In this retrospective cohort study of opioid-naïve patients admitted for an acute medical hospitalization who reported pain, the majority received an opioid. Older patients, Black patients, and patients with multiple admissions during the study period were less likely to receive opioids. Up to 1 year later, only 3% of patients met our criteria for long-term use, and this was associated with opioid receipt at discharge but not opioid receipt during the hospitalization alone. Pain intensity was a strong predictor of opioid receipt, and inclusion of pain in the model partially mediated the association between opioid receipt and long-term use. Documentation by physicians was extremely poor; the majority of charts lacked any documentation of reason for opioid use.
In a national sample, Herzig et al found that more than half of the medical inpatients received an opioid during their stay; however, they were unable to determine what proportion of patients were taking opioids chronically or to identify the indications for opioid use. We found that >75% of patients were opioid experienced preadmission, so most opioids went to those already taking them. Herzig and colleagues also were unable to assess pain. We found that nearly 70% of patients reported pain during their hospitalization, and three-fourths of these (51% overall) received an opioid. This is identical to the proportion observed by Herzig and colleagues. Our study adds to their findings by noting that the high overall rate of opioid prescribing was because most of the patients had pain, and many patients with pain did not receive opioids. Providing appropriate pain control to medical patients is one of the challenges of hospital medicine. Our data suggest that patients describing more pain are more likely to receive opioids, which is appropriate.
A handful of recent studies has found positive associations between opioid exposure in the hospital and long-term use;[19–22] however, these studies also did not measure pain, which is an important confounder. Not surprisingly, we found that patients who have severe pain are both more likely to receive opioids in the hospital and to require them long term. In particular, pain score on the last day before discharge strongly predicted long-term use. Accounting for pain at discharge attenuated the association between opioid exposure and use 6 to 12 months later. Analyses that do not account for pain will likely overestimate the danger of prescribing opioids.
Still, opioids at discharge were predictive of opioid use 6 to 12 months later, even after adjusting for pain; however, the percentage of patients with long-term use was low, which is consistent with two prior studies.[21,22] The absolute increase related to opioid prescribing was <2%, meaning that >50 patients would have to forgo opioids in the hospital to prevent one from using opioids long term. For patients who received opioids in the hospital but not at discharge, the absolute risk was even less. As such, avoiding opioids at discharge may be a better way to reduce long-term use. Although pain at discharge was an important predictor of opioids at discharge, it did not fully explain prescribing at discharge. Understanding why patients are discharged with opioids could help practitioners to understand the nature of the relation with long-term use. Unfortunately, we found documentation to be extremely poor, limiting our ability to understand prescribing rationale.
To our knowledge, ours is the first study to disaggregate the risk of long-term use based on whether patients received opioids during their stay only or also at discharge. Our findings support the growing body of literature[21,22] pointing to discharge as an important exposure point for opioid-naïve patients.
In deciding who should receive an opioid, physicians must balance pain against the risk of addiction or adverse events. Although we found no significant differences in the risk of long-term use by patient demographic characteristics, physicians were less likely to prescribe opioids to Black patients, older patients, and patients with a psychiatric history. Some of these differences may represent concerns about addiction in patients whom doctors suspect are high risk. Psychiatric comorbidities are common among patients with opioid use disorder. The lower rate of prescribing to Black versus White patients is harder to explain, and is likely the result of racial bias. Disparities in pain control between White and Black patients are well documented.[11,18,24,25] We found no difference in odds of long-term use between White and Black patients. To the extent that physicians may be prescribing to Black patients at a lower rate because of concerns about long-term use, this concern is unfounded and could result in inadequate pain control.
Physicians also were less likely to prescribe opioids to older patients, who were not at higher risk of long-term use. Age is independently associated with opioid-related adverse events,[26,27] and physicians may be more worried about adverse events than addiction in older patients. Unfortunately, there are few safer alternatives, and untreated pain also is a risk factor for delirium in older adults.[28,29]
Given the competing issues of patient expectations, fear of causing addiction, and desire to provide appropriate pain management, physicians may need decision support, particularly for patients they suspect may be at risk of addiction or adverse effects. In 2018 the Society of Hospital Medicine published guidelines to guide opioid use for hospitalized patients with noncancer pain. These cover the decision to initiate opioids, type and dosage if opioids are prescribed, and opioid prescription at discharge. The extent to which these new guidelines are used to inform opioid prescribing decision making is unknown. Shared decision making has demonstrated positive results in reducing opioid use in the postsurgical setting;[31,32] however, physicians report many challenges in communicating with patients with chronic pain about opioids.[33,34] The frustration of negotiating with these patients may have contributed to the poor documentation we identified. Regardless of the cause, documentation stands as an area ripe for improvement.
Our study had some limitations. Our data are from hospitals in northeast Ohio. The opioid burden in Ohio is high and may limit the generalizability of this study to other regions. Our prescribing rates were similar to those described elsewhere, however. Although we included many covariates in our analysis, there is potential for confounding by unobserved measures. Our outpatient medication data were limited to prescriptions ordered by physicians. These prescriptions may have not been filled by patients, potentially overestimating long-term opioid use compared with pharmacy fill data. Patients also may have obtained opioids via other health systems or procured them illegally, neither of which we can account for.
In our study of medical inpatients with pain, the majority received an opioid, yet documentation regarding indication was scarce. Physicians were less likely to prescribe opioids to some groups of patients; however, these groups were no more likely to become long-term users. Overall, very few patients were using opioids 6 to 12 months after their hospitalization. Physicians may understandably be reticent to prescribe opioids to opioid-naïve patients for fear of starting a cascade of long-term use. For patients who receive opioids only during their hospitalization, however, the risk appears low. Receipt of opioids at discharge is associated with an increased, albeit still small, risk of long-term use, and this appears to be at least partially explained by higher pain levels. This underscores the necessity of accounting for patient pain when examining the association between opioid exposure and long-term use. Studies that fail to do so will likely overstate the danger of using opioids to treat pain in medical inpatients.
South Med J. 2021;114(10):623-629. © 2021 Lippincott Williams & Wilkins