Abstract
Nationally representative rates of incident prescription opioid use in the United States adult population and selected subpopulations are unknown. Using the National Health Interview Survey (2019–2020) longitudinal cohort, a cohort with 1-year follow-up created using random cluster probability sampling of noninstitutionalized civilian U.S. adults, we estimated rates and predictors of incident opioid use. Of 21,161 baseline (2019) participants randomly chosen for follow-up, the final analytic sample included 10,415 who also participated in 2020. Exposure variables were selected per the socio-behavioral model of health care utilization: predisposing characteristics (sex, age, race, etc), enabling characteristics (socioeconomic status, insurance status), health status (pain, disability, comorbidities, etc), and health care use (office visits, emergency room visits, and hospitalizations). Among adults who did not use prescription opioids in 2019, a 1-year cumulative incidence of 4.1% (95% confidence interval [CI]: 3.5–4.6) was seen in 2020, with an incidence rate (IR) of 32.6 cases of new prescription opioid use per 1,000 person-years (PYs). Cumulative incidence, IR, and adjusted relative risk (RR) varied by participant characteristics. We observed the highest IR in those with ineffective pain treatment (81.6 cases per 1,000 PY) and those who visited the emergency room ≥3 times (93.8 cases per 1,000 PY). Participants reporting ≥4 painful conditions had an adjusted RR of 2.9 (95% CI: 2.0–4.1), while the RR for those with sleep problems was 2.3 (95% CI: 1.7–3.1). Overall, this study presents nationally representative rates of incident prescription opioid use and suggests that some participants are using prescription opioids as an early-resort analgesic contrary to best-practice guidelines.
Keywords: Opioid use, incidence, cohort study, epidemiology, chronic pain
Perspective:
This longitudinal cohort study presents nationally representative rates of incident prescription opioid use in U.S. adults and selected subpopulations. Our data suggest that some participants are using prescription opioids as a first-line or early-resort analgesic, contrary to best-practice guidelines.
As part of a public health effort to improve pain management, the use of opioids increased substantially.1 Inadvertently, this change contributes to a public health crisis in the United States. In 2015, over a third (37.8%) of 91.8 million noninstitutionalized U.S. adults used prescription opioids,2 and at least 2 million people reported opioid use disorder.3 Despite sustained public health efforts, opioid use, dependence, and overdose are persistent and warrant prioritization in public health research.
Studies of prescription opioid use have predominantly focused on prevalence4,5 with few examining incidence (new use).6-10 Prior incidence studies are based on sources not representative of the U.S. adult population, such as insurance claims data and nonsurvey sources. Only one study8 presented relative risk (RR) by sex, age, and race. None of the studies considered Hispanic heritage, socioeconomic status (SES), access to care, and underlying comorbidities. Therefore, in the present study, we provide the first nationally representative rates of incident prescription opioid use in the general U.S. adult population and selected subpopulations using data from the 2019 to 2020 National Health Interview Survey (NHIS) Longitudinal Cohort (NHIS-LC). The members of this cohort were chosen using random cluster probability sampling of noninstitutionalized civilian U.S. adults and interviewed twice, once in 2019 and once in 2020. Those who reported opioid use in 2020 but not in 2019 were considered to have incident opioid use.
To identify factors that predict incident prescription opioid use, we employed the Andersen socio-behavioral model of health care utilization.11-13 This framework considers 5 variable sets posited to interact and influence the use of health services: the external environment, predisposing factors, enabling factors, health need measures, and personal health behaviors. The model predicts that health needs are the most direct cause of health service use, followed by enabling and predisposing factors.14,15 Therefore, we hypothesized that those with greater health care needs will be more likely to use prescription opioids after controlling for other variables in the socio-behavioral model. We also explored the relative contribution of predisposing (age, sex, race, Hispanic heritage, etc), enabling (ability to pay bills, health insurance coverage), external environment (U.S. region, urban vs rural), and personal health behavior (smoking) factors in an individual’s choice to begin to use prescription opioids. Identifying the factors associated with an individual’s decision to start using prescription opioids in a nationally representative population can help inform targeted outreach strategies to optimize pain management plans.
Methods
Ethics
Data collection received approval from the National Center for Health Statistics (NCHS) Research Ethics Review Board and was exempted by the Mayo Clinic Institutional Review Board. Verbal consent was obtained from all respondents.
Study Population
The NHIS is a nationally representative, annual survey of the U.S. civilian, noninstitutionalized population from 50 states and the District of Columbia.16 The NHIS is conducted by the NCHS and uses a multistage clustered sample design, which oversamples Black, Asian, and Hispanic populations. Applying NCHS-derived sampling weights allows an accurate extrapolation of findings to the U.S. civilian, noninstitutionalized population.16
Of the 31,997 NHIS participants in 2019, 21,161 were randomly chosen by the NCHS for possible inclusion in the NHIS-LC. Of these, 1,746 were excluded due to proxy responses or a lack of contact information, and 334 had died or resided in an institutional setting (Fig 1). Of the remaining 19,081, 10,415 participated in 2020 and were included in this analysis. This report follows the STROBE guideline for cohort studies.17
Figure 1.
Of the 31,997 NHIS participants in 2019, 21,161 were randomly chosen by the NCHS for possible inclusion in the NHIS-LC.18 Of these, 1,746 were excluded due to proxy responses or a lack of contact information, and 334 died or resided in an institutional setting. Of the remaining 19,081, 10,415 agreed to participate in 2020 and were included in this analysis.
Operational Definitions of Opioid Use in 2019 and 2020
For NHIS years 2019 to 2020, sample adults were asked 2 questions on opioid use: “During the past 12 months, have you taken any opioid pain relievers prescribed by a doctor, dentist, or other health professional? Respondents who responded ‘yes’ to this question were then asked: “during the past 3 months, have you taken any opioid pain relievers prescribed by a doctor, dentist, or other health professional?”.
Incidence is defined as new health care use (eg, opioid use) that occurs in a population that has not previously used the specified health care service but is at risk. In the present study, incidence estimates were based on the 395 individuals who reported not using any opioids in 2019 but reported opioid use within the previous 3 months in 2020. Missing data are presented as “Unknown.”
All NHIS questions used in this study with final coding for analyses are reported in Supplementary Appendix 1.
Coding of 2019 Variables in the Andersen Model
All questionnaire items included as covariates in our analysis are described in detail in Supplementary Appendix 2. For baseline predisposing factors, we used those previously associated with prevalent opioid use19-21: age, sex, race, graduation status, and Hispanic/Latino heritage. Self-reported race was coded as American Indian/Alaska Native, Asian, Black/African American, White, and Other. We used White race and non-Hispanic/Latino ethnicity as reference groups for comparisons, conceptualizing differences as the result of structural, interpersonal, and internalized racism. Missing data were recoded as “unknown,” which is presented as such in the relevant tables and included in all analyses, unless otherwise noted. No attempt was made to impute missing data. In addition, some analyzed variables were quantitative, such as age, number of comorbidities, number of painful health conditions, number of disabilities, and number of emergency room (ER) visits. However, the distribution of age was not a normal distribution, and certain other quantitative variables were highly skewed toward 0. Therefore, these variables were converted to categorical data, which can lead to loss of some statistical power, but greatly supplement data interpretation.
As surrogate measures of the external environment, we used U.S. region of residence and whether the participant resided in a rural or urban setting.
Our 2 2019 enabling characteristics were: problems paying medical bills and health insurance status. Problems paying bills is a more proximal SES variable than household income, better reflects the material conditions in which a person lives, and is considered a measure of financial stress, also known as financial fragility, financial strain, and financial distress. Difficulty paying bills and other measures of financial stress are related to poor health independent of education attainment or income.22
Health needs in 2019 were defined using the following NHIS variables: sleep quality, ever diagnosed with anxiety, and ever diagnosed with depression. We also coded the number of painful health conditions in the last 3 months (0, 1, 2, 3+) from the following list: arthritis, back pain, mouth or jaw pain, shoulder/arm/hand pain, hips/knees/feet pain, headache/migraine, and abdominal/pelvic/genital pain. As a measure of chronic disease burden, the number of medical comorbidities was counted (0, 1, 2, 3, 4+) from the following list: asthma attack (last 12 months), obesity, hypertension, hypercholesterolemia, diabetes, ever cancer, ever chronic obstructive pulmonary disease, ever coronary artery disease, and ever stroke. Further, we counted the number of disabilities based on responses assessing difficulty with climbing steps, self-care, limitations in work from physical, mental, or emotional problems, and difficulty in participating in social activities.
Measures of health care utilization in 2019 included ER visit (past 12 months), hospitalization at least overnight (past 12 months), number of nonopioid pain treatments (past 3 months), and participant-reported effectiveness of all pain treatments in 2019.
In the 2019 NHIS, all participants who reported pain were asked consecutive questions in the following format: “Over the past three months, did you use any of the following to manage your pain?” The survey then presented several options, including opiates, physical therapy, rehabilitative therapy, occupational therapy, spinal manipulation or chiropractic care, cognitive behavioral therapy, self-management program or workshop, peer-support groups, yoga or tai-chi, massage, meditation, guided imagery or other relaxation techniques, and “other” approaches for pain management. Although we cannot extract exactly what approach led participants to select the “other” category, it seems likely that survey participants were considering use of pain management approaches such as over-the-counter analgesics and nonopioid analgesics that were not specifically mentioned in the preceding questions. All these questions related to management of pain are incorporated into the regression model as a count variable (eg, number of nonopioid pain treatments by cohort participants at baseline [2019]).
Operational Definitions of Nonopioid Medication Use in 2019 and 2020
For comparison purposes, we calculated the 1-year cumulative incidence and rates of use per 1,000 PY for several other groups of prescribed drugs that were captured in both the 2019 and 2020 NHIS: antianxiety drugs, anti-depressant drugs, cholesterol-lowering drugs, anti-hypertensive drugs, and hypoglycemic agents.
Statistical Analysis
In Table 1, we used absolute standard difference values to compare the baseline characteristics for the complete analytic sample (ie, all 10,415 participants in the NHIS-LC) to those 21,582 participants in the 2019 NHIS not included in the longitudinal cohort (ie, those not randomized for invitation, randomized but found ineligible, and randomized but declined).
Table 1.
Baseline Characteristics of 2019 NHIS Adult Participants in the 2019 to 2020 Longitudinal Cohort
| 2019 ENROLLED (LONGITUDINAL WEIGHTS) | 2019 ENROLLED (SAMPLE ADULT WEIGHTS) | |||||
|---|---|---|---|---|---|---|
| VARIABLE | RAW FREQUENCY |
WEIGHTED FREQUENCY (1000's) |
WEIGHTED % (95% CI) |
RAW FREQUENCY |
WEIGHTED FREQUENCY (1000's) |
WEIGHTED % (95% CI) |
| Total | 10,415 | 250,917 | 100 | 10,415 | 75,489 | 100 |
| Age group (yrs) | ||||||
| 18 to 44 | 3,441 | 115,265 | 45.9 (44.4–47.5) | 3,441 | 30,899 | 40.9 (39.5–42.4) |
| 45 to 64 | 3,569 | 82,655 | 32.9 (31.6–34.2) | 3,596 | 26,170 | 34.7 (33.5–35.9) |
| ≥65 | 3,405 | 52,996 | 21.1 (20.1–22.2) | 3,405 | 18,420 | 24.4 (23.3–25.5) |
| Sex | ||||||
| Male | 4,790 | 121,160 | 48.3 (46.9–49.7) | 4,790 | 36,516 | 48.4 (47.1–49.6) |
| Female | 5,624 | 129,741 | 51.7 (50.3–53.1) | 5,624 | 38,966 | 51.6 (50.4–52.9) |
| Unknown | 1 | UR | UR | 1 | UR | UR |
| Race/Ethnicity | ||||||
| Hispanic | 1,153 | 41,506 | 16.5 (14.7–18.4) | 1,153 | 10,785 | 14.3 (12.7–15.8) |
| White, non-Hispanic | 7,495 | 158,534 | 63.2 (60.9–65.4) | 7,495 | 50,733 | 67.2 (65.2–69.2) |
| Black, non-Hispanic | 999 | 29,678 | 11.8 (10.5–13.2) | 999 | 7,694 | 10.2 (9.1–11.3) |
| Other, non-Hispanic* | 768 | 21,198 | 8.4 (7.3–9.6) | 768 | 6,278 | 8.3 (7.3–9.4) |
| Education | ||||||
| < High school | 771 | 29,086 | 11.6 (10.5–12.7) | 771 | 7,469 | 9.9 (9.0–10.8) |
| High school or equivalent | 5,377 | 147,784 | 58.9 (57.5–60.3) | 5,377 | 41,666 | 55.2 (53.9–56.5) |
| College graduate | 4,228 | 72,259 | 28.8 (27.4–30.2) | 4,228 | 25,877 | 34.3 (32.9–35.7) |
| Unknown | 39 | 1,787 | .7 (.4–1.0) | 39 | 478 | .6 (.4–.9) |
| Employment status | ||||||
| Currently employed | 6,197 | 162,210 | 64.6 (63.2–66.0) | 6,197 | 48,009 | 63.6 (62.3–64.9) |
| Previously employed | 4,059 | 82,006 | 32.7 (31.3–34.0) | 4,059 | 25,858 | 34.3 (33.0–35.5) |
| Never worked | 149 | 6,470 | 2.6 (2.0–3.1) | 149 | 1,544 | 2.0 (1.6–2.4) |
| Unknown | 10 | UR | UR | 10 | UR | UR |
| Poverty status | ||||||
| < 100% FPL | 981 | 26,601 | 10.6 (9.6–11.6) | 981 | 7,049 | 9.3 (8.5–10.2) |
| 100% ≤FPL < 200% | 1,674 | 47,132 | 18.8 (17.6–20.0) | 1,674 | 12,493 | 16.5 (15.6–17.5) |
| 200% ≤FPL < 400% | 3,068 | 78,574 | 31.1 (30.0–32.6) | 3,068 | 22,791 | 30.2 (29.0–31.3) |
| ≥400% FPL | 4,692 | 98,610 | 39.3 (37.8–40.8) | 4,692 | 33,157 | 43.9 (42.5–45.4) |
| Veteran | ||||||
| Yes | 1,059 | 20,398 | 8.1 (7.5–8.8) | 1,059 | 6,681 | 8.9 (8.2–9.5) |
| No | 9,355 | 230,514 | 91.9 (91.2–92.5) | 9,355 | 68,806 | 91.1 (90.5–91.8) |
| Unknown | 1 | UR | UR | 1 | UR | UR |
| Health Insurance | ||||||
| None | 726 | 28,595 | 11.4 (10.3–12.5) | 726 | 6,804 | 9.0 (8.2–9.8) |
| Public† | 3,030 | 64,572 | 25.7 (24.4–27.1) | 3,030 | 19,645 | 26.0 (24.8–27.3) |
| Private‡ | 6,641 | 157,269 | 62.7 (61.2–64.2) | 6,641 | 48,916 | 64.9 (63.4–66.2) |
| Unknown | 18 | 480 | .2 (.1–3) | 18 | 124 | .2 (.1–3) |
| Region | ||||||
| Northeast | 1,789 | 44,572 | 17.8 (15.6–19.9) | 1,789 | 13,432 | 17.8 (15.9–16.7) |
| Midwest | 2,439 | 52,760 | 21.0 (18.9–23.2) | 2,439 | 17,051 | 22.6 (20.6–24.6) |
| South | 3,561 | 94,533 | 37.7 (35.0–40.4) | 3,561 | 26,436 | 35.0 (32.7–37.3) |
| West | 2,626 | 59,051 | 23.5 (21.0–26.1) | 2,626 | 18,570 | 24.6 (22.4–26.8) |
| Urban-rural | ||||||
| Urban | 8,687 | 214,938 | 85.7 (83.5–87.8) | 8,687 | 64,223 | 85.1 (83.2–87.0) |
| Rural | 1,728 | 35,978 | 14.3 (12.2–16.5) | 1,728 | 11,266 | 14.9 (13.0–16.8) |
Abbreviations: UR, unreliable; FPL, federal poverty level.
NOTE. Participants were followed up for a mean (SD) of 1.3 (.3) years.
Other includes American Indian, Alaska Native, Native Hawaiian, Other Pacific Islanders, and individuals of mixed races.
Public includes Medicare, Medicaid, military coverage, and other state and federal health plans or programs.
Private includes all individuals with private health insurance, regardless of whether they also have public health insurance.
Incident prescription opioid use is presented as both 1-year cumulative incidence and incidence per 1,000 person-years (PYs). Frequency analyses, incidence rates (IRs), and RR were generated using SAS Survey Procedures (version 9.4, SAS Institute Inc, Cary, NC). All proportion estimates in the text and tables were calculated using SAS Proc SurveyFreq weighted to the U.S. civilian, noninstitutionalized population 18 years of age and older using longitudinal weights supplied by NCHS. Variable levels with cell sample size < 15 resulted in unstable calculated proportions and odds ratios that do not meet NCHS standards for NHIS data analyses23 and were therefore indicated in tables as unreliable.
RR was calculated using SAS Proc GENMOD incorporating NHIS sampling characteristics while running multivariable Poisson regression models with robust standard errors. Adjusted RR controlled for core demographic characteristics (age, sex, race, Hispanic heritage, education, region, urban or rural, and type of health insurance). An alpha value of .05 was selected as the threshold for statistical significance.
Results
Baseline Characteristics
There was excellent balance (absolute standard difference < .1) in the core demographic distributions of 2019 NHIS participants by NHIS-LC participation status (Table 1). Among 10,415 participants, 5,624 (51.7%; 95% confidence interval [CI]: 50.3–53.1) were female; 3,405 (21.1%; 95% CI: 20.1–22.2) were ≥65 years of age. In our sample, 782 participants, representing 16.9 million adults or 6.8% (95% CI: 6.1–7.5) of the U.S. adult population, reported using prescription opioids in 2019. The remaining 9,633 participants, representing 234 million adults or 93.2% (95% CI: 92.5–93.9), had not used prescription opioids and made up our sample for analysis (Supplementary Appendix 3). Baseline characteristics are presented in Table 2. The distribution of core demographic characteristics over time showed very little change, which is expected for immutable demographic variables (Supplementary Appendix 4a-4f).
Table 2.
One-Year Cumulative Incidence of 3-Month Prescription Opioid Use in 2020 for Those Not Using Prescription Opioids in 2019
| NO PRESCRIPTION OPIOID USE IN 2020 | ANY PRESCRIPTION OPIOID USE IN 2020 | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| TYPE OF COVARIATES | VARIABLE | RAW FREQUENCY |
WEIGHTED FREQUENCY (THOUSANDS) |
WEIGHTED SE (THOUSANDS) |
WEIGHTED PY (THOUSANDS) |
RAW FREQUENCY |
WEIGHTED FREQUENCY (THOUSANDS) |
WEIGHTED SE (THOUSANDS) |
WEIGHTED PY (THOUSANDS) |
1-YEAR CUMULATIVE INCIDENCE (%) (95% CI) |
| Predisposing | Total | 9,238 | 224,470 | 6,461 | 279,885 | 395 | 9,505 | 735 | 11,861 | 4.1 (3.5–4.6) |
| Age group (years) | ||||||||||
| 18 to 44 | 3,218 | 107,525 | 3,921 | 134,087 | 102 | 3,674 | 538 | 4,655 | 4.1 (2.4–4.2) | |
| 45 to 64 | 3,087 | 71,331 | 2,519 | 89,106 | 151 | 3,634 | 408 | 4,495 | 4.8 (3.8–5.8) | |
| ≥65 | 2,933 | 45,614 | 1,574 | 56,692 | 142 | 2,197 | 242 | 2,710 | 4.6 (3.7–5.5) | |
| Sex | ||||||||||
| Male | 4,348 | 110,291 | 3,643 | 137,604 | 149 | 3,853 | 468 | 4,805 | 3.9 (2.6–4.1) | |
| Female | 4,889 | 114,163 | 3,675 | 142,266 | 246 | 5,652 | 524 | 7,056 | 4.7 (3.9–5.6) | |
| Unknown | 1 | 16 | UR | UR | 0 | UR | ||||
| Race | ||||||||||
| White | 7,204 | 161,797 | 5,055 | 201,987 | 320 | 6,966 | 630 | 8,738 | 4.1 (3.5–4.9) | |
| Black | 889 | 270,355 | 1,750 | 33,558 | 44 | 1,369 | 275 | 1,734 | 4.8 (3.0–6.6) | |
| Asian | 489 | 13,785 | 1,097 | 17,212 | 10 | 453 | 200 | 562 | 3.2 (.4–5.9) | |
| Native American/Alaskan Native | 160 | 4,015 | 981 | 5,030 | 8 | 269 | 119 | 323 | 6.3 (2.8–9.7) | |
| Other | 99 | 2,813 | 408 | 3,476 | 3 | 50 | 29 | 62 | 1.7 (.0–3.8) | |
| Unknown | 388 | 15,025 | 1,280 | 18,622 | 10 | 398 | 169 | 441 | 2.6 (.5–4.7) | |
| Hispanic heritage | ||||||||||
| Mexican | 616 | 23,152 | 1,984 | 29,126 | 23 | 990 | 277 | 1,227 | 4.1 (2.0–6.2) | |
| Other Hispanic | 423 | 14,919 | 1,383 | 18,513 | 18 | 432 | 127 | 543 | 2.8 (1.3–4.4) | |
| Not Hispanic | 8,190 | 186,307 | 5,691 | 232,129 | 353 | 8,062 | 658 | 10,065 | 4.1 (3.6–4.7) | |
| Unknown | 9 | 91 | 35 | 117 | 1 | 20 | UR | UR | ||
| Education | ||||||||||
| < High school | 647 | 24,863 | 1,494 | 30,597 | 32 | 1,438 | 334 | 1,862 | 5.5 (3.1–7.8) | |
| High school or GED | 4,683 | 130,966 | 4,347 | 163,444 | 212 | 5,669 | 546 | 7,109 | 4.1 (3.4–4.9) | |
| College graduate | 3,874 | 67,075 | 2,357 | 83,845 | 148 | 2,276 | 248 | 2,748 | 3.3 (2.6–4.0) | |
| Unknown | 34 | 1,565 | 333 | 1,999 | 3 | 122 | 72 | 141 | 6.6 (.0–13.9) | |
| Veteran | ||||||||||
| Yes | 929 | 17,797 | 897 | 22,204 | 45 | 844 | 194 | 1,140 | 4.7 (2.8–6.7) | |
| No | 8,308 | 206,668 | 6,047 | 257,676 | 350 | 8,620 | 716 | 10,720 | 4.0 (3.4–4.6) | |
| Unknown | ||||||||||
| Marital status | ||||||||||
| Divorced | 2,317 | 35,115 | 1,351 | 43,739 | 113 | 2,031 | 285 | 2,666 | 2.5 (1.7–3.2) | |
| Never | 1,748 | 50,671 | 2,238 | 63,214 | 63 | 1,953 | 381 | 2,445 | 4.0 (2.9–5.0) | |
| Married or living together | 5,142 | 135,185 | 4,307 | 172,319 | 218 | 5,510 | 514 | 6,850 | 4.9 (3.8–5.9) | |
| Unknown | 31 | 498 | 124 | 613 | 1 | 11 | UR | 11 | UR | |
| Enabling | Problems paying bills | |||||||||
| No | 8,296 | 196,747 | 5,742 | 245,507 | 312 | 7,117 | 575 | 8,803 | 3.5 (3.0–4.0) | |
| Yes | 942 | 27,723 | 1,408 | 34,378 | 83 | 2,388 | 387 | 3,058 | 7.9 (5.6–10.3) | |
| Health insurance | ||||||||||
| Uninsured | 692 | 27,363 | 154 | 33,995 | 14 | 478 | 221 | 632 | 1.7 (.2–3.2) | |
| Public | 2,526 | 54,149 | 2,164 | 67,158 | 134 | 3,001 | 400 | 3,763 | 5.3 (4.0–6.5) | |
| Private | 6,004 | 142,526 | 4,574 | 178,179 | 245 | 5,976 | 539 | 7,396 | 4.0 (3.3–4.7) | |
| Unknown | 16 | 431 | 134 | 554 | 2 | 50 | 36 | 70 | UR | |
| External environment | Region | |||||||||
| Northeast | 1,637 | 42,572 | 2,887 | 51,903 | 51 | 1,059 | 176 | 1,278 | 2.5 (1.7–3.2) | |
| Midwest | 2,174 | 47,408 | 2,753 | 59,135 | 91 | 1,959 | 288 | 2,391 | 4.0 (2.9–5.0) | |
| South | 3,088 | 82,368 | 3,865 | 102,384 | 157 | 4,230 | 522 | 5,346 | 4.9 (3.8–5.9) | |
| West | 2,339 | 53,122 | 3,301 | 66,463 | 96 | 2,258 | 393 | 2,827 | 4.1 (2.8–5.3) | |
| Urban-rural | ||||||||||
| Urban | 7,746 | 193,209 | 5,896 | 241,189 | 327 | 8,017 | 675 | 9,998 | 4.0 (3.4–4.6) | |
| Rural | 1,492 | 31,261 | 2,642 | 38,698 | 68 | 1,488 | 289 | 1,864 | 4.5 (3.1–6.0) | |
| Health behavior | Smoking status | |||||||||
| Current | 1,032 | 27,489 | 1,299 | 34,132 | 64 | 1,692 | 286 | 2,126 | 4.1 (2.0–6.2) | |
| Former | 2,404 | 49,296 | 1,861 | 61,580 | 114 | 2,004 | 226 | 2,477 | 2.8 (1.3–4.4) | |
| Never | 5,794 | 147,493 | 4,655 | 183,932 | 216 | 5,794 | 606 | 7,246 | 4.1 (3.6–4.7) | |
| Unknown | 8 | 192 | UR | 1 | 11 | UR | UR | |||
| Health status | Trouble with sleeping | |||||||||
| No or infrequent | 8,079 | 197,668 | 5,811 | 246,150 | 299 | 7,098 | 597 | 8,793 | 3.5 (2.9–4.0) | |
| Most or every day | 1,131 | 25,927 | 1,246 | 32,311 | 93 | 2,350 | 369 | 3,002 | 8.3 (6.0–10.7) | |
| Unknown | 28 | 874 | 245 | 1,125 | 3 | 58 | UR | 65 | UR | |
| Ever diagnosed with anxiety | ||||||||||
| No | 8,046 | 196,616 | 5,773 | 245,135 | 301 | 7,260 | 612 | 9,015 | 3.6 (3.0–4.1) | |
| Yes | 1,186 | 27,739 | 1,376 | 34,613 | 93 | 2,227 | 323 | 2,828 | 7.4 (5.5–9.4) | |
| Unknown | 6 | 114 | UR | 138 | 1 | 17 | UR | 17 | UR | |
| Ever diagnosed with depression | ||||||||||
| No | 7,795 | 192,188 | 5,681 | 239,723 | 283 | 6,929 | 616 | 8,594 | 3.5 (2.9–4.0) | |
| Yes | 1,434 | 32,139 | 1,446 | 39,984 | 112 | 2,575 | 349 | 3,267 | 7.4 (5.6–9.2) | |
| Unknown | 9 | 143 | UR | 178 | 0 | |||||
| Number of painful health conditions | ||||||||||
| 0 | 3,702 | 100,128 | 3,398 | 125,211 | 82 | 2,527 | 409 | 3,130 | 2.5 (1.7–3.2) | |
| 1 | 1,299 | 28,675 | 1,299 | 35,508 | 44 | 934 | 195 | 1,167 | 3.2 (1.9–4.4) | |
| 2 | 1,350 | 30,668 | 1,502 | 38,349 | 55 | 1,190 | 190 | 1,534 | 3.7 (2.6–4.8) | |
| 3 | 1,276 | 28,907 | 1,231 | 35,840 | 73 | 1,660 | 260 | 2,073 | 5.4 (3.8–7.0) | |
| 4+ | 1,611 | 36,093 | 1,671 | 44,977 | 141 | 3,194 | 393 | 3,956 | 8.1 (6.4–9.9) | |
| Disabilities | ||||||||||
| 0 | 6,393 | 162,104 | 4,963 | 202,085 | 207 | 4,834 | 476 | 6,015 | 2.9 (2.4–3.4) | |
| 1 | 1,788 | 41,006 | 1,745 | 51,181 | 97 | 2,606 | 386 | 3,306 | 6.0 (4.4–7.6) | |
| 2+ | 1,057 | 21,360 | 1,124 | 26,619 | 91 | 2,064 | 314 | 2,540 | 8.8 (6.4–11.2) | |
| Asthma in the last 12 mo | ||||||||||
| No | 8,530 | 207,947 | 6,008 | 259,402 | 340 | 7,830 | 630 | 9,734 | 3.6 (3.1–4.2) | |
| Yes | 698 | 16,340 | 991 | 20,257 | 55 | 1,675 | 351 | 2,127 | 9.3 (5.7–12.9) | |
| Unknown | 10 | 183 | 67 | 0 | UR | |||||
| # of comorbidities | ||||||||||
| None | 3,345 | 94,596 | 3,298 | 118,074 | 97 | 2,992 | 435 | 3,717 | 3.1 (2.2–3.9) | |
| One | 2,579 | 64,861 | 2,418 | 80,831 | 98 | 2,529 | 364 | 3,154 | 3.8 (2.7–4.8) | |
| Two | 1,651 | 33,203 | 1,392 | 41,280 | 83 | 1,672 | 233 | 2,088 | 4.8 (3.6–6.0) | |
| Three | 1,003 | 19,876 | 997 | 24,837 | 55 | 1,164 | 220 | 1,483 | 5.5 (3.6–7.4) | |
| Four or more | 660 | 11,933 | 715 | 14,864 | 62 | 1,148 | 178 | 1,419 | 8.8 (6.2–11.3) | |
| Health care use | Saw a physician in the last 12 mo | |||||||||
| No | 1,179 | 36,982 | 1,840 | 46,540 | 20 | 449 | 127 | 540 | 1.2 (.5–1.9) | |
| Yes | 8,052 | 187,313 | 5,392 | 233,1 16 | 374 | 8,968 | 714 | 11,211 | 4.6 (3.9–5.2) | |
| Unknown | 7 | 174 | UR | 230 | 1 | 88 | UR | 110 | UR | |
| # of emergency department visits | ||||||||||
| 0 | 7,493 | 181,536 | 5,329 | 226,558 | 269 | 6,276 | 571 | 7,825 | 3.3 (2.8–3.9) | |
| 1 | 1,158 | 280,028 | 1,323 | 34,883 | 78 | 1,848 | 261 | 2,723 | 6.2 (4.5–7.9) | |
| 2 | 426 | 10,780 | 708 | 13,370 | 33 | 850 | 195 | 1,076 | 7.3 (4.2–10.5) | |
| 3+ | 154 | 3,971 | 459 | 4,879 | 14 | 521 | 192 | 678 | 11.6 (4.7–18.5) | |
| Unknown | 7 | 155 | 65 | 194 | 1 | 10 | UR | 10 | UR | |
| Hospitalized in the last year | ||||||||||
| No | 8,500 | 209,502 | 6,116 | 261,414 | 339 | 8,157 | 673 | 10,192 | 3.7 (3.2–4.3) | |
| Yes | 734 | 14,829 | 787 | 18,297 | 56 | 1,347 | 236 | 1,668 | 8.3 (5.6–11.0) | |
| Unknown | 4 | 139 | UR | - | UR | |||||
| # of pain treatments | ||||||||||
| None | 5,848 | 152,099 | 4,713 | 189,908 | 196 | 5,266 | 533 | 6,577 | 3.3 (2.7–4.0) | |
| One | 1,954 | 42,621 | 1,810 | 53,095 | 108 | 2,541 | 343 | 3,174 | 5.6 (4.2–7.0) | |
| Two | 837 | 17,089 | 926 | 21,233 | 48 | 881 | 172 | 1,193 | 4.9 (3.1–6.7) | |
| Three or more | 599 | 12,661 | 784 | 15,650 | 43 | 817 | 160 | 1,022 | 6.1 (3.8–8.3) | |
| Effectiveness of pain treatment in 2019 | ||||||||||
| No pain | 3,640 | 96,952 | 3,366 | 121,196 | 85 | 2,506 | 397 | 3,136 | 2.5 (1.8–3.3) | |
| Pain, no 2019 treatment | 1,908 | 48,710 | 1,991 | 60,681 | 86 | 2,102 | 318 | 2,622 | 4.8 (3.7–5.9) | |
| Not effective | 348 | 8,140 | 597 | 10,152 | 34 | 925 | 200 | 1,178 | 10.2 (6.3–14.2) | |
| Somewhat effective | 1,342 | 29,603 | 1,397 | 36,868 | 95 | 1,852 | 262 | 2,236 | 5.9 (4.3–7.5) | |
| Very effective | 1,701 | 34,646 | 1,579 | 42,973 | 70 | 1,461 | 258 | 1,870 | 4.0 (2.7–5.4) | |
| Unknown | 229 | 6,419 | 552 | 8,016 | 25 | 657 | 175 | 818 | 9.3 (4.7–13.9) | |
Abbreviations: SE, standard error; UR, unreliable; GED, general educational diploma.
NOTE. Participants were followed up for a mean (SD) of 1.3 (.3) years. The degrees of freedom in computing the confidence limits is 549.
Incident Prescription Opioid Use in 2020 by Participant 2019 Characteristics
We found a 1-year cumulative incidence of 4.1% (95% CI: 3.5–4.6; all percentages are weighted; Table 2) and an IR of 32.6 cases per 1,000 PY of follow-up. The 1-year cumulative incidence varied considerably by 2019 participant characteristics, from lows of 1.7% (95% CI: .2–3.2) in those without health insurance in 2019 and 1.2% (95% CI: .5–1.9) in those who had not seen a physician in 2019, to highs of 10.2% (95% CI: 6.3–14.2) in those reporting that their pain treatment in 2019 was not effective and 11.6% (95% CI: 4.7–18.5) for those who had gone to the ER at least 3 times in 2019. The IRs per 1,000 PY similarly varied by 2019 participant characteristics.
Unadjusted and adjusted RRs of cases by participant characteristics are presented in Table 3. Many predisposing factors, including sex, race, Hispanic heritage, veteran status, health insurance status, and marital status, were not associated with significant risk. Those with less than a high school education were 80% more likely to report incident prescription opioid use than college graduates (adjusted RR 1.8; 95% CI: 1.1–2.9). Per the external environment, differences were seen by region with those living in the Midwest (adjusted RR 1.5; 95% CI: 1.0–2.3), South (adjusted RR: 2.0; 95% CI: 1.4–2.9), and West (adjusted RR 1.7; 95% CI: 1.1–2.6) at greater risk than those living in the Northeast. However, no differences were seen when comparing an urban environment to a rural setting. Our single measure of health behaviors, smoking status, was not associated with increased risk of incident opioid use. Among enabling factors, health insurance status was not associated with incident opioid use. For our single measure of SES, those having problems paying bills had 2.3 times the risk of incident opioid use as those without such problems (adjusted RR 2.3; 95% CI: 1.7–3.2).
Table 3.
IRs and RR of 3-Month Prescription Opioid Use in 2020 for Those Not Using Prescription Opioids in 2019
| TYPE OF COVARIATES | VARIABLE | IR PER PY | IR PER 1,000 PY (95% CI) | UNADJUSTED RR (95% CI) |
ADJUSTED RR* (95% CI) |
|---|---|---|---|---|---|
| Predisposing | Total | .033 | 32.58 (30.06–35.10) | ||
| Age group (years) | |||||
| 18 to 44 | .026 | 26.48 (22.60–30.36) | Ref | Ref | |
| 45 to 64 | .039 | 38.82 (34.47–43.18) | 1.54 (1.07–2.20) | 1.45 (1.02–2.04) | |
| ≥65 | .037 | 36.99 (32.91–41.06) | 1.43 (1.01–2.03) | 1.13 (.75–1.69) | |
| Sex | |||||
| Male | .027 | 27.06 (23.77–30.34) | .75 (.56–1.00) | .76 (.57–1.00) | |
| Female | .038 | 37.85 (34.34–41.36) | Ref | Ref | |
| Unknown | UR | UR | UR | ||
| Race | |||||
| White | .033 | 33.06 (30.07–36.05) | Ref | Ref | |
| Black | .039 | 38.79 (31.00–46.58) | 1.21 (.81–1.82) | 1.10 (.72–1.69) | |
| Asian | .025 | 25.49 (14.23–36.74) | .79 (.33–1.89) | .85 (.35–2.09) | |
| Native American/Alaskan Native | .050 | 50.25 (28.02–72.48) | 1.56 (.90–2.70) | 1.35 (.79–2.30) | |
| Other | .014 | 14.13 (5.94–22.33) | .43 (.13–1.42) | .45 (.14–1.51) | |
| Unknown | .021 | 20.88 (12.01–29.74) | .66 (.29–1.54) | .67 (.26–1.74) | |
| Hispanic heritage | |||||
| Mexican | .033 | 32.62 (23.49–41.74) | .87 (.51–1.47) | 1.12 (.62–2.03) | |
| Other Hispanic | .023 | 22.67 (16.01–29.33) | .71 (.35–1.45) | .83 (.44–1.56) | |
| Not Hispanic | .033 | 33.29 (30.57–36.00) | Ref | Ref | |
| Unknown | UR | UR | UR | ||
| Education | |||||
| < High school | .044 | 44.30 (34.01–54.59) | 1.72 (1.06–2.77) | 1.78 (1.08–2.92) | |
| High school or GED | .033 | 33.24 (30.04–36.44) | 1.28 (.98–1.66) | 1.30 (.98–1.73) | |
| College graduate | .026 | 26.28 (23.42–29.15) | Ref | Ref | |
| Unknown | .057 | 57.01 (23.36–90.65) | 2.23 (.70–7.14) | 2.61 (.80–8.51) | |
| Veteran | |||||
| Yes | .036 | 36.15 (27.84–44.47) | 1.18 (.76–1.84) | 1.20 (.74–1.94) | |
| No | .032 | 32.12 (29.45–34.78) | Ref | Ref | |
| Unknown | UR | UR | UR | ||
| Marital status | |||||
| Divorced | .044 | 43.77 (37.63–49.91) | 1.36 (1.09–1.70) | 1.21 (.89–1.66) | |
| Never | .030 | 29.74 (23.94–35.55) | .73 (.47–1.15) | 1.02 (.65–1.62) | |
| Married or living together | .031 | 30.75 (27.88–33.62) | Ref | Ref | |
| Unknown | UR | UR | UR | ||
| Enabling | Health insurance | ||||
| Uninsured | .014 | 13.80 (7.42–20.19) | .43 (.17–1.05) | .36 (.14–.89) | |
| Public | .042 | 42.31 (36.67–47.95) | 1.31 (.98–1.77) | 1.16 (.85–1.58) | |
| Private | .032 | 32.20 (29.30–35.11) | Ref | Ref | |
| Unknown | UR | UR | UR | ||
| Problems paying bills | |||||
| No | .028 | 27.84 (25.50–30.18) | Ref | Ref | |
| Yes | .064 | 63.79 (53.45–74.13) | 2.35 (1.69–3.26) | 2.30 (1.65–3.20) | |
| External environment | Region | ||||
| Northeast | .020 | 19.91 (16.60–23.22) | Ref | Ref | |
| Midwest | .032 | 31.84 (27.16–36.52) | 1.56 (1.06–2.31) | 1.54 (1.03–2.30) | |
| South | .039 | 39.26 (34.42–44.11) | 1.97 (1.37–2.83) | 2.00 (1.39–2.89) | |
| West | .033 | 32.59 (26.92–38.26) | 1.64 (1.07–2.52) | 1.67 (1.07–2.62) | |
| Urban-rural | |||||
| Urban | .032 | 31.92 (29.23–34.60) | Ref | Ref | |
| Rural | .037 | 36.68 (29.56–43.81) | 1.16 (.82–1.64) | .99 (.70–1.40) | |
| Health behavior | Smoking status | ||||
| Current | .047 | 46.67 (38.78–54.55) | 1.58 (1.09–2.30) | 1.42 (.93–2.17) | |
| Former | .031 | 31.28 (27.76–34.81) | 1.02 (.75–1.39) | .94 (.69–1.28) | |
| Never | .030 | 30.31 (27.14–33.48) | Ref | Ref | |
| Unknown | UR | UR | UR | ||
| Health status | Trouble with sleeping | ||||
| No or infrequent | .028 | 27.84 (25.50–30.18) | Ref | Ref | |
| Most or every day | .067 | 66.55 (56.10–77.00) | 2.41 (1.76–3.30) | 2.27 (1.68–3.07) | |
| Unknown | UR | UR | UR | ||
| Ever diagnosed with anxiety | |||||
| No | .029 | 28.57 (26.16–30.97) | Ref | Ref | |
| Yes | .059 | 59.48 (50.85–68.11) | 2.07 (1.54–2.79) | 1.98 (1.49–2.63) | |
| Unknown | UR | UR | UR | ||
| Ever diagnosed with depression | |||||
| No | .028 | 27.90 (25.42–30.38) | Ref | Ref | |
| Yes | .060 | 59.54 (51.47–67.61) | 2.12 (1.59–2.84) | 1.96 (1.48–2.59) | |
| Unknown | UR | UR | UR | ||
| Number of painful health conditions | |||||
| 0 | .020 | 19.69 (16.50–22.88) | Ref | Ref | |
| 1 | .025 | 25.47 (20.15–30.78) | 1.18 (.68–2.05) | 1.12 (.66–1.92) | |
| 2 | .030 | 29.84 (25.07–34.60) | 1.50 (.97–2.33) | 1.44 (.96–2.17) | |
| 3 | .044 | 43.78 (36.93–50.64) | 2.14 (1.41–3.23) | 1.96 (1.32–2.91) | |
| 4+ | .065 | 65.27 (57.24–73.30) | 3.26 (2.23–4.76) | 2.86 (2.00–4.09) | |
| Disabilities | |||||
| 0 | .023 | 23.23 (20.94–25.52) | Ref | Ref | |
| 1 | .048 | 47.83 (40.74–54.91) | 2.07 (1.47–2.91) | 1.94 (1.39–2.72) | |
| 2+ | .071 | 70.78 (60.02–81.55) | 3.03 (2.19–4.21) | 2.60 (1.82–3.72) | |
| Asthma in the last 12 mo | |||||
| No | .029 | 29.09 (26.75–31.43) | Ref | Ref | |
| Yes | .075 | 74.83 (59.15–90.51) | 2.58 (1.70–3.92) | 2.39 (1.61–3.53) | |
| Unknown | UR | UR | |||
| # of comorbidities | |||||
| None | .025 | 24.57 (20.99–28.14) | Ref | Ref | |
| One | .030 | 30.1 1 (25.78–34.45) | 1.29 (.87–1.91) | 1.23 (.83–1.81) | |
| Two | .039 | 38.55 (33.18–43.93) | 1.61 (1.10–2.35) | 1.47 (1.02–2.13) | |
| Three | .044 | 44.22 (35.87–52.58) | 1.79 (1.12–2.84) | 1.53 (.95–2.48) | |
| Four or more | .071 | 70.50 (59.57–81.43) | 2.81 (1.85–4.25) | 2.31 (1.47–3.65) | |
| Health care use | Saw a physician in the last 12 mo | ||||
| No | .010 | 9.54 (6.84–12.23) | Ref | Ref | |
| Yes | .037 | 36.70 (33.78–39.63) | 3.76 (2.16–6.55) | 3.04 (1.83–5.06) | |
| Unknown | UR | UR | UR | ||
| # of emergency department visits | |||||
| 0 | .027 | 26.78 (24.34–29.21) | Ref | Ref | |
| 1 | .049 | 49.14 (42.20–56.08) | 1.87 (1.35–2.58) | 1.74 (1.25–2.41) | |
| 2 | .059 | 58.84 (45.34–72.34) | 2.35 (1.48–3.72) | 2.22 (1.38–3.56) | |
| 3+ | .094 | 93.76 (59.20–128.31) | 3.58 (1.96–6.55) | 3.11 (1.75–5.53) | |
| Unknown | UR | UR | UR | ||
| Hospitalized in the last year | |||||
| No | .030 | 30.03 (27.55–32.51) | Ref | Ref | |
| Yes | .067 | 67.47 (55.65–79.29) | 2.30 (1.61–3.28) | 2.06 (1.44–2.95) | |
| Unknown | UR | UR | UR | ||
| # of pain treatments | |||||
| None | .027 | 26.80 (24.09–29.51) | Ref | Ref | |
| One | .045 | 45.16 (39.06–51.25) | 1.66 (1.22–2.27) | 1.57 (1.16–2.13) | |
| Two | .039 | 39.28 (31.62–46.95) | 1.48 (.99–2.22) | 1.48 (1.01–2.16) | |
| Three or more | .049 | 49.00 (39.41–58.60) | 1.89 (1.24–2.87) | 1.92 (1.27–2.88) | |
| Effectiveness of pain treatment in 2019 | |||||
| No pain | .020 | 20.16 (16.96–23.35) | .63 (.40–.99) | .67 (.43–1.03) | |
| Pain, no 2019 treatment | .033 | 33.21 (28.18–38.23) | 1.00 (.63–1.57) | .96 (.60–1.54) | |
| Not effective | .082 | 81.64 (63.99–99.29) | 2.63 (1.56–4.45) | 2.48 (1.47–4.20) | |
| Somewhat effective | .047 | 47.36 (40.66–54.06) | 1.43 (.90–2.25) | 1.40 (.90–2.20) | |
| Very effective | .033 | 32.58 (26.83–38.33) | Ref | Ref | |
| Unknown | .074 | 74.37 (54.56–94.18) | 2.33 (1.26–4.30) | 1.90 (1.02–3.53) |
Abbreviations: UR, unreliable; GED, general educational diploma.
NOTE. Participants were followed up for a mean (SD) of 1.3 (.3) years. The degrees of freedom in computing the confidence limits is 549.
Adjusted RR includes the following core demographic variables in the regression model: age, sex, race, Hispanic heritage, education, region, urban or rural, and type of health insurance.
As suggested by the Andersen model, health status (need) variables were consistent predictors of incident opioid use. Particularly high risks were seen in those with 4 or more painful health conditions in 2019 (adjusted RR 2.9; 95% CI: 2.0–4.1), those with 2 or more disabilities (adjusted RR 2.6; 95% CI: 1.8–3.7), those with asthma in the last 12 months (adjusted RR 2.3; 95% CI: 1.6–3.5), those with problems sleeping most days or every day (adjusted RR 2.3; 95% CI: 1.7–3.1), and those with 4 or more medical comorbidities (adjusted RR 2.3; 95% CI: 1.5–3.7). Further, participants who ever had a diagnosis of depression or a diagnosis of anxiety had a higher RR of incident prescription opioid use (adjusted RR 2.0 [95% CI 1.5–2.6] for both predictors).
Health care use was also strongly associated with incident prescription opioid use: Participants who saw a physician in 2019 were 3 times more likely to have incident prescription opioid use in 2020 than those who had not (adjusted RR 3.0; 95% CI: 1.8–5.1). Also, showing increased risk for new prescription opioid use were individuals who had 3 or more ER visits (adjusted RR 3.1; 95% CI: 1.8–5.5) or were hospitalized overnight in 2019 (adjusted RR 2.1; 95% CI: 1.4–3.0). Participants using 3 or more treatments for their pain in 2019 were almost twice as likely (adjusted RR 1.9; 95% CI: 1.3–2.9) to exhibit incident prescription opioid use in 2020, with those saying the treatments were not effective 2.5 times more likely to newly use prescription opioids compared to those who said the treatment was very effective (adjusted RR 2.5; 95% CI: 1.5–2.2). Supplementary Appendix 5 displays the top-30 adjusted RRs for incident prescription opioid use by participant characteristic.
Table 4 presents the rates of incident use of opioids and several other groups of prescription drugs. The highest rate was 43.9 new cases (per 1,000 PY) of using cholesterol-lowering medications, followed closely by a rate of 43.6 new cases (per 1,000 PY) of using antianxiety drugs. The lowest rate was for new use of hypoglycemic drugs (16.2 new cases per 1,000 PY) followed by 29.3 new cases per 1,000 PY for use of antidepressants, then new cases of opioid use at 32.6 per 1,000 PY.
Table 4.
IRs for Drugs Captured in the 2019 to 2020 NHIS Longitudinal Cohort
| NO DRUG USE IN 2020 | ANY DRUG USE IN 2020 | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| VARIABLE | RAW FREQUENCY |
WEIGHTED FREQUENCY (THOUSANDS) |
WEIGHTED SE (THOUSANDS) |
WEIGHTED PY (THOUSANDS) |
RAW FREQUENCY |
WEIGHTED FREQUENCY (THOUSANDS) |
WEIGHTED SE (THOUSANDS) |
WEIGHTED PY (THOUSANDS) |
1-YEAR CUMULATIVE INCIDENCE (%) (95% CI) |
IR PER 1,000 PY (95% CI) |
| Antianxiety drugs | 8,451 | 206,998 | 6,016 | 258,117 | 479 | 11,909 | 816 | 14,896 | 5.4 (4.7–6.0) | 43.6 (40.6–46.6) |
| Antidepressants | 8,759 | 215,102 | 6,202 | 268,353 | 341 | 8,167 | 646 | 10,133 | 3.6 (3.1–4.1) | 29.3 (27.0–31.6) |
| Cholesterol-lowering drugs | 7,632 | 197,582 | 5,971 | 246,490 | 545 | 11,446 | 742 | 14,385 | 5.5 (4.8–6.1) | 43.9 (41.0–46.7) |
| Antihypertensive drugs | 6,696 | 177,668 | 5,498 | 221,562 | 422 | 9,634 | 715 | 12,116 | 5.1 (4.4–5.8) | 41.2 (38.2–44.3) |
| Hypoglycemic drugs | 9,284 | 226,847 | 6,551 | 282,837 | 194 | 4,680 | 464 | 5,967 | 2.0 (1.6–2.4) | 16.2 (14.6–17.8) |
| Opioids | 9,238 | 224,470 | 6,461 | 279,885 | 395 | 9,505 | 735 | 11,861 | 4.1 (3.5–4.6) | 32.6 (30.1–35.1) |
Abbreviations: SE, standard error; GED, general educational diploma.
Discussion
This nationally representative cohort study showed that the 1-year cumulative incidence and IR of prescription opioid use among U.S. adults were 4.1% and 32.6 cases per 1,000 PY, respectively. Health care needs and utilization, health care status, and enabling factors displayed strong associations with higher prescription opioid IRs, aligning with the socio-behavioral model.11-13 Participants with 3 or more ER visits, physician consultation in 2019, and ineffective pain treatments had the highest prescription opioid IRs. Notably, the IR of new opioid use was in the middle compared with rates of other commonly prescribed nonopioid medications.
Five previous studies examined the incidence of prescription opioid use, each with different data sources and design.6-10 One examined national health insurance claims from a single insurer,8 one examined North Carolina health insurance claims from a single state insurer,10 one examined data from a state-level prescription drug monitoring program,7 and one examined claims from the national Australian Pharmaceutical Benefits program.9 Only one study examined nationally representative U.S. data.6 Similar to our present study, Bernard et al6 required a 12-month opioid-free baseline period. However, they did not report cumulative incidence or IRs. They did report regression coefficients, enabling comparison with our RR data. Similar to our findings, Bernard et al6 did not see significant associations between incident opioid use and race, Hispanic heritage, or rural residence, but did see associations with being uninsured. They also reported associations between opioid use and poor mental health, aligning with our study’s observation. Our findings differed in that we did not see differences by sex or smoking status. Also, while Bernard et al6 included age, education, marital status, and U.S. region in their models, they did not report these data; we found significant variations by age, education, and U.S. region. Our study examined additional characteristics such as pain management approaches and their perceived effectiveness, both of which were strongly associated with incident opioid use. Combined, these data are consistent with best-practice guidelines, in that patients and providers may turn to prescription opioids when other less-aggressive approaches fail. Moreover, participants with multiple disabilities and comorbidities were more likely to exhibit incident prescription opioid use, aligning with evidence highlighting the association between pain and the burden of disabilities and comorbidities.24
Our analysis of health care status covariates showed that incident prescription opioid use was 2.3 times as likely in participants with sleep problems most days or every day and twice as likely in those with depression or anxiety. A potential explanation for the association with trouble sleeping is that respondents may use opioids as a sleep aid due to their sedative properties, despite evidence suggesting opioids can worsen sleep quality,25,26 increase sleep latency,25 and increase the risk for sleep-disordered breathing (eg, central apnea).26 Further, in respondents with sleep issues who already take other sedating medications to treat insomnia (eg, benzodiazepines), the initiation of opioids may have a synergistic effect on sedation and respiratory depression. Regarding the association with anxiety/depression, the bidirectional relationship between pain and mood disorders has been extensively studied.27 Participants with mental health issues may also misuse opioids to treat mood symptoms instead of pain.28 A longitudinal study found that depression predicted later opioid misuse.29
Interestingly, a diagnosis of asthma predicted incident opioid use, possibly due to use of codeine-containing antitussive syrup to suppress coughing during asthma attacks. This is consistent with state-level ER33 and national ambulatory care data34 showing that respiratory diagnoses were among the most common diagnoses where opioids were prescribed, with codeine-containing syrup being the most frequently used.33 Data from the OneFlorida Clinical Research consortium indicate that opioid antitussive use peaked in 2014 to 2015 but has been experiencing a resurgence since 2020 with asthma and COPD as frequent underlying diagnoses.35
The sole SES measure in our enabling factor domain, problems paying bills in 2019, predicted greater risk for incident opioid use in 2020. This may reflect financial hardship and inability to afford other treatments36,37 such as complementary health options and procedural interventions, many of which are often not covered by health insurance. Conversely, lacking health insurance was a protective factor associated with a 64% lower risk of new opioid use; this association may reflect lack of access to health care services, including among participants who may be appropriate candidates for opioid treatment.
Participants with less than a high school education were 80% more likely to report incident prescription opioid use than college graduates, which may be consistent with previous findings linking lower educational attainment to higher chronic pain rates.38 Moreover, participants residing in the South, West, and Midwest were 2, 1.7, and 1.5 times more likely to initiate new prescription opioid use, respectively, compared to those in the Northwest, which reflects geographic variations in opioid prescribing.39
Our findings suggest that some risk factors for incident opioid use are comparable to risk factors for opioid use disorder. Studies have consistently highlighted that risk factors for opioid use disorder include past or current substance abuse, psychiatric disease, young age, and social and familial settings that encourage misuse.40 Since our cohort study captured incident opioid use in the last 3 months, it is likely that the sample included only a few participants with opioid use disorder.
Compounding the opioid crisis, the onset of the COVID-19 pandemic during the current study time period presented another challenge.41 The convergence of 2 concurrent public health emergencies made it difficult for clinicians to address the needs of patients suffering from acute or chronic pain.41 Restrictions on in-person patient visits, disruption to social support systems, cancellation of elective interventional pain cases, and continued implementation of monitoring programs and regulations surrounding opioid prescriptions hindered pain management during the COVID-19 pandemic.41 Wainwright et al42 and Ochalek et al43 suggested that substance use and opioid overdoses during the COVID-19 era increased. Despite these 2 studies, data are lacking on the incidence of prescription opioid use and its predictors during the COVID-19 pandemic.
Limitations
First, self-reported data pose risks of recall bias, poor response rate, and inaccurate responses, and data collected twice during a 2-year period lack detailed information on the trajectory of opioid use. Second, there is potential for type-I statistical error. Third, several variables, including prescription opioid use, were abstracted as binary variables. Incorporating additional measures, including dose, might explain the observed associations more comprehensively. Fourth, while the relationship between psychiatric diagnoses and incident opioid use is important from a public health standpoint, data on some psychiatric diagnoses (eg, personality disorder, schizophrenia, and post-traumatic stress disorder) were not abstracted in the NHIS, as the 1-hour survey allows for only a limited number of questions that were determined by the NCHS, part of the Centers for Disease Control and Prevention. Fifth, data on specific pain diagnoses also were not abstracted in the NHIS. Instead, the NHIS contains global questions capturing pain frequency, severity, and related disability. These questions on pain diagnoses were developed by a broad panel of pain clinicians, scientists, psychologists, and psychiatrists as part of the U.S. Department of Health and Human Services (DHHS) National Pain Strategy and are used to track the national burden of pain through the DHHS Healthy People Initiative44 in healthy participants and as a benchmark for DHHS Best Practice Pain Management guidelines.45 Therefore, while granular details are unavailable on types of pain diagnosis, these generic questions remain a strength of the NHIS and this study as they contribute to several national initiatives to reduce the burden of pain. Sixth, while smoking status was abstracted in the NHIS, alcohol use and use of illicit substances were not abstracted.
Potential sources of bias include selection bias, inadequate capture of uninsured patients, temporal bias, and the healthy cohort effect. The NHIS partly accounts for selection bias by using sampling weights to make the cohort look like the U.S. population. Unlike an insurance claims database, the NHIS does not exclude uninsured patients. However, the NHIS does exclude those in prison, nursing homes, or otherwise institutionalized, who are likely to have worse health than the general population, and those in the military serving outside the United States. The longitudinal cohort design of the NHIS-LC accounts for potential temporal bias. Finally, participants with new serious health issues in 2020 but not in 2019 may have declined participation in the 2020 survey, resulting in a healthy cohort effect and potential underestimation of the true prescription opioid use IR.
Conclusions
To the best of our knowledge, this is the first study to present nationally representative rates of incident prescription opioid use in the U.S. adult population. This study identifies important predictors of new prescription opioid use in U.S. adults, analyzing associations with variables representative of health care utilization, health need measures, personal health behaviors, predisposing factors, and enabling factors. Although NHIS data do not provide enough information to know if the incident opioid use is high risk, our data suggest that some participants are using opioids as a first-line or early-resort analgesic, instead of following various best-practice guidelines that recommend nonpharmacologic modalities, over-the-counter medications, and other nonopioid analgesics as initial treatment for pain.
Supplementary Material
Footnotes
Supplementary data accompanying this article are available online https://doi.org/10.1016/j.jpain.2024.104665 at www.jpain.org and www.sciencedirect.com.
Disclosures
The authors did not receive funding for this project.
R.S.D. does not have any conflicts of interest related to this work. R.S.D. has an investigator-initiated grant with Nevro Corp and Saol Therapeutics. R.L.N. does not have any conflicts of interest related to this work. His activities were done within his role as a Federal Employee.
Supplementary Data
Supplementary data related to this article can be found in the online version at doi:10.1016/j.jpain.2024.104665.
Data Availability
The data are available upon reasonable request to the principal investigator of this study.
References
- 1.Nahin RL, Sayer B, Stussman BJ, Feinberg TM: Eighteen-year trends in the prevalence of, and health care use for, noncancer pain in the United States: data from the medical expenditure panel survey. J Pain 20(7):796–809, 2019. [DOI] [PubMed] [Google Scholar]
- 2.Han B, Compton WM, Blanco C, Crane E, Lee J, Jones CM: Prescription opioid use, misuse, and use disorders in U.S. adults: 2015 national survey on drug use and health. Ann Intern Med 167(5):293–301, 2017. [DOI] [PubMed] [Google Scholar]
- 3.Degenhardt L, Charlson F, Mathers B, et al. The global epidemiology and burden of opioid dependence: results from the global burden of disease 2010 study. Addiction 109(8):1320–1333, 2014. [DOI] [PubMed] [Google Scholar]
- 4.Zajacova A, Grol-Prokopczyk H, Limani M, Schwarz C, Gilron I: Prevalence and correlates of prescription opioid use among US adults, 2019-2020. PLoS One 18(3):e0282536. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Weng L, Huang J, Lou Y, et al. Comparison of the prevalence of opioid use among U.S. adults with cardiac conditions before and during the COVID-19 pandemic. Front Public Health 11:1127636. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Bernard DM, Encinosa W, Cohen J, Fang Z: Patient factors that affect opioid use among adults with and without chronic pain. Res Social Adm Pharm 17(6):1059–1065, 2021. [DOI] [PubMed] [Google Scholar]
- 7.Deyo RA, Hallvik SE, Hildebran C, et al. Association of prescription drug monitoring program use with opioid prescribing and health outcomes: a comparison of program users and nonusers. J Pain 19(2):166–177, 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Goldstick JE, Guy GP, Losby JL, Baldwin G, Myers M, Bohnert ASB: Changes in initial opioid prescribing practices after the 2016 release of the CDC guideline for prescribing opioids for chronic pain. JAMA Netw Open 4(7):e2116860. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Lalic S, Ilomäki J, Bell JS, Korhonen MJ, Gisev N: Prevalence and incidence of prescription opioid analgesic use in Australia. Br J Clin Pharmacol 85(1):202–215, 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Maierhofer CN, Ranapurwala SI, DiPrete BL, et al. Intended and unintended consequences: changes in opioid prescribing practices for postsurgical, acute, and chronic pain indications following two policies in North Carolina, 2012-2018 - controlled and single-series interrupted time series analyses. Drug Alcohol Depend 242:109727. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Andersen R, Newman JF: Societal and individual determinants of medical care utilization in the United States. Milbank Mem Fund Q Health Soc 51(1):95–124, 1973. [PubMed] [Google Scholar]
- 12.Aday LA, Andersen R: A framework for the study of access to medical care. Health Serv Res 9(3):208–220, 1974. [PMC free article] [PubMed] [Google Scholar]
- 13.Andersen RM: Revisiting the behavioral model and access to medical care: does it matter? J Health Soc Behav 36(1):1–10, 1995. [PubMed] [Google Scholar]
- 14.Nahin RL, Dahlhamer JM, Stussman BJ: Health need and the use of alternative medicine among adults who do not use conventional medicine. BMC Health Serv Res 10:220, 2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.de Boer AG, Wijker W, de Haes HC: Predictors of health care utilization in the chronically ill: a review of the literature. Health Policy 42(2):101–115, 1997. [DOI] [PubMed] [Google Scholar]
- 16.Zablotsky B, Lessem SE, Gindi RM, Maitland AK, Dahlhamer JM, Blumberg SJ: Overview of the 2019 national health interview survey questionnaire redesign. Am J Public Health 113(4):408–415, 2023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.von Elm E, Altman DG, Egger M, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ 335(7624):806–808, 2007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Centers for Disease Control & Prevention. National Center for Health Statistics: about the National Health Interview Survey. CDC. Published November 22, 2023. Accessed July 1, 2024. https://www.cdc.gov/nchs/nhis/about_nhis.htm. [Google Scholar]
- 19.Axon DR, Slack M, Warholak T: Identifying predictors of opioid use among older United States' adults with pain. J Opioid Manag 18(2):95–105, 2022. [DOI] [PubMed] [Google Scholar]
- 20.Montiel Ishino FA, McNab PR, Gilreath T, Salmeron B, Williams F: A comprehensive multivariate model of biopsychosocial factors associated with opioid misuse and use disorder in a 2017-2018 United States national survey. BMC Public Health 20(1):1740, 2020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Axon DR, Slack M, Barraza L, Lee JK, Warholak T: Nationally representative health care expenditures of community-based older adults with pain in the United States prescribed opioids vs those not prescribed opioids. Pain Med 22(2):282–291, 2021. [DOI] [PubMed] [Google Scholar]
- 22.Neckerman KM, Garfinkel I, Teitler JO, Waldfogel J, Wimer C: Beyond income poverty: measuring disadvantage in terms of material hardship and health. Acad Pediatr 16(3 Suppl):S52–S59, 2016. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Parker JD, Talih M, Malec DJ, et al. National center for health statistics data presentation standards for proportions. Vital Health Stat 2 175:1–22, 2017. [PubMed] [Google Scholar]
- 24.Davis JA, Robinson RL, Le TK, Xie J: Incidence and impact of pain conditions and comorbid illnesses. J Pain Res 4:331–345, 2011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Chakravorty S, Vandrey RG, He S, Stein MD: Sleep management among patients with substance use disorders. Med Clin North Am 102(4):733–743, 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Wang D, Teichtahl H: Opioids, sleep architecture and sleep-disordered breathing. Sleep Med Rev 11(1):35–46, 2007. [DOI] [PubMed] [Google Scholar]
- 27.Angst F, Benz T, Lehmann S, et al. Extended overview of the longitudinal pain-depression association: a comparison of six cohorts treated for specific chronic pain conditions. J Affect Disord 273:508–516, 2020. [DOI] [PubMed] [Google Scholar]
- 28.Rogers AH, Zvolensky MJ, Ditre JW, Buckner JD, Asmundson GJG: Association of opioid misuse with anxiety and depression: a systematic review of the literature. Clin Psychol Rev 84:101978. [DOI] [PubMed] [Google Scholar]
- 29.Quinn K, Frueh BC, Scheidell J, Schatz D, Scanlon F, Khan MR: Internalizing and externalizing factors on the pathway from adverse experiences in childhood to non-medical prescription opioid use in adulthood. Drug Alcohol Depend 197:212–219, 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Nestler EJ, Carlezon WA: The mesolimbic dopamine reward circuit in depression. Biol Psychiatry 59(12):1151–1159, 2006. [DOI] [PubMed] [Google Scholar]
- 31.Rojas-Corrales MO, Berrocoso E, Gibert-Rahola J, Micó JA: Antidepressant-like effects of tramadol and other central analgesics with activity on monoamines reuptake, in helpless rats. Life Sci 72(2):143–152, 2002. [DOI] [PubMed] [Google Scholar]
- 32.Kross E, Berman MG, Mischel W, Smith EE, Wager TD: Social rejection shares somatosensory representations with physical pain. Proc Natl Acad Sci USA 108(15):6270–6275, 2011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Perry WM, Agala CB, Agala EM: Evaluation of a state law on opioid-prescribing behaviour and the void affecting codeine-containing antitussive syrup. Emerg Med J 38(12):889–894, 2021. [DOI] [PubMed] [Google Scholar]
- 34.Yang S, Hincapie-Castillo JM, Ke X, et al. Evaluation of cough medication use patterns in ambulatory care settings in the United States: 2003-2018. J Clin Med 11(13)2022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Yang S, Huang S, Hincapie-Castillo JM, et al. Patterns of cough medication prescribing among patients with chronic cough in Florida: 2012–2021. J Clin Med 12(19)2023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.D'Souza RS, Langford B, Wilson RE, et al. The state-of-the-art pharmacotherapeutic options for the treatment of chronic non-cancer pain. Expert Opin Pharmacother :1–15, 2022. [DOI] [PubMed] [Google Scholar]
- 37.Weissman JD, Russell D, Taylor J: The relationship between financial stressors, chronic pain, and high-impact chronic pain: findings from the 2019 national health interview survey. Public Health Rep 138(3):438–446, 2023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Dahlhamer J, Lucas J, Zelaya C, et al. Prevalence of chronic pain and high-impact chronic pain among adults - United States, 2016. MMWR Morb Mortal Wkly Rep 67(36):1001–1006, 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Schieber LZ, Guy GP, Seth P, et al. Trends and patterns of geographic variation in opioid prescribing practices by state, United States, 2006-2017. JAMA Netw Open 2(3):e190665. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Webster LR: Risk factors for opioid-use disorder and overdose. Anesth Analg 125(5):1741–1748, 2017. [DOI] [PubMed] [Google Scholar]
- 41.Manchikanti L, Vanaparthy R, Atluri S, Sachdeva H, Kaye AD, Hirsch JA: COVID-19 and the opioid epidemic: two public health emergencies that intersect with chronic pain. Pain Ther 10(1):269–286, 2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Wainwright JJ, Mikre M, Whitley P, et al. Analysis of drug test results before and after the US declaration of a national emergency concerning the COVID-19 outbreak. JAMA 324(16):1674–1677, 2020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Ochalek TA, Cumpston KL, Wills BK, Gal TS, Moeller FG: Nonfatal opioid overdoses at an urban emergency department during the COVID-19 pandemic. JAMA 324(16):1673–1674, 2020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Pronk NP, Kleinman DV, Richmond TS: Healthy people 2030: moving toward equitable health and well-being in the United States. EClinicalMedicine 33:100777. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.U.S. Department of Health and Human Services. Pain Management Best Practices Inter-Agency Task Force Report: Updates, Gaps, Inconsistencies, and Recommendations. U.S. Department of Health and Human services website. Published May 9, 2019. Accessed July 1, 2024. https://www.hhs.gov/sites/default/files/pmtf-final-report-2019-05-23.pdf. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data are available upon reasonable request to the principal investigator of this study.