Pain Severity and Interference and Substance Use among Community Pharmacy Patients Prescribed Opioids: A Secondary Analysis of the PHARMSCREEN Study

Abstract

This secondary analysis examined relationships between pain severity and interference and substance use among patients filling opioid prescriptions in Indiana and Ohio community pharmacies (n = 1,461). We likewise sought to explore the moderating role of gender in pain-substance use relations. We used patient-reported data from a cross-sectional health survey linked with controlled substance dispensing data from statewide prescription drug monitoring programs. Multivariable logistic regression estimated associations between pain severity and interference and various indices of risky prescription opioid use and non-opioid substance use. Exploratory analyses examined whether gender moderated associations. Increased pain severity was associated with increased odds of moderate- to high-risk opioid use (OR: 1.23; 95% CI: 1.16–1.31) and opioid-benzodiazepine co-use (OR: 1.20; 95% CI: 1.03–1.40). Increased pain interference was associated with greater odds of receiving opioids from multiple pharmacies or providers (OR: 1.15; 95% CI: 1.01–1.31). Increased pain severity and interference were associated with higher odds of any tobacco use (severity: OR: 1.13; 95% CI: 1.06–1.21; interference: OR: 1.07; 95% CI: 1.01–1.12) and weekly to daily sedative use (severity: OR: 1.13; 95% CI: 1.03–1.25; interference: OR: 1.13; 95% CI: 1.04–1.22). Increased pain severity was associated with decreased odds of any alcohol use (OR: 0.93; 95% CI: 0.88–0.99). Gender was a significant effect modifier in associations between pain and alcohol, tobacco, and cannabis use.

The study was registered in the database of clinicaltrials.gov (register number NCT03936985).

Introduction

Chronic pain is a serious public health concern in the United States (US) representing a major cause of mental and physical health-related disability.⁵¹ Between 1997–2014, there was a 25% increase in the prevalence of diagnosed chronic pain among US adults.⁴⁶ Recent estimates indicate that chronic pain affects nearly one in five US adults, or around 50 million individuals.¹⁷ In parallel, there has been a significant rise in chronic opioid use in US, mostly driven by greater ambulatory use for chronic non-cancer pain.¹⁰ Approximately 2% to 3.6% of US adults use prescription opioids regularly and approximately half of them are long term users.³⁵

The relationship between pain and aberrant use of opioids and other substances must be understood in the broader context of the opioid epidemic. Several studies have found that pain severity and pain interference, or pain interfering with daily activities, are associated with opioid misuse and dependence.^{9, 12, 48, 49, 54} Notwithstanding, research is lacking regarding how pain severity and interference are linked to specific risky opioid use outcomes, such as overdose, opioid-benzodiazepine co-use, high daily opioid dosage, and obtaining prescriptions from multiple providers or pharmacies. While a body of research has demonstrated that pain severity and interference are closely related to alcohol,^{5, 38, 40, 41} tobacco,^{3, 5, 6, 41, 53} and cannabis use and use disorders,^{18, 36, 41} how those pain factors are related to polysubstance, sedative, and stimulant use is not yet clear. Specifically, in previous research that examined polysubstance use among patients with co-occurring chronic pain and substance use disorders (SUDs), pain interference had no association with any polysubstance use.⁵⁷ In another past study conducted among patients with chronic pain, pain severity was not associated with benzodiazepine use after adjustment for confounders.¹⁶ To our knowledge, no research has examined associations between pain and stimulant use.

Additionally, several studies have identified sex/gender as a moderator of associations between pain severity or interference with alcohol, tobacco, and cannabis use.^{5, 6, 41, 53, 57} However, results from previous research have been variable and inconclusive with some studies reporting significant associations between pain and alcohol and nicotine dependence among men but not women and others reporting significant relationships among women but not men.^{6, 41, 53} Given the differences between men and women in pain experiences⁷ as well as in substance use patterns and consequences,⁴² additional research characterizing sex/gender differences in pain-substance use relations is warranted.

While several studies have investigated pain-substance use relations, none have examined this topic among a community pharmacy population prescribed opioids. Because community pharmacies are widely accessible, they provide important touch points with the health care system for substance use screening and intervention. The main objective of this study was to evaluate the relationship between pain severity and interference and risky opioid use and other substance use behaviors among community pharmacy patients filling opioid prescriptions. We likewise sought to explore the moderating effect of gender in these relationships. Based on previous research, we hypothesized that increased pain would be associated with increased use of opioids, sedatives, alcohol, tobacco, and cannabis and that gender would moderate effects. Because of the mixed results from previous studies, we had no hypothesis about to what extent gender would moderate effects.

Methods

Study Design and Sites

This is a secondary analysis of data from the Validation of a Community Pharmacy-Based Prescription Drug Monitoring Program Risk Screening Tool (PHARMSCREEN) study, funded by the National Institute of Drug Abuse Clinical Trials Network and registered at ClinicalTrials.gov (Identifier: NCT03936985). PHARMSCREEN was a one-group, cross-sectional study conducted in 19 community pharmacies located in the Indiana and Ohio between November 2019 and October 2020. Institutional review boards at University of Cincinnati and University of Utah approved this study, which was supported by the National Institutes of Health HEAL Initiative.

Study Participants

Study design details and procedures as well as the main findings have been published elsewhere.¹⁵ Briefly, we recruited a convenience sample of English-speaking community pharmacy patients aged ≥18 years filling opioid prescriptions. Trained pharmacy staff informed potentially eligible participants about the study opportunity at the point-of-dispensing. Interested patients received a study flyer and computer tablet to complete an interest survey, which once completed, triggered an automatic email containing the link to an electronic informed consent document. After consenting to the study, participants were directed to complete a self-screening assessment, with eligible participants completing a one-time, self-administered health survey. Participants received a $50 gift card compensation for their time.

A total of 2,090 patients completed consent forms and 1,921 completed the self-screening assessment (Supplementary Material, Figure S1). There were 281 self-screening fails. Reasons for exclusion included criminal justice system involvement (n = 8), cancer treatment participation (n = 49), and receipt of solely buprenorphine or buprenorphine combination products (n = 127). Among 1,629 patients who initiated the health survey, 1,461 completed the survey.

Data Sources

Patient-reported outcomes included measures of opioid use and risk behaviors, other substance use, and physical and mental health. Survey data were collected and managed using REDCap electronic data capture tools.²⁹ Patient identifying information (name, address, date of birth, and pharmacy location where their opioid was filled) was used to link survey data to participant records from Indiana and Ohio Prescription Drug Monitoring Program (PDMP) databases containing statewide information on controlled substance (Schedules II-IV) dispensations. After the linkage, our research team performed quality assurance checks to ensure the accuracy of the linked data. A detailed description of the survey and PDMP data linkage has been previously published.¹⁵

Measures

Pain

Patient-reported pain was captured using the Brief Pain Inventory Short Form (BPI-SF), a widely used self-report pain inventory with 11 items to assess pain severity and interference and two additional items to report current treatment regimen.¹³ The reliability and validity of the BPI-SF for assessing non-cancer pain patients has been well documented.³⁴ This study used participant responses from the 4-item pain severity scale measuring worst, least, average, and current pain and 7-item pain interference scale measuring how pain interferes with general activity, mood, walking ability, work, relations, sleep, and enjoyment of life. For each item, participants rated their pain in the past week on a scale from 0 (no pain/interference) to 10 (worst pain/interference). We averaged responses for pain severity and interference items separately to create overall scores for each subscale.

Patient-Reported Substance Use

Moderate- to high-risk prescription opioid use

Moderate- to high-risk prescription opioid use was assessed using an adapted version of the World Health Organization Alcohol, Smoking and Substance Involvement Screening Test (WHO ASSIST, version 3.0), a valid tool to identify lifetime and current substance use and related risks.³⁰ We employed the version developed and validated by McNeely and colleagues⁴³ that includes separate items for prescription and illicit opioids (e.g. heroin) as well as prescription and illicit stimulants (e.g. methamphetamine). We calculated a prescription opioid-specific risk score from WHO ASSIST questions 2–7 asking participants about prescription opioid use for non-medical reasons (taking higher doses, more frequent doses, or for other reasons than prescribed) and categorized responses into low-risk use (0–3) and moderate- to high-risk use (≥4) on the basis of standard ASSIST methodology.²⁷

Non-opioid substance use

Non-opioid substance use was assessed using WHO ASSIST question 2 that asks participants to rate their current (past 3 month) substance use on a 5-point Likert scale from “never” to “daily or almost daily”.³⁰ Measures included any use (vs no use) and weekly to daily use (vs less frequent use) of the following substances: any non-opioid substance (alcohol, tobacco, non-opioid drugs), polysubstance (>1 non-opioid substance), alcohol, tobacco, cannabis, sedatives, stimulants (cocaine, methamphetamine, prescription stimulants), and other drugs (inhalants, hallucinogens, and other non-specific drugs).

Lifetime prescription or illicit overdose

Lifetime prescription or illicit overdose was assessed using the Overdose Experiences, Self and Witnessed—Drug (OESWD) instrument.²⁴ Participants were provided a description of the term ‘overdose’ and asked to report how many lifetime overdose events (0–≥6) they had experienced. We dichotomized OESWD scores into no overdose (0) and one or more overdoses (≥1).

PDMP Measures

High opioid dosage

High opioid dosage was defined as ≥50 average daily morphine milligram equivalents (MME) over 180 days on the basis of Centers for Disease Control and Prevention prescribing guidelines.²⁰ We calculated this measure as the sum of MMEs across all opioid prescriptions obtained during the 180-day observation window divided by 180 days.²⁵

Opioid–benzodiazepine overlap

Opioid–benzodiazepine overlap was defined as concurrent use of prescription opioids and benzodiazepines for ≥30 cumulative days in the past 90 days according to the Pharmacy Quality Alliance (PQA) measure.²

Multiple providers or pharmacies

Multiple providers or pharmacies was defined as obtaining narcotic prescriptions from ≥4 prescribers or ≥4 pharmacies within 180 days. We adapted this measure from the PQA measure for multiple prescribers in persons without cancer (NQF #2950),² which calculates prescriptions obtained from ≥4 prescribers and ≥4 pharmacies, because few individuals in our sample (n = 7) met both criteria.

Physical and Mental Health

Self-rated general health

Self-rated general health was assessed using one Likert-scale item from construct-valid health-related quality-of-life questionnaire, the Short Form-12.⁵⁸ Participants rated their general health from poor to excellent.

Depressive symptoms

Depressive symptoms were assessed with the Patient Health Questionnaire–2 (PHQ-2), a two-item valid and reliable brief screener measuring depressed mood and anhedonia.³⁷ Items were “having little interest or pleasure in doing things” and “feeling down, depressed, or hopeless” in the past two weeks. Responses for each item were rated on a four-point scale, with total scores ranging from 0–6. Consistent with established cutoffs, we defined depressive symptoms as a score of ≥3.³⁷

Sociodemographics

Sociodemographics were captured using PhenX Toolkit measures²⁸ and included age, race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, other), gender, education (high school/GED, some college/associates, bachelors/masters/doctorate), employment (employed, disabled, other), insurance status, and marital status (married/coupled, divorced/widowed/separated, never married).

Statistical Analysis

For descriptive analyses, we assessed the prevalence of substance use among individuals reporting mild (0–4), moderate (5–6), and severe (7–10) pain according to previously established cutpoints for non-cancer pain.⁵⁹ Proportions and 95% confidence intervals (CI) were calculated separately for pain severity and interference.

For primary analyses, we characterized associations between pain and risky prescription opioid use and non-opioid substance use using logistic regression with sequential adjustment for potential confounders. Pain severity and interference were modeled as continuous variables on an 11-point scale to increase statistical power and improve clinical interpretability of the findings. Our crude model included pain as the only predictor variable. Model 1 added sociodemographic variables to the crude model. Model 2 added physical and mental health variables to model 1. The final adjusted model added substance use variables to model 2. Covariates were modeled as continuous or dichotomized to account for small cell counts. We retained all covariates in the models because of their conceptual importance¹⁹ and to enable consistency and comparability across outcome models. We tested the fit of successive logistic regression models using the Pearson goodness-of-fit statistic, and all final models were found to have acceptable model fit (P > 0.05). Multicollinearity between independent variables was assessed using variance inflation factor (VIF) diagnostic statistics with linear regression. VIF values were < 2.5 indicating no multicollinearity.⁴⁴ Multiple imputation technique was not used because there were not more than 5% missing data for any variable.³¹ Results from logistic regression analyses are presented as odds ratios (ORs) and 95% confidence intervals (CIs) and represent the relative change in the odds of the outcome per every one-unit increase in pain.

Additional analyses were conducted to test for effect modification by gender. We included paingender interaction terms in the final adjusted models and fit separate models for men and women when interaction effects were statistically significant. We defined statistical significance as P < 0.05. Because we made no adjustments for multiple comparisons in this secondary analysis, findings should be interpreted as exploratory. We reported findings according to Strengthening of Reporting in Observational Epidemiology (STROBE) guidelines for cross-sectional studies.²² Data were analyzed in R software version 4.0.4 (R Foundation for Statistical Computing).

Results

Participant Characteristics

Among 1,461 participants, mean (standard deviation [SD]) pain severity and interference were 4.9 (2.2) and 4.9 (2.7), respectively (Table 1). Women had higher mean (SD) pain severity (5.0 (2.2) vs 4.6 (2.3); P = 0.006) and pain interference (5.0 (2.7) vs 4.6 (2.8); P = 0.01) scores than men. A majority of participants had a high school degree/GED or lower (40.8%) or some college (39.7%) education, were married/coupled (59.9%), identified as white, non-Hispanic race-ethnicity (93.9%). Women were significantly younger (mean (SD) age, 48.6 (14.8) vs 51.0 (14.6) years; P = 0.002) and more likely to be insured (96.3% vs 93.6%; P = 0.016) than men. A higher proportion of women than men reported having depressive symptoms (23.0% vs 14.8%; P < 0.001).

Table 1.

Characteristics of participants, overall and stratified by gender, Validation of a Community Pharmacy-Based Prescription Drug Monitoring Program Risk Screening Tool (PHARMSCREEN) study.

Characteristics	Overall (n=1461)	Men (n = 552)	Women (n = 909)	P a
Pain b
Pain severity, mean (SD)c	4.9 (2.2)	4.6 (2.3)	5.0 (2.2)	0.006
Pain interference, mean (SD)c	4.9 (2.7)	4.6 (2.8)	5.0 (2.7)	0.010
Sociodemographics
Age, years, mean (SD)c	49.5 (14.7)	51.0 (14.6)	48.6 (14.8)	0.002
Education, n (%)				0.104
High school/GED or less	585 (40.8)	233 (43.0)	352 (39.5)
Some college/associates	569 (39.7)	196 (36.2)	373 (41.8)
Bachelors/masters/doctorate	280 (19.5)	113 (20.9)	167 (18.7)
Employment, n (%)				0.016
Employed full- or part-time	596 (41.2)	252 (46.0)	344 (38.3)
Disabled	501 (34.7)	175 (31.9)	326 (36.3)
Otherd	349 (24.1)	121 (22.1)	228 (25.4)
Insurance status, n (%)				0.016
Uninsured	68 (4.7)	35 (6.4)	33 (3.7)
Insured	1378 (95.3)	510 (93.6)	868 (96.3)
Marital Status, n (%)				<0.001
Married/coupled	871 (59.9)	370 (67.5)	501 (55.4)
Divorced/widowed/ separated	369 (25.4)	102 (18.6)	267 (29.5)
Never married	213 (14.7)	76 (13.9)	137 (15.1)
Race/ethnicity, n (%)				0.787
White, non-Hispanic	1359 (93.9)	515 (94.7)	844 (93.5)
Black, non-Hispanic	63 (4.4)	21 (3.9)	42 (4.7)
Hispanic	11 (0.8)	3 (0.6)	8 (0.9)
Other/unknown race, non-Hispanice	14 (1.0)	5 (0.9)	9 (1.0)
Study site state, n (%)				0.816
Indiana	192 (13.1)	74 (13.4)	118 (13.0)
Ohio	1269 (86.9)	478 (86.6)	791 (87.0)
Physical and mental health
General health, n (%)				0.178
Excellent	54 (3.7)	24 (4.4)	30 (3.3)
Very good	247 (17.0)	100 (18.2)	147 (16.2)
Good	561 (38.5)	224 (40.7)	337 (37.2)
Fair	447 (30.7)	152 (27.6)	295 (32.6)
Poor	148 (10.2)	51 (9.3)	97 (10.7)
Depressive symptoms, n (%)f				<0.001
Yes	285 (19.9)	80 (14.8)	205 (23.0)
No	1147 (80.1)	461 (85.2)	686 (77.0)

Values may not add to 100% due to rounding.

^aObtained using chi-square for categorical data (unless otherwise noted) and statistically significant at P < 0.05.

^bMeasured using the validated Brief Pain Inventory (BPI) consisting of a 4-item pain severity subscale and a 7-item pain interference subscale with scores ranging from 0–10.

^cP value obtained using Student’s t-test for continuous data.

^dOther employment includes retired, keeping house, student status, looking for work, temporary leave, or other employment.

^eOther race includes American Indian or Alaska Native, Asian, Native Hawaiian or Pacific Islander, and other non-specific race.

^fAssessed in the past 2 weeks using the Patient Health Questionnaire (PHQ)-2 and defined as a score of 3 or greater on the PHQ-2.

Prevalence of Substance Use by Pain Category

The prevalence of risky prescription opioid use monotonically increased with increasing pain severity and interference (Table 2). As an example, the proportion of participants reporting moderate- to high-risk prescription opioid use was greater for those rating pain severity as severe (61.2%; 95% CI: 56.5%–65.9%) than as moderate (50.8%; 95% CI: 46.5–55.1) or mild (28.3%; 95% CI: 24.3%–32.3%). With the exception of alcohol, the prevalence of non-opioid substance use largely increased as pain severity and interference increased. For example, the proportion of participants reporting any tobacco use was 27.6% (95% CI: 23.9%–31.4%), 39.0% (95% CI: 34.0%–44.0%), 43.2% (95% CI: 39.0%–47.4%) among those rating pain interference as mild, moderate, and severe, respectively. The prevalence of any and weekly to daily alcohol use decreased as pain severity and interference increased. The proportion of participants reporting weekly to daily alcohol use was 22.2% (95% CI: 18.7%–25.7%) and 12.2% (95% CI: 9.0%–15.0%), respectively, among persons reporting mild and severe pain interference.

Table 2.

Prevalence of risky opioid use and non-opioid substance use, by pain severity and interference category, PHARMSCREEN study (n=1,461).

Substance Use Outcomes	Pain Severitya % (95% CI)			Pain Interferencea % (95% CI)
	Mild (n = 495)	Moderate (n = 523)	Severe (n = 436)	Mild (n = 544)	Moderate (n = 360)	Severe (n = 538)
Risky opioid use
Moderate- to high-risk Rx opioid useb	28.3 (24.3–32.3)	50.8 (46.5–55.1)	61.2 (56.5–65.9)	34.7 (30.7–38.7)	47.6 (42.4–52.8)	56.5 (52.3–60.8)
Lifetime Rx or illicit drug overdosec	8.1 (5.7–10.5)	9.6 (7.1–12.2)	11.5 (8.5–14.5)	5.3 (3.5–7.2)	11.5 (8.2–14.8)	13.0 (10.2–15.9)
High opioid dosaged	2.2 (0.9–3.5)	6.7 (4.6–8.8)	9.6 (6.9–12.4)	3.3 (1.8–4.8)	6.1 (3.6–8.6)	8.7 (6.4–11.1)
Opioid-benzodiazepine overlape	1.6 (0.5–2.7)	5.4 (3.4–7.3)	7.8 (5.3–10.3)	2.2 (1.0–3.4)	5.8 (3.4–8.3)	6.5 (4.4–8.6)
Multiple pharmacies or providersf	1.8 (0.6–3.0)	4.8 (2.3–6.7)	7.1 (4.7–9.5)	2.0 (0.8–3.2)	5.0 (2.8–7.3)	6.7 (4.6–8.8)
4 distinct pharmacies	0.2 (0.0–0.6)	1.7 (0.6–2.8)	1.6 (0.4–2.8)	0.4 (0.0–0.9)	1.7 (0.3–3.0)	1.7 (0.6–2.8)
4 distinct providers	1.8 (0.6–3.0)	3.8 (2.2–5.5)	6.0 (3.7–8.2)	2.0 (0.8–3.2)	3.9 (1.9–5.9)	5.6 (3.6–7.5)
Non-opioid substance useg
Any non-opioid substance
Any use	76.7 (73.0–80.5)	75.9 (72.2–80.0)	78.7 (74.8–82.5)	74.4 (70.7–78.1)	76.7 (72.3–81.0)	79.7 (76.3–83.1)
Weekly to daily use	47.8 (43.4–52.2)	56.4 (52.2–60.7)	62.6 (58.1–67.2)	47.7 (43.5–51.9)	58.6 (53.5–63.7)	61.2 (57.0–65.3)
Polysubstanceh
Any use	31.6 (27.5–35.7)	38.1 (33.9–42.2)	41.7 (37.1–46.4)	30.6 (26.7–34.5)	43.3 (38.2–48.5)	35.1 (34.4–43.3)
Weekly to daily use	14.6 (11.5–17.7)	14.5 (11.5–17.6)	20.4 (16.6–24.2)	13.4 (10.6–16.3)	19.4 (15.4–23.5)	17.1 (13.9–20.3)
Alcohol
Any use	64.0 (59.7–68.2)	52.9 (48.6–57.2)	47.9 (43.2–52.6)	59.1 (54.9–63.2)	59.4 (54.3–64.5)	48.1 (43.9–52.4)
Weekly to daily use	26.7 (22.8–30.6)	16.7 (13.5–19.9)	12.1 (9.0–15.2)	22.2 (18.7–25.7)	23.3 (18.9–27.7)	12.2 (9.0–15.0)
Tobacco
Any use	26.8 (22.9–30.7)	38.0 (33.8–42.2)	44.3 (39.7–49.0)	27.6 (23.9–31.4)	39.0 (34.0–44.0)	43.2 (39.0–47.4)
Weekly to daily use	22.2 (18.6–25.9)	34.8 (30.7–38.9)	40.6 (36.0–45.2)	24.3 (20.6–27.9)	34.4 (29.4–39.3)	39.3 (35.1–43.4)
Cannabis
Any use	12.3 (9.4–15.2)	13.7 (10.7–16.6)	19.9 (16.1–23.7)	11.0 (8.3–13.6)	19.4 (15.3–23.5)	16.7 (13.6–19.9)
Weekly to daily use	5.3 (3.3–7.3)	4.5 (2.7–6.2)	10.0 (7.2–12.8)	4.1 (2.4–5.8)	9.0 (6.0–12.0)	7.2 (5.0–9.4)
Sedatives
Any use	12.6 (9.7–15.6)	17.6 (14.3–20.9)	28.3 (24.1–32.6)	12.2 (9.4–15.0)	18.1 (14.1–22.0)	26.5 (22.8–30.3)
Weekly to daily use	7.7 (5.4–10.1)	10.9 (8.2–13.6)	21.5 (17.6–25.4)	7.4 (5.2–9.6)	11.4 (8.1–14.7)	19.7 (16.3–23.0)
Stimulantsi
Any use	3.5 (1.9–5.2)	7.4 (5.1–9.7)	8.3 (5.6–10.9)	4.5 (2.8–6.3)	8.2 (5.3–11.1)	6.7 (4.5–8.8)
Weekly to daily use	1.7 (0.5–2.8)	4.3 (2.5–6.0)	4.3 (2.3–6.2)	3.4 (1.9–5.0)	3.1 (1.3–4.9)	3.4 (1.9–5.0)
Other drugsj
Any use	1.4 (0.4–2.5)	3.1 (1.6–4.5)	2.1 (0.7–3.4)	2.4 (1.1–3.7)	1.9 (0.5–3.4)	2.2 (1.0–3.5)
Weekly to daily use	0.8 (0.0–1.6)	2.3 (1.0–3.6)	1.4 (0.3–2.5)	2.0 (0.8–3.2)	1.1 (0.0–2.2)	1.3 (0.3–2.3)

Abbreviations: Rx, prescription; MME, morphine milligram equivalents.

^aPain severity and pain interference were categorized into mild (0–4), moderate (>4–6), and severe pain (>6–10) based on common cutpoints for chronic non-cancer pain from review of the literature.

^bAssessed in the using the World Health Organization Alcohol, Smoking and Substance Involvement Screening Test (ASSIST). Measure was constructed based on responses to items asking about non-medical use of Rx opioids (e.g. morphine, codeine, fentanyl, oxycodone). Response scores were summed and categorized into low-risk use (ASSIST risk score ≤3) and moderate- to high-risk use (ASSIST risk score ≥4).

^cAssessed with the Overdose Experiences, Self and Witnessed—Drug instrument (OESWD), a 1-item questionnaire with scores ranging from 0–≥6 and dichotomized into yes vs no.

^dDefined as ≥50 average daily MME over 180 days. Average daily MME calculation was the sum of MMEs across all opioid prescriptions obtained during the 180-day observation window divided by 180 days.

^eDefined as concurrent use of prescription opioids and benzodiazepines for ≥30 cumulative days in the past 90 days.

^fDefined as obtaining opioid prescriptions from ≥4 distinct pharmacies or ≥4 distinct providers in the past 180 days.

^gNon-opioid substance use assessed in the past 3 months using the WHO ASSIST and dichotomized into any use (vs no use) and weekly to daily use (vs less frequent use).

^hDefined as any use or weekly to daily use of >1 of the following non-opioid substances: alcohol, tobacco cannabis, sedatives, stimulants, and other drugs.

ⁱIncludes crack/cocaine, non-medical use of Rx stimulants (e.g. Adderall, Ritalin), and methamphetamine.

^jIncludes inhalants, hallucinogens, and other non-specific drugs.

Associations of Pain with Substance Use

In final adjusted models, each unit increase in pain severity was associated with a 23% and 20% increase in odds of moderate- to high-risk prescription opioid use (OR: 1.23; 95% CI: 1.16–1.31) and opioid-benzodiazepine overlap (OR: 1.20; 95% CI: 1.03–1.40), respectively, while each unit increase in pain interference was associated with a 15% increase in odds of receiving prescription opioids from multiple pharmacies or providers (OR: 1.15; 95% CI: 1.01–1.31) (Table 3). Higher pain interference was associated with increased odds of moderate- to high-risk prescription opioid use and lifetime prescription or illicit drug overdose in models adjusted for sociodemographic, self-rated general health, and depression (model 2), but associations did not remain significant after adjustment for substance use confounders (final adjusted models) (Supplementary Material, Table S1).

Table 3.

Adjusted associations between pain and risky opioid use and non-opioid substance use, PHARMSCREEN study (n = 1,461).

Substance Use Outcomes	Pain Severitya OR (95% CI)	Pain Interferencea OR (95% CI)
Risky opioid useb
Moderate- to high-risk opioid use (vs low-risk)	1.23 (1.16–1.31)	1.05 (1.00–1.10)
Lifetime Rx or illicit drug overdose (vs no overdose)	1.02 (0.92–1.13)	1.09 (1.00–1.19)
≥50 average daily MME over 180 days (vs <50)	1.08 (0.95–1.24)	1.04 (0.93–1.17)
≥30 days opioid-benzodiazepine overlap in 90 days (vs <30)	1.20 (1.03–1.40)	1.10 (0.97–1.25)
≥4 distinct pharmacies or providers in 180 days (vs <4)	1.15 (0.98–1.34)	1.15 (1.01–1.31)
Non-opioid substance usec
Any non-opioid substance
Any use vs no use	1.01 (0.95–1.08)	1.03 (0.98–1.09)
Weekly to daily use vs less frequent use	1.07 (1.01–1.13)	1.06 (1.01–1.11)
Polysubstance
Any use vs no use	1.06 (0.99–1.12)	1.02 (0.97–1.07)
Weekly to daily use vs less frequent use	1.05 (0.97–1.14)	1.03 (0.96–1.10)
Alcohol
Any use vs no use	0.93 (0.88–0.99)	0.97 (0.92–1.02)
Weekly to daily use vs less frequent use	0.89 (0.83–0.96)	0.97 (0.92–1.03)
Tobacco
Any use vs no use	1.13 (1.06–1.21)	1.07 (1.01–1.12)
Weekly to daily use vs less frequent use	1.14 (1.06–1.22)	1.05 (1.00–1.11)
Cannabis
Any use vs no use	1.04 (0.95–1.14)	1.00 (0.93–1.08)
Weekly to daily use vs less frequent use	1.12 (0.99–1.27)	1.09 (0.99–1.21)
Sedatives
Any use vs no use	1.10 (1.02–1.20)	1.08 (1.00–1.15)
Weekly to daily use vs less frequent use	1.13 (1.03–1.25)	1.13 (1.04–1.22)
Stimulants/other drugsd
Any use vs no use	1.04 (0.92–1.17)	1.00 (0.91–1.10)
Weekly to daily use vs less frequent use	1.04 (0.90–1.20)	0.96 (0.86–1.09)

Abbreviations: OR, odds ratio; CI, confidence interval; Rx, prescription; MME, morphine milligram equivalents.

NOTE. Final adjusted models included the following covariates: age, education, employment, gender (non-stratified only), insurance status, marital status, pharmacy site, race/ethnicity, self-rated general health, depressive symptoms, moderate- to high-risk non-medical Rx opioid use (excluding outcome model), and non-opioid substance use (excluding outcome model substance). Bold typeface indicates statistically significant findings.

^aPain severity and pain interference were measured using the Brief Pain Inventory (see Table 1 footnotes) and modeled as continuous on an 11-point scale.

^bRisky opioid use assessment and definitions are described in the Table 2 footnotes.

^cNon-opioid substance use assessment and outcomes are explained in the Table 2 footnotes.

^dIncludes crack/cocaine, non-medical use of Rx stimulants (e.g. Adderall, Ritalin), methamphetamine, inhalants, hallucinogens, and non-specific drugs.

Significant positive associations were observed for weekly to daily use of non-opioid substances with pain severity (OR: 1.07; 95% CI: 1.01–1.13) and interference (OR: 1.06; 95% CI: 1.01–1.11). While increased pain severity was associated with higher odds of any tobacco use (OR: 1.13; 95% CI: 1.06–1.21), weekly to daily tobacco use (OR: 1.14; 95% CI: 1.06–1.22), any sedatives use (OR: 1.10; 95% CI: 1.02–1.20), and weekly to daily sedatives use (OR: 1.13; 95% CI: 1.03–1.25), increased pain interference was only associated with any tobacco use (OR: 1.07; 95% CI: 1.01–1.12) and weekly to daily sedatives use (OR: 1.13; 95% CI: 1.04–1.22). There was a significant inverse relationship between increased pain severity and any alcohol use (OR: 0.93; 95% CI: 0.88–0.99) as well as weekly to daily alcohol use (OR: 0.89; 95% CI: 0.83–0.96). Increased pain severity was associated with higher odds of any polysubstance, cannabis, and stimulant/other drug use in model 2, however associations were not significant in final adjusted models (Supplementary Material, Table S1).

Effect Modification by Gender

Gender was a significant effect modifier in associations between: pain severity and weekly to daily alcohol use; and, pain interference and weekly to daily alcohol use, any tobacco use, weekly to daily tobacco use, and any cannabis use (P < .05 for all) (Table 4). There was a significant negative association between pain severity and weekly to daily alcohol use among women (OR: 0.86; 95% CI: 0.76–0.96) but not men (OR: 0.92; 95% CI: 0.83–1.02). Also, among women, each unit increase in pain interference was associated with a 10% and 11% increase in odds of tobacco use (OR: 1.10; 95% CI: 1.03–1.18) and daily to weekly tobacco use (OR: 1.11; 95% CI: 1.03–1.19), respectively. However, increased pain interference was not associated with any tobacco use (OR: 1.2; 95% CI: 0.94–1.11) nor daily to weekly tobacco use (OR: 0.99; 95% CI: 0.91–1.08) among men. Although associations between pain interference and any cannabis use were different between women and men, these associations were not significant in final adjusted models.

Table 4.

Gender-stratified associations between pain and alcohol, tobacco, and cannabis use, PHARMSCREEN study (n=1,461).

Substance Use Outcomes	Pain Severitya		Pain Interferencea
	Men OR (95% CI)	Women OR (95% CI)	Men OR (95% CI)	Women OR (95% CI)
Alcoholb
Weekly to daily use vs less frequent use	0.92 (0.83–1.02)	0.86 (0.76–0.96)	1.00 (0.92–1.09)	0.93 (0.85–1.01)
Tobaccob
Any use vs no use	--	--	1.02 (0.94–1.11)	1.10 (1.03–1.18)
Weekly to daily use vs less frequent use	--	--	0.99 (0.91–1.08)	1.11 (1.03–1.19)
Cannabisb
Any use vs no use	--	--	1.05 (0.94–1.18)	0.96 (0.87–1.06)

Abbreviations: OR, odds ratio; CI, confidence interval

NOTE. Effect modification by gender was tested by including a pain*gender interaction term in overall models. All interactions were significant at P< .05. Final adjusted models included the following covariates: age, education, employment, gender (non-stratified only), insurance status, marital status, pharmacy site, race/ethnicity, self-rated general health, depressive symptoms, moderate- to high-risk non-medical Rx opioid use (excluding outcome model), and non-opioid substance use (excluding outcome model substance). Bold typeface indicates statistically significant findings.

^aPain severity and pain interference were measured using the Brief Pain Inventory (see Table 1 footnotes) and modeled as continuous on an 11-point scale.

^bAlcohol, tobacco, and cannabis use assessment and outcomes are explained in the Table 2 footnotes.

Discussion

This study leveraged linked patient-reported and controlled substance dispensing data to examine associations between pain and substance use among a sample of community pharmacy patients prescribed opioids. We found that the prevalence of risky prescription opioid use and non-opioid substance use, except for alcohol, increased with worsening pain severity and interference. After adjusting analyses for sociodemographic, physical and mental health, and substance use variables, increasing pain severity and interference were positively associated with use of non-medical prescription opioids, sedatives, and tobacco; increasing pain severity was inversely associated with alcohol use. Gender moderated associations of pain with alcohol, tobacco, and cannabis use.

Average pain severity and interference scores reported in this study are not dissimilar to those reported in other studies conducted among individuals prescribed opioids. In a study of post-9/11 Veterans reporting any opioid use, mean pain severity and interference were 5.2 and 4.5, respectively, compared to 4.9 and 4.9, respectively, in our sample.¹⁴ In another US study conducted among persons with chronic pain taking opioids, mean pain severity was 5.1.⁴⁹ Pain is closely linked to the rise of prescription opioid misuse in the US,²⁶ and thus it is not surprising that more than half of participants in our sample who reported severe pain also reported moderate-to high-risk prescription opioid use. Our finding that pain interference was associated with obtaining opioids from multiple providers or pharmacies extends prior research as no studies, to our knowledge, have linked pain with obtaining opioids from multiple pharmacies or doctors. It should be noted that our study cannot shed light on the reasons underlying this association. Pain is the primary indication and reason for opioid use among US adults.⁵² While it is probable that patients were obtaining opioids from multiple providers or pharmacies for pain relief purposes, it is also possible that participants were seeking opioids for other reasons, such as coping with emotional distress. Given the relatively small number of persons in our sample with this combination of risky behaviors, this novel finding should be further explored in future studies with larger sample sizes.

Another important finding of this study was that the prevalence of sedative use, including co-use of opioids and benzodiazepines, increased with increasing pain. Pain severity and interference were positively associated with weekly to daily sedative use and opioid-benzodiazepine overlap. Anxiety, tension, insomnia, and depression, which are common indications for sedative prescriptions, especially benzodiazepines,⁵⁰ occur at high rates with acute or chronic painful conditions³² and may contribute to these findings. Notably, nearly one-fifth of adults with severe pain reported weekly to daily sedative use. While the high proportion of individuals with concurrent opioid and regular sedative use is concerning because of the well-documented overdose risks of mixing these prescription drugs,²¹ our finding is consistent with current patterns of concomitant opioid-sedative prescribing and use in the US. Between 2003–2015, co-prescribing of benzodiazepines and other sedative-hypnotics with opioids quadrupled and doubled, respectively.¹

Consistent with evidence that smoking rates are higher among persons with than those without co-occurring pain,³⁹ we found that the prevalence of tobacco use increased with increasing pain. Pain severity and interference were associated with increased likelihood of tobacco after adjustment for confounders. These data agree with established evidence that tobacco and pain interact in a bidirectional relationship: tobacco use may confer short-term analgesia but may be a precursor to chronic pain, which may drive use of tobacco as a coping mechanism.^{19, 39} We additionally found that gender moderated effects of pain with tobacco use such that pain interference was associated with tobacco use among women but not men. We propose two hypotheses that may explain this association for women, but not men. Smit et al⁵³ hypothesized women are more susceptible to tobacco use with increasing pain due to greater sensitivity to the antinociceptive effects of nicotine compared to men. Alternatively, it may be that men with increasing pain interference are more successful in abstaining from tobacco than their female counterparts, as women have more difficulty quitting smoking than men.⁵⁵ Nevertheless, inconsistencies in findings related to sex, pain, and tobacco use highlight the need for further study on the role of gender in pain-tobacco relations given the high disease burden associated with tobacco use and chronic pain.

Contrary to our hypothesis, alcohol use prevalence decreased as pain increased. While another study among patients on chronic opioid therapy reported a similar finding,¹¹ our results were unexpected given that alcohol is a powerful analgesic that has been found to provide moderate to large reductions in pain intensity rating.⁵⁶ It may be that patients with severe pain are taking higher opioid doses and thus reduce alcohol use in response to increased opioid use, potentially due to concern of overdose risk. Like benzodiazepines, alcohol in high doses or in combination with other central nervous system (CNS) depressants can cause respiratory depression and has been implicated in large number of non-fatal and fatal overdoses involving opioids.³³ Because alcohol often has harmful interactions with prescription medications, prescription drug packaging has long contained warning labels cautioning against using medications in combination with alcohol. Our assumption is that pharmacists and prescribers may help patients understand the risks of taking opioids with alcohol but may fail to mention similar risks posed by other specific CNS depressants.

Practice Considerations

Our findings underscore the critical need for advancements in multidisciplinary pain and substance use management programs into non-specialized clinical settings. Community pharmacies are ideally situated to deliver such interventions given that pharmacist are trusted, skilled health care professionals and 91% of Americans live within five miles of a pharmacy.⁴⁷ Community pharmacies are widely accessible, often offering walk-in appointments, and more commonly visited than primary care physicians,⁸ suggesting that pharmacists may have frequent opportunities to screen for pain and substance use and potentially deliver targeted motivational interventions. Nonetheless, pharmacy is an underutilized resource and community pharmacists do not regularly engage in pain or substance use screening among their patient populations. If pharmacists had more actionable information about patient pain, it could benefit the continuum of care for persons living with chronic pain with respect to opioid dose adjustments and other substance use.

To reduce the long-term risk of substance abuse and dependence among patients with persistent pain, pain management strategies should not only reduce pain symptoms but also directly target mechanisms – decreased hedonic capacity, increased motivation towards drug seeking, and elevated stress²³ – through which pain and substance misuse may overlap. Mental health comorbidities, including anxiety and depression, and psychological processes underlying mental health conditions (i.e. transdiagnostic factors), such as anxiety sensitivity, pain-related anxiety, and pain catastrophizing, may interact with pain and substance use to hinder effective treatment and should also be targeted.¹⁹ Additionally, because findings from our and other studies suggest that men and women use substances differently in the context of pain, one-size fits all treatment strategies may not be appropriate. Considering underlying sex/gender differences in pain and substance use as well as in pharmacologic response to pain or substance use treatments may be helpful when developing intervention targets. Evidence-based strategies that have been demonstrated to be successful in concurrently treating pain, substance use, and psychiatric comorbidity, such as acceptance and commitment therapy, mindfulness-oriented recovery enhancement, and cognitive behavioral therapy,⁴ could be more effective by incorporating gender-specific components than taking a gender-neutral approach. Future research in this area is warranted.

Strengths and Limitations

This study has strengths and limitations. First, the cross-sectional study design limits our ability to understand the temporal order of pain and substance use outcomes, thus no causal inferences can be made. However, our results are consistent with past research findings that pain is associated with substance use and may inform future longitudinal studies assessing causal relations between these two conditions. Second, because several outcomes were patient-reported, results are subject to potential reporting bias. However, the linked survey and PDMP database provided a unique source of information that allowed for evaluation of pain and indices of risky prescription opioid use not traditionally captured together in administrative or survey data. Third, participants were predominately non-Hispanic white patients of midwestern US pharmacies. While the racial/ethnic composition of our sample is consistent with the demographic of persons prescribed opioids in the US,⁴⁵ this study’s findings may not extend to urban community pharmacy patients of other races/ethnicities. Future studies are needed to establish the generalizability of our findings to different samples. Fourth, although many potential confounding factors were adjusted, we did not have information on several factors, including cause of pain, length of time on opioids, use of other medications, such as anti-depressants, or relevant transdiagnostic factors, including pain-related anxiety or anxiety sensitivity, known to influence pain and substance-related outcomes.¹⁹ It is possible that the observed relationships would differ in the presence of these factors, thus future studies extending this research would benefit from measuring a broader range potentially confounding factors. Further, our pain measures did not reveal cause of pain or distinguish between acute vs chronic pain, which is meaningful in that substance use is more strongly related to chronic than acute pain. Fifth, our study had a larger sample size as compared to other studies using primary data collection, but several outcomes were relatively rare in the sample and stratifications resulted in small cell counts. In addition, we may not have detected all relevant interactions due to limited power as PHARMSCREEN was not designed to test effect modification.

Conclusion

In a community pharmacy population prescribed opioids, pain severity and interference were independently associated with increased use of non-medical opioids, sedatives, and tobacco as well as decreased use of alcohol. Moreover, pain severity was associated with decreased odds of alcohol use and increased odds of tobacco use among women but not men. If replicated, these findings can be used to guide the development of gender-sensitive, evidence-based strategies to ameliorate or prevent substance abuse or dependence among individuals living with painful conditions.

Highlights.

• Use of non-medical opioids, sedatives, and tobacco increased with increasing pain

• Alcohol use decreased with increasing pain

• Pain severity was linked to reduced alcohol use among women

• Pain interference was linked to increased tobacco use among women

Perspective.

This study suggests that pain severity and interference are associated with increased use of non-medical prescription opioids, sedatives, and tobacco and decreased use of alcohol, in ways that are different between women and men. Findings may guide the development of gender-sensitive evidence-based strategies to ameliorate or prevent substance misuse among patients living with pain.