The Cost of Opioid Abuse: A Review of Recent Estimates

May 9, 2018

It is well known that the United States is staggering under the burden of an opioids epidemic. This has led to several efforts over the last 20 years to quantify the overall cost of opioid addiction and abuse. The various estimates agree that the cost is extremely high, but there are important differences—in some cases dramatic ones—between the estimates. The sources of these differences lie in methodological choices.

The Opioid Crisis

Opioids are a class of drugs that bind to the body’s opiate receptors and tend to induce sleep and reduce pain. There are natural and semisynthetic opioids (such as morphine, heroin, and oxycodone, the active ingredient in OxyContin), which are derived from the opium poppy, and synthetic opioids (such as fentanyl). Most prescription opioids are indicated for pain relief, and they are widely prescribed for that purpose.

Over the past 20 years, the incidence of opioid misuse and addiction has increased rapidly. In 2016, 11.8 million people in the United States misused opioids—4.4 percent of the population of people over the age of 12. In the great majority of cases, these people were misusing prescription opioids.[1] Concomitant with the rise of opioid addiction and abuse, there has been an alarming increase in the number of overdose deaths, which went from 8,050 in 1999 to 42,249 in 2016 (see Figure 1).

Figure 1

CDC, National Center for Health Statistics.[2]

The upward trend in opioid overdose deaths has been remarkably persistent. In recent years, drug companies have introduced “abuse resistant” formulations of opioids intended to deter abuse, but the data indicates that this effort has not reversed the trend. In fact, the release of an abuse-deterrent version of OxyContin in 2010 was followed by a sharp increase in the number of deaths from overdoses of heroin and the synthetic opioid, fentanyl, as represented in the graph above.[3]

One response to this crisis has been a spate of lawsuits filed by states, counties, cities, and the Cherokee Nation against opioid drug manufacturers and others in the pharmaceutical supply chain. These suits allege that the opioid manufacturers contributed to the epidemic by deceptively marketing their highly addictive products in a way that minimized their risks, and the plaintiffs seek damages for the costs stemming from this behavior.

A specific allegation in some of these lawsuits is that the opioid manufacturers promoted the use of opioids for non-cancer, chronic pain (such as the common conditions of back pain and joint pain), a use not approved by the FDA. The hypothesis that the increase in opioid abuse can be traced to such “off-label” use gains support from epidemiological studies. An article by researchers at Johns Hopkins University, the Mayo Clinic, and Stanford University studied the medical management of non-cancer pain in the United States between 2000 and 2010. It found that the percentage of cases treated with opioids nearly doubled over the time period studied, while the percentage of cases treated with non-opioid therapies did not change.[4]

Cost Estimates

Over the past 15 years, as awareness of this epidemic has grown, various studies have generated estimates of the cost of opioid addiction and abuse in the United States. Five of these studies have attempted a comprehensive estimate of the social cost of opioid addiction and abuse: Birnbaum et al. 2006, Hansen et al. 2011, Birnbaum et al. 2011, Florence et al. 2016, and Council of Economic Advisors 2017.[5] Each of these studies relies on various types of data (e.g., data on health insurance claims, epidemiology, employment, and the criminal justice system) to estimate four components of the total cost: excess healthcare costs, workplace costs, criminal justice costs, and fatality costs. The estimates of annual costs range from $12 billion in 2001 to $504 billion in 2015, as shown in Figure 2.[6]

Figure 2

Some of the difference between these estimates is attributable to the increases in both the number of cases of opioid addiction/abuse and the number of deaths from opioid overdose. And some of the difference is attributable to methodological choices.

Fatality cost

The cost component most affected by methodology is the fatality cost. In the first four estimates, the fatality cost is a moderate percentage of the total (between 10% and 27%), but in the CEA estimate, the fatality cost is 86% of the total. In all five studies, the fatality cost was estimated by first determining the number of opioid-related deaths and second multiplying that number by a measure of the economic losses attributable to those deaths.[7] However, three methodological differences explain why the CEA estimated higher fatality costs than the others.

First, the CEA used a different methodology than the other four to determine the economic cost attributable to each death. In the first four studies, this cost was determined by estimating the person’s lost earnings due to early death (based on average earnings data and the decedent’s sex and age).[8] This is a common methodology for estimating losses from fatalities and is sometimes used in cost-of-illness models. Using this approach, Florence et al. (2016) estimated $1.3 million in lost earnings per opioid overdose fatality.[9] The CEA used a different methodology, based on the value of a statistical life (VSL). This approach is often used by the U.S. government to set a value on the loss of life when conducting policy analysis, cost-benefit analysis, and regulatory impact analyses.[10] Economists estimate VSL based on revealed preference studies and stated preference studies. Revealed preference studies evaluate how employees trade off personal safety for wages; for example, they may compare the wages of jobs that are similar but have different levels of fatality risk. Stated preference studies use surveys to collect data on how people trade off money for risk of death.[11] Estimates of VSL may have a wide confidence interval, but most have a central value around $9 million to $10 million. The U.S. Department of Transportation recommends using a VSL of $9.6 million; the EPA recommends a VSL of $10.1 million, and the Department of Health and Human Services recommends using a range from $4.4 million to $14.3 million, with a central value of $9.4 million.[12] Using the VSL approach, CEA estimated a cost of $10.5 million per opioid overdose fatality[13] (much higher than the $1.5 million estimated by Florence et al.).

Second, when computing the number of relevant overdose deaths, the CEA includes deaths from heroin overdose, whereas the four other studies limit their analysis to the number of prescription opioid deaths. The CEA’s choice is based on recent research. It is well known that prescription opioid abuse and addiction often leads to heroin use, and studies indicate that the rise of heroin overdose deaths has its roots in the prescription opioid crisis.[14] The inclusion of heroin deaths significantly increases the estimated fatality costs. For example, using the CDC data, Florence et al. determined that the number of prescription opioid overdose deaths in 2013 was 16,235. If they had included all opioid overdose deaths, the number would have been 25,052, an increase of 54%.

The third reason CEA’s estimate of fatality costs is higher than the others is that it corrects for undercounting in the CDC mortality data. Research shows that opioid overdoses are underreported on death certificates, and based on that research, CEA adds 24% to the CDC’s count of opioid overdose deaths.[15]

Costs of non-fatal opioid abuse

Setting aside fatality costs, the five estimates are comparable, for the most part, and the increase in costs over time reflects the growth in the number of cases of opioid addiction and abuse. Figure 3 replicates the values in Figure 2, but excludes fatality costs.

Figure 3

 

Excess healthcare costs

As represented in Figure 3, the five estimates of excess healthcare costs show a substantial increase in these costs over time, from $4 billion in 2001 to $37 billion in 2015. The one outlier is Hansen et al., which uses a methodology different than that used by the other four. But setting aside Hansen et al., the increase in costs over time is mostly a function of the growth in opioid abuse.

The studies by Birnbaum et al., Florence et al., and CEA use similar methodologies to estimate excess healthcare costs. They conduct a regression analysis on a large set of insurance claims data which includes identifiers for particular patients and provides data on each patient’s demographic characteristics, diagnoses, treatments, procedures, and drug prescriptions, including the dollar value of each insurance claim. The researchers identify those patients with a diagnosis of opioid abuse or dependence (using standard diagnosis codes), they control for demographic and therapeutic variables that might confound the analysis, and they compare the annual healthcare costs of the opioid-abusing patients with a control group of patients who do not have a diagnosis of opioid abuse or dependence. The cost differential between opioids abusers and non-opioid abusers represents the excess healthcare cost attributable to opioid abuse.

The excess cost per opioid-abusing patient is fairly consistent across the studies. Birnbaum et al. (2006) estimate the excess cost for patients with private health insurance at $13,363, Florence et al. estimate it at $16,257, and CEA sets it at $15,474 (for comparison, all amounts are converted to 2015 dollars) (this particular value is not reported in the 2011 articles by Birnbaum et al. and Hansen et al.).[16] Furthermore, there are other studies that conduct similar regression analyses to estimate the excess cost per opioid-abusing patient, and the results are fairly consistent, especially for the more recent studies. Figure 4 shows the results of 9 studies that used regression analysis to estimate the excess annual healthcare cost per opioid-abusing patient among the privately insured.[17] For comparison, all of the estimates are converted to 2015 dollars.[18]

Figure 4

Some of the lower estimates in this graph can be explained by differences in methodology. The estimate of Birnbaum et al. (2006) is lower than other estimates from the same time period in part because it excludes the costs of drug abuse treatment (because it estimated treatment costs using a different methodology). The estimates of Rice et al. (2014a and 2014b) are lower than previous estimates for two reasons (by the authors’ account): (1) they controlled for more underlying conditions than previous studies did and (2) abuse-related costs might have decreased.[19]

As mentioned above, Hansen et al. (2011) use a different methodology to compute excess healthcare costs for opioid abusers. They do not conduct a regression analysis using claims data; instead, they limit their analysis to the costs of a few pre-specified conditions known to be suffered by opioid abusers.[20] This method is unable to capture all excess healthcare costs, so, not surprisingly, it returns a much lower estimate than the more common method (see Figure 3).

Workplace and criminal justice costs

The five studies used different methodologies to estimate workplace and criminal justice costs. Workplace costs are essentially losses due to reduced productivity. In three of the studies, the researchers estimated the number of productive hours lost to (1) disability and absenteeism due to opioid abuse and dependence and (2) opioid-related incarceration, and they multiplied this by the estimated production value of each hour lost.[21] In two other studies, the researchers relied upon a prior study of workplace costs from drug abuse and adjusted that study’s results to fit the current conditions.[22] In either case, the total workplace cost is partly a function of the number of opioid abusers and the number of people incarcerated due to opioid abuse or addiction. For four out of the five studies, the estimates are comparable, but for Hansen et al., the estimate is much higher (see Figure 3). This is primarily due to a higher estimate of lost productivity attributable to incarceration.[23]

Criminal justice costs include the additional costs for police protection, legal adjudication, correctional facilities, and the losses of crime victims. The five studies all used apportionment methods to estimate these costs. In most cases, they obtained data on the total U.S. costs for policing, legal and adjudication costs, correctional facilities, and property losses to crime victims. Then they estimated the portion of these amounts attributable to opioid abuse using one of two metrics: (1) the percentage of arrests that are attributable to opioid abuse or (2) the percentage of incarcerations that are attributable to opioid abuse.[24] For all five studies, criminal justice costs represent a relatively small portion of the total estimated costs, and the differences are mostly a result of increases in the number of opioid abusers over time.

Application of Economic Analysis to the Opioid Crisis

Economic analysis—in conjunction with data analysis from related fields such as epidemiology and scientific network analysis—is invaluable for addressing the opioid crisis, whether through policymaking or litigation.

Economic analyses such as those reviewed above should have an important place in policymaking. These estimates show that the total social cost of opioid abuse is enormous. Thus, they justify government expenditures on policies and programs that effectively reduce opioid abuse, addiction, and overdose. To the extent that these analyses break down the expenditures by payer (including government entities, private-sector businesses, and individuals), they aid in the design and funding of targeted interventions.

In litigation, one of the main roles of economics is to prove and calculate damages. Here, although estimates of the total cost of the opioids epidemic would certainly be relevant, the damage analysis may call for more targeted cost estimates and a model that accounts for the interplay between various damage components as well as the effect of differences across time and geography in factors such as tighter regulation, rehab access, naloxone distribution, and marketing of the opioids themselves. These analyses can be decomposed to isolate the portion of the economic injury suffered by particular groups of plaintiffs. They can be further decomposed to isolate the portion of that injury causatively linked to a particular defendant or group of defendants.

In a number of opioids lawsuits, the plaintiffs allege that opioid manufacturers have promoted the use of prescription opioids for off-label use—specifically for use in chronic, non-cancer pain. They allege that the manufacturers have intentionally misled the medical community about the risk of addiction from opioids, and that they have effected this by various means: influencing publications in medical journals, paying physicians as “opinion leaders,” and offering continuing medical education (CME) courses.

These allegations implicitly raise questions about the factors that influence physician prescribing, analyses of the effect of drug promotion on physician decision-making, and econometric studies of the effect of opioid promotion on sales. The approaches to these different, but related, questions must be harmonized in the comprehensive damage analysis. For example, whether physician prescribing was affected by intentionally deceptive journal publications might be addressed through citation analysis. One brief analysis traces the way a non-peer-reviewed letter in the New England Journal of Medicine was cited 439 times to support the conclusion that addiction is rare in patients treated with opioids.[25] While this analysis presents a helpful description of the influence of one article, it does not take the next logical step by attempting to trace that phenomenon back to the actions of the opioid manufacturers or by observing the multiplicative effect of second, third and even fourth generation citations.

The relationship between prescription opioids and use of non-prescription drugs such as fentanyl and heroin, between prescriber conduct and distributor over-stocking, and the indirect effects on school children or other family members, pose similarly important questions of causation and damage.[26]  Resolving these questions in a single, multi-faceted model, requires the coordination of many disciplines.

[1] Substance Abuse and Mental Health Services Administration (SAMHSA), “SAMHSA Shares Latest Behavioral Health Data, Including Opioid Misuse,” SAMHSA News (October 12, 2017).

[2] CDC, National Center for Health Statistics, “Drug Overdose Deaths in the United States, 1999–2016,” NCHS Data Brief No. 294, by Holly Hedegaard, Margaret Warner, and Arialdi M. Miniño (December 2017).

[3] Abby Alpert, David Powell, and Rosalie Liccardo Pacula, “Supply-side drug policy in the presence of substitutes: Evidence from the introduction of abuse-deterrent opioids,” NBER Working Paper Series, Working Paper 23031 (January 2017).

[4] Matthew Daubresse, Hsien-Yen Chang, Yuping Yu, Shilpa Viswanathan, Nilay D. Shah, Randall S. Stafford, Stefan P. Kruszewski, and G. Caleb Alexander, “Ambulatory diagnosis and treatment of nonmalignant pain in the United States, 2000-2010,” Med Care (2013) 51(10).

[5] Howard G. Birnbaum, Alan G. White, Jennifer L. Reynolds, Paul E. Greenberg, Mingliang Zhang, Sue Vallow, Jeff R. Schein, and Nathaniel P. Katz, “Estimated costs of prescription opioid analgesic abuse in the United States in 2001: A societal perspective,” Clinical Journal of Pain (2006) 22(8): 667–676; R.N. Hansen, G. Oster, J. Edelsberg, et al. “Economic costs of nonmedical use of prescription opioids,” Clinical Journal of Pain (2011) 27: 194–202; Howard G. Birnbaum, Alan G. White, Matt Schiller, Tracy Waldman, Jody Cleveland, and Carl L. Roland, “Societal costs of prescription opioid abuse, dependence, and misuse in the United States,” Pain Medicine (2011) 12: 657–667; Curtis S. Florence, Chao Zhou, Feijun Luo, and Likang Xu, “The economic burden of prescription opioid overdose, abuse, and dependence in the United States, 2013,” Medical Care 54(10): 901–906; Council of Economic Advisers (CEA), The Underestimated Cost of the Opioid Crisis (November 2017).

[6] Note: Three of these five studies received funding or other support from opioid manufacturers. Birnbaum et al. 2006 was funded by Ortho-McNeil Janssen / Johnson & Johnson and included authors employed by those companies. Hansen et al. (2011) was funded by Purdue. Birnbaum et al. (2011) was funded by King Pharmaceuticals and included authors employed by King.

[7] Two different sources were used to determine the number of relevant opioid-related deaths. Birnbaum et al. used data from Drug Abuse Warning Network, and Hansen et al., Florence et al., and the CEA used the Center for Disease Control and Prevention (CDC) data on “Multiple Causes of Death.”

[8] Birnbaum et al. 2006, p. 672; Hansen et al. 2011, p. 196; Birnbaum et al. 2011, p. 661; Florence et al. 2016, p. 903.

[9] Florence et al. 2016, p. 904.

[10] CEA, 2017, p. 3.

[11] CEA, 2017, pp. 3–4.

[12] CEA, 2017, p. 4.

[13] CEA, 2017, p. 6.

[14] Abby Alpert, David Powell, and Rosalie Liccardo Pacula, “Supply-side drug policy in the presence of substitutes: Evidence from the introduction of abuse-deterrent opioids,” NBER Working Paper Series, Working Paper 23031 (January 2017).

[15] CEA, 2017, p. 6.

[16] Birnbaum et al. 2006, p. 671; Birnbaum et al. 2011; Florence et al. 2016, p. 904. The estimate used by CEA is very similar to that computed by Florence et al. because it is explicitly derived from the analysis of Florence et al. (CEA 2017, p. 7.

[17] A. White, H. Birnbaum, M. Mareva, et al., “Direct costs of opioid abuse in an insured population in the United States,” Journal of Managed Care Pharmacy (2005) 11(6): 469–479; Howard G. Birnbaum, Alan G. White, Jennifer L. Reynolds, Paul E. Greenberg, Mingliang Zhang, Sue Vallow, Jeff R. Schein, and Nathaniel P. Katz, “Estimated costs of prescription opioid analgesic abuse in the United States in 2001: A societal perspective,” Clinical Journal of Pain (2006) 22(8): 667–676; Alan G. White; Howard G. Birnbaum; Matt Schiller; Tracy Waldman; Jody M. Cleveland; and Carl L. Roland, “The economic impact of opioid abuse, dependence, and misuse,” American Journal of Pharmacy Benefits (2011) 3(4): e59–e70; J. Bradford Rice, Noam Y. Kirson, Amie Shei, Alice Kate G. Cummings, Katharine Bodnar, Howard G. Birnbaum, Rami Ben-Joseph, “Estimating the costs of opioid abuse and dependence from an employer perspective: A retrospective analysis using administrative claims data,” Applied Health Economics and Health Policy (2014) 12(4): 435–46; J. Bradford Rice, Noam Y. Kirson, Amie Shei, Caroline J. Enloe, Alice Kate G. Cummings, Howard G. Birnbaum, Pamela Holly, and Rami Ben-Joseph, “The economic burden of diagnosed opioid abuse among commercially insured individuals,” Postgraduate Medicine (2014) 126(4): 53–58; Curtis S. Florence, Chao Zhou, Feijun Luo, and Likang Xu, “The economic burden of prescription opioid overdose, abuse, and dependence in the United States, 2013,” Medical Care 54(10): 901–906; FAIR Health, “The impact of the opioid crisis on the healthcare system: A study of privately billed services,” A FAIR Health White Paper (September 2016); Noam Y. Kirson, Lauren M. Scarpati, Caroline J. Enloe, Aliya P. Dincer, Howard G. Birnbaum, and Tracy J. Mayne, “The economic burden of opioid abuse: Updated findings,” Journal of Managed Care & Specialty Pharmacy (2017) 23(4): 427–445; Council of Economic Advisers (CEA), The Underestimated Cost of the Opioid Crisis (November 2017).

[18] Six of these nine studies received funding or other support from opioid manufacturers. White et al. (2005) was funded by Janssen Medical Affairs and included authors employed by Janssen. Birnbaum et al. 2006 was funded by Ortho-McNeil Janssen / Johnson & Johnson and included authors employed by those companies. White et al. (2011) included authors employed by King Pharmaceuticals and authors who had received grants from King and Pfizer. Rice et al. (2014a and 2014b) were funded by Purdue and included authors employed by Purdue. Kirson et al. (2017) was funded by Purdue and included authors employed by Purdue.

[19] Rice et al. 2014a, p. 436.

[20] Hansen et al. 2011, pp. 195–96.

[21] See Florence et al. (2016), p. 903, and Birnbaum et al. (2011), p. 661. CEA (2017) relied on the analysis of Florence et al. for this estimate.

[22] See Hansen et al. (2011), p. 196; Birnbaum et al. (2006), p. 672–673.

[23] Compare Hansen et al. (2011), p. 199, to Birnbaum et al. (2011), p. 661. The exact reason for the divergence in estimates is not evident in the text.

[24] Hansen et al. (2011), p. 197–98; Birnbaum et al. (2011), p. 661; Florence et al. (2016), p. 903. CEA (2017) relied on the analysis of Florence et al. for this estimate. Birnbaum et al. (2006) collected data on drug-related arrests and incarcerations and then estimated the portion related specifically to opioids (p. 672).

[25] Pamela T.M. Leung, Erin M. Macdonald, Matthew B. Stanbrook, Irfan A. Dhalla, and David N. Juurlink, “A 1980 Letter on the Risk of Opioid Addiction,” New England Journal of Medicine (2017) 376(22).

[26] This question has been addressed recently by economists. Abby Alpert, David Powell, and Rosalie Liccardo Pacula, “Supply-side drug policy in the presence of substitutes: Evidence from the introduction of abuse-deterrent opioids,” NBER Working Paper Series, Working Paper 23031 (January 2017).