Impact statements

  • Adults with cerebral palsy (CP) are often medically complex, have insufficient clinical-pharmacy care coordination, and can experience pain symptoms from various, but little understood etiologies.

  • This multi-level, nationwide study from the United States found that adults with versus without CP were more likely to be exposed to opioids, for a higher number of days per month, and for a longer duration after the initial opioid exposure.

  • The effects of greater opioid exposure for adults with CP is unknown, but may lead to adverse side effects in the short-term or long-term.

  • Clinical care may need to consider history of opioid exposure when making current health assessments and developing preventive strategies to counterbalance potential adverse effects of greater opioid exposure.

Introduction

Individuals with cerebral palsy (CP) experience many factors that increase the risk of opioid exposure across the life course, such as high rates of pain symptoms from an array of pain generators (e.g., surgeries, injuries, musculoskeletal abnormalities) and etiologies [1,2,3,4,5,6,7]. Yet, the risk-benefit ratio of opioid therapy for acute and chronic pain is unknown for adults with CP. This calls into question if various opioid use patterns lead to a net benefit or net harm, in light of the lack of guidelines for opioid prescribing specific to adults with CP.

Clinically, opioids can be prescribed for acute and long-term therapeutic purposes. Long-term opioid therapy for chronic pain management is associated with mental health and substance abuse disorders, fractures, and cardiovascular events [8,9,10]. Acute opioid therapy can lead to respiratory depression via neural mechanisms resulting in brain damage or death if left untreated [11, 12]. Mental health disorders, bone fragility, and cardiometabolic disease are common health issues for adults with CP [13,14,15,16], and long-term opioid exposure may exacerbate declines in these physiological systems. Individuals with CP also have an altered neural anatomy and physiology, brain stem depression even without opioid therapy, and high risk of respiratory morbidity and mortality [17, 18], which may be further complicated by acute or long-term opioid exposure. On the other hand, opioid therapy may be necessary in certain circumstances, potentially leading to more favorable outcomes than not treating the individual with an appropriate opioid regime. However, there are non-opioid alternatives for pain management that may be better positioned to optimize risk-benefit for a particular situation, such as analgesics, physical therapy, and behavioral therapy. Therefore, surveillance efforts are needed to understand opioid prescription patterns among adults with CP, which may serve as a basis for optimizing opioid and non-opioid therapeutic strategies.

Emerging research is documenting opioid prescription patterns among children and adults with CP [19,20,21,22], which is necessary to understand how opioids are being prescribed to this population. Such knowledge can inform health systems to prepare for complications that arise from various opioid therapeutic strategies, such as short- and long-term side effects and their disease sequelae. However, inferences on opioid prescription patterns are limited as the available studies have been cross-sectional, included small sample sizes, or provided estimates over broad time intervals (e.g., yearly), which does not fully capture the duration, regularity, or volume of opioid exposure at the population- or individual-level.

Aim

The first aim was at the population-level and designed to quantify opioid exposure monthly over a 7-year period for adults with versus without CP. The second aim was at the individual-level and designed to track opioid exposure monthly for 1-year after the first opioid exposure month.

Ethics approval

The University’s Institutional Review Board confirmed that this study was exempt (HUM00158800) given the de-identified nature of the data and that patient consent was not required.

Method

Design, database, and setting

This retrospective study accessed patient-level claims data from 01/01/2011-12/31/2017 from Optum’s de-identified Clinformatics® Data Mart Database, which contains medical and outpatient pharmacy claims with representation across the United States. [23]. Individuals that have enrollment in a private payer plan either pay for their coverage, are covered by their employer or spouse’s employer, or are covered by their caregiver’s insurance up to 26 years old. Claims are used for billing reimbursement of healthcare services.

Claims submitted by providers are primarily used for billing reimbursement of healthcare services. Claims can be viewed as “real-world” clinical data. For research, medical conditions can be identified by searching for unique codes attached to claims, which are presented in Supplementary material 1, and outpatient pharmacy prescriptions can be identified by searching for medication variables of interest (e.g., name, fill date, days supplied). Healthcare services with clinical information in the form of claims are not “carried forward” in time. Rather, the clinical diagnoses and outpatient pharmacy prescriptions can only be detected during the sampling period.

The time period analyzed does not necessarily reflect the current landscape of opioid prescription patterns, given major events that may impact opioid prescriptions, such as opioid prescribing guidelines and COVID-19. However, estimates may provide reasonable “baseline” information to inform future research.

Sample selection

For the first aim, adults ≥ 18 years old with CP with a study entry date between 01/01/2011 and 2/31/2016, that had ≥ 12-months of continuous health plan enrollment 01/01/2011–12/31/2017 starting from their study entry date, and without missing data on key variables (i.e., age, sex, and region) were included (flow chart is presented in Supplementary material 2). For the second aim, those from the first aim’s cohort that had opioid exposure were followed from their first opioid exposure month. To obtain a sufficient baseline period and follow-up time, individuals were required to have continuous enrollment for 3-months prior to and 11-months following the first opioid exposure month. The allowable start date was 04/01/2011 to 01/01/2015 to avoid follow-up time overlapping with the March 2016 release of the opioid prescribing guidelines from the Centers for Disease Control and Prevention, which may impact the individual-level patterns that cross this time period. To optimize sensitivity and specificity, adults with CP were identified by ≥ 2 claims containing a pertinent code for CP, where each claim had to be on a separate day within 12-months of one another.

Adults without CP were defined as those with 0 claims containing a pertinent code for CP. For the first aim, adults without CP were matched to each person with CP 1:20 (case:control) for age (± 2 years), sex, state, and study entry year. For the second aim, adults without CP were matched to each person with CP 1:5 for age (± 2 years), sex, state, and month/year of the first opioid exposure month.

Opioids

Exposure to opioids was captured using prescription information available in the pharmacy claims (e.g., medication name, fill date, days supplied). Opioid exposure was examined at monthly intervals from each person’s study entry date to their drop in health plan enrollment or end of study period, whichever came first. Individuals were included in the analysis per month if they had enrollment for that entire month. Opioid exposure was determined by an outpatient pharmacy fill for a product containing hydrocodone, oxycodone, tramadol, codeine, morphine, fentanyl, and “other”, which included hydromorphone, buprenorphine, propoxyphene, oxymorphone, methadone, dihydrocodeine, levorphanol tartrate, meperidine hydrochloride, opium, pentazocine, and tapentadol.

The timing of opioid exposure was based on the date of prescription and number of days supplied, which allowed for the determination of monthly exposure as binary (yes/no) and number of days supplied. To standardize the number of days supplied per month, the proportion of each month exposed was calculated as the days supplied divided by the number of days in that month. This study did not standardize doses across opioid types (e.g., oral morphine equivalents) given the variability in suggested conversion factors [24]. Further, it is unknown if the proportion of opioid prescriptions by type differs for adults with and without CP, and how variation in conversion factors could impact interpretations. Therefore, this study focused on measures of opioid exposure as exposed/not exposed, the number of days supplied, and the proportion of opioid prescriptions by type.

Characteristics

Sex, race, and U.S. region of residence were retrieved from the baseline period. Age was determined based on the study start date, which was assessed separately for each aim. Co-occurring intellectual disabilities and epilepsy were determined in the same manner as CP, and were examined based on their clinical relevance [25, 26].

Statistical analysis

For the first aim, baseline characteristics were described for adults with CP and matched adults without CP. Monthly opioid exposure was graphically depicted as the prevalence of any opioids (with 95% confidence intervals (CI)) and the median number of days supplied (with the 25th and 75th percentile). The proportion of opioid prescriptions by type was graphically depicted quarterly.

For the second aim, group-based trajectory modelling (GBTM) was used for adults with and without CP separately to identify 1-year trajectories of opioid exposure as the monthly proportion of days supplied starting with the first opioid exposure month. The GBTM technique is a semi-parametric application of finite mixture modeling that identifies groups of individuals based on inter-individual similarities of the baseline value and trajectory pattern of the outcome under investigation [27]. In the context of this study, GBTM may help to identify various opioid prescription patterns based on chronicity and volume of opioid exposure over time [27]. The trajectory groups were modelled using a normal distribution with cubic polynomial structures [28]. Recommendations were followed for fitting an optimal model to derive the number of trajectory groups based on statistical, visual, and interpretable factors [29, 30], as previously described [31]. The adequacy of selected models was evaluated based on recommendations [29], including the average posterior probability of group membership where > 0.70 is considered sufficient group assignment (1.0 indicates no ambiguity in assigning individuals to groups) and the odds of correct classification where > 5.0 is considered sufficient (1.0 being similar to random guessing). The proportion of opioid prescriptions by type was graphically depicted monthly for adults with and without CP to assess for differential medications. Basic demographics were described for each trajectory group.

Estimates with n ≤ 11 cases were suppressed as part of the Data Use Agreement to preserve patient de-identification. Analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA).

Results

The baseline descriptive characteristics of adults with CP (n = 13,929) and matched adults without CP (n = 278,538) for the first aim (population-level) are presented in Table 1. The majority (99.96%) of the cohort with CP had suitable matches at a ratio of 1:20, while five adults with CP had < 20 matches: three adults with CP had 17–18 matches each and two had 2–3 matches each. These five adults with CP and their corresponding matches were included in the analysis.

Table 1 Baseline descriptive characteristics of adults with cerebral palsy (CP) and matched* adults without CP (w/oCP)
Full size table

During the follow-up, 49.2% (n = 6,848) of adults with CP and 44.1% (n = 122,816) of matched adults without CP were exposed to opioids. This estimate for matched adults without CP was similar to the full, unmatched non-CP cohort (44.1% from n = 18.4 million).

Population-level opioid prescription patterns

Adults with CP had a higher prevalence of exposure to any opioids for each month across the 7-year study period. The slope of monthly opioid exposure over time was ~ 2x steeper for adults with versus without CP (slope, 0.0006 vs. 0.0003, respectively) (Fig. 1A). The median number of days supplied with any opioids was also higher for each month across the study period, and the slopes were positive for both cohorts; however, the slope was ~ 1.9x steeper for adults without versus with CP (Fig. 1B). For both cohorts, the upper quartile (75th percentile) was nearly 100% for each month, while the lower quartile (25th percentile) increased over time for both cohorts. The change in sample size per month for adults with CP mirrored that of the matched cohort without CP across the entire study period (Fig. 1C).

Fig. 1
figure 1

Population-level estimates of monthly opioid exposure. A Prevalence of exposure to any opioid monthly with 95% confidence intervals (vertical lines) from April, 2011 to December, 2017 for adults with cerebral palsy (CP, black) and matched adults without CP (w/oCP, gray). B Among those exposed to opioids, the median (solid lines) and interquartile range (dashed lines) of the number of days supplied opioids as percent of month for standardization. C The sample size included in the analysis per month. The red lines in A and B show the linear regression. The “slope” value reported is the slope from each cohort’s linear regression

Full size image

The majority of opioid prescriptions for both cohorts were from hydrocodone, followed by oxycodone and tramadol (Fig. 2).

Fig. 2
figure 2

Population-level estimates of opioid type. The bar graphs represent the proportion of total opioid prescriptions by the type of opioid per quarter for A adults with cerebral palsy (CP) and B matched adults without CP. The “other” opioid category includes hydromorphone, methadone, buprenorphine, oxymorphone, dihydrocodeine, meperidine hydrochloride, opium, pentazocine, and tapentadol

Full size image

Individual-level trajectories of opioid prescription patterns

Among those with opioid exposure, 2,099 adults with CP met the eligibility criteria (Supplementary material 2) and there were 10,361 matched adults without CP. The majority (97.67%) of the cohort with CP had suitable matches at a ratio of 1:5, while 49 adults with CP had < 5 matches: 19 adults with CP had 3–4 matches each, 26 had 1–2 matches each, and 4 had no suitable matches. These 49 adults with CP and their corresponding matches were included in the analysis.

For adults with CP, 6 trajectory groups were identified of monthly days supplied over 1-year from the initial opioid exposure month (Fig. 3A). Group 1 (55.7% of the cohort) was characterized as having absent exposure (0 opioid days) from months 3–12. Group 2 (30.4%) had consistently low exposure (0–1 day/month) from months 3–12. Group 3 (3.1%) had a “U” shaped exposure of an initial decrease (months 1–4) followed by an increase (months 7–12). Group 4 (5.1%) had a moderate start (~ 2 weeks/month) with a slow decline that never reached 0 days (~ 2–3 days in month 12). Group 5 (2.6%) had a moderate start that increased from months 5–12 (~ 3–4 weeks in month 12). Group 6 (3.0%) had a moderate-to-high start (~ 2–3 weeks/month) and then consistently high exposure from months 3–12 (~ 3–4 weeks/month).

Fig. 3
figure 3

Trajectories of monthly opioid exposure over 1-year. Trajectories of monthly exposure to opioids as days supplied (standardized to proportion of month) over 1-year from the initial opioid exposure month for A adults with cerebral palsy (CP) and B matched adults without CP. The percent value represents the proportion of the cohort assigned to that trajectory group

Full size image

For the matched cohort without CP, 5 trajectory groups were identified, some of which had a similar pattern as Groups 1, 2, and 6 from adults with CP (Fig. 3B). Notably, for the consistently high trajectory group (Group 6 for CP; Group 5 for non-CP), the exposure level (peaked ~ 90% vs. ~80%) and cohort proportion (3.0% vs. 1.5%) was higher for adults with versus without CP. There were also differences in trajectories and patterns of relatively similar trajectories. For example, adults without CP did not exhibit a “U” shaped pattern like Group 3 from CP. The moderate start followed by a decline in exposure (Group 4 for non-CP) exhibited a more rapid decline and went to 0 days which differed from Group 4 for CP. The moderate start that did not decline (Group 3 for non-CP) did not increase at later months like Group 5 for CP.

The model diagnostics suggests that the model was sufficient for adults with CP (average posterior probability, 0.88–0.98; odds of correct classification, 5.7–1,881.9). The model was sufficient for matched adults without CP for all groups (average posterior probability, 0.88–0.98; odds of correct classification, 82.6–332.7) except Group 1 that had an average posterior probability of 0.85 and an odds of correct classification of 3.3.

The distribution of opioid prescriptions by the opioid type was relatively similar for both cohorts across the 12-month follow-up (Fig. 4). The absent to minimal trajectory groups for both cohorts (Groups 1 and 2 for both cohorts) were on average younger than the other groups with greater monthly exposure (Table 2). The consistently high trajectory group had a greater proportion of those with co-occurring epilepsy and/or intellectual disabilities compared to the other trajectory groups, but only for the cohort with CP. Other basic demographics were similar among the comparable trajectory groups between adults with CP (Groups 1, 2, and 6) and without CP (Groups 1, 2, and 5).

Fig. 4
figure 4

Opioid exposure by type. The bar graphs represent the proportion of total opioid prescriptions by the type of opioid per month among starting from the first opioid exposure month for A adults with cerebral palsy (CP) and B matched adults without CP. The “other” opioid category includes hydromorphone, methadone, buprenorphine, oxymorphone, dihydrocodeine, meperidine hydrochloride, opium, pentazocine, and tapentadol

Full size image
Table 2 Basic baseline demographic characteristics of adults with cerebral palsy (CP) and matched* adults without CP based on 1-year trajectory group of opioid exposure (as days supplied per month)
Full size table

Discussion

The findings from the 7-year population-level analysis suggest that adults with versus without CP were more likely to be exposed to opioids and, among those exposed, had a higher number of days supplied with opioids per month. The findings from the individual-level analysis suggest that, among those with opioid exposure, adults with versus without CP were more likely to be prescribed a higher monthly volume and for longer periods of time over the course of 1-year following the initial opioid exposure month. This multi-level study provides novel epidemiologic evidence of the extent of possible opioid exposure patterns among adults with CP during this period, which has implications for clinical practice. For example, even if an optimal opioid regime was prescribed that effectively managed the patient’s pain, inadvertent side effects and resultant sequalae may create new or exacerbate existing health issues, thus complicating long-term clinical management of the individual with CP.

The population-level monthly proportion exposed to opioids increased over 7-years for adults with and without CP, but the rate was ~ 2x steeper for adults with versus without CP. Although the increasing rate of the median number of days supplied with opioids was steeper for adults without versus with CP, adults with CP had a higher number of days supplied across the 7-year period. In 2017, the median number of days supplied was ~ 25 days for adults with CP and ~ 21 days for adults without CP. The proportion of opioid prescriptions by the type of opioid was largely similar between cohorts over time, with the majority being from products containing hydrocodone, oxycodone, and tramadol. However, for adults with CP, the hydrocodone proportion decreased slightly over time while the tramadol proportion slightly increased.

At the individual-level, the majority of adults with (~ 86%) and without (~ 92%) CP exposed to opioids were prescribed a relatively low monthly volume for a short duration of time (~ 1–3 months), suggesting predominance of acute, as opposed to long-term, opioid therapy. However, the other ~ 14% and ~ 8% of adults with and without CP, respectively, were prescribed a higher monthly volume of opioids for extended periods (higher volume and duration for CP), which may suggest that opioid patterns range from acute-repeated to chronic-continuous. Moreover, the patterns differed among those with repeated monthly exposure, which may reflect different medical strategies to manage pain amidst a constellation of other complications. Indeed, the high-consistent trajectory group for adults with CP contained a higher proportion of co-occurring epilepsy and/or intellectual disabilities, which is associated with greater medical complexity for the individual with CP [25, 26]. The trajectory patterns observed in the cohort without CP was similar to other studies [32,33,34], suggesting representation of this database and methodology to identify trajectories and differences for adults with versu without CP. Taken together, these findings point to the variability in how opioids were being prescribed in the population of adults with CP.

Evidence is beginning to emerge showing that individuals with CP are more likely to receive opioids and at higher volumes [19,20,21,22]. The main benefit of opioid therapy is providing acute pain relief. This is important as pain is negatively associated with quality of life, daily function, mental health, and sleep among individuals with CP [3, 4, 35,36,37,38]. However, long-term opioid therapy may be less effective for managing chronic versus acute pain [39], and may be the wrong therapeutic choice for the ~ 1 in 3 children and ~ 3 in 4 adults with CP who experience chronic and/or nociplastic pain [3, 4, 6, 7]. Future studies are needed to identify the risk-benefit balance of various opioid therapeutic strategies.

There are limitations to this study that may influence interpretations. First, the cohorts were derived from a private insurance database, which may reflect a slightly healthier segment of the adult population with CP with possible selection bias. Findings should be interpreted within the context of this privately insured sample. Second, this study did not have access to inpatient pharmacy prescriptions and may have missed important inpatient opioid prescriptions that may influence opioid exposure trajectories after discharge. Third, this study quantified prescription patterns as it is not possible to determine if the patient took the medication and adhered to the treatment protocol. Thus, we were unable to quantify actual consumption of opioids. Fourth, the reason for opioid prescriptions (e.g., for CP or other conditions) or who prescribed the opioids were not examined. Evidence suggests that 1% of healthcare providers that prescribe opioids account for a disproportionate volume of doses and prescriptions [40]. It is unknown if the lack of clinician knowledge on how to manage pain and other related symptoms for adults with CP results in inappropriately higher opioid prescribing. Lastly, claims data does not provide information about the severity of CP using common clinical measurements, such as the gross motor function classification system.

Conclusion

This multi-level study provides novel epidemiologic evidence of opioid prescription patterns among adults with CP. In general, adults with versus without CP were more likely to be exposed to opioids, for a higher number of days per month, and for a longer duration from 2011 to 2017. Future studies are needed to identify risks and benefits of various opioid and non-opioid therapeutic strategies and long-term effects.