary to language discordance in pain assessments; although all patients who preferred English were only asked about pain in their preferred language, 61% of patients with LEP were asked about their pain at least once in their nonpreferred language. Multiple studies have documented that trained interpreters are often not used for patients with LEP despite the perceived need by patients and clinicians.^25–28,45  This failure to use interpretation services often leads to poorer understanding of a patient’s disease and treatment plan, and in our study, may have also negatively impacted the communication of pain scores.

Hospitalized patients with LEP were also less likely to receive pain medication at the time of their worst pain and received less opioid analgesia compared with English-proficient patients. This is consistent with previous studies showing that pediatric patients and/or their parents with LEP have longer time to analgesia after surgery and receive less analgesia in the postoperative period than English-proficient patients.^23,46  Additional studies by Goyal et al and others examining pediatric emergency department visits found children from minority racial/ethnic groups were less likely to receive any analgesia and/or opioid analgesia than non-Hispanic White children.^37–41  Although being of a minority racial or ethnic group does not necessarily correlate with LEP (eg, African Americans), our results suggest that LEP may partially explain analgesia discrepancies in these studies when considering some minority populations (eg, Hispanic ethnicities).

The underlying reasons for these disparities are complex and may involve the patient, clinician, and health care system.^35,47  At the individual level, culture and language play a large role in how patients express pain.^19,20,42  Patients may have unique pain behaviors that are not recognized by nurses, and nurses’ preconceived notions regarding certain patient groups may influence their assessment of pain.^20,47  It is well documented that clinicians may have both conscious and unconscious biases when diagnosing and treating minority children with LEP, with less diagnostic imaging and analgesia frequently ordered for these children.^{23,37–41,45,47,48}  At the level of the health care system, there are often time and convenience barriers to utilizing interpretation services, with many clinicians frequently relying on untrained bilingual staff members for communication needs, resulting in poorer health outcomes for patients with LEP.^25,45 

This study had several limitations. The study was conducted at a single center, so it may not be representative of other pediatric populations in the United States. Although our hospital has a 24-hour telephone language line available for interpretation services, there are no formally trained, in-person interpreters available. Twenty-four-hour patient recall of pain has been established as a reliable method to assess pain prevalence and has been used in previous audits; however, there are inherent limitations of recall bias with this method.^5–9,18  Nursing assessments of pain may not happen at the exact moment of patients’ self-reported worst pain, which may account for the discrepancies in pain scores. There were significant differences in the demographics between the 2 exposure groups; we attempted to adjust for these differences using multivariable logistic regression incorporating age, sex, and survey respondent as covariates because these were statistically different between the 2 groups. Because our sample size was relatively small, we were limited in the number of predictors that we could include in the multivariable model to avoid overfitting, and therefore did not include race/ethnicity as a covariate because it is highly correlated with language. Furthermore, language is just 1 of many components that may affect pain outcomes in hospitalized children. Although the child’s preferred language was asked about during the survey, some parental proxies that answered on behalf of their children may prefer a different language, and future studies may consider incorporating other facets of LEP, including preferred language of parents, language spoken at home, comfort level with English, and the satisfaction with interpreter services.

Pain continues to be underrecognized and undertreated in pediatric inpatient populations. Nurses consistently documented lower pain scores than patients self-reported, which may reflect a lack of procedural pain documentation and/or miscommunication about pain. Children with LEP were more likely to be asked about their pain in a nonpreferred language, have discrepant nurse-documented pain scores, and receive less medications, especially opioids, at the time of their worst pain. Educational and quality improvement initiatives are warranted to improve pain assessment and management in this vulnerable pediatric population.