1. Introduction

Pain is the most common complication of sickle cell disease (SCD) and is associated with significant morbidity,14 poor health-related quality of life,19 and premature mortality.7,24 Acute pain episodes are the major cause of adult and pediatric SCD emergency department (ED) visits15 and hospitalizations. Adolescence is associated with an increasing frequency of complications in SCD, particularly pain.13 Children aged 15 to 18 years with SCD make up the largest group of children admitted for pain,12 and approximately 40% of children aged 16 to 18 years have at least 1 episode of health care utilization for pain over a 1-year period.13 The period around adolescence is associated with sex differences in pain21,22 in other painful conditions, but sex differences in acute pain during adolescence are not well described in SCD. In addition, a subgroup of adolescents has high health care utilization (HCU) for pain,13 but it is not known whether their acute pain experience and trajectories differ from those who do not have frequent HCU for pain.

In this study, we sought to determine whether there were differences in acute pain trajectories by sex and frequency of pain episodes among adolescents with SCD who presented to the ED. Given the higher pain burden experienced by adolescents aged 15 to 18 years compared with their younger counterparts, we focused our study on adolescents in this age group.

2. Methods

This was a retrospective study completed at a large, urban, multicampus, academic pediatric SCD program in the United States using existing electronic health record (EHR) data. This study was approved by the Institutional Review Board at the Children's Healthcare of Atlanta.

2.1. Inclusion criteria

All adolescents of 15 to 18 years of age with SCD who previously had either an outpatient or inpatient encounter at least once in the prior 2 years and who had a subsequent ED visit for SCD-related pain in 2019 were identified. A priori, we selected the last ED visit for SCD-related pain in 2019 for each patient as the index ED visit under study, regardless of pain severity or type of treatment received. In identifying the last ED visit in 2019, if there were back-to back ED visits for pain, defined as ED visits for SCD pain that occurred within 7 days of each other, we considered the multiple ED visits to be part of the same pain episode and used the initial ED visit of the series of visits as the index ED visit under study.

2.2. Exclusion criteria

Patients with SCD with a co-occurring pain condition where pain exacerbations could be similar to SCD pain were excluded. We excluded ED visits at which (1) ≤1 pain intensity score was recorded during the visit, (2) adolescents received analgesia in the ED or en route to ED through emergency medical services before documentation of the first pain intensity score, (3) adolescent was transferred from or to another facility before or after presentation at our center, or (4) for visits where pain was explained by an alternative diagnoses or according to study investigators was unlikely to be an SCD-related acute pain episode.

2.3. Demographic and clinical characteristics

Demographic and clinical data were abstracted from the EHR. We calculated past health care utilization (HCU) for pain as the sum of all ED visits and hospitalizations for pain in the 12 months prior to the index ED visit under study. We defined a patient as having high HCU if they had ≥3 visits in the 12 months for pain prior to the index ED visit. The threshold of ≥3 visits in the 12 months for pain is frequently used as a marker of severity in SCD.23,27 Medication history was obtained by review of both outpatient prescriptions and clinical notes. Medications of interest included opioids, nonsteroidal anti-inflammatory drugs (NSAIDs), and adjunctive analgesics for pain. A past diagnosis of avascular necrosis (AVN) was determined based on the review of radiology reports. To calculate steady-state hemoglobin values, we took an average of 2 outpatient steady-state hemoglobin values (if available) in the past year. Pain-related variables collected in the context of clinical care, including pain intensity and the number of locations of reported pain, and medications (opioids, NSAIDs) received were abstracted from the medical record for ED visit under study. We calculated time to receiving time to receiving opioid analgesia referent to the time of arrival to the ED. Length of stay (LOS) was calculated amongst those discharged home as the time between arrival and discharge from the ED. We calculated longitudinal trajectories of pain intensity scores referent to the time of assessment and value of the initial pain intensity score in the ED up to the time of assessment and value of the final pain intensity score in the ED prior to discharge home or admission to the inpatient floor. This period was designated as the “follow-up time” period. We analyzed up to the first 360 minutes (6 hours) from the initial pain score because this represented approximately the 96th percentile (354 minutes) of length of time from the first pain score for the entire cohort. We also counted the number of pain assessments available per individual. Opioids from all routes (intranasal, intravenous, or oral) administered up to the assessment of the final pain score before 360 minutes or up to 360 minutes (to correspond with trajectories of pain intensity) were converted to oral morphine milligram equivalents (MME) as per standard conversion ratios1,18 (1 mg intravenous [IV] morphine = 3 MME, 1 mg IV hydromorphone = 20 MME, 1 μg intranasal [IN] fentanyl = 0.16 MME, 1 μg IV fentanyl = 0.3 MME, 1 mg IV nalbuphine = 3 MME, 1 mg oral hydrocodone = 1 MME, 1 mg oral oxycodone = 1.5 MME), and thereafter, MME was calculated per kilogram of body weight (MME [kg]).

2.4. Statistical analysis

All statistical analyses were performed in CRAN v.4.0.2 (Vienna, Austria) and SAS v.9.4 (Cary, NC). We used descriptive statistics to describe clinical, demographic, and pain-related variables of interest overall, by sex and by history of high HCU for pain. Differences were tested for statistical significance using 2-sample t-tests and χ2 tests of independence or their nonparametric equivalents (ie, Wilcoxon rank–sum, Kolmogorov–Smirnov, and Fisher exact tests). To study pain trajectories over time, we used general linear mixed models through the PROC MIXED procedure in SAS. We first examined a bivariable model, considering the relationship between follow-up time and pain score for the entire cohort. Subsequently, we considered main effects for sex, history of high HCU for pain, and the combination of sex with high HCU for pain. Each of these main effects were statistically interacted with follow-up time, with significant slopes indicating differences in pain score trajectories. All relationships were considered unadjusted for confounders, adjusted alone for MME (in kilograms), and adjusted both for MME (in kilograms) and prescription of hydroxyurea in the last 12 months. For all general linear mixed models, follow-up time, patient sex, HCU, and their various interactions were treated as fixed effects; concurrently, participant-level intercepts and slopes for follow-up time were treated as random effects (ie, random intercepts and random slopes), to account for heterogeneity in pain scores, as well as heterogeneity in change over time between individuals. Time was modeled as a linear term, and degrees of freedom were estimated using the Kenward–Roger method. An unstructured covariance matrix was used in all regression models and calculated separately for main effects when possible. Results from mixed models are presented as β estimates (ie, slope) with 95% confidence intervals.

3. Results

One hundred thirteen patients had ED visits that met study inclusion criteria.

3.1. Demographic and clinical characteristics

Demographic and clinical characteristics of the 113 patients included are presented in Table 1 for the entire cohort, as well as by sex and history of high HCU for pain. The mean age was 16.6 (SD: 0.9) years, and 41.6% were female. The mean body mass index (BMI) was higher among female patients (P = 0.006) and those with high HCU (P = 0.004). The majority (n = 88, 77.9%) had HbSS or a similarly severe genotype (HbSβ0 thalassemia and Hemoglobin SO-Arab), and there were no differences by sex or history of high HCU for pain. The median number of episodes of HCU for pain in 12 months was 2 (interquartile range [IQR] 1–5). Of the total, 43.4% (n = 49) patients had ≥3 episodes of HCU for pain in the 12 months prior to the index ED visit, and there were no sex differences in the history of high HCU for pain. Patients with known AVN comprised 15% (n = 17) of the total cohort with no differences in the history of known AVN between sexes or with history of high HCU for pain.

Table 1 - Demographic and clinical characteristics. Characteristic Overall Sex HCU P Male Female P Low HCU High HCU n 113 66 47 64 49 Age (mean [SD]) 16.6 (0.9) 16.7 (0.9) 16.6 (0.9) 0.466 16.5 (0.9) 16.8 (0.8) 0.170 Sex (n, %)  Male 66 (58.4) 40 (62.5) 26 (53.1) 0.414  Female 47 (41.6) 24 (37.5) 23 (46.9) Weight in kilograms (mean [SD]) 62.9 (16.5) 61.8 (15.4) 64.4 (18) 0.412 59.7 (14) 67 (18.6) 0.018 Body mass index (mean [SD])* 22.3 (5.3) 21.1 (4.1) 24 (6.4) 0.006 21 (3.9) 23.9 (6.4) 0.004 Genotype (n, %)  Hemoglobin SS/Hemoglobin S-β0 thalassemia/Hemoglobin S-OArab 88 (77.9) 49 (74.2) 39 (83.0) 0.383 46 (71.9) 42 (85.7) 0.127  Hemoglobin SC/Hemoglobin S-β+ thalassemia 25 (22.1) 17 (25.8) 8 (17.0) 18 (28.1) 7 (14.3) Presence of avascular necrosis (n, %) 17 (15.0) 10 (15.2) 7 (14.9) 1 7 (10.9) 10 (20.4) 0.258 HCU for pain in past 12 mo (median [IQR]) 2 [1, 5] 1[0, 4.8] 2 [1, 5.5] 0.152 1 [0, 1] 6 [4, 9] <0.001 Presence of 3 or more visits in the part 12 mo (n, %) 49 (43.4) 26 (39.4) 23 (48.9) 0.414 Disease-modifying therapy (hydroxyurea or l-glutamine) in past 12 mo (n, %) 73 (64.6) 42 (63.6) 31 (66) 0.956 33 (51.6) 40 (81.6) 0.002  Hydroxyurea 72 (63.7) 41 (62.1) 31 (66.0) 0.826 32 (50.0) 40 (81.6) 0.001   l-Glutamine 11 (9.7) 4 (6.1) 7 (14.9) 0.196 2 (3.1) 9 (18.4) 0.009 Chronic transfusion therapy in past 3 mo 5 (4.4) 4 (6.1) 1 (2.1) 0.399 4 (6.2) 1 (2.0) 0.386 Home medications (n, %)  Short-acting opioids (hydrocodone, oxycodone, hydromorphone, tramadol, or oral morphine-immediate release) 109 (96.5) 63 (95.5) 46 (97.9) 0.639 60 (93.8) 49 (100.0) 0.131  Long-acting opioids (methadone or oral morphine-controlled release) 5 (4.4) 2 (3.0) 3 (6.4) 0.647 0 (0.0) 5 (10.2) 0.013  NSAID  Ibuprofen/naproxen 100 (88.5) 60 (90.9) 40 (85.1) 0.513 56 (87.5) 44 (89.8) 0.935  Celecoxib/meloxicam 9 (8.0) 4 (6.1) 5 (10.6) 0.486 0 (0.0) 9 (18.4) <0.001  Adjunctive pain medications 39 (34.5) 21 (31.8) 18 (38.3) 0.608 12 (18.8) 27 (55.1) <0.001  Clonidine 9 (8.0) 5 (7.6) 4 (8.5) 1 1 (1.6) 8 (16.3) 0.009  Gabapentin/pregabalin 9 (8.0) 3 (4.5) 6 (12.8) 0.16 0 (0.0) 9 (18.4) <0.001  Amitriptyline 1 (0.9) 1 (1.5) 0 (0.0) 1 0 (0.0) 1 (2.0) 0.433  Muscle relaxants (methocarbamol, cyclobenzaprine, or tizanidine) 37 (32.7) 21 (31.8) 16 (34.0) 0.964 12 (18.8) 25 (51.0) 0.001 Baseline hemoglobin g/dL (mean [SD])† 9.8 (1.6) 10 (1.8) 9.5 (1.4) 0.153 9.7 (1.8) 9.8 (1.4) 0.78

Hypothesis testing done by t test with equal variances or non-parametric equivalent (ie, Wilcoxon rank sum test) for continuous variables and χ2 with Yates continuity correction/Fisher exact test for categorical variables.

*n = 110.

†n = 105.

HCU, health care utilization; IQR, interquartile range; NSAID, nonsteroidal anti-inflammatory drugs.

Most patients (64.6%, n = 73) were prescribed disease-modifying therapies for SCD, but patients with high HCU were more likely to be prescribed hydroxyurea (P = 0.001) or l-glutamine (P = 0.009). On review of prescribed medications at home, almost all patients were prescribed short-acting opioids and NSAIDs, but those with high-HCU for pain were more likely to receive celecoxib or meloxicam (P < 0.001). A minority (n = 5, 4.4%) were prescribed long-acting opioids or methadone, but all of these individuals had high HCU for pain (P = 0.013). Approximately one-third (n = 39, 34.5%) were prescribed adjunctive medications for pain, which were more likely to be prescribed in those with high HCU for pain (P < 0.001).

3.2. Outcomes in emergency department

Characteristics of the index ED visit under study are presented in Table 2 for the entire cohort, as well as by sex and history of high HCU. Most participants (n = 86, 76.1%) had taken either an opioid or an NSAID before presentation to the ED. There was no difference in mean initial pain intensity scores at presentation by sex, but mean pain intensity score was slightly higher in those with high HCU for pain. The mean number of pain locations was approximately 2 (SD 1), and there were no differences between the sexes or by history of high HCU for pain. The mean time to opioid analgesia was 36.2 minutes (SD 25.5), and there were no sex differences or differences by history of high HCU for pain. There were no differences in total dose of opioid analgesia per kilogram body weight administered in the ED between sex and history of high HCU for pain, although those with high HCU for pain were more likely to receive IV opioids (P < 0.001). More than half (n = 65, 57.5%) received IN fentanyl, 80.5% (n = 91) received IV ketorolac, 52.2% (n = 59) received PO opioids, almost all received an opioid medication (98.2%) and all received either an opioid or IV ketorolac. Most (n = 86, 76.1%) received IV fluids, with no differences based on sex or history of high HCU for pain.

Table 2 - Pain and analgesia received in the emergency department (ED). Characteristic Overall Sex HCU P Male Female P Low HCU High HCU n 113 66 47 64 49 Opioid use prior to ED visit (n, %) 57 (50.4) 37 (56.1) 20 (42.6) 0.221 35 (54.7) 22 (44.9) 0.400 NSAID use prior to ED visit (n, %) 59 (52.2) 32 (48.5) 27 (57.4) 0.454 31 (48.4) 28 (57.1) 0.467 Opioid or NSAID use prior to ED visit (n, %) 86 (76.1) 51 (77.3) 35 (74.5) 0.904 46 (71.9) 40 (81.6) 0.326 Initial pain score (mean [SD]) 7.8 (2) 7.6 (2.2) 8.1 (1.7) 0.207 7.4 (2.2) 8.3 (1.6) 0.023 No. of pain locations (mean [SD]) 1.9 (1) 2 (1.1) 1.8 (1) 0.234 1.9 (1) 1.9 (1.1) 0.887 Time to opioid analgesia in minutes (mean [SD])* 36.2 (25.5) 33.1 (18.6) 40.8 (32.8) 0.119 33.4 (22.3) 39.8 (29) 0.188 Medications received in ED (n, %)  Intravenous opioid (morphine, hydromorphone, nalbuphine, or fentanyl) 93 (82.3) 53 (80.3) 40 (85.1) 0.682 45 (70.3) 48 (98.0) <0.001  Intranasal fentanyl 65 (57.5) 42 (63.6) 23 (48.9) 0.172 38 (59.4) 27 (55.1) 0.792  Intravenous opioid or intranasal fentanyl 106 (93.8) 63 (95.5) 43 (91.5) 0.447 57 (89.1) 49 (100.0) 0.018  Intravenous ketorolac 91 (80.5) 52 (78.8) 39 (83.0) 0.754 51 (79.7) 40 (81.6) 0.985  Intravenous opioid or intranasal fentanyl or intravenous ketorolac 113 (100) 66 (100.0) 47 (100.0) NA 64 (100.0) 49 (100.0) NA  Oral opioid (hydrocodone, oxycodone or hydromorphone) 59 (52.2) 34 (51.5) 25 (53.2) 1 30 (46.9) 29 (59.2) 0.268  Any intravenous/intranasal/oral opioid 111 (98.2) 66 (100.0) 45 (95.7) 0.171 62 (96.9) 49 (100.0) 0.504 Oral morphine milligram equivalents (MME) per kilogram body weight† 0.70 (0.37) 0.73 (0.35) 0.65 (0.40) 0.240 0.64 (0.37) 0.77 (0.38) 0.055 Received intravenous fluids 86 (76.1) 50 (75.8) 36 (76.6) 1 45 (70.3) 41 (83.7) 0.153 Laboratory values  White blood cell count (X 10–9/L) 12.3 (4.8) 12.6 (4.8) 12 (4.9) 0.571 12.3 (4.7) 12.3 (4.9) 0.988  Hemoglobin (g/dL) 9.8 (1.8) 10.2 (1.9) 9.3 (1.4) 0.011 9.9 (2) 9.7 (1.3) 0.432  Platelets (X 10–9/L)‡ 388.2 (176) 365.2 (162.7) 421.2 (190.5) 0.098 367.1 (172.9) 415.4 (178) 0.151 Admitted to hospital from ED (n, %) 52 (46.0) 31 (47.0) 21 (44.7) 0.961 22 (34.4) 30 (61.2) 0.008 Length of stay if discharged home from ED (mean [SD]) 235.8 (72) 220.5 (70.2) 256.5 (70.3) 0.052 221.6 (71.6) 267.3 (63.6) 0.020

Hypothesis testing done by 2-sample t test with equal variance for continuous variables and χ2 with Yates continuity correction/Fisher exact tests for categorical variables.

*n = 111.

†MME calculated up to assessment of the final pain score before 360 minutes or up to 360 minutes. Home methadone dosing not incorporated in calculation.

‡n = 112.

HCU, health care utilization; NSAID, nonsteroidal anti-inflammatory drugs.

Patients discharged home from the ED had a mean LOS of 235.8 minutes (SD 72). Amongst those discharged home, female patients had a trend towards longer LOS (P = 0.052) and those with high HCU for pain had a longer LOS (P = 0.020). Those with high HCU for pain were also more likely to be admitted to the hospital.

3.3. Longitudinal pain trajectories

The mean follow-up time of assessment of the last pain score (referent to the time of the first pain score) was 214.3 minutes (SD 73.2). The mean number of pain score assessments was 6.34 (SD 2.47). Although the follow-up time of assessment of the last pain score was longer amongst those with high HCU (P = 0.019), there were no significant differences in the number of pain score assessments in those with high HCU as compared with those with low HCU. Results from a bivariable linear mixed model, examining trend in pain score over time, found a significant decrease in pain score for the entire cohort, with an average 0.5-point drop for every 30 minutes (each 1-minute slope = −0.016, 95% CI: −0.018, −0.013, P < 0.001; Table 3, Fig. 1A). Similar models adjusting for MME (kg) alone and MME (kg) plus use of hydroxyurea each found a common relationship between follow-up time and pain scores (each 1-minutes slopes = −0.015 and −0.015, respectively, both P < 0.001; Table 3). Using this same base linear mixed model and adding both patient sex as a main effect and the statistical interaction between follow-up time and patient sex, the change in pain score over follow-up between male and female patients was found to be insignificant (male 1-minute slope = −0.017, female 1-minute slope = −0.015, P = 0.511; Table 4, Fig. 1B). After adjustment for MME (in kilograms) alone and MME (in kilograms) plus hydroxyurea, the statistical interactions between follow-up time and patient sex remained insignificant (P = 0.508 and P = 0.509 respectively; Table 4). Implementing these same methods, and substituting patient sex with HCU, the change in pain score over follow-up between low HCU and high HCU was found to be significant (low HCU 1-minute slope = −0.019, high HCU 1-minute slope = −0.011, P = 0.001; Table 5, Fig. 1C). After adjustment for MME (in kilograms) alone and MME (in kilograms) plus hydroxyurea, the statistical interactions between follow-up time and HCU remained significant (both P = 0.001; Table 5). Finally, retaining both patient sex and HCU as main effects, and considering the 3-way interaction between patient sex, HCU, and follow-up, a significant interaction was found with female-low HCU as having the steepest slope in pain score over follow-up (1-minute slope = −0.021), followed by male-low HCU (1-minute slope = −0.019) and male-high HCU (1-minute slope = −0.013). Female-high HCU had the flattest slope (1-minute slope = −0.009), meaning that their change in pain scores over follow-up was the least of the 4 patient combinations. The trajectory for female-high HCU was significantly different from female-low HCU and male-low HCU (both P = 0.005; Table 6, Fig. 1D). After adjustment for MME (in kilograms) alone and MME (in kilograms) plus hydroxyurea (both statistically significant), the same statistical interactions remained significant (Table 6). Specifically, compared with the referent slope for female-high HCU, the slope for female-low HCU was 2.5 times greater (−0.020 vs −0.008) and for male-low HCU, it was 2.25 times greater (−0.018 vs −0.008).

Table 3 - All results unadjusted and adjusted for (1) morphine milligram equivalents (in kilograms) and (2) morphine milligram equivalents (in kilograms) and hydroxyurea. Effect Unadjusted
β (95% CI) P Adjusted 1
β (95% CI) P Adjusted 2
β (95% CI) P Intercept 7.49 (7.11, 7.86) <0.001 5.88 (5.14, 6.62) <0.001 5.68 (4.84, 6.53) <0.001 Follow-up (min) −0.016 (−0.018, −0.013) <0.001 −0.015 (−0.018, −0.013) <0.001 −0.015 (−0.018, −0.013) <0.001 MME (kg) — — 2.28 (1.35, 3.21) <0.001 2.37 (1.42, 3.31) <0.001 Hydroxyurea  No — — — — 0.38 (−0.35, 1.12) 0.305  Yes — — — — Reference

95% CI, 95% confidence interval; MME, morphine milligram equivalents.

Figure 1.:

(A) Unadjusted trends in pain score over follow-up with 95% CI. (B) Unadjusted trends in pain score over follow-up by sex, with interaction of slopes P value and 95% CI. (C) Unadjusted trends in pain score over follow-up by health care utilization, with interaction of slopes P value and 95% CI. (D) Unadjusted trends in pain score over follow-up by sex and health care utilization, with interaction of slopes P value. 95% CI, confidence interval; HCU, health care utilization.

Table 4 - Results by sex, unadjusted and adjusted for (1) morphine milligram equivalents (in kilogram) and (2) morphine milligram equivalents (in kilogram) and hydroxyurea. Effect Unadjusted
β (95% CI) P Adjusted 1
β (95% CI) P Adjusted 2
β (95% CI) P Intercept 7.91 (7.40, 8.42) <0.001 6.36 (5.57, 7.15)