https://orcid.org/0000-0002-1785-223X
Figures
Abstract
Background
Long-term opioid use is an increasingly important problem related to the ongoing opioid epidemic. The purpose of this study was to identify patient, hospitalization and system-level determinants of long term opioid therapy (LTOT) among patients recently discharged from hospital.
Design
To be eligible for this study, patient needed to have filled at least one opioid prescription three-months post-discharge. We retrieved data from the provincial health insurance agency to measure medical service and prescription drug use in the year prior to and after hospitalization. A multivariable Cox Proportional Hazards model was utilized to determine factors associated with time to the first LTOT occurrence, defined as time-varying cumulative opioid duration of ≥ 60 days.
Results
Overall, 22.4% of the 1,551 study patients were classified as LTOT, who had a mean age of 66.3 years (SD = 14.3). Having no drug copay status (adjusted hazard ratio (aHR) 1.91, 95% CI: 1.40–2.60), being a LTOT user before the index hospitalization (aHR 6.05, 95% CI: 4.22–8.68) or having history of benzodiazepine use (aHR 1.43, 95% CI: 1.12–1.83) were all associated with an increased likelihood of LTOT. Cardiothoracic surgical patients had a 40% lower LTOT risk (aHR 0.55, 95% CI: 0.31–0.96) as compared to medical patients. Initial opioid dispensation of > 90 milligram morphine equivalents (MME) was also associated with higher likelihood of LTOT (aHR 2.08, 95% CI: 1.17–3.69).
Conclusions and relevance
Several patient-level characteristics associated with an increased risk of ≥ 60 days of cumulative opioid use. The results could be used to help identify patients who are at high-risk of continuing opioids beyond guideline recommendations and inform policies to curb excessive opioid prescribing.
Citation: Kurteva S, Abrahamowicz M, Weir D, Gomes T, Tamblyn R (2022) Determinants of long-term opioid use in hospitalized patients. PLoS ONE 17(12): e0278992. https://doi.org/10.1371/journal.pone.0278992
Editor: Vijayaprakash Suppiah, University of South Australia, AUSTRALIA
Received: September 14, 2021; Accepted: November 7, 2022; Published: December 15, 2022
Copyright: © 2022 Kurteva et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: We will not be able to share the governmental provincial administrative (RAMQ) data used in these analyses publicly. Our Research Ethics Board (REB) approval for this project limits data access to Robyn Tamblyn’s team. To provide access to other researchers, they would need to obtain permission from the Institut de la Statistique du Québec for their project that would use RAMQ data. More details regarding eligibility to request data can be found here: https://statistique.quebec.ca/research/#/a-propos/utilisation-guichet. To use the clinical data, they would a) have to apply for REB approval for the project, and request that a de-identified copy of the data be made available as consent was provided by patients for only Robyn Tamblyn (consent form appended). More information regarding access to data can be found here: https://statistique.quebec.ca/research/#/demarche/etape-par-etape. A link to submit a data request can be found here: https://statistique.quebec.ca/zone-chercheur/#/login.
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
LTOT has been associated with an increased risk of opioid-related adverse events [1–3]. While opioids may be appropriate for short-term treatment for pain, LTOT has not been shown to improve pain relief [2, 4, 5]. The Centers for Disease Control and Prevention (CDC) guideline defines long-term opioid therapy (LTOT) as use of opioids on most days for more than 3 months [2], but in a recent empirical study we have found that the risk of opioid-related adverse events does not increase considerably beyond 60 days (Appendix D.1.a in S1 File) [6]. We observed only moderate increases in risk above 60 days of cumulative use with no evidence of further impact of durations longer than 100 days of use. There is a gap in evidence related to determinants of risk of LTOT using this empirically- derived definition of 60 days.
Previous studies have reported that the most common predictors associated with an increased risk for LTOT included age [7–11], sex [7–11], arthritis [9, 10, 12], race [8, 9], the presence of chronic pain [8–10] and mental illnesses such as anxiety and depression [8, 11–13]. Increased LTOT risk was also observed with certain characteristics of the first opioid prescription such as opioid doses of >90MME and longer duration (days’ supply) [8, 9, 12, 14]. However, the evidence is limited to the data elements available in administrative data, such as patient and prescription characteristics in the community, and these databases do not typically have data on medications used in hospital or discharge information [15].
Hospitalization itself may inadvertently be a risk factor to initiating opioids [8], and inadequate communication of changes in medication at the time of hospital discharge is also a well-established problem [16]. As a result, community physicians may continue opioids started in the hospital, for acute pain relief, as they have no information about the treatment indication nor the expected duration of therapy [17, 18].
There is a large variation in prescribing of opioids between providers [19], which is unrelated to patient characteristics. The association between organizational /in-hospital prescriber characteristics and their contribution to the initiation, maintenance or prevention of LTOT is poorly understood. In addition, the type of opioid prescribed at hospital discharge varies across different attending physicians and residents [20, 21]. Thus, physician-prescribing behavior for post-discharge analgesia may be associated with an increased risk of LTOT. In this study, in addition to patient and medical characteristics, we were able to incorporate information on provider characteristics to better understand their contribution to the development of post-discharge LTOT. The purpose of this study was a) to estimate the proportion of hospitalized patients with LTOT in the one year after hospital discharge, and b) to identify modifiable patient, prescriber and system-level risk factors for long-term prescription opioid use compared to episodic use.
Methods
Setting
We carried out a secondary analysis of a cluster-randomized trial on discharge medication reconciliation conducted at the McGill University Health Centre (MUHC) [22]. The MUHC is an over 1000-bed quaternary care teaching hospital in Montreal (Canada) that operates within the universal health care plan of the province of Quebec (RAMQ). This plan covers all hospitalizations, and essential medical care for provincial residents. It also provides drug insurance for registrants 65 years of age and older, income security recipients, and those not insured through their employer (approximately 50% of the Quebec population). Ethics approval was provided by the MUHC Research Ethics Board. Privacy Commissioner approval was obtained to link clinical and administrative data from the Commission d’accès à l’information du Quebec.
Participants
A prospective cohort of medical and surgical hospitalized patients discharged from the MUHC between October 2014 and November 2016 were followed 12 months’ post-discharge. To be eligible for the original trial, patients had to be 18 years of age or older at admission, admitted from the community or transferred from another hospital, with at least one-year continuous provincial healthcare coverage prior to hospital admission. To be included in this study, patients needed to fill at least one opioid prescription during the 90 days following their hospital discharge.
Data sources
Multiple data sources were assembled and linked to address the study objectives. For each patient, demographic, clinical, healthcare use and prescription data were retrieved from admission notes as well as provincial healthcare administrative databases (RAMQ medical services and prescription claims) in the year prior to and after the hospitalization, for which the patient was enrolled. Dates of admission/discharge, admitting/discharge unit, patient demographics, diagnoses at admission and discharge, major procedures (surgeries, treatment interventions), were retrieved from the MED-ECHO hospitalization database. Medications at admission, in-hospital as well as those prescribed at discharge were abstracted from the MUHC Data Warehouse.
Study measurements and predictors of long-term opioid use
Long Term Opioid Use (LTOT).
Opioid use in the one-year post-discharge period was ascertained using RAMQ pharmacy administrative claims (see Appendix A.2 in S1 File for detailed description of LTOT calculation). As definitions of LTOT vary in different studies, standardized and evidence-driven definitions are needed [15]. In our recent study, the estimated non-linear effect of cumulative opioid duration showed no further increases in risk of opioid-related adverse events beyond 2 months of use (Appendix D.1.a in S1 File) [6]. Thus, in this study we used this empirically-defined 60 days’ threshold to define LTOT.
Patient-related characteristics.
RAMQ drug programs and hospital charts were used to collected information on age at admission, sex and socioeconomic status. Pain disorders, cancer diagnosis, mental health diagnoses, and conditions associated with abuse were measured at the time of hospital admission and during the hospital stay as fixed-in-time covariates identified by using ICD-9 from medical service claims and ICD-10 codes from hospitalization data. Other co-existing illnesses were measured and adjusted for using the Charlson comorbidity index (CCI), using information collected during the one-year baseline period (see Appendix 5-B in S1 File for a full list of covariates included in the model).
Drug and healthcare utilization.
Psychoactive drugs such as antidepressants, benzodiazepines, Z-drugs or antipsychotics, when used together with opioids, have been associated with an increased risk of opioid-dependency and rates of adverse events [12, 13, 23]. Using RAMQ medical services databases in the one year prior to admission, we measured previous number of emergency department (ED) visits, hospitalizations as well as the distinct number of prescribing physicians.
Medication use and hospitalization characteristics.
Opioid and non-opioid medications administered as part of the in-hospital pain regimen as well as other medications, which may increase patients’ risk of LTOT were extracted from the hospital pharmacy system using corresponding ATC codes. For surgical patients, information on the type of surgery received (thoracic vs upper-gastrointestinal) was retrieved from the MED-ECHO hospitalization database.
Prescribing physician and system-level characteristics.
Information on the attending physician’s gender, language, training status (resident, attending physician) and number of years since licensure were abstracted from the patient medical chart, the medication reconciliation software databases and the hospital data warehouse.
Opioid discharge prescription and initial dispensation.
Treatment changes made to opioid medications from the community were evaluated by using patients’ discharge prescriptions data in comparison to their community drug list. A categorical variable for whether a given opioid was stopped, continued or newly prescribed was derived. A binary variable for the presence of an opioid as part of the discharge pain regimen was constructed: patients with newly added or continued opioids were flagged as having an opioid prescription. In addition, dose, duration and type of initial opioid dispensed (e.g. oxycodone, hydromorphone) were extracted from the RAMQ pharmacy claims.
Statistical analyses
Descriptive statistics were used to compare LTOT patients versus episodic users with respect to patient, provider and hospital unit characteristics. Main analyses relied on time-to-event methodology [24]. A multivariable Cox proportional hazards (PH) model was utilized to determine which factors were associated with the development of LTOT within the one year post-discharge period. Start of follow-up corresponded to the date of the first opioid dispensation. End of follow-up corresponded to the day when the patient first met the criteria for LTOT, or to right censoring at the end of follow-up or death, whichever came first. Moreover, since a patient is considered exposed based on periods of medication possession, patients were temporarily censored during subsequent re-admissions. Since we explored independent associations of various factors potentially related to LTOT, all a priori selected covariates were retained in the model. In addition, since patients could have subsequent hospitalizations and ED visits that could influence their risk of LTOT, a time-varying variable for the cumulative number of past post-discharge hospitalizations and ED visits, updated during the follow-up period, was included to adjust for any changes made to patient’s medications. For each covariate, the results were presented as adjusted hazard ratios (aHR), with 95% confidence intervals (CI). We tested the PH assumption, both globally and for each covariate in the multivariable Cox model, using the Grambsch and Therneau approach [25], incorporated in Survival package. For covariates for which the PH hypothesis was violated (p<0.05), we relied on smooth residual plots to assess how the corresponding adjusted hazard ratio varied over the follow-up (Appendix E in S1 File) [26, 27].
Sensitivity analyses
To account for the fact that patient medications and medical history would most likely change over the course of one year, we updated information on selected co-morbidities and represented them as additional time-varying covariates. We also adjusted for a time-varying count of distinct prescribers, from discharge until a given day, as an indicator of fragmentation in care, which may be due to increased opioid seeking behavior [28, 29]. These analyses were additionally adjusted for a time-varying indicator of being currently exposed to ≥2 opioid products. Moreover, to account for the fact that the association of interest may vary depending on previous opioid use, we stratified the analyses by (i) previous LTOT and (ii) new opioid users. Due to small sample sizes within each stratum, variable selection was necessary, and was based on a combination of substantive knowledge and backward selection. To account for the cluster randomized design of the original trial and, thus, to assess potential clustering, i.e. interdependence, of the outcomes within the hospital units, we re-estimated the main multivariable Cox model while adding unit-specific frailties, i.e, random intercepts, using R. Appendix F in S1 File summarizes the power calculations.
All MSM Cox PH models were implemented with SAS version 9.4 (SAS Institute, Cary, NC). All non-linear relationships were tested using customized programs in R [30].
Results
Overall, 1511 patients were discharged alive from study units and filled an opioid prescription within 3 months’ post-discharge. The proportion of patients who went on to become LTOT by accumulating more than 60 days’ supply of opioids was 22.4% (n = 338) (Table 1), for the incidence rate of 26.8 (95% CI: 24.0–29.9) per 100 person-years. Among those 338 patients, the mean time to LTOT was 115.5 (SD = 76.8) days.
Fig 1 shows the breakdown of patients with respect to their previous history of opioid use, opioid administration during the hospitalization and the receipt of an opioid prescription at hospital discharge. (Fig 1) Among patients who filled at least one opioid dispensation three months’ post-discharge, more than half were opioid-naïve with no history of documented opioid use one year prior to their admission (n = 884, 58.5%).
The average age for LTOT patients was 66.3 (SD = 14.3) and more than half were female (179, 52.9%) (Table 2). LTOT patients were less likely to receive an opioid prescription at discharge, and more likely to have more than a 7-day supply on their first dispensation post-discharge (Table 3).
In multivariable analysis (Table 4), patients who had no copayment status had almost twice higher hazard of becoming LTOT users as compared to patients who had a ‘full’ copay drug insurance status (aHR 1.91, 95% CI: 1.40–2.60). As expected, patients who were previous LTOT users were several times more likely to also meet the criteria for LTOT in one-year post-discharge (aHR: 6.05, 95% CI: 4.22–8.68). History of benzodiazepine use (aHR: 1.43, 95% CI: 1.12–1.83), having a higher CCI (aHR: 1.77, 95% CI: 1.06–2.98)) and a starting daily opioid dose of >90 MME (aHR: 2.08, 95% CI: 1.17–3.69) were all independently associated with increased likelihood of becoming a LTOT user. Having undergone cardiothoracic surgery, as compared to internal medicine patients, was associated a 45% lower risk of LTOT in the post-discharge period (aHR 0.55, 95% CI: 0.31–0.96).
The PH hypothesis was rejected by the global test (p = 0.0046 and for four covariates (p<0.05 for each), for which Appendix E in S1 File describes how the corresponding aHR’s varied with increasing follow-up.
Results from sensitivity analyses that adjusted for selected time-varying characteristics were similar, with a few exceptions. First, we found an association between recent cancer diagnoses and the risk of LTOT, showing an increased risk (aHR: 1.39, 95% CI: 1.07–1.81) (Appendix C.1 in S1 File). Second, after having adjusted for a time-varying count of distinct prescribing physicians and an indicator for using ≥2 opioid products, there was a 34% increased risk of LTOT use associated with having an initial days’ supply of >7 days (aHR: 1.34, 95% CI: 1.05–1.72). Finally, having between 2 and 3 prescribing physicians led to more than doubling in the risk of becoming a LTOT user, relative to patients with only 1 prescriber (aHR: 2.43, 95% CI: 1.85–3.19) (Appendix C.2 in S1 File). In stratified analyses, out of all opioid-naïve patients (n = 1050), 117 (11.1%) went on to become LTOT users. On the other hand, among patients who were LTOT users in the year prior to admission (n = 192), only 68 (35.4%) were LTOT during the follow-up period. The risk of LTOT was 54% higher among patients with a history of pain syndromes compared to patients with no history (aHR: 1.54, 95% CI: 1.01–2.23) (Appendix C.4 in S1 File). The risk of LTOT was more than doubling when initial days’ supply post-hospitalization exceeded >7 (aHR: 2.62, 95% CI: 1.52–4.53). There was no evidence of the clustering of the outcomes within individual hospital units (p = 0.96 for the frailty test).
Discussion
Our study showed that 22.4% of hospitalized patients were characterized as LTOT users. There was an increased risk of LTOT among patients with no drug copay status, history of opioid use, history of benzodiazepine use, higher comorbidity index and higher starting daily dose in the first opioid dispensation post-discharge, whereas, surgical compared to medical patients had a decreased risk of LTOT.
Other studies have also confirmed that previous opioid use leads to greater risks of developing LTOT [7, 9–11, 13]. Previous research, which used a 90-day definition of LTOT, also found that mental health diagnoses, history of pain diagnoses and benzodiazepine use to be associated with an increased risk of LTOT [7–9]. In our study, these patients’ characteristics were also associated with an increased risk, albeit some of these associations were non-significant. Having the first opioid prescription written by an in-hospital prescriber was not associated with an increased risk of LTOT. In addition, it has been frequently argued that prescribing behavior and physicians’ characteristics may contribute to the opioid epidemic [31, 32]. In our study, however, none of the physicians’ characteristics such as years of practice, sex or having a resident approve the discharge prescription were associated with greater risks of LTOT post-discharge. One study found that LTOT patients of residents were more likely to receive early refills following their primary care clinic visits when compared to attending physicians’ patients [20]. In our study, the lack of difference in risk of LTOT associated with the status of the in-hospital prescriber might be attributed to the fact that residents provided similar care with respect to opioid prescribing when compared to attending physicians because they are being trained and monitored by the same physicians. These findings show that in-hospital prescribers may have the opportunity to reduce the occurrence of LTOT by providing patients with adequate pain treatment strategies, without contributing to the development of LTOT.
We found that initial dose and days’ supply of the opioid dispensation both lead to an increased likelihood of LTOT, with initial doses having a greater impact on the risk of LTOT than opioid days’ supply. Previous research has also found that both initial dose and duration of opioid use were associated with an increased risk of LTOT [33–35]. A few of the studies, which examined the initial opioid prescription characteristics and the likelihood of LTOT, found initial duration to be associated with a higher likelihood of LTOT than initial doses [14, 34, 36]. We observed a similar trend in stratified analyses, when looking at the risk among previous LTOT users only. However, like most other studies, opioid dose was associated with greater increases in risk of LTOT for opioid-naïve patients [8, 12]. These findings suggest that selecting an optimal initial opioid duration may be more important than initial dose to reduce the risk of subsequent use, especially for previous users. Moreover, initial opioid exposure characteristics should be used to profile patients who might be at risk of transitioning into LTOT.
This study’s strength is its ability to link data on medication use prior to admission, during the hospitalization and dispensations post-discharge. Using multiple data sources enhances the internal validity of the study by providing detailed covariate information. This allowed us to consider not only patient-, but also provider- and system-level predictors of LTOT. Accounting for healthcare services use during the follow-up and including measures such as number of unique prescribers allowed us to examine the effect of poor coordination of care, flare-ups and complications requiring walk-in and ED visits, on the risk of LTOT. The risk for LTOT has been considered as a central component of quality care assessment [37, 38]. Measuring and defining LTOT is key to understanding potential risk factors, monitoring prevalence and incidence of LTOT, and improving clinical practices [39]. Yet, in previous studies, arbitrarily-defined measures were used without addressing the appropriateness of their cut-offs. In this study, we apply a novel and evidence-driven approach to defining LTOT. In addition, definitions in previous studies only relied on having prescriptions filled during a specified window [9, 40–43]. In this study, we used a time-varying definition of opioid duration constructed based on days’ supply and fill dates. This allowed us to account for gaps between prescriptions and overlapping dispensations, and capture more accurately consistent opioid use.
Some limitations of our work merit emphasis. First, we used prescription duration as recorded by the pharmacist, but since opioids are usually prescribed on a prn basis, exposure mismeasurement is possible. We did not capture actual opioid consumption. Nevertheless, subsequent prescription fills are suggestive of patient’s continual opioid consumption. Future research should use data from multiple healthcare systems to incorporates measures on patients’ healthcare providers and practice environments, and replicate our findings in larger populations cohorts. There is also a possibility of residual confounding due to unmeasured confounders, such as pain severity. Our decision to include only patients with at least one opioid dispensation post-discharge, as well as to include time-varying measures of selected comorbities and patient’s healthcare utilization, reduces concerns about potential bias due to confounding by indication.
Conclusions
We found increases in the risk of LTOT with multiple patient-level characteristics. Quantifying factors associated with the development of LTOT post-discharge is an important step in identifying and targeting patients who need more frequent clinical vigilance and better pain treatment strategy.
Acknowledgments
We thank Dr. Marie-Eve Beauchamp for providing statistical input and technical help with the analyses.
References
- 1. Campbell C. I. et al., "Age and gender trends in long-term opioid analgesic use for noncancer pain," (in eng), American journal of public health, vol. 100, no. 12, pp. 2541–7, Dec 2010, pmid:20724688
- 2. Dowell D., Haegerich T. M., and Chou R., "CDC Guideline for Prescribing Opioids for Chronic Pain—United States, 2016," (in eng), Jama, vol. 315, no. 15, pp. 1624–45, Apr 19 2016, pmid:26977696
- 3. Paulozzi L. J., Zhang K., Jones C. M., and Mack K. A., "Risk of adverse health outcomes with increasing duration and regularity of opioid therapy," (in eng), Journal of the American Board of Family Medicine: JABFM, vol. 27, no. 3, pp. 329–38, May-Jun 2014, pmid:24808111
- 4. Busse J. W. et al., "Opioids for Chronic Noncancer Pain: A Systematic Review and Meta-analysis," (in eng), Jama, vol. 320, no. 23, pp. 2448–2460, Dec 18 2018, pmid:30561481
- 5. Busse J. W. et al., "Guideline for opioid therapy and chronic noncancer pain," (in eng), CMAJ: Canadian Medical Association journal = journal de l’Association medicale canadienne, vol. 189, no. 18, pp. E659–e666, May 8 2017, pmid:28483845
- 6.
A. M. Kurteva S, Beauchamp ME, Tamblyn R, "Flexible modeling of opioid exposure provides new insights in its association with adverse outcomes.," [Prepared for journal submission].
- 7. Johnson S. P. et al., "Risk of Prolonged Opioid Use Among Opioid-Naive Patients Following Common Hand Surgery Procedures," (in eng), The Journal of hand surgery, Sep 07 2016, pmid:27692801
- 8. Mosher H. J., Hofmeyer B. A., Hadlandsmyth K., Richardson K. K., and Lund B. C., "Predictors of Long-Term Opioid Use After Opioid Initiation at Discharge From Medical and Surgical Hospitalizations," (in eng), Journal of hospital medicine, vol. 13, no. 4, pp. 243–248, Apr 2018, pmid:29624187
- 9. Thornton J. D. et al., "Predictors of Transitioning to Incident Chronic Opioid Therapy Among Working-Age Adults in the United States," (in eng), American health & drug benefits, vol. 11, no. 1, pp. 12–21, Feb 2018. pmid:29692877
- 10. Ray G. T., Bahorik A. L., VanVeldhuisen P. C., Weisner C. M., Rubinstein A. L., and Campbell C. I., "Prescription opioid registry protocol in an integrated health system," (in eng), The American journal of managed care, vol. 23, no. 5, pp. e146–e155, May 1 2017.
- 11. Alghnam S. and Castillo R., "Traumatic injuries and persistent opioid use in the USA: findings from a nationally representative survey," (in eng), Injury prevention: journal of the International Society for Child and Adolescent Injury Prevention, vol. 23, no. 2, pp. 87–92, Apr 2017, pmid:27597400
- 12. Brummett C. M. et al., "New Persistent Opioid Use After Minor and Major Surgical Procedures in US Adults," (in eng), JAMA surgery, p. e170504, Apr 12 2017, pmid:28403427
- 13. Fritz J. M., King J. B., and McAdams-Marx C., "Associations Between Early Care Decisions and the Risk for Long-term Opioid Use for Patients With Low Back Pain With a New Physician Consultation and Initiation of Opioid Therapy," (in eng), The Clinical journal of pain, vol. 34, no. 6, pp. 552–558, Jun 2018, pmid:29135698
- 14. Shah A., Hayes C. J., and Martin B. C., "Factors Influencing Long-Term Opioid Use Among Opioid Naive Patients: An Examination of Initial Prescription Characteristics and Pain Etiologies," (in eng), The journal of pain: official journal of the American Pain Society, vol. 18, no. 11, pp. 1374–1383, Nov 2017, pmid:28711636
- 15. Karmali R. N., Bush C., Raman S. R., Campbell C. I., Skinner A. C., and Roberts A. W., "Long-term opioid therapy definitions and predictors: A systematic review," (in eng), Pharmacoepidemiology and drug safety, vol. 29, no. 3, pp. 252–269, Mar 2020, pmid:31851773
- 16. H A. Tamblyn R, Meguerditchian A, et al. Using Novel Canadian Resources to Improve Medication Reconciliation at Discharge: Study Protocol For a Randomized Controlled Trial. Trials 2012; 13:150. pmid:22920446
- 17. Fregoso G., Wang A., Tseng K., and Wang J., "Transition from Acute to Chronic Pain: Evaluating Risk for Chronic Postsurgical Pain," (in eng), Pain physician, vol. 22, no. 5, pp. 479–488, Sep 2019. pmid:31561647
- 18. Glare P., Aubrey K. R., and Myles P. S., "Transition from acute to chronic pain after surgery," (in eng), Lancet (London, England), vol. 393, no. 10180, pp. 1537–1546, Apr 13 2019, pmid:30983589
- 19. Herzig S. J. et al., "Safe Opioid Prescribing for Acute Noncancer Pain in Hospitalized Adults: A Systematic Review of Existing Guidelines," (in eng), Journal of hospital medicine, vol. 13, no. 4, pp. 256–262, Apr 2018, pmid:29624188
- 20. Khalid L. et al., "Adherence to prescription opioid monitoring guidelines among residents and attending physicians in the primary care setting," (in eng), Pain medicine (Malden, Mass.), vol. 16, no. 3, pp. 480–487, 2015, pmid:25529863
- 21. Kahl L. K., Stevens M. W., Gielen A. C., McDonald E. M., and Ryan L., "Characteristics of opioid prescriptions for discharged pediatric emergency department patients with acute injuries," (in eng), Journal of investigative medicine: the official publication of the American Federation for Clinical Research, vol. 67, no. 6, pp. 1024–1027, Aug 2019, pmid:31109930
- 22. Tamblyn R. et al., "Effect of an Electronic Medication Reconciliation Intervention on Adverse Drug Events: A Cluster Randomized Trial," (in eng), JAMA Netw Open, vol. 2, no. 9, p. e1910756, Sep 4 2019, pmid:31539073
- 23. Park T. W., Saitz R., Ganoczy D., Ilgen M. A., and Bohnert A. S., "Benzodiazepine prescribing patterns and deaths from drug overdose among US veterans receiving opioid analgesics: case-cohort study," (in eng), BMJ (Clinical research ed.), vol. 350, p. h2698, Jun 10 2015, pmid:26063215
- 24. Kragh Andersen P. et al., "Analysis of time-to-event for observational studies: Guidance to the use of intensity models," (in eng), Statistics in medicine, vol. 40, no. 1, pp. 185–211, Jan 15 2021, pmid:33043497
- 25. Therneau T. M., Grambsch P. M., and Pankratz V. S., "Penalized Survival Models and Frailty," Journal of Computational and Graphical Statistics, vol. 12, no. 1, pp. 156–175, 2003/03/01 2003,
- 26. Quantin C. et al., "Variation over time of the effects of prognostic factors in a population-based study of colon cancer: comparison of statistical models," (in eng), American journal of epidemiology, vol. 150, no. 11, pp. 1188–200, Dec 1 1999, pmid:10588079
- 27. Abrahamowicz M., Mackenzie T., and Esdaile J. M., "Time-Dependent Hazard Ratio: Modeling and Hypothesis Testing with Application in Lupus Nephritis," Journal of the American Statistical Association, vol. 91, no. 436, pp. 1432–1439, 1996/12/01 1996,
- 28.
Dyer, "Canada’s prescription opioid epidemic grows despite tamper proof pills," 2015.
- 29. Chou R. et al., "Management of Postoperative Pain: A Clinical Practice Guideline From the American Pain Society, the American Society of Regional Anesthesia and Pain Medicine, and the American Society of Anesthesiologists’ Committee on Regional Anesthesia, Executive Committee, and Administrative Council," (in eng), The journal of pain: official journal of the American Pain Society, vol. 17, no. 2, pp. 131–57, Feb 2016, pmid:26827847
- 30.
"Sylvestre MP, Beauchamp ME, Kyle R, Abrahamowicz M. (2018). WCE: Weighted Cumulative Exposure Models. R package, version 1.0.2. https://cran.r-project.org/package=WCE."
- 31.
"Frank R. Targeting the opioid drug crisis: a Health and Human Services initiative. Health Affairs Blog. April 3, 2015 (http://healthaffairs.org/blog/2015/04/03/targeting-the-opioid-drug-crisis-a-health-and-human-services-initiative).".
- 32.
"Department of Health and Human Services Behavioral Health Coordinating Committee. Addressing prescription drug abuse in the United States: current activities and future opportunities. 2014 (https://www.cdc.gov/drugoverdose/pdf/hhs_prescription_drug_abuse_report_09.2013.pdf.).".
- 33. Deyo R. A. et al., "Association Between Initial Opioid Prescribing Patterns and Subsequent Long-Term Use Among Opioid-Naïve Patients: A Statewide Retrospective Cohort Study," (in eng), J Gen Intern Med, vol. 32, no. 1, pp. 21–27, Jan 2017, pmid:27484682
- 34. Hadlandsmyth K., Lund B. C., and Mosher H. J., "Associations between initial opioid exposure and the likelihood for long-term use," (in eng), Journal of the American Pharmacists Association: JAPhA, vol. 59, no. 1, pp. 17–22, Jan–Feb 2019, pmid:30409501
- 35. Shah A., Hayes C. J., and Martin B. C., "Characteristics of Initial Prescription Episodes and Likelihood of Long-Term Opioid Use—United States, 2006–2015," (in eng), MMWR. Morbidity and mortality weekly report, vol. 66, no. 10, pp. 265–269, Mar 17 2017, pmid:28301454
- 36. Miller M. et al., "Prescription opioid duration of action and the risk of unintentional overdose among patients receiving opioid therapy," (in eng), JAMA Intern Med, vol. 175, no. 4, pp. 608–15, Apr 2015, pmid:25686208
- 37.
HEDIS., "NCQA Updates Quality Measures for HEDIS® 2019," https://www.ncqa.org/news/ncqa-updates-quality-measures-forhedis-2019, Accessed November 25, 2020.
- 38.
Centers for Disease Control and Prevention. Quality Improvement and Care Coordination: Implementing the CDC Guideline for Prescribing Opioids for Chronic Pain. 2018. National Center for Injury Prevention and Control.
- 39. Von Korff M. R., "Long-term use of opioids for complex chronic pain," (in eng), Best practice & research. Clinical rheumatology, vol. 27, no. 5, pp. 663–72, Oct 2013, pmid:24315147
- 40. Goesling J. et al., "Trends and predictors of opioid use after total knee and total hip arthroplasty," (in eng), Pain, vol. 157, no. 6, pp. 1259–65, Jun 2016, pmid:26871536
- 41. Rosenbloom B. N., McCartney C. J. L., Canzian S., Kreder H. J., and Katz J., "Predictors of Prescription Opioid Use 4 Months After Traumatic Musculoskeletal Injury and Corrective Surgery: A Prospective Study," (in eng), The journal of pain: official journal of the American Pain Society, Mar 25 2017, pmid:28347798
- 42. Anciano Granadillo V., Cancienne J. M., Gwathmey F. W., and Werner B. C., "Perioperative Opioid Analgesics and Hip Arthroscopy: Trends, Risk Factors for Prolonged Use, and Complications," (in eng), Arthroscopy: the journal of arthroscopic & related surgery: official publication of the Arthroscopy Association of North America and the International Arthroscopy Association, vol. 34, no. 8, pp. 2359–2367, Aug 2018, pmid:29730217
- 43. Pugely A. J. et al., "Opioid use following cervical spine surgery: trends and factors associated with long-term use," (in eng), The spine journal: official journal of the North American Spine Society, vol. 18, no. 11, pp. 1974–1981, Nov 2018, pmid:29653244