Opioid risk tool validation

Pain Med. Predicting aberrant behaviors in opioid-treated patients: preliminary validation of the Opioid Risk Tool. Salmond S, Allread V. Orthop Nurs. Development and validation of the Current Opioid Misuse Measure. Development of a self-report screening instrument for assessing potential opioid medication misuse in chronic pain patients. J Pain Symptom Manage. Opioid risk assessment in palliative medicine. J Community Support Oncol.

Opioid Epidemic. Opioid epidemic in the United States. Pain Physician. Anesth Analg. Opioid, Risk Tool. In: StatPearls [Internet]. In this Page. Related information. Similar articles in PubMed. Review Biased Opioid Ligands. Epub Sep Comparison of the in vitro efficacy of mu, delta, kappa and ORL1 receptor agonists and non-selective opioid agonists in dog brain membranes.

Brain Res. Epub Jan Activity of the delta-opioid receptor is partially reduced, whereas activity of the kappa-receptor is maintained in mice lacking the mu-receptor.

Overall, it appears that the BRI is a valid risk assessment tool that, after a brief training session, can be used effectively by pain clinicians. Further study is needed in other practice settings and with larger sample sizes.

J Pain Symp Manag. Passik SD, Kirsh KL: The interface between pain and drug abuse and the evolution of strategies to optimize pain management while minimizing drug abuse. Exp Clin Psychopharmacol. Ann Intern Med. Ellingwood S: How to curb prescription drug abuse.

For purposes of this study we determined that a score of 2 or greater would be a more conservative means of capturing incidences of positive opioid misuse.

Self-report of patient status at follow-up was obtained using the PDUQ. Information on urine toxicology screens for any illicit medications was obtained directly from ordered urine screens and the medical records of the patients who were participating in this study.

Each report included evidence of amphetamines, barbiturates, benzodiazepines, cocaine metabolites, ethanol, methadone, other opiates, propoxyphene, cannabinoids, methaqualone, and phencyclidine. All patients were requested to give a urine sample. The participants did not know when a sample would be requested and if and when the test would be conducted.

The patients were requested to give a urine sample during their clinic visits and they were not observed when giving the sample. Specifically, a patient was classified as positive for drug misuse if the person had a positive urine toxicology result.

As in previous studies, 4 , 16 , 41 this triangulation of data allowed for identification of patients by combining objective indicators with self-report measures to maximize the chances of accurately determining opioid misuse. In previous studies the incidence of a positive DMI ranged between Patients entering treatment at the clinic were approached for participation.

Those who agreed and provided consent completed the self-report baseline questionnaires and the initial OCC items. Chart reviews were conducted to ascertain toxicology results. All patients were tracked for at least one year after the conclusion of the study and all patients had given at least one urine sample.

Relations among demographic data, interview items, and questionnaire data were analyzed using correlations, t-tests, computations of coefficient alpha, and receiver operating characteristic ROC curve analysis, as appropriate. We used logistic regression models to assess the impact of OCC baseline variables on the outcome. We started with a univaraite logistic regression model by considering one of the 12 OCC baseline variables each time, in order to assess individual predictive capability of each variable.

Then, we considered multivariate logistic regression models with a different number of predictors OCC baseline variables , from a model with 2 predictors to a model with all 12 predictors. We identified the best prediction model with a certain fixed number of predictors e. The area under the ROC curve AUC , which is a rank-based measure of binary classification performance, is frequently used in evaluation of medical diagnostic tests.

If we look at the AUC of a model fitted on the full data set, we will find that the more predictors in a model, the higher the AUC, even if an individual predictor may not be useful in predicting the outcome. Twelve self-report items were developed for the OCC by the research team based on the literature search and consensus from clinical experience in using a clinic-based opioid agreement. Each item could be answered by patients on a yes and no scale.

The 12 items were created so that at least one item reflected the content of each of the categories identified in the literature review. One hundred fifty-seven patients who were taking long-term opioid medication for chronic noncancer pain were recruited for this study Table 1.

The average age of the patients was Electronic medical record data were assessed for up to 3 years after the conclusion of the study to examine urine toxicology screen data. Ten subjects who were not followed were either lost to contact or refused to participate. The patients were prescribed immediate-release Twenty seven percent were taking both long and short-acting opioids for pain.

Urine toxicology results were available for all the patients and data from the urine screens resulted in multiple screens over a 5-year period average number of screens was Of the initial subjects, 70 Sixty six No differences were found between positive and negative DMI groups on age, gender, race, disability status, employment status, anxiety or depression HADS , pain site, pain duration, pain intensity, and whether or not they were taking short-acting opioids.

Analyses were conducted to determine which of the 12 items should be included in the OCC Prediction Score. Among the original subjects, 70 or We excluded the subjects with missing values in OCC baseline variables in order to calculate five-fold cross-validated AUC. After excluding observations with missing values in OCC baseline measures, we randomly split the observations into five groups with 25 subjects in each group.

Using four out of five groups as the training dataset, we implemented each of the proposed estimation procedures. We then applied the final fitted regression model to the group that was omitted from the training set and calculated the AUC for that subset. We repeated this procedure five times so that each subject served as the validation set once. The model with higher cross-validated AUC was determined to have better prediction performance.

We calculated the five-fold cross-validated AUC from univariate logistic regression models. Among the 12 OCC baseline variables, item 5 ran out of pain medication early had the highest prediction capability, with a cross-validated AUC of 0.

Table 4 summarizes the best prediction model with the different number of predictor variables. For example, among all the prediction models with 2 predictor variables, the model with OCC baseline variables v2 and v5 had the best prediction performance, with cross-validated AUC 0. Comparing the 12 best prediction models with different number of predictors, the prediction model with 5 OCC baseline variables item number 4, 5, 6, 8, and 12 was the best among all the models, which had the highest cross-validated AUC of 0.

After identifying the model with best prediction performance, we then obtained the prediction score by fitting the model on the full data set, which is applicable to a future patient. The prediction score is based on a logistic regression model with OCC baseline variables item numbers 4, 5, 6, 8, and 12 fitted on all the subjects without missing values in these five OCC baseline variables. The AUC for this prediction score was 0. Importantly, given that opioid compliance is a heterogeneous construct that may involve a wide range of behaviors that are not necessarily intercorrelated, it is not surprising to obtain a relatively low internal consistency estimate.

In terms of test-retest reliability, given that opioid compliance is likely to fluctuate over time, estimates of test-retest coefficients should not be expected to be high.

Fitting a multivariate logistic regression model on all the subjects without missing values in these five OCC baseline variables, the resulting prediction score was found to have an AUC of. Using the cutoff value. Alternatively, using a score equal to the simple sum of the 5 items, the AUC for the prediction was.

As seen, the performance of the prediction using the simple sum of the 5 items is comparable to the performance of the optimal prediction score based on a multivariate logistic regression model with the 5 items. These combined results suggest that one positive response on the OCC smallest cutoff value was the best predictor of likelihood of opioid misuse. Although statistically the 5-item model yields the best prediction results Table 4 , 10 of the 12 items were found to be valuable in predicting opioid misuse and may be clinically useful.

Since a difference of 0. Thus, the item model, which has the cross-validated AUC slightly lower than 0. Using SUM to represents the sum of scores for the 10 items all except item numbers 9 and 11 , the data set suggests that Thus, the greater the number of items endorsed in the direction of misuse the greater the chances that person will develop a problem with misuse of opioids. While, overall, any positive response on the 5-item OCC is predictive of future opioid misuse, our findings suggest adequate stability and clinical utility with using 10 items, especially given the possibility that situations tapped by the different OCC questions are likely to change over an extended time interval.

Despite the growing use of opioid agreements designed to list responsibilities among consenting patients prescribed opioids for pain, there have not been any tools specifically designed to periodically monitor adherence with these agreements. The present study reports on a consensus-based effort to develop and test such a brief compliance checklist. The measure is easily understood by patients, takes very little time to administer and score 1 minute , and taps information believed to be important for determining adherence among chronic pain patients who are prescribed long-term opioid medication.

Data collected on a sample of chronic pain patients suggests the 5-item OCC may be a useful and valid checklist of opioid adherence among persons with chronic pain. Opioid therapy has been shown to be effective for some patients with chronic pain. While the 5-item OCC requires additional research, the findings in this study suggest that this brief checklist may be a useful means to help monitor use of opioids in the clinic. At a minimum, this self-report checklist can be used to alert the treating physician to potential risks that might help avert future problems.

The internal consistency of the 5-item OCC was found to be. Some authors 23 suggest that clinical measures should have alphas in excess of. Perspective: The revised ORT is the first tool developed on a unique cohort to predict the risk of developing an OUD in patients with CNMP receiving opioid therapy, as opposed to aberrant drug-related behaviors that can reflect a number of other issues.

The revised ORT has clinical usefulness in providing clinicians a simple, validated method to rapidly screen for the risk of developing OUD in patients on or being considered for opioid therapy.