Predictors of Postdeployment Prescription Opioid Receipt and Long-term Prescription Opioid Utilization Among Army Active Duty Soldiers

Abstract

Introduction

Little is known about long-term prescription opioid utilization in the Military Health System. The objectives of this study were to examine predictors of any prescription opioid receipt, and predictors of long-term opioid utilization among active duty soldiers in the year following deployment.

Materials and Methods

The analytic sample consisted of Army active duty soldiers returning from deployment to Operation Enduring Freedom, Operation Iraqi Freedom, or Operation New Dawn in fiscal years 2008–2014 (N = 540,738). The Heckman probit procedure was used to jointly examine predictors of any opioid prescription receipt and long-term opioid utilization (i.e., an episode of 90 days or longer where days-supply covered at least two-thirds of days) in the postdeployment year. Predictors were based on diagnoses and characteristics of opioid prescriptions.

Results

More than one-third of soldiers (34.8%, n = 188,211) had opioid receipt, and among those soldiers, 3.3% had long-term opioid utilization (or 1.1% of the cohort, n = 6,188). The largest magnitude predictors of long-term opioid utilization were receiving a long-acting opioid within the first 30 days of the episode, diagnoses of chronic pain (no specified source), back/neck pain, or peripheral/central nervous system pain, and severe pain score in vital records.

Conclusions

Soldiers returning from deployment were more likely to receive an opioid prescription than the overall active duty population, and 1.1% initiated a long-term opioid episode. We report a declining rate of opioid receipt and long-term opioid utilization among Army members from fiscal years 2008–2014. This study demonstrates that the most important predictors of opioid receipt were not demographic factors, but generally clinical indicators of acute pain or physical trauma.

INTRODUCTION

The risks of long-term prescription opioid utilization include an increased drug tolerance, risk of opioid use disorder, overdose, and suicide behaviors.^1–9 The drug overdose death rate has been increasing in the USA to 16.3 per 100,000 of the population in 2015, with 63% of all drug overdose-related deaths involving an opioid.¹⁰ The Department of Defense (DoD) has mounted initiatives to find and use non-pharmacologic approaches to treat the increasing number of military members with acute and chronic pain stemming from non-combat and deployment-related injuries, in efforts to improve retention and to reduce morbidity and mortality associated with prescription opioid use.^11–17

Predictors and risks associated with prescription opioid receipt among Veterans who have utilized care at the Veterans Health Administration (VHA) are well studied,^18–21 however, much less is known about prescription opioid receipt among military members treated within the Military Health System (MHS) and those returning from deployments associated with the Afghanistan and Iraq operations (i.e., Operation Enduring Freedom/Operation Iraqi Freedom/Operation New Dawn [OEF/OIF/OND]). A study of active duty members in fiscal year (FY) 2010 found that 26.4% received at least one opioid prescription, including 0.7% who utilized opioids long-term.²² Predictors of long-term opioid utilization included having a combat injury and receiving a substance use diagnosis, however, this study did not examine a postdeployment cohort that may have unique deployment-related injuries.

The most widely cited study that examined prescription opioid receipt after combat deployment was based on survey data collected from Army soldiers in one infantry brigade three months after returning from an Afghanistan deployment in 2011.²³ Over 15% of the brigade self-reported receiving an opioid prescription in the past month, with most (63%) receiving only a few or several days of opioids. This study did not examine long-term opioid utilization.

The objectives of this study were to examine (1) predictors of any prescription opioid receipt and (2) predictors of long-term prescription opioid utilization among soldiers in the year after returning from an OEF/OIF/OND deployment. Research that informs our understanding of risk factors for long-term prescription opioid utilization among soldiers will provide the DoD with targeted information to incorporate into its efforts to implement interdisciplinary pain management approaches aimed at reducing reliance on opioids and improving the wellbeing and readiness of its soldiers.^24–26

METHODS

Analytic Sample and Data Sources

We used data from the Substance Use and Psychological Injury Combat (SUPIC) study, a longitudinal, observational study of postdeployment health including pain management treatment and outcomes.²⁷ The SUPIC cohort includes Army soldiers who returned from an index deployment associated with OEF/OIF/OND ending in FYs 2008–2014 (N = 865,640). From the SUPIC active duty sample (n = 573,453), step-wise during the first 6-months postdeployment, we excluded soldiers who died (n = 250), soldiers with a cancer diagnosis (n = 16,880), and those with a pregnancy (n = 7,655). We restricted our analysis to soldiers who were enrolled in the MHS during the first quarter postdeployment and who had at least 1 month of enrollment in the second quarter; 7,785 soldiers who did not meet these criteria were excluded, as well as 145 soldiers with missing data. The analytic sample consisted of 540,738 active duty soldiers.

Data were drawn from the DoD’s Military Data Repository and the Contingency Tracking System (deployment data), and demographic and military characteristics were from the Defense Enrollment Eligibility Reporting System. MHS health care utilization data included outpatient and inpatient care provided at military facilities and care “purchased” by the DoD from civilian providers. Opioid prescription data were from the Pharmacy Data Transaction Service (PDTS), and self-reported pain severity ratings were based on the 0–10 Numerical Rating Scale from the vital records of the Clinical Data Repository.

Measures

Dependent Variables

Opioid receipt was defined as any opioid prescription filled during the postdeployment year. Long-term opioid utilization was defined as an episode of 90 days or longer in which the total opioid days was at least two-thirds of the episode duration. Thus, long-term opioid receipt required a minimum of 60 days-supply. The episode spanned all days covered by the first and subsequent opioid fills until there was a 60-day gap with no fill after the calculated end of the last fill. We checked the quantity dispensed and days-supply for outliers. For scripts with a quantity of one with a days-supply ≥7 (less than 0.5%), the days-supply was set to one. All long-term opioid episodes started in the postdeployment year, but were measured until their end, even beyond the postdeployment year. Opioid prescriptions included codeine, dihydrocodeine, fentanyl, hydrocodone, hydromorphone, meperidine, methadone (excluding the liquid form), morphine, oxycodone, oxymorphone, and tapentadol.²⁸ To ensure that we captured all National Drug Codes for opioids over the analysis period, we identified all PDTS claims with an AHFS Pharmacologic-Therapeutic Classification of 28.08.08 (opiates) or 28.08.12 (opiate partial agonist). We excluded injectables. The product name value was used to classify the opiate subclasses (e.g., Hydrocodone, Hydromorphone, Oxycodone), and to exclude medications for the treatment of opioid dependence (e.g., buprenorphine). The code list of analysis opioids is available from the authors upon request.

Covariates

Ten pain category covariates developed by the SUPIC project²⁹ included peripheral/central nervous system (CNS); osteoarthritis; back/neck; headache/migraine; non-traumatic joint disorders; other musculoskeletal; visceral/pelvic; wounds/injury/fractures; acute postoperative/trauma pain; and chronic pain (no specified source). Pain severity ratings were based on the highest score on the 0–10 scale administered by clinicians during outpatient medical encounters in military facilities, with cut-points for missing, none/mild (0–3), moderate (4–6), and severe pain (7–10).³⁰ Other health care comorbidities included any inpatient stay (yes/no), and indicator variables for behavioral health comorbidity (adjustment, anxiety, posttraumatic stress disorder [PTSD], depression, alcohol use disorder, and substance use disorder), and traumatic brain injury comorbidity. A count of seven of the most common Major Diagnostic Categories was collapsed to 1, 2, or 3+.

Because we used two-stage modeling, described below, definitions for the covariates differed by stage. In the first stage, pain and health care comorbidities were assessed within −30 to 183 days after the end of the index deployment and diagnoses required at least two outpatient diagnoses at least 30 days apart or one inpatient claim. In the second stage, pain and health care comorbidities were assessed within ±30 days of the first opioid prescription fill date and required at least one outpatient diagnosis or inpatient claim.

Characteristics of Initial Opioid Receipt

Initial days-supply was the sum of days-supply of opioids on the first day of the episode. Initial daily morphine milligram equivalents (MME) was the sum of total MME for all fills divided by the sum of total days-supply on the first day of the episode. Long-acting opioid was defined based on the product form of the National Drug Code for all fills within the first 30 days of the episode.

Demographic covariates were measured during the first month after the end of the index deployment and consisted of gender, race/ethnicity, marital status, and age groupings. Military characteristics included rank, number of previous deployments, and the FY of the deployment end.

Statistical Analysis

We used Stata’s Heckman probit procedure to model prescription opioid receipt in the first stage, and long-term opioid utilization in the second stage with the sample restricted to soldiers that received a prescription opioid. We report marginal effects at the mean (MEM), standard errors, and two-sided p-values at the p ≤ 0.001 level to account for the large sample size. The MEM coefficient is the predicted change in the likelihood of the outcome between a soldier having the characteristic and not having the characteristic, setting all other variable values to their means. The MEM is a common way to estimate the average effect of a characteristic across all sample subjects.

In the first regression model, we included pain and health covariates, as well as demographic and military characteristics. In the second regression model, restricted to soldiers with opioid receipt, we included the same measures and added three measures characterizing initial opioid prescriptions in the episode.

Brandeis University’s Committee for Protection of Human Subjects and the Human Research Protection Program at the Office of the Assistant Secretary of Defense for Health Affairs/Defense Health Agency conducted the human subjects review. The Defense Health Agency’s Privacy and Civil Liberties Office executed the data use agreements.

RESULTS

Most active duty soldiers returning from OEF/OIF/OND deployments in FYs 2008–2014 were male (91.5%), White, non-Hispanic (57.3%), married (58.6%), and age 18–24 (40.5%; Table I). Approximately half were junior enlisted (E1–E4), and most were returning from their first deployment (67.2%).

Table I.

Sociodemographic and Military Characteristics of Active Duty Soldiers Returning from OEF/OIF/OND Deployments in FYs 2008–2014, by Subgroups with Prescription Opioid Receipt and Long-term Prescription Opioid Utilization

	Active Duty Soldiers	Soldiers with Prescription Opioid Receipta		Soldiers with Long-term Prescription Opioid Utilizationb
	N = 540,738	n = 188,211 (34.8%)		n = 6,188 (1.1%)
	n	n	% of AD Subgroupc	n	% of AD Subgroupc
Gender
Female	46,231	19,810	42.9	423	0.9
Male	494,507	168,401	34.1	5,765	1.2
Race/Ethnicity
Asian or Pacific Islander	60,670	21,852	36.0	1,158	1.9
Black, non-Hispanic	96,511	33,299	34.5	524	0.5
Hispanic	61,824	21,383	34.6	455	0.7
White, non-Hispanic	310,104	107,756	34.8	3,939	1.3
Other	11,629	3,921	33.7	112	1.0
Age-groups
18–24	218,964	80,947	37.0	2,223	1.0
25–29	144,760	47,997	33.2	1,790	1.2
30–34	76,066	24,993	32.9	943	1.2
35–39	56,144	18,789	33.5	654	1.2
40+	44,804	15,485	34.6	578	1.3
Marital status
Married	317,023	113,189	35.7	4,248	1.3
Never married	194,902	64,596	33.1	1,616	0.8
Separated	28,813	10,426	36.2	324	1.1
Member rank
Junior enlisted (E1–E3)	73,866	30,601	41.4	1,201	1.6
Junior enlisted (E4)	199,401	75,091	37.7	2,578	1.3
Senior enlisted (E5)	88,591	29,899	33.8	1,013	1.1
Senior enlisted (E6–E9)	98,599	32,250	32.7	1,033	1.1
Warrant officer	12,812	3,695	28.8	82	0.6
Junior officer	46,535	11,393	24.5	163	0.4
Senior officer	20,934	5,282	25.2	118	0.6
FY of index deployment end date
2008	81,330	28,027	34.5	1,406	1.7
2009	127,552	44,143	34.6	1,359	1.1
2010	123,296	44,430	36.0	1,521	1.2
2011	73,899	26,308	35.6	963	1.3
2012	62,772	22,282	35.5	590	0.9
2013	40,837	13,288	32.5	223	0.6
2014	31,052	9,733	31.3	126	0.4
Number of previous deployments to index
0	363,588	131,739	36.2	4,352	1.2
1	140,672	45,253	32.2	1,542	1.1
2+	36,478	11,219	30.8	294	0.8

^aPrescription opioid receipt is an opioid prescription fill within −30 days to 364 days after return from deployment.

^bLong-term prescription opioid utilization is at least 90 days of opioid claims in which the total days-supply was at least two-thirds of the opioid episode.

^cRow percentages.

AD, active duty.

Note. Chi-square statistics showed that all relationships between characteristics of the AD sample and prescription opioid receipt were significant at p ≤ 0.001 (e.g., the distribution of marital status was significantly different between soldiers with an opioid prescription and soldiers without). Chi-square statistics also showed that all relationships between characteristics of the AD sample and long-term prescription opioid utilization were significant at p ≤ 0.001.

Among all active duty soldiers, 34.8% (n = 188,211) received an opioid prescription in the postdeployment year (Table I). Across all ages and ranks, the median days-supply for the initial fill was between 3 and 4 days (data not shown). Women were more likely than males to receive an opioid (42.9% versus 34.1%, respectively), and soldiers who were junior enlisted (E1–E3) were more likely to receive an opioid than enlisted soldiers from a higher rank or officers. Among all active duty soldiers, 1.1% (n = 6,188) had a long-term opioid episode in the postdeployment year, representing 3.3% of soldiers who received an opioid prescription. The mean duration for days-supply during the episode was 213.5 days among those with a long-term opioid episode (maximum duration post-coded to 365 days), and was 19.6 days among soldiers with any opioid receipt (data not shown). Males were more likely to have a long-term opioid episode than females (1.2% versus 0.9%, respectively). The likelihood of having a long-term opioid episode decreased as rank group increased. There was a time trend with the prevalence of long-term opioid utilization decreasing from 1.7% of soldiers in FY 2008 to 0.4% among those returning in FY 2014 (Table I). Among soldiers with opioid receipt, the prevalence of long-term opioid utilization fell from 5.0% in FY 2008 to 1.3% in FY 2014 (Fig. 1). Almost 2% of soldiers who received an opioid prescription had a long-acting opioid in the first 30 days of an opioid fill, compared with 17.1% of long-term opioid users (data not shown).

FIGURE 1.

Prevalence of postdeployment prescription opioid receipt and long-term opioid utilization among opioid users, by FY of index deployment end date.

The receipt of an opioid prescription was the highest among soldiers with acute postoperative/trauma pain (93.7%), an inpatient stay (75.1%), and chronic pain no specified source (72.5%; Table II). The likelihood of receiving an opioid increased as pain severity ratings increased, and with each increase in the number of comorbid MDC diagnoses. Appendix 1 shows bivariate associations of health care comorbidities and complexity, assessed within ±30 days of starting on an opioid prescription, with the likelihood of having a long-term opioid episode.

Table II.

Health Care Comorbidities and Complexity During the first 6-months Postdeployment Among the Active Duty Sample and Subgroup with Prescription Opioid Receipt in the Postdeployment Year

Characteristics	Active Duty Soldiersa		Soldiers with Prescription Opioid Receiptb
	N = 540,738		N = 188,211 (34.8%)c
	n	% of AD Sample	n	% of AD Subgroup
Inpatient stay, yes	11,503	2.1	8,641	75.1
Highest pain severity score
Missing	195,039	36.1	51,383	26.3
None/low (0–3)	177,365	32.8	53,849	30.4
Moderate (4–6)	92,372	17.1	40,734	44.1
Severe (7–10)	75,962	14.1	42,245	55.6
Type of pain
Peripheral/CNS, yes	4,593	0.9	2,833	61.7
Osteoarthritis, yes	4,534	0.8	2,690	59.3
Back/neck, yes	65,995	12.2	33,372	50.6
Headache/migraine, yes	20,937	3.9	11,089	53.0
Non-traumatic joint disorders, yes	89,841	16.6	45,243	50.4
Other musculoskeletal, yes	96,046	17.8	49,190	51.2
Visceral/pelvic, yes	16,558	3.1	10,367	62.6
Wounds/injuries/fractures, yes	31,667	5.9	20,109	63.5
Acute postoperative/trauma, yes	3,015	0.6	2,824	93.7
Chronic pain no specified source, Yes	3,840	0.7	2,784	72.5
Number of non-pain major diagnostic category comorbiditiesd
1	171,187	31.7	67,244	39.3
2	67,716	12.5	33,146	48.9
3+	29,017	5.4	17,330	59.7
Behavioral health diagnosis,e yes	42,075	7.8	20,833	49.5
Traumatic brain injury diagnosis, yes	10,785	2.0	6,257	58.0

^aClinical characteristics were assessed within −30 to 183 days after the end of the index deployment. Diagnoses required at least two outpatient diagnoses at least 30 days apart or one inpatient claim.

^bClinical characteristics were assessed within ±30 days from opioid start date.

^cRow percentages.

^dA count of the following MDCs: eye; ear, nose, throat; respiratory, circulatory, skin and breast, kidney, infectious and parasitic, max value set to three. MDCs were based on the presence of at least one claim.

^eBehavioral health includes: adjustment, anxiety, PTSD, depression, alcohol use disorder, and substance use disorder.

Note. Chi-square statistics showed that all relationships between characteristics of the AD sample and soldiers with prescription opioid receipt were significant at p ≤ 0.001.

Multivariate Predictors of Any Opioid Receipt

The largest predictors of opioid receipt were three pain diagnoses: acute postoperative/trauma pain (50% more likely), chronic pain no specified source (21% more likely), or wound/injuries/fracture pain (18% more likely; Table III). Soldiers who had an inpatient stay were 23% more likely than others to have an opioid fill. Soldiers with severe pain scores were 11% more likely than soldiers with only a mild pain score, and soldiers returning in FY 2014 were 6% less likely that soldiers who returned in FY 2008 to fill an opioid prescription. Other factors had smaller influences even though statistically significant.

Table III.

Predictors of Prescription Opioid Receipt Among Active Duty Soldiers Returning from an OEF/OIF/OND Deployment in FYs 2008–2014 (n = 540,738)

Characteristicsa	Marginal Effect at the Meanb	Standard Error
FY of Index Deployment End Date (ref = 2008)
2009	−0.028	0.002
2010	−0.021	0.002
2011	−0.033	0.003
2012	−0.033	0.003
2013	−0.056	0.003
2014	−0.058	0.003
Gender, female	0.048	0.003
Member rank (ref E1–E3)
Junior enlisted (E4)	−0.032	0.002
Junior enlisted (E5)	−0.059	0.003
Junior enlisted (E6–E9)	−0.076	0.003
Warrant officer	−0.109	0.005
Junior officer	−0.123	0.003
Senior officer	−0.146	0.004
Number of previous deployments to index (ref = 0)
1	−0.019	0.002
2+	−0.019	0.003
Inpatient stay, yes	0.233	0.005
Highest NRS pain severity score (ref = mild/1–3)
Missing	−0.013	0.002
Moderate (4–6)	0.040	0.002
Severe (7–10)	0.114	0.002
Peripheral/CNS pain, yes	0.098	0.008
Osteoarthritis pain, yes	0.132	0.008
Back and neck pain, yes	0.095	0.002
Headache/migraine pain, yes	0.048	0.004
Non-traumatic joint disorders pain, yes	0.088	0.002
Other musculoskeletal pain, yes	0.101	0.002
Visceral/pelvic pain, yes	0.132	0.004
Wounds/injuries/factures pain, yes	0.177	0.003
Acute postoperative/trauma pain, yes	0.500	0.009
Chronic pain no specified source, yes	0.206	0.009
Behavioral health diagnosis,c yes	−0.002	0.003
Traumatic brian injury diagnosis, yes	0.025	0.005

Note. This model also included measures for race/ethnicity, marital status, age-groups, and a count of major diagnostic categories. All characteristics were significant at the p ≤ 0.001 except behavioral health diagnosis.

^aClinical characteristics were assessed within −30 to 183 days around the end of the index deployment. Diagnoses required at least two outpatient diagnoses or one inpatient claim at least 30 days apart. Outpatient diagnoses for the pain groups were required to be primary and face-to-face.

^bMarginal effect at the mean (MEM) – the difference in probability estimated by the model between having the characteristic and not having the characteristic with all other independent variables set to their mean values.

^cBehavioral health includes: adjustment, anxiety, PTSD, depression, alcohol use disorder, and substance use disorder.

Multivariate Predictors of Long-term Opioid Utilization

The largest predictor of long-term opioid utilization was receiving a long-acting opioid within the first 30 days of opioid start (Table IV); these soldiers were 14.2% more likely to have a long-term opioid episode. Soldiers with a primary pain diagnosis of chronic pain no specified source, or back/neck pain were 8.8% and 7.9% more likely to have a long-term opioid episode, respectively. Soldiers who self-reported severe pain were 4% more likely to be long-term opioid users compared to those with only mild pain. Soldiers returning in FY 2014 were 4% less likely than those returning in FY 2008 to initiate long-term opioid use.

Table IV.

Predictors of Long-term Prescription Opioid Utilization Among Active Duty Soldiers Who Returned from and OEF/OIF/OND Deployment in FYs 2008–2014 (n = 188,211)

Characteristicsa	Marginal Effect at the Meanb	Standard Error
FY of index deployment end date (ref = 2008)
2009	−0.027***	0.003
2010	−0.026***	0.003
2011	−0.022***	0.003
2012	−0.035***	0.003
2013	−0.040***	0.004
2014	−0.042***	0.004
Gender, female	−0.026***	0.002
Member rank (ref = E1–E3)
Junior enlisted (E4)	−0.005	0.002
Junior enlisted (E5)	−0.006	0.003
Junior enlisted (E6–E9)	−0.012***	0.003
Warrant officer	−0.021***	0.005
Junior officer	−0.022***	0.004
Senior officer	−0.018***	0.005
Number of previous deployments to index (ref = 0)
1	−0.001	0.002
2+	−0.010***	0.003
Inpatient stay, yes	0.001	0.003
Highest NRS pain severity score (ref = mild/1–3)
Missing (0–3)	0.024***	0.002
Moderate (4–6)	0.021***	0.002
Severe (7–10)	0.041***	0.003
Peripheral/CNS pain, yes	0.035***	0.005
Osteoarthritis pain, yes	0.003	0.004
Back/neck pain, yes	0.079***	0.003
Headache/migraine pain, yes	−0.004	0.002
Non-traumatic joint disorders pain, yes	0.011***	0.002
Other musculoskeletal pain, yes	−0.000	0.002
Visceral/pelvic pain, yes	0.004	0.003
Wounds/injuries/factures pain, yes	−0.005	0.002
Acute postoperative/trauma pain, yes	−0.004	0.003
Chronic pain no specified source, yes	0.088***	0.006
Behavioral health diagnosis,c yes	0.029***	0.002
Traumatic brian injury diagnosis, yes	0.017***	0.003
Long-acting opioid,d yes	0.142***	0.008
Initial days-supplye	0.003***	0.000
Initial daily MMEf	0.001***	0.000

Note. This model also included measures for race/ethnicity, marital status, and age-groups. *** p ≤ 0.001.

^aClinical characteristics were assessed within ±30 days from opioid start date. Pain group diagnoses required a primary diagnosis.

^bMarginal effect at the mean (MEM) – the difference in probability estimated by the model between having the characteristic and not having the characteristic with all other independent variables set to their mean values.

^cBehavioral health includes: adjustment, anxiety, PTSD, depression, alcohol use disorder, and substance use disorder.

^dReceived a long-acting opioid within 30 days of the opioid start date.

^eSumming the days-supply of opioids on the first day of an opioid prescription.

^fAverage daily MME on the start date of the opioid episode.

The opioid days-supply on the first day was associated with a slight increase in the likelihood of long-term opioid utilization For example, 14-days-supply compared with 7-days-supply, increased the likelihood of long-term opioid utilization by 2.1%. Additionally, receipt of 30 MME on the first day of an opioid episode, compared with 20 MME, would have a 1.0% increase in the likelihood of long-term opioid episode.

DISCUSSION

Unlike most prior studies of prescription opioid receipt, this study took a population-based approach and had comprehensive clinical measures that came from health care records and pain severity scores from vital records. We examined a high-risk population, soldiers returning from deployments, who might experience injuries and stressors that could increase opioid requirement. In addition to its population focus, a strength of this study is that it spans 7-years of data, the same time period when opioid overdose deaths in the USA were increasing.¹⁰ During this period, the DoD established the Army Surgeon General’s Pain Management Task Force, which published its 2010 comprehensive pain management strategy with recommendations aimed at reducing reliance on opioids and use of long-term opioid therapy, and the Army embarked on several initiatives to improve pain management approaches including the development of an interdisciplinary pain management network within the MHS.^19,22

We found that 34.8% of soldiers received at least one opioid prescription within the postdeployment year. Our annual rate is higher than the self-reported postdeployment estimate of 23.2% of past month opioid prescription receipt among one brigade of soldiers,²³ perhaps reflecting the longer observation period of our study. Consistent with a high level of acute and chronic pain in our study population, the opioid exposure rate is higher than among the active duty population (including non-deployed members) and the general population. Specifically, a population-based study of active duty members not specific to a postdeployment cohort, reported that 26.4% of members from all Services received at least one opioid prescription during FY 2010.²² Further, the Centers for Disease Control and Prevention reported that approximately 20.7% of the U.S. population received at least one opioid prescription in 2014.³¹ The relatively high population “base rate” in our study implies other factors in addition to deployment are contributing to the use of opioids among service members.

One important contribution of our study is that we report a similar time trend in prescription opioid receipt among Army members compared to the U.S. population, in which opioid prescribing rates were increasing (during 2008–2010), remaining constant (2010–2012), and declining starting in 2012.³² The prevalence of receiving an opioid prescription in the postdeployment year decreased from 34.5% in FY 2008 to 31.3% in FY 2014, and the prevalence of long-term opioid prescription decreased from 1.7% in FY 2008 to 0.6% among those returning in FY 2014. We also report that the average days of opioid exposure was less than 20 for those with opioid receipt that was not long-term, consistent with acute care treatment. This paper did not examine if the declining rate in prescription opioid utilization observed during the study was associated with a substitution of non-pharmacological approaches to pain management within the MHS encouraged by new guidelines and military reports.^2,12

This study examines clinical factors not available in the previous military literature. We confirm the disparities reported in Toblin et al²³ and found that being female and from a junior enlisted pay grade each contributed to increased likelihood to receive an opioid prescription. Further, soldiers returning from first deployments were more likely than others to receive opioid treatment. These results are consistent with the healthy-warrior effect,^33,34 which suggests that less healthy military members leave the military after deployment, and that those who remain within the military after deployments are more fit for future war-time service. One important contribution of this study is our finding that these disparities remained when controlling for clinical characteristics. Future studies could explore how soldiers returning from first deployments differ from those who have had multiple deployments in terms of postdeployment health, treatment utilization, and reasons for early separation from the military. Being female predicted greater probability of receiving a prescription opioid than males, but females were less likely to use opioids long-term. These findings are similar to previous studies with military members.^22,23

Indeed, this study demonstrates that the most important predictors of opioid receipt were not demographic factors, but generally clinical indicators of acute pain or physical trauma: having an inpatient stay, moderate or severe pain score, treatment for acute postoperative/trauma pain, or wound/injuries/fracture pain. The indicator for chronic pain no specified source had a similarly large impact but other indicators associated with chronic pain had smaller influence on receipt of an opioid.

In this study, 3.3% of soldiers who initiated opioid use, or 1.1% of all study soldiers (n = 6,188), met the definition for long-term opioid utilization. This estimate is higher than the previously reported population estimate of 0.7% of active duty members among all Services, using a definition of opioid use for at least 90 days.²² Importantly, the soldiers who met our study definition of long-term opioid utilization received, on average, opioids that covered 213 days, an underestimate because we post-coded all episodes to no greater than 365 days-supply. By comparison, in a VHA study population, 5% of opioid users initiated an episode of chronic opioid therapy.³⁵ Long-term opioid use has been associated with low discontinuation rates, and other adverse health events,³⁶ as well as increased risk for developing an opioid use disorder.³⁷ In a military population, long-term opioid use may affect future career plans as soldiers maintained on opioids are not deployable.³⁸ Our multivariate analyses found that initiation of long-term opioid use was declining over the years of the study, and was 0.4% in 2014. The largest predictor of long-term opioid receipt was receiving a long-acting opioid within the first 30 days of an opioid episode. This finding is consistent with a study of opioid naïve adults during 2006–2015, which found that patients who initiated on long-acting opioids had the highest likelihood of long-term opioid utilization.³⁹ In 2016, the Centers for Disease Control and Prevention released updated guidelines for prescribing opioids for chronic pain, which included a recommendation for clinicians to prescribe immediate-release opioids when starting opioid therapy instead of extended-release/long-acting opioids.¹ While this study cannot establish causation, this finding suggests future research is warranted about discontinuation rates for patients on extended-release opioids.

We examined the role of having a behavioral health diagnosis on opioid receipt and long-term opioid use, given that prior literature has established that chronic pain is often comorbid with PTSD among those who served in OEF/OIF/OND,^18,40–43 and that military members are at heightened risk for behavioral health problems during the postdeployment year.^44,45 Unlike a study with VHA utilizers,¹⁸ we found that a behavioral health diagnosis was not associated with the likelihood of receiving an opioid, yet, we did find that a behavioral health diagnosis recorded near the start of an opioid prescription increased the probability of long-term opioid use. Previous studies have also shown that Veterans and civilians with behavioral health conditions are more likely to receive long-term opioid therapy.^21,46

This study had several limitations. First, MHS data do not incorporate in-theater health care utilization (i.e., during deployment) or dispensed medications. It is plausible that injuries and prescriptions received while deployed would influence postdeployment opioid receipt and long-term opioid utilization, but these predictors were unavailable. Second, literature is emerging in civilian healthcare settings about the variation among prescribers and prescriber settings that may influence receipt of opioids.⁴⁷ This topic was beyond the scope of these analyses. We did not have access to data to examine the role of opioids received during dental procedures as a predictor of long-term prescription opioid utilization. We cannot generalize from this sample to other services or Army cohorts. Finally, because of the observational study design, this study cannot infer a causal relationship of any of the predictors with the dependent variables.

In 2017, the DoD and Veterans Administration updated the VA/DoD Clinical Practice Guideline for Opioid Therapy for Chronic Pain.^2,48 Among many recommendations, the guidelines advised against initiation of long-term opioid therapy, and included risk mitigation strategies for those currently on opioids. The report included a strong recommendation for initiating alternatives to opioid therapy including self-management strategies and non-pharmacological treatments. Additional research is underway by study team members to examine the role of non-pharmacological treatments as a substitute or complement to opioids in relation to long-term outcomes (e.g., ongoing opioid use, opioid use disorder, overdose, pain levels, and functional status) among soldiers who remain in the military and among cohort members who leave the military and utilize care in the VHA.

Funding

This study was funded by the National Center for Complementary and Integrative Health (NCCIH; R01 AT008404), with support to develop the original study cohort from the National Institute on Drug Abuse (NIDA; R01 DA030150).

Previous Presentation

This study had been presented orally in the Addiction Health Services Research Conference in October 19, 2017 in Wisconsin.