Authors: Megan E. Feeney (1Department of Pharmacy, Boston Medical Center, Boston, MA), Anica C. Law (2Department of Medicine, The Pulmonary Center, Boston University Chobanian & Avedisian School of Medicine School of Medicine, Boston, MA), Allan J. Walkey (2Department of Medicine, The Pulmonary Center, Boston University Chobanian & Avedisian School of Medicine School of Medicine, Boston, MA), Nicholas A. Bosch (2Department of Medicine, The Pulmonary Center, Boston University Chobanian & Avedisian School of Medicine School of Medicine, Boston, MA)
Categories: Article, critical illness, intensive care units, drug therapy, methadone, buprenorphine
Source: Critical care medicine
Authors: Megan E. Feeney, Anica C. Law, Allan J. Walkey, Nicholas A. Bosch
To describe practice patterns surrounding the use of medications to treat opioid use disorder (MOUD) in critically ill patients.
Retrospective, multicenter, observational study using the Premier Inc. AI Healthcare Database.
United States intensive care units (ICUs).
Adult (≥18 years old) patients with a history of opioid use disorder (OUD) admitted to an ICU between 2016–2020.
None.
Of 108,189 ICU patients (658 hospitals) with a history of OUD, 20,508 patients (19.0%) received MOUD. Of patients receiving MOUD, 13,745 (67.0%) received methadone, 2,950 (14.4%) received buprenorphine, and 4,227 (20.6%) received buprenorphine/naloxone. MOUD use occurred in 37.9% of patients who received invasive mechanical ventilation. The median day of MOUD initiation was hospital day 2 (IQR 1–3) and the median duration of MOUD use was 4 (IQR 2–8) days. MOUD use per hospital was highly variable (median 16.0%, IQR 10–24, range 0–70.0%); admitting hospital explained 8.9% of variation in MOUD use. A primary admitting diagnosis of unintentional poisoning (aOR 0.41, 95% CI 0.38–0.45), presence of an additional substance use disorder (aOR 0.66, 95% CI 0.64–0.68), and factors indicating a greater severity of illness were associated with reduced odds of receiving MOUD in the ICU.
In a large multicenter, retrospective study, there was large variation in the use of MOUD among ICU patients with a history of OUD. These results inform future studies seeking to optimize the approach to MOUD use during critical illness.
The opioid epidemic is a public health crisis that affects patients across the spectrum of health care. Between 2019 and 2020, overdose deaths increased by nearly 30%, with 75% of overdose deaths being attributed to opioids.(1) In the intensive care unit (ICU), admissions linked to opioid use (e.g., acute overdose) and patients with comorbid opioid use disorder (OUD) are increasing and are associated with high mortality.(2)
Critical care clinicians are increasingly challenged to consider aspects of care unique to patients with OUD or OUD-related comorbidities. Medications to treat opioid use disorder (MOUD: methadone, buprenorphine, and naltrexone) reduce substance use, increase retention in treatment programs, and reduce substance use-related mortality in the outpatient setting,(3–7) but both continuation of outpatient use(8) or new initiation of MOUD in the ICU is understudied. Further, MOUD therapy in the ICU introduces additional complexities; opioid agonist therapy (methadone or buprenorphine) and antagonist therapy (naltrexone) directly interact and affect receptors targeted by analgesia used in patients receiving invasive mechanical ventilation (IMV)(7) and acute organ dysfunction complicates pharmacokinetic and pharmacodynamic properties of MOUD. Because hospitalization and poorly managed acute pain is associated with an increased risk of earlier MOUD discontinuation and reduced retention in OUD treatment programs,(9–11) expert opinion recommends continuing MOUD for perioperative patients(12–15) and through management of acute pain(9, 11, 12, 16–19). However, it is unclear if limited data outside the ICU generalize to patients who are critically ill; currently no guidelines specifically address use of MOUD in the ICU.(20–24) To begin to address knowledge gaps concerning MOUD use in the ICU, we sought to characterize practice patterns and factors associated with the use of MOUD in ICU patients using a large, multicenter database of United States hospital admissions.
We used the Premier Inc. AI Healthcare Database (PINC AI; 2016–2020)(25), an enhanced claims-based database containing patient- and service-level data (e.g., demographics and International Classification of Diseases diagnosis and procedure codes), to identify the study population. PINC AI contains data from approximately 25% of United States hospitalizations, including patient demographics, diagnosis and procedure codes, and billing and charge code information with minimal missing data (<0.01 of variable fields are missing)(26), and facilitates capturing detailed treatment information to the level of the calendar day. Self-reported hospital characteristics (e.g., bed capacity, urban and rural populations served, teaching status) are also recorded for all contributing hospitals. We included adult patients (age 18–89 years) with a history of OUD present on admission who were admitted to an ICU on hospital day one. Patients with a history of OUD were identified using the International Classification of Diseases, Tenth Revision (ICD-10) code for Opioid-related disorders (F11x) with a code identifier of “present on admission”.(27) We chose to exclude patients with opioid-related disorders in remission (ICD-10 code F11.x1) to more accurately reflect a population of active opioid use and/or dependence, as outlined in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5)(28); we included patients with OUD in remission in a sensitivity analyses. For patients with more than one admission, a random hospital admission was selected for inclusion.
The primary practice assessed was the use of MOUD (yes/no) during hospitalization. Receipt of MOUD was defined as at least one charge for methadone (enteral, parenteral), buprenorphine (sublingual, transdermal), and/or buprenorphine/naloxone (sublingual, enteral) and was ascertained through pharmacy billing codes.(7) Extended release injectable naltrexone was excluded from the study definition of MOUD as it is used primarily for the prevention of relapse in patients with OUD no longer physically dependent on opioids. Because the PINC AI is limited to inpatient data, the proportion of patients receiving MOUD in the outpatient setting was not known and thus the outcome included possible new initiations of MOUD in addition to continuations of outpatient therapy. Secondary practices were type of MOUD received (methadone, buprenorphine, and/or buprenorphine/naloxone), duration of MOUD use, MOUD use while receiving IMV, use of parenteral sedatives and analgesics, use of non-opioid analgesics, use of parenteral opioids on first MOUD day, hospital day of MOUD initiation, and incidence of new ventricular arrhythmias (defined as ICD-10 codes I49.01, I49.02, and I47.21). Duration of MOUD use was defined as the number of total calendar days of use during the hospitalization and may not have occurred consecutively.
We referenced the Andersen Behavioral Model of Health Services Use(29) and prior studies(8, 30) in non-critically ill patients to identify characteristics potentially associated with MOUD use for inclusion in study models (eTable 1). Predisposing factors were age, sex, race, and ethnicity. Enabling factors were insurance type, US census region, teaching status, hospital size, urban location, and safety-net status(31) (hospitals in the highest quartile of each United States census region, ranked by proportion of hospitalized patients who are uninsured or on Medicaid).(30) Need factors included pregnancy, comorbid disease (by Gagne(32, 33)), diagnosis of a non-OUD substance use disorder (SUD), major depressive disorder, or chronic pain present on admission. We included validated measures of organ dysfunction or sepsis present on admission (by Angus(34)), ICU type, primary admitting diagnosis, and use of IMV as other clinical factors, and vasopressors, parenteral opioids (fentanyl, hydromorphone, morphine), parenteral sedatives (propofol, benzodiazepines, dexmedetomidine, ketamine), non-opioid analgesics (acetaminophen, ibuprofen, gabapentin), QT-prolonging medications (amiodarone, azithromycin, clarithromycin, levofloxacin, ondansetron, or citalopram), and antipsychotics (defined as haloperidol, olanzapine, risperidone, or quetiapine fumarate) as medication factors.(8, 30, 35–37)
Candidate characteristics and MOUD practices were summarized overall, by discharge month, and by hospital. Patients without charge codes or ICD-10 codes for a specific study variable were coded as not having the variable of interest. For missing race and ethnicity data, a separate category of “unknown race/ethnicity” was used. Analysis of MOUD use by hospital was limited to hospitals with greater than 25 patients to stabilize model effect sizes and improve convergence. We used multivariate hierarchical logistic regression models with hospital of admission as a random intercept to identify characteristics associated with MOUD use. In the model, patients categorized as receiving MOUD could have received a single MOUD therapy or any combination of methadone, buprenorphine, and/or buprenorphine/naloxone. From these models, we calculated adjusted odds ratios (aOR) to quantify the association between each characteristic and MOUD use and calculated the adjusted intraclass correlation coefficient (ICC) to describe the proportion of variance in MOUD use that was attributable to hospital of admission.(38) The median odds ratio (MOR) is the median value of the odds ratio for admission hospital moving from a randomly selected hospital with higher MOUD use to a hospital with lower use and facilitates comparisons between fixed effect odds ratio and the effect of admission hospital.(39)
We conducted a sensitivity analysis including patients with OUD in remission (ICD-10 code F11.x1) in the study population definition to assess the robustness and generalizability of our definition of OUD. As methadone and buprenorphine both carry alternative prescribing indications that could confound our results, we conducted a post-hoc analysis of MOUD practices excluding patients with a history of chronic pain (ICD-10 G89.21, G89.22, G89.28, G89.29, G89.4, G89.3).
A p-value of <0.05 was considered statistically significant. We did not adjust for multiple comparisons; results of all secondary analyses should be considered as hypothesis-generating. R version 4.0.2 software (R Foundation for Statistical Computing) was used for analyses. This study was deemed not human subjects research by the Boston University Institutional Review Board (#H-41795).
We identified 108,189 patients with a history of OUD present on admission who were admitted to an ICU at 658 hospitals (Figure 1) over the 5-year study period. Included hospitals admitted a median of 101 (IQR, 54–192) patients who received MOUD. The median patient age was 51 (IQR, 36–61 years) and most patients were admitted to ‘general’ ICUs (76.7%) (eTable 2). The most common primary admitting diagnoses were sepsis (18.3%) and unintentional poisoning (8.6%). On admission, patients had an average of three comorbidities (IQR, 1–5) and respiratory dysfunction was the most common form of acute organ injury, occurring in 52.1% of patients. Sepsis criteria(34) was present on admission in 41.9% of patients and 31.8% of patients received vasopressors. A total of 42,378 (39.2%) patients had another SUD (other than OUD) at the time of admission and 44,014 (40.7%) patients had a history of chronic pain. Admitting hospital location was an urban setting for 88.3% of patients; 49.2% of patients were admitted to a teaching hospital and 35.8% were admitted to a safety net hospital.
Of the included patients, 20,508 (19.0%) had at least one charge for MOUD therapy after ICU admission. The median duration of MOUD use during hospitalization was 4 days (IQR, 2–8 days), with the median first day of use occurring on day 2 (IQR, day 1–3) (Table 1). The proportion of patients discharged who received MOUD was 17% (95% CI 15.2–18.7%) in January 2016 and 25.5% (95% CI 23.0–28.0%) in December 2020 at the end of the study period (Figure 2).
Methadone was the most frequently used MOUD therapy (67.0% vs. 14.4% buprenorphine vs. 20.6% buprenorphine/naloxone [1.98% received more than one type of MOUD therapy]) (Table 1). The median duration of therapy was 5 days in patients receiving methadone (IQR 3–8 days), 3 days (IQR 2–6 days) for buprenorphine, and 4 days (IQR 2–6 days) for buprenorphine/naloxone (Figure 3). Within the methadone group, 40.5% of patients received IMV. Use of IMV in patients who received buprenorphine and buprenorphine/naloxone was 32.1% and 34.1%, respectively. Fentanyl was received by 52.3% of patients who received methadone, 42.3% who received buprenorphine, and 43.7% who received buprenorphine/naloxone. Parenteral opioids were administered on the first day of MOUD therapy in 45.4% of patients who received methadone, 24.9% of patients receiving buprenorphine, and 22.6% of patients receiving buprenorphine/naloxone. Rates of new ventricular arrhythmias were low overall, occurring in 0.3% of the methadone and buprenorphine groups and in 0.2% of the buprenorphine/naloxone group (Table 1).
The median percentage of patients admitted to an ICU with OUD who received MOUD per hospital was 16% (IQR, 10–24) and ranged from 0–70% (Figure 4a). Types of MOUD used by hospital was variable; 44.2% of hospitals used a combination of methadone and either buprenorphine or buprenorphine/naloxone, 18.5% of institutions exclusively used methadone, and 34% used a combination of all three agents. Only 3% of institutions exclusively used buprenorphine or buprenorphine/naloxone (Figure 4b). The hospital of admission explained 8.9% of the total variation in MOUD use after accounting for measured characteristics.
In the multivariable model adjusting for covariates, older age (aOR 0.89, 95% CI 0.87–0.91), a primary admitting diagnosis of unintentional poisoning (aOR 0.41,95% CI 0.38–0.45), additional SUD(aOR 0.66, 95% CI 0.64–0.68), and several factors associated with acute illness including the presence of hepatic dysfunction (aOR 0.63, 95% CI 0.57–0.71) and renal dysfunction (aOR 0.75, 95% CI 0.72–0.78) were associated with reduced odds of receiving MOUD in the ICU (eTable 3). Trauma, as an admitting diagnosis, was also associated with lower odds of receiving MOUD, compared to patients admitted for sepsis (aOR 0.75, 95% CI 0.68–0.83). Pregnancy during ICU admission was associated with the greatest odds of receiving MOUD (aOR 4.23, 95% CI 3.50–5.11). Meeting sepsis criteria on admission (aOR 1.25, 95% CI 1.19–1.31), admission to urban (aOR 1.32, 95% CI 1.14–1.52), teaching (aOR 1.21, 95% CI 1.07–1.36), and safety-net hospitals (aOR 1.26, 95% CI 1.12–1.41) were also associated with increased odds of receiving MOUD. The MOR for admission hospital was 1.74 (95% CI 1.72–1.76).
Results of sensitivity analyses (eTable 4) were largely similar to the primary analyses. There were 3,854 patients with OUD in remission, with 858 (22.2%) having at least one charge for MOUD therapy documented during ICU admission. Amongst the 112,043 patients that included patients with OUD in remission, the proportion receiving MOUD was 19% (21,366 patients). Practices by type of MOUD were also similar to the primary analysis, with methadone being the predominant agent. In the group excluding patients with a history of chronic pain (n=64,175 patients), the proportion that received MOUD was 22% (14,270) with a similar distribution between MOUD therapies to that of the overall population.
We used a large, multicenter database of US inpatient admissions to examine practice patterns for MOUD therapies. Less than a quarter of patients with a history of OUD received MOUD while admitted to the ICU, but over time, the proportion of patients receiving MOUD in the ICU increased. We also found substantial variation in ICU MOUD use and type of MOUD use between hospitals. Our results inform the importance of future studies evaluating the impact of MOUD initiation or continuation on short-term and post-ICU outcomes, the epidemiology of potential complications associated with MOUD in the ICU, and studies seeking to understand the effects of MOUD use on analgesia and sedative dosing.
Our results should be considered in the context of prior studies. National guidelines(20, 21) do not currently address MOUD use in the ICU or in the setting of concomitant sedation and analgesia and a single recent publication addresses considerations for MOUD use during critical illness.(24) A survey of U.S. ICUs reported only 7% (3/43) of institutional sedation guidelines addressed the needs of patients with OUD and only a single institution offered guidance for the continuation of outpatient MOUD in ICU patients.(23) Of 55 respondents, 78% reported sometimes or always continuing patients on their outpatient MOUD. Buprenorphine use and associated outcomes have been described in a small cohort of 153 ICU patients; among the 49% of patients who were prescribed buprenorphine prior to admission, 61% were continued on their home therapy within 24 hours of ICU admission.(40) Another single center study reported 44% of patients receiving buprenorphine within three months of hospital admission received buprenorphine during the ICU phase of care, compared to 68% in the post-ICU phase of care.(36) In the same study, the odds of receiving opioids in the ICU was more than six times higher in patients not receiving buprenorphine compared to patients who were. In our larger study, we demonstrated reduced odds of receiving any form of MOUD in patients receiving fentanyl (aOR 0.92), parenteral hydromorphone (aOR 0.83), and parenteral morphine (aOR 0.77). Taken together, these results may have multiple potential (1) clinicians have concerns with using concomitant opioids with MOUD; (2) when used in the ICU, MOUD therapy may be initiated once critical illness has passed.
Studies of MOUD use in non-hospitalized adults with OUD reported similar MOUD use to our study (~20%).(8) However, unlike in non-hospitalized patients, we found no race or sex disparities in MOUD use among ICU patients.(8) Similar to our findings, among non-critically ill hospitalized patients, the odds of receiving MOUD decrease with an increasing number of co-occurring SUDs.(37) Limitations in care models for patients with OUD and polysubstance use, avoidance of opioid agonist therapy in patients with active alcohol use disorder, or clinician attitudes for treatment adherence may contribute to this finding.(22, 37, 41, 42) Also consistent with our findings, there is an inverse relationship between MOUD use and comorbidity burden in non-hospitalized patients.(37) In an ICU setting, this may be related to concerns of augmented absorption or clearance of MOUD in the setting of acute organ dysfunction. Last, women who were pregnant in our study were more than four times as likely to receive MOUD in the ICU compared with non-pregnant women despite prior work suggesting that women who are pregnant have low access to MOUD.(30, 43)
Our results prompt several additional questions for further investigation. The high utilization of parenteral opioids and sedatives in patients receiving MOUD but inverse correlation suggests a need to identify interactions between ICU MOUD use, pain control, and opioid dosing for analgesia used during IMV. There is also a need for comparative effectiveness studies of MOUD continuation versus discontinuation, of different types of MOUD use in the ICU, and of MOUD initiation timing (i.e., during critical illness or post-critical illness hospitalization) across the spectrum of acute illness and comorbidities. Future multicenter studies are also needed to explore associations between MOUD use in the ICU and short- (e.g., pain scores, self-directed discharge rates) and long-term (e.g., abstinence, MOUD adherence, opioid-related mortality) outcomes.(44–47) Future studies should also compare outcomes and patient characteristics between hospitals that routinely continue outpatient MOUD in the ICU to hospitals that routinely stop MOUD in the ICU.
This study has limitations. The Gagne Combined Score(32) used to describe comorbid disease burden was developed and validated in a population aged 65 and older, which may limit generalizability to our study population. Included patients were identified using ICD-10 codes for diagnoses of OUD. Prior studies validating OUD ICD-10 codes have found low (10–40%) sensitivity but high (>95%) specificity and thus, our study may exclude some patients with OUD.(48, 49) The PINC AI database provides charge claims only for medications received during inpatient admission. Due to our inability to differentiate new MOUD use from continued MOUD use, we could not reliably compare outcomes between MOUD therapies or between MOUD and no MOUD therapy in this study population. Duration of pre-hospital MOUD use likely also impacts associations with MOUD therapy, which cannot be accounted for through use of PINC AI. Similarly, we were unable to determine pre-admission use of other opioid therapy (e.g., fentanyl, heroin). It is possible that some methadone or buprenorphine use defined as MOUD therapy in our study was indicated for the management of other disease states, such as chronic pain. Reassuringly, MOUD use in a post-hoc analysis of patients without a chronic pain diagnosis was similar to the primary analysis. We were unable to identify the specialty of MOUD prescribing providers and therefore could not differentiate the proportion of MOUD that was managed by ICU providers compared to consulting providers (e.g., psychiatry, addiction medicine). Doses or dose changes for MOUD therapies could not be quantified, nor could we determine if patients experienced withdrawal symptoms based on type of MOUD or changes to MOUD therapies; future studies are needed to clarify current ICU MOUD dosing strategies.
In a large retrospective multicenter study, we found substantial variation among practices with MOUD therapies in adult patients admitted ICUs across the United States and found that MOUD, when initiated, was started early in the ICU course and in patients with few comorbidities and low acuity of illness. These results inform several future studies seeking to optimize the management of patients with OUD who are admitted to the ICU.