Perioperative Opioid Prescribing and Iatrogenic Opioid Use Disorder and Overdose: A State-of-the-Art Narrative Review

Abstract

Background/Importance:

Considerable attention has been paid to identifying and mitigating perioperative opioid-related harms. However, rates of post-surgical OUD and overdose, along with associated risk factors, have not been clearly defined.

Objective:

Evaluate the evidence connecting perioperative opioid prescribing with postoperative opioid use disorder (OUD) and overdose, compare these data with evidence from the addiction literature, discuss the clinical impact of these conditions, and make recommendations for further study.

Evidence Review:

State-of-the-art narrative review.

Findings:

Nearly all evidence is from large retrospective studies of insurance claims and Veterans Health Administration (VHA) data. Incidence rates of new OUD within the first year after surgery ranged from 0.1–0.8%, while rates of overdose events ranged from 0.01–0.8%. Higher rates were seen among VHA patients, which may reflect differences in data completeness and/or risk factors. Identified risk factors included those related to substance use (preoperative opioid use; non-opioid substance use disorders; preoperative sedative, anxiolytic, antidepressant, and gabapentinoid use; and postoperative new persistent opioid use); demographic attributes (chiefly male sex, younger age, white race, and Medicaid or no insurance coverage); psychiatric comorbidities such as depression, bipolar disorder, and PTSD; and certain medical and surgical factors. Several challenges related to the use of administrative claims data were identified; there is a need for more granular retrospective studies and, ideally, prospective cohorts to assess postoperative OUD and overdose incidence with greater accuracy.

Conclusions:

Retrospective data suggest an incidence of new postoperative OUD and overdose of up to 0.8% during the first year after surgery, but prospective studies are lacking.

INTRODUCTION

Most opioid misuse and about 25% of the 80,000 annual overdose deaths in the U.S. involve prescription opioids, half of all patients with opioid use disorder (OUD, the Diagnostic and Statistical Manual for Mental Disorders 5^th edition [DSM-5] diagnostic term for opioid addiction,¹ Table 1) are initiated to opioids via a prescription, and the majority of individuals with heroin use disorder initially misuse prescribed opioids.^2–5 Opioid prescribing for acute pain, and specifically postoperative pain, contributes substantially to these outcomes. Postoperative opioid prescribing is often patients’ first opioid exposure, and the large percentage of unused postoperative opioids are a reservoir for diversion and accidental poisoning.^6–8 In response, numerous guidelines and best practice recommendations to improve postoperative opioid prescribing to mitigate these harms have been published.^{9; 10} Implementation of such guidelines has not led to worsened pain, decreased satisfaction, or increased need for refills.^{11; 12}

Table 1:

OUD diagnostic criteria (DSM-5). At least 2 of the following should be observed over a 12-month period.² Emphasis added.

1. Opioids are often taken in larger amounts or over a longer period than was intended.

2. There is a persistent desire or unsuccessful efforts to cut down or control opioid use.

3. A great deal of time is spent in activities necessary to obtain the opioid, use the opioid, or recover from its effects.

4. Craving, or a strong desire or urge to use opioids.

5. Recurrent opioid use resulting in a failure to fulfill major role obligations at work, school, or home.

6. Continued opioid use despite having persistent or recurrent social or interpersonal problems caused or exacerbated by the effects of opioids.

7. Important social, occupational, or recreational activities are given up or reduced because of opioid use.

8. Recurrent opioid use in situations in which it is physically hazardous.

9. Continued opioid use despite knowledge of having a persistent or recurrent physical or psychological problem that is likely to have been caused or exacerbated by the substance.

10. Exhibits tolerance* (either sub-criterion)

a. A need for markedly increased amounts of opioids to achieve intoxication or desired effect.

b. A markedly diminished effect with continued use of the same amount of an opioid.

11. Exhibits withdrawal* (either sub-criterion)

a. The characteristic opioid withdrawal syndrome.

b. Opioids (or a closely related substance) are taken to relieve or avoid withdrawal symptoms.

^*Tolerance and withdrawal criteria are not applicable to individuals taking opioids solely under appropriate medical supervision.

Although increasing attention has been given to the approximately 10% of opioid-naïve surgical patients who develop new persistent opioid use (NPOU) following surgery, the greatest impacts on morbidity and mortality are seen with frank OUD and overdose. To date, postoperative rates of these conditions are not clearly defined. To optimize post-surgical opioid prescribing and minimize iatrogenic OUD and overdose, the incidence of and risk factors for these conditions in the post-surgical context must be understood. This focused narrative review evaluates literature connecting perioperative opioid prescribing with postoperative OUD and overdose, addresses clinical impact of these conditions, and makes recommendations for further study.

INCIDENCE OF POSTOPERATIVE OUD AND OVERDOSE

New postoperative OUD

Existing postoperative OUD incidence data derive largely from examining diagnostic codes within large insurance claims databases (Table 2), with all but one solely examining U.S. data. Several U.S. studies have examined broad surgical populations. Brat evaluated 568,612 patients undergoing nearly all types of inpatient and outpatient surgeries.¹³ The composite outcome was a new International Classification of Diseases (ICD)-9 diagnosis of opioid abuse, dependence, or overdose, excluding those with preexisting OUD diagnosis.¹⁴ The composite outcome was observed in 0.6% of patients overall (5,906; median 2.7 years follow-up), but only 0.2% (1,857) within one year postoperatively. Another broad retrospective cohort study by Wylie included 35,335 opioid-naïve patients undergoing any type of surgery at a single academic medical center.¹⁵ All included patients were followed for a median 3.5 years postoperatively. The authors used ICD-9 and −10 codes followed by manual chart review to identify cases of incident OUD within the single-center cohort with greater specificity. Results indicated that 0.2% (63 patients) displayed incident postoperative OUD. Additionally, surgical patients not prescribed a postoperative opioid had a significantly lower rate of OUD or overdose compared with those who did receive an opioid (1.7 vs. 5.6 events per 10,000 person-years of follow-up). Another relatively broad study (Kim) evaluated 11,713 patients undergoing nine major surgeries within a large nationwide commercial claims database.¹⁶ The OUD outcome reflected new ICD-9 codes for opioid abuse or dependence. During the up-to-5-year follow-up period, 0.5% (61) developed incident OUD.

Table 2:

Studies of postoperative OUD incidence.

First author	Year	Sample size	Postoperative timeframe	Opioid-naïve	Prior OUD exclusion	Continuous insurance coverage required	Evidence of postoperative opioid fill required	F11.90 not used	Excludes overdose as outcome	OUD Incidence	Notes
Shah	2017	675,527	1 year	No	Yes	No	No	Yes	No	0.09%	Urological surgeries only
Brat	2018	568,612	1 year	Yes	Yes	Yes	Yes	Yes	No	0.2%
Aalberg	2022	304,780	1 year	Yes	Yes	Yes	No	No	Yes	0.7%	VHA population
Siglin	2020	261,208	1 year	No	Yes	Yes	No	Yes	Yes	0.8%	VHA population; colonoscopy included
Huang	2020	301,871	180 days	Yes	No	Yes	Yes	Unknown	No	0.5%
Jivraj *	2020	162,830	1 year	Yes	No	Yes	No	No	Yes	0.1%	Ontario population
Wiese	2021	56,723	1 year	Yes	Yes	Yes	Yes	Yes	Yes	0.7%
Kim	2020	11,713	Up to 5 years	Yes	Yes	Yes	Yes	Yes	Yes	0.5%
Wylie	2022	35,335	Median 3.5 years	Yes	Yes	No	Yes	Unknown	Yes	0.2%	Single center; included manual record review

^*Canadian population

Three studies have focused specifically on urologic or gynecologic surgery populations. In the largest postoperative OUD study to date, Shah examined a composite outcome of ICD-9 codes reflecting postoperative opioid abuse, dependence, or overdose in 675,527 California patients undergoing various urologic surgeries.¹⁴ Results indicated that 0.05% of patients received new OUD or overdose diagnoses within 90 days of surgery (372 patients) and 0.09% within 1 year of surgery (590 patients). In another study, Huang examined 301,871 patients undergoing nine common benign gynecologic surgeries within large commercial insurance and Medicaid databases covering 12 states.¹⁷ The OUD outcome comprised hospitalizations or emergency department visits with ICD-9 or −10 codes for “opioid dependence, misuse, or overdose.” While overall OUD incidence is not directly reported for the full sample, it can be calculated that about 0.5% of patients (1,539) met the composite OUD outcome within 180 days of surgery (Supplemental Digital Content 1).

Another study by Wiese analyzed 65,170 opioid-naïve Tennessee patients with continuous Medicaid enrollment who underwent cesarean delivery.¹⁸ “Serious opioid-related events” was the primary outcome, reflecting earliest development of: (1) NPOU (defined conservatively as filling a 90-day opioid supply within a 180-day period, (2) OUD (a prescription fill for methadone or buprenorphine, or ICD-9 code for opioid abuse or dependence), (3) opioid-related overdose, or (4) ICD-10 death certificate code for opioid-related death. OUD was the first qualifying event in 0.7% (462 patients) within the first year postoperatively. Of the 1,136 patients whose first qualifying endpoint was NPOU, 40 eventually met study criteria for OUD within the first follow-up year (A. Wiese, personal communication, October 19, 2022). This would increase overall OUD incidence to at least 0.8% (502 patients).

Two studies have targeted the Veterans Health Administration (VHA) population. In a VHA retrospective database study, Aalberg examined 304,780 opioid-naïve patients undergoing common major and minor surgeries.¹⁹ The primary OUD outcome was appearance of new ICD-9/10 codes for opioid abuse or dependence. Overall, 2.9% of surgical patients (8,950) developed new postoperative OUD over 5.6 years (median) follow-up. However, only 0.7% (2,006) did so during the first postsurgical year. Another study by Siglin using the same VHA database included 261,208 patients undergoing 26 common procedures (including 16.1% undergoing colonoscopy).²⁰ The composite outcome used the same ICD-9 codes for opioid abuse and dependence as in the work described above, as well as ICD-10 codes for opioid abuse and dependence. OUD incidence within the first postsurgical year was 0.8% (2,155 patients), compared with 0.6% of a propensity-matched cohort of VHA patients not undergoing surgery (p<0.001). Among patients with no prior OUD history, having a procedure was associated with 13% increased odds for new OUD diagnosis during the first postoperative year (OR 1.13 [1.02–1.24]).

Only one study was identified that addressed postoperative OUD rates in a non-U.S. population, a study of 162,830 opioid-naïve patients in Ontario, Canada undergoing 18 common surgical procedures (Jivraj).²¹ The single-payer nature of healthcare in Ontario likely renders more complete data than U.S. insurance databases, although postoperative prescribing practices differ between the U.S. and Canada.²² The primary OUD outcome was an emergency department, acute care facility, or mental health institution presentation with an ICD-10 code of F11.XX (encompassing codes for OUD-adjacent “opioid abuse” and “opioid dependence” but also “opioid use”) or a new prescription fill for buprenorphine or methadone. Overall, 0.1% (146 patients) developed postoperative OUD within the first postoperative year.

Synthesis of the evidence: the incidence of opioid use disorder after surgery

Overall, incidence rates of new OUD within the first postoperative year ranged from approximately 0.1 to 0.8%. Extrapolating based on the 44.7 million individual opioid prescriptions written yearly by surgeons and dentists in the United States,²³ there may be 45,000 to 358,000 new post-surgical OUD cases annually. One study evaluated OUD rates in surgical patients compared with a control group that did not undergo surgery, and another compared with a control group that underwent surgery without receiving a postoperative opioid.^{15; 20} Both identified significantly higher OUD incidence among surgical patients receiving opioids, suggesting specific links between opioid-related OUD susceptibility and perioperative prescribing without proving causation. The lone study conducted outside of the U.S. (Jivraj—Canada) reported a lower OUD rate than most of the U.S. studies examined.²¹ This lower OUD incidence might reflect the impact of more cautious postoperative opioid prescribing in Canada vs. the U.S.²²

Methodological limitations in existing studies of postoperative OUD

Multiple potential methodological issues are present in the studies reviewed above that may limit interpretability of reported OUD incidence rates. All nine studies reviewed were retrospective and, with one exception (Wylie), depended exclusively on ICD codes in the electronic health record to identify incident OUD. Consequently, true OUD rates could be higher that reported if inadequate surveillance and reporting of OUD by healthcare providers is assumed; of note, OUD in general is underdiagnosed in clinical contexts, with existing prevalence estimates likely undercounting the number of people with OUD by at least 3–5 times.²⁴ Alternatively, OUD incidence may be lower than reported due to inaccuracies in coding. Concern for this latter possibility is heightened by the results of the lone study employing confirmatory chart review (Wylie): Of 231 patients with OUD or opioid overdose diagnostic codes, 164 (71%) were excluded following medical record review.¹⁵ This considerable number of OUD false positives indicates that larger database-only studies may contain similar inaccuracies.

Whether OUD-related ICD codes in electronic health records indicated that patients met formal OUD criteria (DSM-5) is unknown in retrospective database studies, raising additional questions about true OUD rates in the postoperative population. An additional issue is that ICD codes for OUD and related non-OUD diagnoses overlap. The “opioid abuse” and “opioid dependence” terms used as DSM analogs of OUD by these studies may not be congruent with actual OUD criteria. As few as 58–62% of such patients met DSM-5 OUD criteria in one study directly examining this issue.^{25; 26} Another study evaluating 307 patients across four primary care clinics in granular detail found similarly limited correspondence between these OUD-related codes and confirmed OUD: 45% for F11.23 (“opioid dependence with intoxication with withdrawal”), 51% for F11.20 (“opioid dependence, uncomplicated”), and 64% for F11.10 (“opioid abuse, uncomplicated”). Each of these codes is commonly used in administrative claims studies as evidence of OUD, a methodology likely to inflate OUD incidence rates in retrospective database studies.

Other ICD codes used to define OUD in some studies are even more problematic. At least two studies (Aalberg, Jivraj; unclear for Huang and Wylie) used the F11.90 ICD-10 code for “opioid use, unspecified, uncomplicated” to indicate OUD.^{15; 17; 19; 21} This diagnosis has particularly limited overlap with DSM-5 OUD criteria: 43% in one study.²⁷

Continuous insurance coverage throughout the study period is required to capture all pertinent events and provide reliable incidence rates. Failure to require this may have led to underestimation of incident OUD rates in two studies (Shah, Wylie).^{14; 15} Not requiring evidence of a postoperative opioid prescription fill for inclusion may complicate the evaluation of associations between postoperative opioid use and subsequent OUD (and attribution of incident OUD to perioperative opioid exposure) in four studies (Shah, Aalberg, Siglin, Jivraj).^{14; 19–21} Of particular concern for this issue is Siglin, for which 16.1% of the cohort underwent colonoscopy, a procedure that rarely entails postoperative opioid prescription. This aspect of the latter study may lead to underestimation of postoperative OUD incidence. Two studies (Shah, Siglin) included patients who were not opioid-naïve, which may lead to the conflating of undiagnosed pre-surgical OUD with new postoperative OUD.^{14; 20} The same concern applies to Huang and Jivraj, which did not exclude patients with prior OUD.^{17; 21} Additionally, three studies (Shah, Brat, Huang) utilized a composite outcome including overdose, which can occur in the absence of OUD.^{13; 14; 17} Finally, Aalberg and Jivraj included use of certain medications as indicators of OUD even though they are sometimes prescribed for pain management in the absence of OUD.^{19; 21} For example, Aalberg included new prescriptions for buprenorphine/naloxone, methadone, buprenorphine, and naltrexone (potential medications for OUD; MOUD) as OUD surrogates, despite these medications sometimes being prescribed for chronic pain without coexisting OUD. These issues may also potentially lead to overreporting of OUD. In summary, existing studies of postoperative OUD incidence are based on retrospective database studies and suffer from multiple methodological issues that could potentially lead to significant overreporting or underreporting of true postoperative OUD incidence.

Postoperative overdose

Several studies directly report overdose rates in postoperative patients following discharge. All except one include U.S. patients exclusively. The largest such study (Ladha) examined 1,305,715 patients undergoing 22 common surgeries within a nationwide commercial insurance database.²⁸ Overdose was defined as hospitalization or emergency department visit with an ICD-9 code for “poisoning” or “accidental poisoning” by non-heroin opioids within 30 days post-discharge. The 134 (0.01%) patients with overdose yielded a rate of 10.3 overdoses per 100,000 surgeries (95% CI 8.7–12.2). This rate decreased to 3.2 overdoses (95% CI 2.3–4.3) per 100,000 surgeries between 61–90 days postoperatively.

The previously discussed study by Jivraj also assessed overdose through ICD-10 codes.²¹ Of 162,830 Canadian surgical patients, 0.01% (20 patients) had an overdose event within the first postoperative year (a much longer timeframe than the 30-day window in the largest overdose study). Among Wiese’s 65,170 Tennessee Medicaid post-cesarean-section patients, 0.05% (35 patients) had overdose events identified by codes during the first year postoperatively.¹⁸ Of note, this latter study may understate the actual overdose rate given that the study protocol stopped analysis as soon as any endpoint was met. A similar overdose rate was seen in the single-center academic medical center study (Wylie; n=35,335) that defined outcomes based on codes that were confirmed by manual chart review: 0.03% (10 overdose events).¹⁵ However, the latter study was over a much longer follow-up (approximately 3.5 years) than other studies. Differences between the academic medical center patient population in Wylie and the Medicaid population in Wiese may contribute to the discrepancy in incidence.

Three VHA studies identified much higher overdose rates compared with the studies described above. The two VHA database studies discussed in the OUD section, Aalberg and Siglin, also independently assessed overdose.^{19; 20} In Aalberg (n=304,780), 1.26% of opioid-naïve surgical patients (3,850) had an overdose event as assessed by ICD-9 and −10 codes across all follow-up (median 5.56 years). When analysis was restricted to the first postoperative year, 0.68% (2,069 patients) had overdose events. Siglin (n=261,208) showed a similar overdose rate, with 0.7% (1,893 patients) experiencing overdose events within the first postoperative year. Notably, this latter study observed overdose events in only 0.1% of propensity-matched controls (518 patients) who did not undergo surgery during the same timeframe (OR 6.71 for overdose in surgical patients vs. controls [95% CI 5.80–7.75]). The discrepancy in overdose rates between VHA surgical patients and matched controls, as well as the fact that the highest incidence of postoperative overdose in the smaller Siglin study occurred within one month of surgery (median 12 days), supports a potential relationship between postoperative opioid prescribing and overdose. A third VHA database study by Mudumbai et al. also looked at overdoses by ICD-9 codes within 64,391 patients undergoing inpatient non-cancer surgery who were prescribed postoperative opioids.²⁹ Overdose occurred in 0.1% (68 patients) within the first 30 days after surgery and in 0.8% (544 patients) within the first year. Taken together, it is possible that the VHA population is at greater risk of postoperative overdose than the patients represented in other datasets, or alternatively, that the nationwide VHA network does a better job of capturing overdose events.

Synthesis of the evidence: the incidence of opioid overdose after surgery

Table 3 lists the studies of postoperative overdose discussed above with pertinent characteristics highlighted. Incidence appears to be highly variable based on the studied population, with the highest rates seen in studies of veterans (0.7–0.8%) and, to a lesser degree, Medicaid recipients (0.05%). Of note, opioid amounts prescribed after surgery have been shown to be lower for VHA and Medicaid patients than for patients at academic centers, suggesting that these higher rates may not be due simply to increased opioid availability.³⁰ Overall, overdose incidence rates in community settings (non-VHA) of 0.01–0.03% were observed. If extrapolated across all post-procedural prescribing, this would yield between 4,470 and 13,410 overdoses per year. As noted below, overdoses in these data likely underreport the actual incidence given the absence of high-quality information on broader community overdose events related to prescribed postoperative opioids.

Table 3:

Studies of postoperative opioid overdose incidence.

First author	Year	Sample size	Postoperative timeframe	Opioid naïve	Continuous insurance coverage required	Evidence of postoperative opioid fill required	Overdose Incidence	Notes
Ladha	2018	1,305,715	30 days	No*	No	Yes	0.01%
Aalberg	2022	304,780	1 year	Yes	Yes	No	0.68%	VHA population
Siglin	2020	261,208	1 year	No	Yes	No	0.7%	VHA population; colonoscopy included
Mudumbai	2019	64,391	1 year	No*	Yes	Yes	0.8%	VHA population
Jivraj **	2020	162,830	1 year	Yes	Yes	No	0.01%	Ontario population
Wiese	2021	56,723	1 year	Yes	Yes	Yes	0.05%
Wylie	2022	35,335	Median 3.5 years	Yes	No	Yes	0.03%	Single center; included manual record review

^*Assessed separately

^**Canadian population

Methodological limitations in existing studies of postoperative overdose

Several issues are apparent in reviewing the above studies. Perhaps the most notable is the risk that claims data may not fully capture overdose events. For example, overdoses in the community not leading to hospital admission may be missed, including fatal overdoses at home or instances in which emergency medical personnel administer naloxone in the field but patients refuse transport to the emergency department. This may occur in up to 35% of community overdose cases.³¹ Additionally, the accuracy of overdose coding is unclear (e.g., overdose following surgery might be coded as respiratory failure). However, there appears to be greater concordance between administrative codes for opioid overdose and actual overdose than seen for OUD (76% agreement in one study of emergency department patients).³² This may be due to the generally less complex nature of overdose diagnosis compared with OUD diagnosis. Three further issues discussed above for OUD are also relevant to the overdose literature. Two of the studies (Ladha and Wylie) did not require continuous insurance coverage across the studied timeframe,^{15; 28} Siglin did not require that patients be opioid-naïve for study inclusion (potentially overestimating the overdose rate),²⁰ and evidence of a postoperative opioid fill was not an inclusion criterion in Aalberg, Siglin, and Jivraj (potentially underestimating the postoperative overdose rate).^19–21

RISK FACTORS FOR POSTOPERATIVE OUD AND OPIOID OVERDOSE

Several of the studies of perioperative OUD and overdose incidence referenced above also evaluated risk factors for these outcomes (Figure 1). These are reviewed briefly below and are presented in greater detail in Supplemental Digital Content 2.

Figure 1:

Risk factors for postoperative opioid use disorder and overdose. OUD = opioid use disorder, SUD = substance use disorder, PTSD = post-traumatic stress disorder, LOS = length of stay

*Risk factor for overdose only

Several risk factors related to substance use were identified. Although most studies of new perioperative OUD and overdose included only opioid-naïve patients, Siglin observed progressively increasing odds for both new OUD and overdose events as the extent of preoperative opioid use increased.²⁰ Ladha similarly observed that overdose frequency increased in correspondence with greater amounts of preoperative daily opioid use.²⁸ Of note, both studies relied on prescription database fill data to identify preoperative opioid use, which misses diverted/illicit use and may also fail to identify many prescription fills.³³ Similarly, one study identified preoperative opioid use disorder (OUD) as a risk factor for overdose specifically, and history of prior overdose as a risk factor for both new postoperative OUD and subsequent overdose.²⁰ Recent literature suggests that preoperative risky substance use is common and screening for it is advisable.³⁴ Four studies observed associations between preexisting non-opioid substance use disorders and both new postoperative OUD and overdose.^{14; 15; 17; 20} One study noted increased risk among patients taking preoperative benzodiazepines (OUD and overdose), gabapentinoids (overdose), and antidepressants (OUD).²⁰ Finally, elevated rates of OUD^{15; 17; 19} and overdose^{15; 19} were seen in patients with postoperative new persistent opioid use (NPOU).

Demographic attributes associated with OUD or overdose risk in the summarized studies include male sex²⁰ (OUD only), younger age,^{14; 15; 20} white race,^{14; 20} and Medicaid insurance (a surrogate marker for socioeconomic status).^{14; 17} Several psychiatric conditions also had significant associations with OUD and overdose risk: depression,¹⁴ PTSD,²⁰ and (for OUD only) bipolar disorder.²⁰ Among medical/surgical factors, both hepatitis C²⁰ and liver disease¹⁴ showed significant associations with OUD alone and OUD or overdose, respectively, likely secondary to a history of intravenous substance use. One study identified vitamin D deficiency as a risk factor for postoperative OUD,¹⁶ while another observed associations for OUD and overdose with COPD and peptic ulcer disease diagnoses.¹⁴ Lastly, one study observed higher OUD or overdose risk in patients undergoing thoracic and maxillofacial surgery¹⁵ while another found that patients with longer postoperative length of stay were at greater risk for both these outcomes.¹⁴

The primary limitation of these opioid risk data is that all reflect associations rather than causation due to the retrospective nature of the relevant studies. Many, but not all, of these risk factors are seen in the non-surgical addiction literature, which is reviewed briefly in Supplemental Digital Content 3.

CLINICAL IMPACT OF POSTOPERATIVE OUD

A new diagnosis of OUD after surgery represents a significant complication arising during postoperative recovery, with calls in the surgical community to consider this an event that should never happen.³⁵ Other surgery-related sentinel events such as retention of a foreign object or wrong-sided surgery occur at rates ranging from 0.01% to 0.2%, which fall below or appear comparable to conservative estimates of the incidence of OUD after surgery.^{36; 37} Consequently, this is an issue that warrants continued attention and further study even though complete elimination of postoperative OUD is unlikely. Well-recognized barriers to care for patients with OUD include stigma toward both persons who have a diagnosis of OUD and toward medications to treat OUD, inadequate training of clinicians to recognize and treat the condition, and fragmented delivery of care for those diagnosed with OUD.³⁸ Consequently, gaps in the receipt of medications for OUD persist, with more than 8 in 10 persons with OUD not receiving evidence-based MOUD treatment.³⁹ These challenges underscore the importance of effective strategies to identify patients with or at risk of OUD in perioperative settings and link them to appropriate care which, although more established in non-surgical acute care settings like emergency medicine, is gaining recognition in the context of surgery.⁴⁰

RESEARCH PATH FORWARD

Retrospective administrative claims data is particularly problematic for studies of OUD and overdose incidence following. For OUD, the most salient issues are capturing the complexity of the DSM OUD diagnostic criteria in available ICD codes, the unfamiliarity and/or discomfort many physicians may have with applying OUD criteria, and the incomplete overlap between the various ICD-9 and −10 administrative codes used as surrogates for OUD diagnosis. The chief research limitation for studies of postoperative opioid overdose is the likely failure to capture fatal overdoses and/or those occurring in the community without an emergency department or hospital charge, although published data are lacking for this contention.

Technological approaches like the application of natural language processing to large clinical datasets have been utilized as potential solutions to the problem of OUD diagnostic inaccuracy.⁴¹ While such approaches may prove beneficial and improve future large-scale retrospective analyses, it is unclear whether further claims data studies will provide the diagnostic sensitivity and specificity required to determine postoperative OUD incidence with greater certainty and accurately define the scope (and predictors) of the clinical problem to be addressed. Consequently, there is a need for alternative approaches to this issue. One option is to conduct more granular retrospective studies incorporating individual chart review in concert with review of claims data (like Wylie).¹⁵ Populations with single-payer health systems and robust clinical informatics databases may be ideal settings for such studies. Given that opioid prescribing and consumption patterns vary widely worldwide, this would also provide an opportunity to assess the generalizability of the existing retrospective literature that is chiefly based on U.S. data.⁴²

The ideal approach, however, would involve prospective cohort studies of surgical patients followed longitudinally postoperatively and evaluated periodically for opioid misuse and incident OUD using validated measurement approaches. One validated measure of interest would be the Current Opioid Misuse Measure-9 (COMM-9), which assesses aberrant prescription-opioid-related behavior (e.g., taking more than prescribed, from multiple sources) and may capture behavior indicating opioid misuse.⁴³ A more rigorous standard for determining OUD diagnosis that balances the need for in-depth interviews with the efficiency required for large sample sizes would be the use of Module C of the Quick Structured Clinical Interview for DSM-5 Disorders (QuickSCID-5), which provides accurate, criterion-based diagnosis of new-onset OUD.⁴⁴ Although more labor intensive than other approaches, structured interviews can be conducted remotely (phone, telehealth) making this approach potentially feasible for use in prospective studies with relatively large samples.

Prospective designs are the only approach that will be able to accurately detect OUD and overdose through uniform follow-up and determine the extent to which findings from retrospective studies may reflect over- or underdiagnosis of OUD incidence rates. Additionally, prospective studies incorporating careful baseline phenotyping may have utility for the creation of more accurate risk assessment algorithms, potentially permitting precision care. Future prospective studies would likely need to be large given the relatively low incidence of postoperative OUD and overdose. For example, a sample size of at least 2,215 opioid-naïve surgical patients prescribed postoperative opioids would be needed to detect a true OUD incidence between 0.4% and 1.2% with 95% confidence, an incidence in the range suggested by retrospective OUD data reviewed above. Collaboration with investigators trained in the assessment of opioid-related adverse outcomes including OUD will be necessary. Finally, given the considerable differences between VHA and non-VHA patient populations and the discrepant OUD and overdose incidence rates seen between them in the existing literature, an ideal prospective study would include diverse populations, including both civilians and veterans.

CONCLUSION

Despite increasing attention on perioperative opioid prescribing in the surgical literature and a reported 34% prevalence of opioid misuse among surgical patients,⁴⁵ robust data on new postoperative OUD and overdose are lacking. Extant studies are retrospective, relying almost exclusively on the use of codes that often reference outdated diagnoses (e.g., opioid “abuse”) or bear little relation to actual OUD (e.g., uncomplicated opioid use). These studies have identified an incidence for new postoperative OUD ranging between 0.1–0.8% and for post-surgical overdose between 0.01–0.8% within the first year after surgery. Rates of both outcomes appear considerably higher among VHA patients than in the general postoperative population. These rates extrapolate to as many as 358,000 new postprocedural OUD cases and 13,410 postprocedural opioid overdoses yearly in the U.S. Identified risk factors for postoperative OUD and overdose generally accord with those seen in non-surgical populations, with factors related to substance use, co-prescribed sedative-hypnotics, male sex, younger age, and psychiatric comorbidities most prominent. Given the vagaries of claims data and the clinical importance of opioid-related harms, there is a pressing need for more prospective work in this area, particularly studies using improved analytical methodology, structured clinical chart reviews, and even prospective interviews of at-risk patients to improve the accuracy of postoperative OUD and overdose diagnoses.