Abstract
Objectives. We estimated trends in drug poisoning death rates by educational attainment and investigated educational inequalities in drug poisoning mortality by race, gender, and region.
Methods. We linked drug poisoning death counts from the National Vital Statistics System to population denominators from the Current Population Survey to estimate drug poisoning rates by gender, race, region, and educational attainment (less than high school degree, high school degree, some college, college degree) from 1994 to 2010.
Results. There were 372 485 drug poisoning deaths. Education-related inequalities increased during the study among all demographic groups and varied by region. Absolute increases in educational inequalities were higher among Whites than Blacks and men than women. The age-adjusted rate difference between White men with less than a high school degree increased from 8.7 per 100 000 in 1994 to 27.4 in 2010 (change = 18.7). Among Black men, the corresponding increases were 11.7 and 18.3, respectively (change = 6.6).
Conclusions. We found strong educational patterning in drug poisoning rates, chiefly by region and race. Rates are highest and increasing the fastest among groups with less education.
Drug poisoning death rates have doubled over the past decade in the United States 1 and recently overtook motor vehicle accidents as the leading cause of death from accidental injury.2 Drug poisoning death rates are not increasing uniformly but instead show pronounced differences by sociodemographic indicators such as race,3 age,3 gender,3,4 rural location,5,6 and region.4,5 For instance, there was nearly a fivefold variation in drug poisoning mortality rates across US states in 2008, ranging from 5.5 per 100 000 in Nebraska to 27.0 per 100 000 in New Mexico.4
Although there are striking differences in drug poisoning rates by demographic factors, to our knowledge no studies have examined national trends in drug poisoning deaths by socioeconomic position. Not only does describing trends by socioeconomic position add to the growing body of knowledge about factors associated with risk of drug poisoning death, but drug poisoning deaths may be an exception to the well-established inverse gradient between socioeconomic position and all-cause mortality.7 For instance, prescription medications are involved in the majority of drug poisoning deaths8 and may be more accessible to those with more education because of better health care coverage9,10 and less accessible to Black Americans, who are consistently less likely to be prescribed opioid analgesics than are Whites.11–13 Therefore, counter to the pattern observed for all-cause mortality, drug poisoning rates may be higher or increasing faster among individuals of higher education, and educational inequalities in rates may be increasing more among White Americans than among Black Americans.
We investigated trends in drug poisoning death rates by educational attainment from 1994 to 2010 using data from the National Vital Statistics System (NVSS) linked to population denominators from the Current Population Survey (CPS). We also investigated educational trends in drug poisoning mortality by race, gender, and region.
METHODS
Decedent information was recorded on death certificates and collected by the NVSS.14 More than 99% of all deaths occurring in the United States are reported to the NVSS, and quality control procedures ensure the accuracy of death classification.15 We used NVSS data to obtain age, gender, race, educational attainment, state of residence, year, and cause of death for all individuals who died owing to drug poisoning. We classified causes of death using the International Classification of Diseases, Ninth Revision (ICD-9)16 coding for 1994 to 1998 and the 10th Revision (ICD-10)17 coding for 1999 to 2010.
We identified all deaths with underlying causes related to drug poisonings (ICD-9: E850.0–E858.9, E950.0–E950.5, E962.0, E980.0–E980.5; ICD-10: X40–X44, X60–X64, X85, Y10–Y14). We used the same drug poisoning scheme as does the Centers for Disease Control and Prevention, which excludes poisonings in which alcohol is the underlying cause of death.3 Similar to other studies,18 we excluded states missing educational information for more than 20% of decedents in any year (Rhode Island, Oklahoma, South Dakota, Kentucky, and Georgia).
We generated population denominators by gender, age, race, years of educational attainment, state, and year from the monthly CPS.19 We aggregated monthly demographic totals within states and averaged them over each year. We used corrected CPS weighting estimates for the years 2000 to 2002 from the National Bureau of Economic Research.20 When population counts were missing (9.2% of all cells), we computed state-specific annual population estimates within each age, gender, race, and education category using linear interpolation (7.7% of cells) or extrapolation (1.5% of cells) from available survey estimates between 1980 and 2013.
In 0.2% of strata, extrapolation resulted in negative population estimates, which we recoded as missing values. These population denominators were linked to the NVSS data to provide rates by year (1994–2010), race (Black, White, other or mixed race), gender, educational attainment (less than high school degree, high school degree, some college, college degree), age (25–34, 35–44, 45–54, 55–64, > 64 years), and state. We estimated regional trends in rates by combining states into US census divisions. A map of these census divisions is provided as a supplement to the online version of this article at http://www.ajph.org.
In 2003, the NVSS educational attainment coding scheme changed from highest grade completed to highest educational degree obtained, although adoption of this new coding scheme varied by state.21 To account for potential differences in recording educational attainment because of coding scheme, we calculated the percentage of deaths using the old or new coding scheme each year for each state. For each state and year, we created a dichotomous variable to adjust for whether the NVSS classified the majority of deaths using the old or new coding system.
Statistical Analysis
Poisson regression models for mortality rates demonstrated overdispersion, so we modeled rates using a 3-parameter generalized negative binomial regression.22,23 Negative binomial models assume that the variance is a function of a parameter power of the mean, and the negative binomial regression model estimates this parameter rather than assuming either constant dispersion (parameter power = 1) or mean dispersion (parameter power = 2). Likelihood ratio tests indicated that the negative binomial regression model provided a better fit (parameter power = 1.7) than either the NB-1 or NB-2 models (both P ≤ .001).
We estimated drug poisoning rates adjusted for age and education coding scheme from 1994 to 2010 by educational attainment, gender, and race. We estimated separate models for men and women and allowed trends in rates to vary by race by including a race-by-year product term in our models. We used the margins postestimation command in Stata, version 12.0 (StataCorp, College Station, TX) to estimate marginal predicted incidence rates and SEs after negative binomial regression. For predicted rates we standardized to the age and education coding distribution for the entire population. All models used robust variance estimators clustered at the state level. We also estimated these models for the subset of poisonings classified as accidental (ICD-9: E850.0–E858.9; ICD–10: X40–X44) and intentional (ICD-9: E950.0–E950.5, E962.0; ICD-10: X60–X64, X85) to see whether accidental and intentional death rates followed the same patterning as all drug poisoning death rates.
We estimated educational differences in rates in each survey year by calculating the slope index of inequality (SII).24,25 The SII estimates the magnitude of health inequalities across the entire distribution of socioeconomic position and accounts for changes in the distribution of educational attainment over time.25 We derived the SII by ranking the population by education, calculating the cumulative percentage of the population in each education group, and assigning each education group the midpoint of their position in the cumulative distribution (ridit score). Thus, a 1-unit change in the education ranking variable estimates the absolute difference in mortality between the bottom and the top of the education distribution. We estimated the SII separately for each demographic group and by census division. Demographic group–specific models adjusted for age and educational coding scheme, and the census division models adjusted for age, race, gender, and educational coding scheme.
Sensitivity Analysis
NVSS educational attainment information is collected from death certificates that are completed by individuals such as funeral directors who collect educational information from next of kin. Previous research indicates that educational attainment is misreported about one quarter of the time, and misreporting varies by educational level and socioeconomic indicators.21 A previous study estimated correction ratios by linking the “gold standard” of self-reported education during a CPS interview with the subsequent educational level reported on death certificates.21 We constructed correction ratios within each stratum of educational attainment and gender by dividing the total number of decedents classified in the CPS by the total number of decedents classified using death certificate information. We used these ratios to correct for bias in reporting educational attainment on death certificates by multiplying the correction ratio by the observed death count in the NVSS within each stratum of educational attainment and gender.
RESULTS
Among US adults aged 25 years or older, between 1994 and 2010 there were 372 485 drug poisoning deaths. Restriction to the 45 states with at least 80% complete reporting on death certificates retained 94% of deaths (n = 349 155). We excluded an additional 4% of deaths (n = 12 822) because of missing decedent education, resulting in inclusion of 90% of all drug poisoning deaths. The majority of poisonings (71%) were classified accidental, 17% were classified intentional, and 12% were of undetermined intent. Between 1994 and 2010, crude drug poisoning rates increased in every region and demographic group (data available as a supplement to the online version of this article at http://www.ajph.org).
Age-adjusted drug poisoning rates showed an educational gradient throughout the study period (Figure 1). In 2010 age-adjusted rates per 100 000 population were highest among those with less than a high school degree (26.0; 95% CI = 22.6, 29.4), and among this demographic group the absolute increase in rates since 1994 was 16.3 (95% CI = 12.7, 19.9). Unadjusted rates had similar point estimates (data available as a supplement to the online version of this article at http://www.ajph.org). This same gradient was observed for the subset of accidental poisonings; however, in the subset of deaths classified as intentional, there was no strong educational gradient and deaths did not increase substantially over the study period (data available as a supplement to the online version of this article at http://www.ajph.org).
FIGURE 1—
Drug poisoning rates by educational attainment: National Vital Statistics System, Current Population Survey; United States; 1994–2010.
Note. HS = high school. Rates are adjusted to the pooled 1994–2010 age distribution.
Adjusted drug poisoning rates by educational attainment, gender, and race showed strong socioeconomic patterning (Table 1). Overall, 2010 rates were highest in Whites, men, and those with less education. In 2010 rates were highest among White men with less than a high school degree (34.0), and the lowest rates were among college-educated women of other or mixed race (0.8). However, between 1994 and 2010, the largest relative increase in rates was among less-educated White women. In this demographic group, rates increased more than 300% among those with less than a college education. Although rates increased in every educational group among Whites, there was virtually no increase among college-educated Blacks or other or mixed race individuals.
TABLE 1—
Drug Poisoning Rates by Gender, Race, and Educational Attainment: National Vital Statistics System, Current Population Survey; United States; 1994–2010
| Characteristic | 1994, Rate (95% CI) | 2010, Rate (95% CI) | Rate Difference, 2010 vs 1994 (95% CI) | Percentage Rate Increase, 2010 vs 1994 (95% CI) |
| Women | ||||
| White | ||||
| < HS degree | 6.3 (5.2, 7.4) | 28.8 (23.4, 34.2) | 22.6 (17.1, 28.0) | 359 (243, 475) |
| HS degree | 4.6 (3.8, 5.5) | 21.7 (19.6, 23.8) | 17.1 (14.8, 19.3) | 367 (273, 461) |
| Some college | 3.0 (2.5, 3.4) | 13.1 (11.5, 14.7) | 10.1 (8.5, 11.7) | 338 (257, 419) |
| College degree | 2.8 (2.3, 3.3) | 5.9 (5.1, 6.7) | 3.1 (2.3, 3.9) | 112 (72, 152) |
| Black | ||||
| < HS degree | 6.0 (4.0, 8.1) | 12.1 (9.2, 15.1) | 6.1 (3.3, 8.8) | 101 (37, 164) |
| HS degree | 5.8 (3.6, 8.0) | 11.3 (8.8, 13.9) | 5.5 (3.3, 7.7) | 95 (35, 155) |
| Some college | 2.5 (1.8, 3.2) | 6.6 (4.9, 8.2) | 4.1 (2.7, 5.4) | 160 (101, 220) |
| College degree | 2.0 (1.1, 3.0) | 2.5 (1.9, 3.1) | 0.5 (–0.6, 1.6) | 24 (–41, 89) |
| Other or mixed race | ||||
| < HS degree | 1.4 (0.6, 2.3) | 4.9 (2.9, 6.9) | 3.4 (1.8, 5.1) | 240 (67, 414) |
| HS degree | 2.5 (0.9, 4.2) | 5.4 (3.6, 7.2) | 2.8 (1.0, 4.7) | 112 (–12, 235) |
| Some college | 0.9 (0.3, 1.5) | 3.8 (2.7, 5.0) | 2.9 (1.9, 3.9) | 320 (82, 559) |
| College degree | 0.7 (0.3, 1.2) | 0.8 (0.4, 1.2) | 0.1 (–0.5, 0.6) | 7 (–76, 90) |
| Men | ||||
| White | ||||
| < HS degree | 11.9 (9.6, 14.2) | 34.0 (29.1, 39.0) | 22.2 (16.5, 27.9) | 187 (114, 260) |
| HS degree | 9.2 (7.2, 11.1) | 30.0 (27.1, 33.0) | 20.9 (18.0, 23.8) | 228 (163, 292) |
| Some college | 4.4 (3.6, 5.2) | 15.5 (13.7, 17.3) | 11.1 (9.4, 12.9) | 255 (191, 319) |
| College degree | 3.2 (2.6, 3.7) | 6.7 (5.9, 7.5) | 3.5 (2.8, 4.2) | 111 (77, 145) |
| Black | ||||
| < HS degree | 15.3 (10.5, 20.0) | 22.2 (16.7, 27.7) | 6.9 (3.2, 10.6) | 45 (16, 75) |
| HS degree | 16.4 (11.7, 21.2) | 22.2 (17.1, 27.3) | 5.8 (1.7, 9.8) | 35 (6, 64) |
| Some college | 6.9 (5.0, 8.9) | 9.7 (7.5, 11.8) | 2.7 (0.4, 5.1) | 39 (–2, 81) |
| College degree | 3.6 (2.2, 5.0) | 3.9 (2.8, 4.9) | 0.3 (–0.9, 1.6) | 9 (–28, 47) |
| Other or mixed race | ||||
| < HS degree | 2.8 (1.4, 4.3) | 5.7 (3.4, 8.0) | 2.8 (0.9, 4.8) | 100 (8, 191) |
| HS degree | 4.4 (2.8, 6.1) | 7.6 (4.8, 10.4) | 3.1 (0.6, 5.7) | 71 (3, 138) |
| Some college | 2.5 (1.3, 3.7) | 4.2 (2.6, 5.7) | 1.7 (0.2, 3.2) | 68 (–9, 146) |
| College degree | 0.8 (0.3, 1.2) | 1.2 (0.7, 1.6) | 0.4 (–0.2, 1.0) | 55 (–44, 154) |
Note. CI = confidence interval; HS = high school. We calculated rates by linking drug poisoning death counts from the National Vital Statistics System to population denominators from the Current Population Survey. Rates are per 100 000 population. We modeled rates separately for men and women and adjusted them for age (24–34, 35–44, 45–54, 55–64, > 64 years) and National Vital Statistics System educational attainment coding scheme (1989 or 2003).
Rate differences between the least and most educated groups increased substantially over the study period for all race groups (Table 2). Across all educational levels, drug poisoning rates increased more for Whites than for other race groups on both relative and absolute scales (Table 2; data available as a supplement to the online version of this article at http://www.ajph.org). Within race groups the absolute rate increases were generally similar for men and women.
TABLE 2—
Educational Differences in Adjusted Drug Poisoning Rates by Gender and Race: National Vital Statistics System, Current Population Survey; United States; 1994–2010
| Characteristic | 1994 Rate (95% CI) | 2010 Rate (95% CI) | Rate Difference, 2010 vs 1994 (95% CI) | Percentage Rate Increase, 2010 vs 1994 (95% CI) |
| Women | ||||
| White | ||||
| < HS degree vs college degree | 3.5 (2.5, 4.5) | 22.9 (17.6, 28.3) | 19.4 (14.1, 24.8) | 555 (327, 783) |
| HS degree vs college degree | 1.9 (1.2, 2.5) | 15.8 (14.2, 17.4) | 13.9 (12.2, 15.6) | 746 (434, 1058) |
| Some college vs college degree | 0.2 (–0.2, 0.6) | 7.2 (6.1, 8.3) | 7.0 (5.7, 8.2) | 3342 (–2831, 9515) |
| SII | 4.5 (3.2, 5.8) | 36.7 (30.6, 42.9) | 32.3 (26.0, 38.5) | 719 (450, 989) |
| Black | ||||
| < HS degree vs college degree | 4.0 (2.1, 5.9) | 9.6 (6.7, 12.5) | 5.6 (2.8, 8.4) | 139 (30, 248) |
| HS degree vs college degree | 3.8 (1.9, 5.7) | 8.8 (6.5, 11.2) | 5.0 (2.8, 7.3) | 133 (27, 238) |
| Some college vs college degree | 0.5 (–0.6, 1.6) | 4.1 (2.6, 5.5) | 3.6 (1.7, 5.5) | 702 (–1038, 2443) |
| SII | 6.3 (3.4, 9.3) | 13.3 (8.7, 17.8) | 6.9 (2.1, 11.7) | 110 (3, 217) |
| Other or mixed race | ||||
| < HS degree vs college degree | 0.7 (–0.3, 1.7) | 4.1 (2.1, 6.1) | 3.4 (1.6, 5.2) | 485 (–251, 1220) |
| HS degree vs college degree | 1.8 (0.1, 3.6) | 4.6 (2.7, 6.5) | 2.8 (0.8, 4.8) | 154 (–70, 379) |
| Some college vs college degree | 0.2 (–0.5, 0.9) | 3.0 (1.9, 4.2) | 2.9 (1.8, 4.0) | 1609 (–4980, 8198) |
| SII | 1.7 (–0.2, 3.5) | 8.6 (4.3, 12.9) | 7.0 (3.2, 10.7) | 418 (–78, 915) |
| Men | ||||
| White | ||||
| < HS degree vs college degree | 8.7 (6.6, 10.8) | 27.4 (22.6, 32.1) | 18.7 (13.3, 24.1) | 215 (117, 312) |
| HS degree vs college degree | 6.0 (4.4, 7.6) | 23.3 (21.0, 25.7) | 17.3 (14.9, 19.8) | 289 (191, 388) |
| Some college vs college degree | 1.2 (0.6, 1.8) | 8.8 (7.5, 10.1) | 7.6 (6.2, 9.0) | 631 (265, 997) |
| SII | 12.9 (9.8, 15.9) | 47.9 (41.5, 54.4) | 35.1 (27.8, 42.3) | 272 (169, 376) |
| Black | ||||
| < HS degree vs college degree | 11.7 (7.4, 16.1) | 18.3 (13.1, 23.6) | 6.6 (3.0, 10.2) | 56 (17, 96) |
| HS degree vs college degree | 12.9 (8.9, 16.9) | 18.3 (13.6, 23.0) | 5.4 (1.7, 9.2) | 42 (7, 77) |
| Some college vs college degree | 3.4 (1.6, 5.1) | 5.8 (3.7, 7.8) | 2.4 (0.0, 4.8) | 71 (–25, 168) |
| SII | 19.2 (12.2, 26.1) | 29.0 (19.5, 38.6) | 9.8 (2.6, 17.1) | 51 (7, 95) |
| Other or mixed race | ||||
| < HS degree vs college degree | 2.1 (0.8, 3.3) | 4.5 (2.3, 6.6) | 2.4 (0.4, 4.4) | 116 (–11, 243) |
| HS degree vs college degree | 3.7 (2.2, 5.1) | 6.4 (3.8, 9.0) | 2.7 (0.2, 5.2) | 74 (–5, 153) |
| Some college vs college degree | 1.7 (0.6, 2.8) | 3.0 (1.4, 4.5) | 1.3 (–0.4, 2.9) | 74 (–51, 199) |
| SII | 4.4 (2.4, 6.4) | 11.2 (6.0, 16.3) | 6.8 (2.0, 11.6) | 156 (24, 289) |
Note. CI = confidence interval; HS = high school; SII = slope index of inequality. We calculated rates by linking drug poisoning death counts from the National Vital Statistics System to population denominators from the Current Population Survey. Rates are per 100 000 population. We adjusted all models for age (24–34, 35–44, 45–54, 55–64, > 64 years) and National Vital Statistics System educational attainment coding scheme (1989 or 2003). We modeled SII separately for each gender and race.
Among White men, the adjusted difference in drug poisoning rates between men with less than a high school degree and men with a college degree was 8.7 per 100 000 in 1994 and 27.4 per 100 000 in 2010 (change = 18.7; 95% CI = 13.3, 24.1). Among White women this difference was 3.5 per 100 000 in 1994 and 22.9 per 100 000 in 2010 (change = 19.4; 95% CI = 14.1, 24.8). Among Black men, the corresponding gap between men with less than a high school degree and men with a college degree was 11.7 per 100 000 in 1994 and 18.3 per 100 000 in 2010 (change = 6.6; 95% CI = 3.0, 10.2). Among Black women this difference was 4.0 per 100 000 in 1994 and 9.6 per 100 000 in 2010 (change = 5.6; 95% CI = 2.8, 8.4).
Summary measures of education-related inequality (SIIs) also showed large differences between Blacks and Whites (Table 2). The SII, which can be interpreted as the absolute difference in mortality between the bottom and the top of the education distribution, was much higher in 2010 among White than among Black men (47.9 vs 29.0, respectively) and among White than among Black women (36.7 vs 13.3, respectively). Additionally, both absolute and relative increases in education-related inequality from 1994 to 2010 were greater among Whites than among Blacks.
Among Whites, the SII increased by 35.1 per 100 000 (95% CI = 27.8, 42.3) among men and 32.3 per 100 000 (95% CI = 26.0, 38.5) among women. Among Blacks, the change in the SII between 1994 and 2010 was 9.8 per 100 000 (95% CI = 2.6, 17.1) among men and 6.9 per 100 000 (95% CI = 2.1, 11.7) among women. Relative increases in the SII were much higher among White men (272%) and women (719%) than among Black men (51%) and Black women (110%). We found similar patterning when analyses were restricted to accidental drug poisonings (data available as a supplement to the online version of this article at http://www.ajph.org).
We also found substantial variation in education-related inequalities in drug poisoning rates by census division (Table 3). In 2010, the East North Central division (Wisconsin, Michigan, Illinois, Indiana, Ohio) had the largest difference in mortality between the top and the bottom of the education distribution (33.3 per 100 000), and the West South Central division (Arkansas, Louisiana, Texas) had the smallest difference (17.0 per 100 000). Between 1994 and 2010 educational inequality increased in all divisions. The largest increase since 1994 (28.6 per 100 000) was in the East South Central division (Tennessee, Mississippi, Alabama), and the smallest increase (12.8 per 100 000) was in the West South Central census division.
TABLE 3—
Slope Index of Inequality by Census Division: National Vital Statistics System, Current Population Survey; United States; 1994–2010
| Census Division | 1994, SII (95% CI) | 2010, SII (95% CI) | Rate Difference, 2010 vs 1994 (95% CI) |
| New England (ME, VT, NH, MA, CT) | 10.7 (4.9, 16.5) | 25.1 (19.2, 31.0) | 14.4 (9.1, 19.7) |
| Mid-Atlantic (NY, PA, NJ) | 8.8 (3.1, 14.5) | 22.5 (16.0, 29.0) | 13.6 (11.4, 15.8) |
| East North Central (WI, MI, IL, IN, OH) | 5.9 (3.1, 8.8) | 33.3 (26.2, 40.3) | 27.3 (18.8, 35.9) |
| West North Central (ND, MN, NE, IA, KS, MO) | 5.0 (3.2, 6.8) | 26.7 (17.0, 36.4) | 21.7 (13.5, 30.0) |
| South Atlantic (DE, MD, WV, VA, NC, SC, FL) | 5.0 (1.0, 9.0) | 29.0 (21.7, 36.4) | 24.0 (18.1, 30.0) |
| East South Central (TN, MS, AL) | 3.2 (1.9, 4.5) | 31.8 (19.6, 44.0) | 28.6 (17.5, 39.7) |
| West South Central (AR, LA, TX) | 4.2 (2.8, 5.5) | 17.0 (9.1, 24.8) | 12.8 (5.7, 19.9) |
| Mountain (MT, ID, WY, NV, UT, CO, AZ, NM) | 12.0 (7.8, 16.2) | 24.9 (20.0, 29.8) | 12.9 (5.0, 20.8) |
| Pacific (WA, OR, CA) | 14.1 (12.3, 15.9) | 32.9 (24.6, 41.2) | 18.8 (11.5, 26.0) |
Note. CI = confidence interval; SII = slope index of inequality. We calculated rates by linking drug poisoning death counts from the National Vital Statistics System to population denominators from the Current Population Survey. Rates are per 100 000 population. We estimated SII separately for each region and adjusted models for age, gender, race, and educational coding system.
In a sensitivity analysis, we applied correction ratios to the NVSS educational attainment classification. These correction ratios did not substantially alter results, although they did produce more pronounced differences in rates between those with less than a high school degree and those with a high school degree (data available as a supplement to the online version of this article at http://www.ajph.org).
DISCUSSION
Gradients in health behaviors,7,26 health conditions,9 all-cause mortality,7 and life expectancy27 have been observed by educational level, and the gap in morbidity and mortality between the least and most educated appears to be increasing.9,28 Our results add to this body of knowledge by describing, for the first time to our knowledge, national inequalities in drug poisoning deaths by educational level as well as by showing a substantial increase in this inequality since 1994.
We found strong educational inequality patterning in drug poisoning deaths, which were driven primarily by accidental drug poisoning deaths. Educational inequalities varied substantially by race and region. Drug poisoning rates were highest and increased the fastest among Whites. Rates increased within every educational level for Whites but did not increase among college-educated non-Whites. There was also substantial variation in the difference in rates between the least and most educated across census divisions, and this variation changed over time. In 1994 the largest educational difference was in the Pacific division (Washington, Oregon, California), but by 2010 the largest difference was in the East North Central division, where the magnitude of education-related inequality increased more than fivefold over the study period.
The causes of drug poisoning patterning by race and region may involve numerous factors such as regional and race differences in illicit drug use,29 geographic differences in the availability of medications such as Naloxone to reverse the effects of overdose from heroin and other opioids,30 and contextual factors such as emergency response times and the reluctance of bystanders to call for help because of mistrust of authority.31 In addition, drug poisoning rates may be influenced by access to and use of prescription medications.
The number of deaths attributed to prescription medications has risen dramatically in the past decade. Prescription medications are involved in the majority of drug poisoning deaths,8 and poisonings from prescription opioids have tripled over the past decade,4 currently accounting for three quarters of drug poisoning deaths.4,32 Although we were not able to specifically examine deaths related to prescription medications, the high percentage of deaths attributed to prescription medications indicates that access to these medications could be a key driver of the observed patterning.
Access to prescription medications—either directly from a physician or through diversion (the act of being given or purchasing prescription medications that were not prescribed by a physician)—may not be uniform across race and region. Whites may have greater access to opioids because of better health insurance coverage,9,10 especially because about one fifth of nonmedically used opioids are obtained through physicians.33 Even when they have access to health care, Blacks are consistently less likely to be prescribed opioid analgesics, particularly in emergency department settings.11–13
Whites’ potentially greater access to prescription medications through physicians may also lead to greater access to diverted medications, because social networks are typically racially homogenous34and the majority of some diverted medications, such as opioids, are obtained through friends and family.33 Racial differences in access to prescription medications align with Whites’ higher self-report of past year nonmedical use of pain relievers compared with that of Blacks (5.1% vs 3.6%).29 Blacks’ restricted access to prescription medications may thus inadvertently protect them from drug poisoning deaths and may contribute to their slower increase in drug poisoning mortality rates relative to Whites, although it should be noted that drug poisoning deaths generally rose for all racial groups.
Prescription medication access may also vary by region. There is a fourfold variation in opioid consumption across US states35 and large differences in prescribing the potent opioid oxycodone.36 State-specific characteristics may also affect access to prescription medications, such as prevalence of uninsured individuals (Massachusetts, 4.8%; Arizona, Texas, and South Carolina, 20.9%).37 Therefore, differential access to prescription medications owing to race or region may partially explain our observed patterning and merits further investigation.
Because of the myriad factors that likely affect drug poisoning patterning by education and other sociodemographic indicators, there are many opportunities for future research. Better understanding the behavioral and environmental pathways linking Whites’ dramatic rise in drug poisoning rates compared with non-Whites could reveal important points of intervention. Examining trends at the state level may prove more informative because many policies that may affect this patterning occur at the state level. State-specific analyses could uncover variation in the difference between the least and the most educated, which could guide additional investigation into effective state policies influencing this disparity. Finally, additional research could investigate the specific drugs involved in poisoning deaths, which would likely reveal distinct poisoning patterns by educational level, race, gender, and location.
Limitations
Perhaps the biggest limitation of this study was our inability to examine specific drugs implicated in drug poisoning deaths. The Underlying Causes of Death Database combines poisoning from disparate drugs into single categories (e.g., ICD-10 code × 62: intentional self-poisoning by and exposure to narcotics and psychodysleptics [hallucinogens]), which precluded us from examining poisonings owing to specific drug classes, such as prescription opioids. Although a related database, the Multiple Causes of Death Database, contains more granular drug poisoning categories, it does not provide educational attainment information.
Although drug poisoning rates vary by rural and urban locations,6,38,39 ethnicity,3 and race groups other than Black and White,3 we could not examine these factors. We could not examine rural versus urban location because the NVSS and CPS use incompatible definitions to classify urban residence. National data show some of the highest and lowest drug poisoning rates in rural locations,39 which could uncover important patterning, especially because rural location is associated with lower educational attainment.40
We did not examine rates by ethnicity because of concerns that the NVSS ethnicity classification is not considered reliable for all states during this study period.41 However, other studies have observed distinctive drug poisoning rates in Hispanics compared with non-Hispanics.3 In addition, the small population size of race groups other than Blacks and Whites potentially obscured important differences in drug poisoning rates in other racial groups. For instance, nationally American Indians/Alaska Natives have very high drug poisoning rates compared with other racial groups,3 yet because of the small population size we could not examine these rates. In addition, the racial makeup of the other or multiracial category has likely changed over our study period. Because different racial groups have very different drug poisoning rates,3 the observed trends in the other or multiracial category could partially reflect changes in the racial composition of this group.
Misclassification of educational attainment in the NVSS is an additional limitation. We attempted to correct for this misclassification through analytic control of the NVSS version (1989 or 2003) used in each state for each year. Additionally, in a sensitivity analysis we applied gender-specific educational attainment misclassification ratios provided in a validation study21 and found the results were consistent with our main findings. However, this same study noted educational misclassification by race and age, although correction ratios have not yet been developed.21 Additional corrections for education misclassification by age and race could lead to different results.
Additionally, drug poisoning classification changed from ICD-9 (1994–1998) to ICD-10 (1999–2010) coding during our study period. Compatibility between these 2 coding schemes is a concern because cause of death classification could vary by coding scheme, which could lead to artificial inflation (or deflation) of drug poisoning death rates. This incompatibility could be indicated by a discontinuity in drug poisoning rates immediately after adoption of the new coding scheme, but we saw little evidence for this (Figure 1; data available as a supplement to the online version of this article at http://www.ajph.org).
Conclusions
We found national inequalities in drug poisoning deaths by educational level and a substantial increase in this inequality since 1994. With passage of the Affordable Care Act, this inequality has the potential to become even more pronounced because those of lower educational attainment may soon have better access to prescription medications. Research into the underlying causes of educational differences in poisoning mortality are urgently needed to inform public health efforts to reduce drug poisoning deaths and promote health equity.
Acknowledgments
This work was supported by the Canadian Institutes of Health Research (grant 115214) and the Canadian Institutes of Health Research Interdisciplinary Capacity Enhanced Team (grant HOA-80072). Sam Harper was supported by a Chercheur-boursier Junior2 from the Fonds de recherche du Québec–Santé.
Human Participant Protection
Institutional review board approval was not necessary because participant data were de-identified.
References
- 1.Centers for Disease Control and Prevention. QuickStats: number of poisoning deaths involving opioid analgesics and other drugs or substances—United States, 1999–2010. MMWR Morb Mortal Wkly Rep. 2013;62(12):234. [Google Scholar]
- 2.Centers for Disease Control and Prevention. Underlying cause of death 1999–2013. Available at: http://wonder.cdc.gov/wonder/help/ucd.html. Accepted April 20, 2015.
- 3.Warner M, Chen LH, Makuc DM, Anderson RN, Miniño AM. Drug poisoning deaths in the United States, 1980–2008. NCHS Data Brief. 2011;(81):1–8. [PubMed] [Google Scholar]
- 4.Centers for Disease Control and Prevention. Vital signs: overdoses of prescription opioid pain relievers—United States, 1999–2008. MMWR Morb Mortal Wkly Rep. 2011;60(43):1487–1492. [PubMed] [Google Scholar]
- 5.Rossen LM, Khan D, Warner M. Trends and geographic patterns in drug-poisoning death rates in the U.S., 1999–2009. Am J Prev Med. 2013;45(6):e19–e25. doi: 10.1016/j.amepre.2013.07.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Wunsch MJ, Nakamoto K, Behonick G, Massello W. Opioid deaths in rural Virginia: a description of the high prevalence of accidental fatalities involving prescribed medications. Am J Addict. 2009;18(1):5–14. doi: 10.1080/10550490802544938. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Stringhini S, Sabia S, Shipley M et al. Association of socioeconomic position with health behaviors and mortality. JAMA. 2010;303(12):1159–1166. doi: 10.1001/jama.2010.297. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Jones CM, Mack KA, Paulozzi LJ. Pharmaceutical overdose deaths, United States, 2010. JAMA. 2013;309(7):657–659. doi: 10.1001/jama.2013.272. [DOI] [PubMed] [Google Scholar]
- 9.Goldman D, Smith JP. The increasing value of education to health. Soc Sci Med. 2011;72(10):1728–1737. doi: 10.1016/j.socscimed.2011.02.047. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Zuvekas SH, Taliaferro GS. Pathways to access: health insurance, the health care delivery system, and racial/ethnic disparities, 1996–1999. Health Aff (Milwood) 2003;22(2):139–153. doi: 10.1377/hlthaff.22.2.139. [DOI] [PubMed] [Google Scholar]
- 11.Mazer-Amirshahi M, Mullins PM, Rasooly I, van den Anker J, Pines JM. Rising opioid prescribing in adult U.S. emergency department visits: 2001–2010. Acad Emerg Med. 2014;21(3):236–243. doi: 10.1111/acem.12328. [DOI] [PubMed] [Google Scholar]
- 12.Tamayo-Sarver JH, Hinze SW, Cydulka RK, Baker DW. Racial and ethnic disparities in emergency department analgesic prescription. Am J Public Health. 2003;93(12):2067–2073. doi: 10.2105/ajph.93.12.2067. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Todd KH. Influence of ethnicity on emergency department pain management. Emerg Med (Fremantle) 2001;13(3):274–278. doi: 10.1046/j.1035-6851.2001.00229.x. [DOI] [PubMed] [Google Scholar]
- 14.Centers for Disease Control and Prevention. Vital statistics data. 1980–2010. Available at: http://www.cdc.gov/nchs/data_access/Vitalstatsonline.htm. Accessed April 20, 2015.
- 15.National Center for Health Statistics. Vital Statistics of the United States: Mortality, 1999. Hyattsville, MD: 2004. [Google Scholar]
- 16.International Classification of Diseases, Ninth Revision. Geneva, Switzerland: World Health Organization; 1980. [Google Scholar]
- 17.International Classification of Diseases, 10th Revision. Geneva, Switzerland: World Health Organization; 1992. [Google Scholar]
- 18.Jemal A, Ward E, Anderson RN, Murray T, Thun MJ. Widening of socioeconomic inequalities in U.S. death rates, 1993–2001. PLoS ONE. 2008;3(5):e2181. doi: 10.1371/journal.pone.0002181. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.US Census Bureau. DataFerrett. Available at: http://dataferrett.census.gov. Accessed April 20, 2015.
- 20.National Bureau of Economic Research. Reading current population survey (CPS) basic monthly data 2000–2002 extract files with SAS, SPSS, or Stata. 2004. Available at: http://www.nber.org/data/cps_extract.html. Accessed February 1, 2014.
- 21.Rostron BL, Boies JL, Arias E. Education reporting and classification on death certificates in the United States. Vital Health Stat 2. 2010;151:1–21. [PubMed] [Google Scholar]
- 22.Greene W. Functional forms for the negative binomial model for count data. Econ Lett. 2008;99(3):585–590. [Google Scholar]
- 23.Hardin JW, Hilbe JM. Regression models for count data based on the negative binomial (p) distribution. Stata J. 2014;14(2):280–291. [Google Scholar]
- 24.Pamuk ER. Social class inequality in mortality from 1921 to 1972 in England and Wales. Popul Stud (Camb) 1985;39(1):17–31. doi: 10.1080/0032472031000141256. [DOI] [PubMed] [Google Scholar]
- 25.Regidor E. Measures of health inequalities: part 2. J Epidemiol Community Health. 2004;58(11):900–903. doi: 10.1136/jech.2004.023036. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Ross CE, Wu C-L. The links between education and health. Am Sociol Rev. 1995;60(5):719–745. [Google Scholar]
- 27.Olshansky SJ, Antonucci T, Berkman L et al. Differences in life expectancy due to race and educational differences are widening, and many may not catch up. Health Aff (Millwood) 2012;31(8):1803–1813. doi: 10.1377/hlthaff.2011.0746. [DOI] [PubMed] [Google Scholar]
- 28.Meara ER, Richards S, Cutler DM. The gap gets bigger: changes in mortality and life expectancy, by education, 1981–2000. Health Aff (Millwood) 2008;27(2):350–360. doi: 10.1377/hlthaff.27.2.350. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Substance Abuse and Mental Health Services Administration. Results From the 2010 National Survey on Drug Use and Health: Summary of National Findings. Rockville, MD: 2011. [Google Scholar]
- 30.Centers for Disease Control and Prevention. Community-based opioid overdose prevention programs providing naloxone—United States, 2010. MMWR Morb Mortal Wkly Rep. 2012;61(6):101–105. [PMC free article] [PubMed] [Google Scholar]
- 31.Galea S, Ahern J, Vlahov D et al. Income distribution and risk of fatal drug overdose in New York City neighborhoods. Drug Alcohol Depend. 2003;70(2):139–148. doi: 10.1016/s0376-8716(02)00342-3. [DOI] [PubMed] [Google Scholar]
- 32.Chen LH, Hedegaard H, Warner M. Drug-poisoning deaths involving opioid analgesics: United States, 1999–2011. NCHS Data Brief. 2014;(166):1–8. [PubMed] [Google Scholar]
- 33.Jones CM, Paulozzi LJ, Mack KA. Sources of prescription opioid pain relievers by frequency of past-year nonmedical use: United States, 2008–2011. JAMA Intern Med. 2014;174(5):802–803. doi: 10.1001/jamainternmed.2013.12809. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.McPherson M, Smith-Lovin L, Cook JM. Birds of a feather: homophily in social networks. Annu Rev Sociol. 2001;27:415–444. [Google Scholar]
- 35.Paulozzi LJ, Ryan GW. Opioid analgesics and rates of fatal drug poisoning in the United States. Am J Prev Med. 2006;31(6):506–511. doi: 10.1016/j.amepre.2006.08.017. [DOI] [PubMed] [Google Scholar]
- 36.Morden NE, Munson JC, Colla CH et al. Prescription opioid use among disabled Medicare beneficiaries: intensity, trends, and regional variation. Med Care. 2014;52(9):852–859. doi: 10.1097/MLR.0000000000000183. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Cohen RA, Martinez ME. Health Insurance Coverage: Early Release of Estimates From the National Health Interview Survey, 2012. Hyattsville, MD: National Center for Health Statistics; 2013. [Google Scholar]
- 38.Paulozzi LJ, Xi Y. Recent changes in drug poisoning mortality in the United States by urban–rural status and by drug type. Pharmacoepidemiol Drug Saf. 2008;17(10):997–1005. doi: 10.1002/pds.1626. [DOI] [PubMed] [Google Scholar]
- 39.Rossen LM, Khan D, Warner M. Hot spots in mortality from drug poisoning in the United States, 2007–2009. Health Place. 2014;26:14–20. doi: 10.1016/j.healthplace.2013.11.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Byun SY, Meece JL, Irvin MJ. Rural–nonrural disparities in postsecondary educational attainment revisited. Am Educ Res J. 2012;49(3):412–437. doi: 10.3102/0002831211416344. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.National Cancer Institute. Policy for calculating Hispanic mortality for 1990+ data. Available at: http://seer.cancer.gov/seerstat/variables/mort/origin_recode_1990+. Accessed August 25, 2014.