Strategies to Identify Patient Risks of Prescription Opioid Addiction When Initiating Opioids for Pain

Key Points Question How can physicians identify patients with pain for whom prescription opioids can be safely prescribed? Findings This systematic review found that a history of opioid use disorder or other substance use disorder, a mental health diagnosis, and concomitant prescription of certain psychiatric medications may be associated with an increased risk of prescription opioid addiction. However, only the absence of a mood disorder appeared useful for identifying lower risk, and assessment tools incorporating combinations of patient characteristics and risk factors were not useful. Meaning This study suggests that there are few valid ways to identify patients who can be safely prescribed opioid analgesics.


eAppendix. Search methods and search strategy
Eligible studies compared patient symptoms and signs among patients being newly prescribed opioids for pain who did or did not subsequently develop prescription OUD. Studies assessing screening tools that utilized combinations of symptoms and signs were also eligible. To identify relevant articles, MEDLINE and EMBASE from January 1946 to October 2017 were searched. Search strategy terms included opioid-related disorders, MESH terms substance related disorders, pain, analgesics, and terms previously found to be useful for retrieving diagnostic studies (see Search Strategy). 1 Additional studies were identified by searching reference lists of original studies and review articles.

Study Selection
Two reviewers (LG and JK) independently screened abstracts for inclusion. Studies that evaluated prescription characteristics, patient characteristics, past substance use disorders, mental health disorders and screening tools assessing the risk of prescription opioid addiction in the context of pain management were included. Articles not reporting original data (i.e. review articles) were also excluded. To be eligible for the present review, we also restricted to studies of opioid naïve patients newly starting opioid medications for pain and excluded studies assessing for a diagnosis of OUD among patients already on opioid-based medications.

Outcome measures
The following outcomes were assessed: symptoms, signs, risk factors, and scores on screening tools of patients who subsequently did and did not develop prescription OUD. As there is currently no gold standard for the diagnosis of OUD in pain patients that has been described in the literature, 2 and since the diagnostic criteria for OUD have evolved over time. 3 We allowed for the definitions that have been used in the literature including a diagnosis of OUD using the Diagnostic and Statistical Manual (DSM), and diagnoses of opioid "abuse" and "dependence" using the DSM-III, DSM-IV, ICD-9, or ICD-10. In addition, we included eligible studies where the presence of aberrant drug-related behaviors and failed urine drug screens was taken as a valid proxy for the above in articles of diagnostic screening in pain care.

Data extraction
All citations identified by searches were independently screened based on title and abstract by two reviewers (LG, JK). Each potentially relevant study was then reviewed in full text and assessed for all inclusion criteria. Any disagreements were resolved by discussion among reviewers and senior authors (JK, EW). Relevant data from eligible articles (i.e., patient and treatment characteristics, outcomes, etc.) were then extracted.

Quality Assessment
Two reviewers (LG and LA/JK) rated study quality using a five-level Hierarchy of Evidence rating scale by Simel and Rennie used as part of the Journal of the American Medical Association's Rational Clinical Examination series (2008). 4 Using this schema, Level 1 indicated the highest quality and was assigned to studies that had independent blinded comparison of the symptoms or signs with a valid criterion standard in a large number of consecutive patients (for this review defined as greater than 150). 4 Level 2 studies were similar to level 1 studies but enrolled fewer than 150 patients. Level 3 studies enrolled nonconsecutive patients. Level 4 studies used non-independent comparisons among a "convenience" sample of patients at risk of having the prescription OUD. Consistent with prior reviews in this series, 4 only studies that met the quality standards of Level 1, 2, or 3 were included. In accordance with the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA-DTA), and Standards for Reporting Diagnostic Accuracy (STARD), sources of bias were also evaluated with the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) Tool. 5  Was the spectrum of patients representative of the patients who will receive the test in practice? Patients at risk of opioid addiction (condition) = yes. If no risk of OUD = no.

2.
Were selection criteria clearly described? If reproducible = yes.

3.
Is the reference standard likely to correctly classify the OUD? If standard laboratory techniques used to diagnose opioid addiction = yes. If ambiguous = no.

4.
Is the time period between reference standard and index test short enough to be reasonably sure that the target condition did not change between the two tests? If OUD testing and assessment done as part of the same consultation or research study site visit = yes. If reported duration between assessment and OUD testing more than 2 days = no.

5.
Did the whole sample or a random selection of the sample, receive verification using a reference standard of diagnosis? Yes or no.

6.
Did patients receive the same reference standard regardless of the index test result? Yes or no.

7.
Was the reference standard independent of the index test (i.e. the index test did not form part of the reference standard)? Yes or no.

8.
Was the execution of the index test described in sufficient detail to permit replication of the test? If description adequately described to allow for replication, including a symptom definition, = yes.

9.
Was the execution of the reference standard described in sufficient detail to permit its replication? If laboratory approach to diagnosing OUD described then = yes. 10 Cochran et al, 2014 55 U Y Y n/a Y n/a n/a n/a Y n/a n/a U Y Y Y n/a Edlund et al, 2010 56 Y Y Y n/a Y Y n/a n/a U n/a n/a Y Y N Y n/a Jones et al, 2015 20

Studies excluded from quantitative synthesis
Clarke et al, 2014 57 N Y N n/a Y n/a n/a n/a Y n/a n/a Y Y Y Y n/a Hooten et al, 2015 58 Y Y N n/a Y n/a n/a n/a Y n/a n/a Y Y Y Y U n/a indicates a study in which there was no index test. These were retrospective cohort studies that looked at the characteristics of patients that did vs did not develop OUD following an opioid prescription.      a The LR range includes two disparate populations, 1) one national, commercially insured population (HealthCore in the West, Mid-West, and South-East regions of the U.S.) and 2) one state-based, publicly insured (Arkansas Medicaid serves "a disadvantaged and vulnerable population with the highest opioid use in the U.S.). Any mental health disorder was derived from the presence of adjustment disorder, anxiety disorder, mood disorder, personality disorder, and miscellaneous disorders (such as an eating disorder or somatoform disorder). For results on an ordinal scale (0, 1, 2 mental health disorders) the sensitivity, specificity, and LR-no longer apply. The LR represents the LR at increasing numbers of mental health disorders from 0 to ≥2. *Positive test indicated by the presence of more 'medium', 'medium high' 'high' and 'very high' ratings (high risk) than 'low' and 'low medium' ratings (low risk) on 12 risk categories. § Although this study 20 did not report high specificity (LR+), it is likely the most accessible of the reported tools as it can be accessed on a US government (.gov) website and has no copyright. # Patients received at least 30 days supply of any opioid, i.e., Schedule III or IV AND short-acting schedule II AND longacting schedule II opioids within a 6-month period. ¥ Total N = 397, but only 155/397 of the total participants had Urine Drug Screening information available. Moreover, only those patients who were suspected of "misusing" opioids underwent urine drug screening. # Patients received at least 30 days supply of any opioid, i.e., Schedule III or IV AND short-acting schedule II AND longacting schedule II opioids within a 6-month period. a The LR range is derived from two separate databases described in this study. 56 Any mental health disorder was derived from the presence of adjustment disorder, anxiety disorder, mood disorder, personality disorder, and miscellaneous disorders (such as an eating disorder or somatoform disorder). For results on an ordinal scale (0, 1, 2 mental health disorders) the sensitivity, specificity, and LR-no longer apply. The LR represents the LR at increasing numbers of mental health disorders from 0 to ≥2.