The Impact of Standardized Recovery Pathways on Language Barriers and Inpatient Pain Management

This retrospective cohort study was conducted among AIS-PSF patients who were surgically managed at the Children’s Hospital At Montefiore from February 2017 to December 2021. All patients were aged between 10 and 20 years, which represents the predominant range for surgical management of AIS. Patients were excluded if they had congenital disorders, comorbidities reliant upon analgesia for treatment, intellectual or behavioral disorders that could interfere with care team communication, a history of cancer, or unusual hospital courses consisting of returns to the operating room or rapid response events. These exclusions were made because of the potential of confounding our outcome measures or a presumed inability to adhere to the postoperative protocol.

Children’s Hospital At Montefiore is a large, academic tertiary care hospital in Bronx, New York, where 58% of residents aged 5 years and older live in households where non-English languages are primarily spoken.⁵⁴  At our hospital, daytime in-person Spanish interpreters are available upon request by providers. Other language and nondaytime Spanish interpretations are done via video or telephonic services, and patients and families are asked throughout their care episode which language they prefer to communicate.

In 2017, our hospital adopted a postoperative protocol for AIS-PSF patients similar to what has been described in literature.^46–50  In our hospital-specific protocol, patients use intravenous (IV) patient-controlled analgesia (PCA) for primary pain control from arrival in the pediatric ICU after surgery to postoperative day (POD) 1 or 2 depending on patient progress. Patients are typically placed on a PCA pump that is demand-driven, meaning patients receive medication when they self-push the pump. However, if patients are in significant pain, they may be placed on a basal or continuous rate of medication or given boluses for additional relief.

After discontinuation of PCA, patients have an IV or oral opioid, usually oxycodone, ordered pro re nata (PRN) until discharge. For pain specifically related to muscle spasms, patients receive an IV or oral benzodiazepine, usually diazepam, PRN. For around-the-clock relief, patients have scheduled IV ketorolac (followed by scheduled oral ibuprofen), as well as scheduled IV and oral acetaminophen throughout the hospitalization. The patient’s clinical status determines whether a medication is given IV or oral, and the amount given is weight-based. Upon discharge, patients are typically prescribed ibuprofen, acetaminophen, and short courses of oxycodone and diazepam. If additional medications are needed beyond these, the patient–family are told to call the physician’s office, or they can address these needs at the 2-week follow-up visit. Details of our hospital-specific protocol are outlined in Supplemental Table 4.

As part of the pathway, patient–families are given instructions, available in English or Spanish, explaining what to expect during their hospitalization and after discharge.

Language of the patient–family was determined through review of electronic medical records. Patient–families were considered LOE if:

1. the surgical informed consent was performed in a non-English language; and/or

2. use of a language interpreter was documented during the hospitalization.

Our primary outcome measures pertain to inpatient pain management. These include total time spent on PCA, number of PCA doses per hour, whether a patient was placed on a basal dosing of PCA or given additional PCA boluses, number of PRN pain medications requested per POD (opioids, benzodiazepines, and both), and total opioids consumed (oral morphine equivalents per kilogram per POD). If patients were erroneously ordered scheduled dosing of 2 or more consecutive opioid or benzodiazepine medications, instead of PRN, they were excluded from our PRN medication analyses. We also assessed need for postoperative ondansetron and diphenhydramine, because the pathway has decreased the occurrence of opioid-induced nausea and vomiting and pruritis.⁵⁵  These were collected from the medication and intake logs from the hospitalization record.

We looked at average pain scores, number of pain assessments, and the proportion of behavioral pain assessments. Nursing elicits pain through verbal and behavioral methods. Verbal options included numeric or verbal rating scales. Behavioral options included the Wong-Baker FACES scale, which asks the patient to point to a face reflecting their level of pain,⁵⁶  or the Face, Legs, Activity, Cry, and Consolability behavior scale, for which nurses designate a score on the basis of patient positioning and body language.⁵⁷  These were collected from the events log of the hospitalization record. Patients who underwent surgery before implementation of the current pain assessment methods in October 2017 were excluded from pain score analyses.

Additional variables measured include postoperative length of stay, which the pathway has been shown to reduce^{49–53,58–60} ; emergency department revisits or hospital readmissions within 6 months, which can be higher for LOE patients more broadly^12,18,34 ; and follow-up pain at 6 weeks and 6 months, when postoperative patient-reported outcomes are assessed. Outpatient pain level was determined using patient responses to surveys designed by the Scoliosis Research Society (SRS) to elicit patient-reported outcomes during treatment. Patients completed the SRS-30 or the SRS-22r surveys; the SRS-22r survey has a translated version that has been validated for Spanish-speaking patients.^61–63  Patients were asked the same questions related to pain regardless of which survey they completed. The survey is scored out of 5.0; higher scores correlate to better responses. Pain scores <4.0 were considered clinically significant, which included the presence of mild back pain, a continued need for nonnarcotic pain medications, and/or limitations to everyday activities because of pain.⁶⁴ 

Demographic and operative variables were collected to ensure there were nonsignificant dissimilarities between patient cohorts and to account for potential confounders in outcomes. Demographic variables included age, sex, and BMI percentile. Operative variables included year of surgery, preoperative pain, American Society of Anesthesiology score, major curve angle, number of vertebral levels fused, number of Ponte osteotomies, length of the operation, estimated blood loss, and whether the patient needed a transfusion intra- or postoperatively.

Univariate analyses (student’s t tests, Wilcoxon ranked sum tests, and χ² tests) examined associations between LOE status or covariates and outcome measures. Associations that were significant (P < .05) were included in multivariable regression models (logistic or linear regression depending on the outcome).

This study was approved by the institutional review board. All analyses were performed using Stata (Version 17.0, StataCorp, College Station, TX).

A total of 160 patients were included, 39 of whom were LOE (24%). Spanish was the only LOE represented. Demographic and operative variables are described in Table 1 by LOE status. There were no significant differences between groups.

Table 2 describes the relationships between LOE status and outcomes. There was no difference in the number of PCA doses per hour (1.01 [0.74–1.33] vs 0.76 [0.49–1.18]; P = .07), but LOE patients were more likely to be given basally dosed PCA or additional boluses (10 of 39 [26%] vs 12 of 121 [12%]; P = .03). Meanwhile, LOE patients received fewer benzodiazepine PRNs per day (0.51 [0.20–1.02] vs 1.04 [0.34–1.68], P = .01). The number of patients excluded from the PRN frequency analyses because of scheduled orders are found in Table 2 footnotes. More patients were excluded from the analysis of total PRNs (opioids and benzodiazepines) than from the analysis of opioid PRNs, which is why our results suggest higher median numbers of opioid PRNs than total PRNs consumed per day.

There were no differences in consumed oral morphine equivalents per kilogram per POD (1.44 [1.02–2.31] vs 1.24 [0.83–1.77]; P = .18) or the likelihood of needing medication for nausea and vomiting (29 of 39 [74%] vs 75 of 121 [62%]; P = .16). Further, there were no significant differences in the number of pain assessments, average pain scores, or percentage of behavioral pain assessments (Table 2). Lastly, among patients who had completed follow-up surveys, LOE status was associated with higher likelihood of pain at 6 weeks (18 of 27 [67%] vs 36 of 84 [43%], P = .03) and 6 months postoperatively (8 of 22 [36%] vs 11 of 69 [16%]; P = .04,), despite nondifferent preoperative pain (Table 1: 12 of 29 [41%] vs 32 of 95 [34%]; P = .45).

On multivariable regression, LOE status was associated with likelihood of being given basally dosed PCA or additional boluses (odds ratio 3.19 [95% confidence interval 1.15–8.85]; P = .03), but was not associated with benzodiazepines per POD or significant pain at either follow-up (Table 3). On univariate analysis, there were no covariates significantly associated with pain at 6 months postoperatively. Because of this, we used the same covariates as in our 6-week multivariable model under the assumption that there would be similar associative pathways for pain at each period.

Caring for LOE patients presents unique challenges in health care, particularly when care is reliant upon clear, real-time communication between patient–families and providers. In our study, we did not detect significant differences in the inpatient pain management of LOE patients except when it came to augmenting a patient’s PCA regimen with a basal dosing or additional boluses. Our findings suggest that standardized protocols, such as our protocol for AIS-PSF patients, can mitigate disparities in care that LOE patients may receive, but there are elements that require further review. To our knowledge, no other studies have examined the influence of language discordance on postoperative recovery protocols.

LOE patients were more likely to be given basally dosed or additional boluses of PCA. This decision is made early in the postoperative course and is typically the result of extreme levels of pain. Interestingly, there were no differences in the reported pain scores or frequency of pain assessments between patients who received these additional pain-control measures and those who did not. Because of the retrospective nature of our study, we cannot comment with certainty as to why these differences were observed; however, it is evident that there are elements of the postoperative pathway that may be susceptible to disparities and thus require further monitoring as we continue utilizing this protocol. Further, we found that female sex was associated with lower odds of receiving these additional forms of PCA. Similarly, we cannot comment with certainty as to why this is, but this finding warrants more investigation.

Although our other results did not demonstrate significance, there are interesting findings to acknowledge. When differentiating between the types of PRN medications dispensed to patients, LOE patients on univariate analysis received fewer benzodiazepines for pain control, though this significance was lost in multivariable analysis. Benzodiazepines are part of the multimodal pain control strategy that is integral to postoperative scoliosis pathways and are used specifically for muscle spasms.^49,50  With language discordance, it may be difficult to articulate muscle spasm-related pain. Furthermore, care teams may be more likely to rely on opioids when LOE patients report any type of pain if the quality and type of pain are not discussed. This hypothesis could also explain why older age was associated with higher benzodiazepine usage (Table 3). Older patients may be more likely to articulate specific symptoms with the care team, leading to a more targeted therapeutic approach. It is also possible that language discordance influences a patient–family’s ability to request medication, but our study did not have the power to detect such a difference.

LOE patients demonstrated higher rates of PCA doses, though nonsignificantly. With PCA, patients can dispense medication on their own without needing to communicate with a provider. It is possible that language discordance contributed to LOE patients receiving more doses of the self-administered PCA, while receiving fewer doses of the communication-dependent PRN medications. However, because of the lack of statistical significance in our findings, we can only speculate on this theory which should be investigated further with a higher-powered study.

In our study, both cohorts had similar frequencies of pain assessments and levels of inpatient pain. Although other studies have found disparities in these measures among LOE patients,^21,23,24  our findings parallel what Plancarte et al found when studying patients within the same hospital system.²⁶  Likewise, LOE patients were not more likely to have behavioral pain assessments as Jimenez et al demonstrated.²³  Although it is possible that our postoperative protocol has led to more standardized pain evaluation, our specific hospital may also have an infrastructure that is less vulnerable to disparities in these metrics. We discuss hospital-specific factors later in the discussion.

We identified that LOE patients were more likely to experience significant pain at routine follow-ups. It is thus possible that the influence of language discordance extends beyond the inpatient setting and can impact patients as they recover and return to normal functioning. LOE patient–families may be more reluctant to engage providers with questions after discharge, as suggested by Khan et al with regard to medical decisions and safety concerns.⁶⁵  Other studies have similarly demonstrated a disconnect with comprehensive discharge communication with LOE patient–families.^3,66  These disparities, however, were not seen when controlling for other variables associated with pain at follow-up, including significant preoperative pain and major scoliotic curve.

Regarding our hospital, it is possible there are setting-specific factors that influenced our results. Several studies have demonstrated the importance of language-concordant care on LOE patients and families.^2,5,67–73  Several members of our care teams, including nurses with whom patient–families interact the most, live in Bronx, New York and come from bilingual backgrounds. Further, as previously mentioned, a large proportion of the hospital’s overall patient population come from households where English is not the primary language, so our providers may be more equipped to care for LOE patients more broadly. Formally, we provide resources for non-English–speaking families, including postoperative information for scoliosis patients available in non-English languages and 24/7 interpreting services. It is possible that our findings would not be replicated in health systems in different geographic locations and with different workforces.

Next, there are limitations to acknowledge. First, our study was limited by size from the relative recency of our postoperative scoliosis pathway. It is possible that some of our results would have achieved significance with a larger cohort. As such, we will continue to monitor this study in years to come. Additionally, all LOE patients were Spanish-speaking, the second most spoken language in the United States. It is possible our results would differ if other languages were represented. The US Spanish-speaking population is incredibly heterogeneous, so it is also possible our results would differ in areas with different backgrounds of LOE patients. Further, our method of determining language status did not account for language proficiency of the pediatric patient or if guardians of the same patient had varying levels of English proficiency because any use of interpreters during the hospitalization or surgical consent process led to LOE classification. This could bias our results toward the null if an English-speaking patient or family member could converse with providers more comfortably than the LOE parent. Although parental language is critical in all aspects of caring for the pediatric patient, future iterations of this study will attempt to examine the impact of language proficiency of the patients themselves, as well as differences within a family unit, to see if results are influenced. We excluded patients with comorbidities that could interfere with care team communication. Future studies should examine disparities in standardized pathways within this subgroup. Next, several of our variables, namely those related to PCA consumption and pain assessments, were collected through chart review of patient intake and event logs. It is possible that errors were made during data entry at the time of hospitalization that would have been unknown during data collection. Last, our study is subject to limitations of retrospective analyses. We cannot conclude if language discordance was the cause of any of our findings; rather, we can only acknowledge the association, or lack thereof, between language discordance and our outcomes.

There are key takeaways to glean from our analyses regarding the management of LOE patients. First, standardized protocols can help deliver more standardized care to patients. However, it is critical to evaluate aspects of protocols that may be vulnerable to language discordance. In our study, we found that LOE patients were more likely to receive augmented PCA regimens, despite seemingly similar measures of pain. Further, LOE patients may request fewer PRN medications while consuming more self-administered analgesia (ie, PCA). They also may face difficulties describing complex symptoms, such as muscle spasms, and thus receive fewer targeted therapeutics, such as our protocol’s use of benzodiazepines. Lastly, postdischarge management may also be hindered by language discordance. Thus, efforts, such as those that went into the development of inpatient standardization protocols, may be beneficial in the postdischarge setting to ensure that quality care continues once patients leave the hospital.