Recent Patterns of Medication Use in the Ambulatory Adult Population of the United States: The Slone Survey

Context Data on the range of prescription and over-the-counter drug use in the United States are not available.

Objective To provide recent population-based information on use of all medications, including prescription and over-the-counter drugs, vitamins and minerals, and herbal preparations/natural supplements in the United States.

Design, Setting, and Participants Ongoing telephone survey of a random sample of the noninstitutionalized US population in the 48 continental states and the District of Columbia; data analyzed here were collected from February 1998 through December 1999.

Main Outcome Measure Use of medications, by type, during the preceding week, compared by demographic characteristics.

Results Among 2590 participants aged at least 18 years, 81% used at least 1 medication in the preceding week; 50% took at least 1 prescription drug; and 7% took 5 or more. The highest overall prevalence of medication use was among women aged at least 65 years, of whom 12% took at least 10 medications and 23% took at least 5 prescription drugs. Herbals/supplements were taken by 14% of the population. Among prescription drug users, 16% also took an herbal/supplement; the rate of concurrent use was highest for fluoxetine users, at 22%. Reasons for drug use varied widely, with hypertension and headache mentioned most often (9% for each). Vitamins/minerals were frequently used for nonspecific reasons such as "health" (35%); herbals/supplements were also most commonly used for "health" (16%).

Conclusions In any given week, most US adults take at least 1 medication, and many take multiple agents. The substantial overlap between use of prescription medications and herbals/supplements raises concern about unintended interactions. Documentation of usage patterns can provide a basis for improving the safety of medication use.

A large number and wide variety of medications approved for use by the US Food and Drug Administration (FDA) are available to the US population, and expenditures on drugs have increased dramatically in recent years.^1-3 New prescription drugs are continually introduced, and older drugs are increasingly available over the counter (OTC), making self-medication commonplace. Adverse reactions to drugs are among the leading causes of hospitalization and death in this country.^4,5 At the same time, there has been a considerable increase in the use of herbal products and other natural supplements^6,7 (henceforth referred to as "herbals/supplements"), which by law are not subject to FDA regulation. Although these products may be taken concurrently with regulated medications, health care professionals are often not informed of such use by their patients.⁷ Evidence is growing that many herbals/supplements have pharmacologic activity that can lead to clinically serious adverse interactions when they are taken together with regulated drugs,⁸ but there is little information available to estimate the potential magnitude of this problem.

More generally, despite the substantial commitment of resources from government, industry, and others to evaluate drug effects, no ongoing information exists on the degree to which the US population is using the broad range of medications, including prescription and OTC drugs, vitamins/minerals, and herbals/supplements. To help meet this need, the Slone Epidemiology Unit of Boston University is conducting an ongoing telephone survey of a random sample of the noninstitutionalized continental US population. In this first report, we focus on general patterns of medication use in the ambulatory US adult population in 1998 and 1999.

Residents of households in the 48 contiguous states and the District of Columbia are eligible for inclusion in the survey. Not eligible are those without telephones; individuals residing in vacation homes for less than 3 consecutive months, nursing homes, rehabilitation hospitals, or "group homes" (eg, halfway houses); and individuals in prisons, military barracks, or college/university dormitories without telephones in individual rooms.

Subjects are identified by random digit dialing (RDD); a 2-stage sampling procedure⁹ is used to generate the telephone numbers to be called. At least 20 attempts are made to contact the targeted numbers over a 1- to 2-month period before a number is considered unreachable. At each contacted number, 1 individual in the household is selected for interview by a computer-generated random number procedure. Persons 14 to 17 years old are interviewed only with the permission of a parent or guardian; for children younger than 14 years and subjects incapable of responding to the questions because of conditions such as Alzheimer disease, a surrogate (parent, spouse, or caretaker) who has knowledge of all the subject's medications is interviewed. Currently, interviews are conducted in English or Spanish. For subjects who only speak another language, the interview is conducted in English if someone is present who can act as an interpreter. In the future, the use of an interpreter service is planned.

Information is recorded on prescription and nonprescription drugs, vitamins/minerals, and herbals/supplements. The last group includes plant extracts except those marketed in regulated drugs (eg, senna laxatives), amino acids, animal extracts, enzymes, and other unclassified agents (eg, glucosamine).

The interviewer explains that information is being sought on any amount of use of all medications taken during the preceding 7 days and asks the subject to gather the relevant bottles or packages. A list of reasons for use (eg, pain/headache/backache, depression/tension/emotional disorders) is then read to elicit recall of medications not covered by the bottles. In addition, a short list of trade names of selected drugs is read. After the names of all reported medications have been recorded, the following information is obtained for each: reason for use, route of administration, number of days taken in the week before the interview, and total duration of the current episode of use. For products containing aspirin, acetaminophen, ibuprofen, or conjugated estrogens, the dose and number of pills per day are also obtained.

Other information elicited includes age, sex, race (using the 1990 US Census categories¹⁰), Hispanic origin, years of education, income (in ranges), whether the subject has prescription coverage through health insurance, ZIP code of residence, and for women aged 18 to 50 years, pregnancy status (including due date or last menstrual period).

Medication names are coded for analysis using a dictionary developed and maintained by the Slone Epidemiology Unit. The dictionary is a computerized linkage system composed of individual agents and multicomponent products, including herbals/supplements, each assigned specific code numbers. All combination products are linked to their individual components. Thus, groupings of drugs that contain a particular entity (eg, aspirin-containing products) can be automatically generated.

The survey was initiated in February 1998; this article covers data collection through the end of December 1999. During that time, we attempted to contact 10 354 telephone numbers, of which 5279 were clearly not eligible residential numbers, and 562 could not be contacted; we assumed that 10% of the latter (ie, 56 numbers) were residences, leaving 4569 eligible numbers. From these, there were 783 refusals, 207 unsuccessful attempts to interview for reasons other than refusal (eg, hearing problems or unable to interview within the specified time limit), the 56 not-contacted numbers assumed to be eligible, 318 with unresolved status when the data file was closed for analysis, and 3205 interviews. Some information was available about the characteristics of the 990 refusals and "unsuccessful interviews" in terms of age (known for 10%), sex (27%), number in household (32%), and region (100%). These nonparticipating subjects were somewhat more likely to be men than the participating subjects (54% vs 46%), but the distributions according to the other factors were similar. A total of 71 interviews were eventually conducted from the group of 318 with unresolved status, but are not included in this report. Thus, the overall participation rate was 72% of the eligible subjects (3276/4569).

From the 3205 interviews in the analytical file, we excluded 22 that were conducted early in the project because the random selection within the household was not done properly, and 3 in which the subject's age or sex was not recorded. Thus, 3180 interviews were available for analysis, of which 370 were completed in 1998 when the survey was conducted on a smaller (but still national) scale, and 2810 in 1999. Among 2590 subjects at least 18 years of age, 183 interviews were conducted with surrogates. Among 590 younger subjects, 493 interviews were conducted with parents.

We compared demographic information on all 3180 interviews to data from US Census projections for 1998 and 2000.^11-15 For the comparison of the racial/ethnic distribution, the 1990 Census data¹⁰ were used because the categories were defined differently in the 1998 and 2000 projections. The probability of selection within each household, which is inversely related to household size, was allowed for by weighting the survey data according to that factor. The subjects included in the Slone Survey generally reflected the US population. The distribution according to race was similar (white: 77% in the survey with 60 refusing to answer, 80% in the United States) and the proportion living within each of 4 broad regions (Northeast, Midwest, South, and West) was within 2% of the US Census figures. The age and sex distributions of study participants were similar to the United States (median age, 34 years; 35 years in the United States; 53% women, 51% in the United States). The Slone Survey included somewhat fewer single-person households (22% vs 26%) and fewer individuals of Hispanic origin (9% vs 12%). The only material differences were that survey subjects had more education (30% college graduate vs 21%) and higher annual household incomes (58% with at least $35 000 vs 55%). Thirteen percent of study participants refused to answer the income questions.

Our results are based on the 2590 subjects who were at least 18 years of age. One-week prevalence is reported, ie, any use during the 1-week period preceding the interview. We elected not to include topical vitamins/minerals and topical herbals/supplements in this analysis because of the likelihood of incomplete reporting of the use of such products. Topical drugs and eyedrops were included. All percentages are adjusted for household size. As examples of the precision of the results, in the full sample of 2590, estimates of 1% are accurate to ±0.4%, estimates of 10% to ±1.2%, estimates of 20% to ±1.5%, and estimates of 50% to ±2%. In the smaller subgroups (eg, women aged 45-64 years), the accuracy is lower.

As shown in Figure 1, 81% of the subjects had taken at least 1 medication (prescription or OTC drug, vitamin/mineral, or herbal/supplement) during the preceding week. Rates of use increased with age and were greater in women than men in every age group, with the exception that among persons 65 years or older, the proportions who took at least 10 medications were the same. The highest overall prevalence was among older women, of whom 94% had taken at least 1 medication, 57% took 5 or more, and 12% took 10 or more. The lowest prevalence was among 18- to 44-year-old men, of whom 68% took at least 1, 7% took at least 5, and fewer than 1% took 10 or more drugs in the preceding week. The overall prevalence of medication use varied somewhat by race/ethnicity: 84% among both whites and Native Americans, 76% among those who identified their race as black or African American, 75% among those of Hispanic origin, and 57% among Asian/Pacific islanders.

After adjustment for sex and household size, the overall prevalences for use of any medications, at least 5 medications, and 10 or more medications were within 2% of the results in Figure 1. The corresponding figures adjusting for education in addition to household size were within 1%. The exclusion of 183 surrogate interviews did not change the estimates.

Prescription drug use (excluding drugs that are available both by prescription and OTC, such as naproxen) is shown in Figure 2. Fifty percent of subjects reported taking at least 1 prescription drug during the previous week and 7% took 5 or more. Prescription drugs were used more frequently by women than men, and by older than younger persons. Among older women, 23% took 5 or more prescription drugs during the preceding week; among older men, 19% used at least 5 prescription drugs.

The 40 most commonly used prescription and OTC drug entities, taken either as single- or multiple-component products, are listed in Table 1. Although almost two thirds (27) are available only by prescription, 6 of the 10 most frequently used drugs, including the top 4, are available OTC. The most frequently used entities were acetaminophen, ibuprofen, and aspirin. Multiple-ingredient preparations accounted for 45% of acetaminophen and 26% of aspirin use. By contrast, only 2% of ibuprofen products were combination products. The most frequently used prescription-only drugs were conjugated estrogens (fifth overall), thyroid supplement (seventh), an estrogen contained in oral contraceptives (eighth), and a diuretic (10th).

Some of the drugs had strong sex- or age-specific patterns of use. For example, aspirin was most commonly taken by older men, among whom 58% of the users took it for cardiovascular prophylaxis. Aspirin use was also quite common among older women: 51% of those users took it for cardiovascular prophylaxis. Among middle-aged women who reported taking conjugated estrogens, 45% used medroxyprogesterone concomitantly. Atorvastatin, furosemide, warfarin, and digoxin were most frequently used by older men; levothyroxine, amlodipine, hydrochlorothiazide, triamterene, and diltiazem were most frequently taken by older women; diuretics and antihypertensives were more commonly used in the older age groups; ibuprofen and caffeine (not including caffeine in beverages or food) were more frequently used by younger subjects.

Vitamins/minerals were taken in the week before interview by 40% of the population. Table 2 presents rates for the 10 most commonly used entities. Multivitamins, defined as products containing at least 4 different vitamins (B vitamins are counted only once), were the most common, taken by about one fourth of the individuals. The listed specific vitamins and minerals reflect the prevalence reported for products other than multivitamins. The most common of these were vitamin E, vitamin C, and calcium, either alone or in combination with 2 or fewer other components. The total prevalence of use of specific compounds contained in multivitamins is the sum of the use of the individual agent and the use of multivitamins. Thus, for example, the total usage of vitamin E was approximately 36% (26% multivitamin plus 10% other vitamin E use), and of folic acid, 28%. The prevalence of vitamin and mineral use was higher among middle-aged and older subjects of both sexes, and generally higher among women than among men.

Fourteen percent of the population had taken at least 1 herbal/supplement in the preceding week. As shown in Table 3, overall use for each of the 10 most common entities ranged between 0.9% and 3.3%. Among both men and women, the most frequent users of these preparations tended to be middle-aged. Exceptions include creatine, which was most commonly taken by younger men; glucosamine and Serenoa repens, taken by older men; and Ginkgo biloba, taken by older women.

Table 4 provides the 10 most commonly reported reasons for use of prescription and OTC drugs, vitamins/minerals, and herbals/supplements. For regulated drugs, these were hypertension, headache, "heart," allergy, and unspecified pain. Three percent of the reasons for use were reported as "don't know/no reason specified." By far the most common reason given for use of vitamins/minerals was "health/good for you"; the next most common was "diet supplement." Prophylaxis was a frequently cited reason, most often for osteoporosis and colds/influenza. For 3% of vitamin/mineral use, the reason was "don't know/no reason specified." For herbals/supplements, the most frequently reported reasons were "health/good for you," arthritis, and memory improvement; "don't know/no reason specified" was reported for 2% of the use.

Lacking diagnostic documentation, the survey data do not provide medically confirmed information about the appropriateness of drug use. Rather, the survey was designed to reflect respondents' perceived reasons for why they were taking specific medications. Nonetheless, insights about whether drugs are being used for their intended purposes can be gained. As one example, 95% of the drugs taken for hypertension were products approved for that indication (eg, atenolol, amlodipine). Examined from a different perspective, among individuals who had used lisinopril (the most frequently reported antihypertensive drug), 78% gave hypertension as the reason, 12% reported indications possibly related to hypertension but less clearly so (eg, unspecified heart disease), and the remaining 11% either did not know the reason or stated that their physician had recommended the drug. Patterns observed for other reported reasons for use consistently reflected that patients' perceptions generally correlated with the approved indications for the drugs they were taking.

Overall, 16% of prescription drug users also reported use of 1 or more herbals/supplements. Among users of the 40 most common prescription and OTC drugs (Table 1), concomitant use of herbals/supplements during the same 1-week period ranged from 7% among paroxetine users to 20% or more for users of simvastatin (20%), conjugated estrogens (21%), and fluoxetine (22%).

With rapidly increasing medical care costs, concerns are growing about the benefits and risks of the wide range of prescription and OTC medications. Dramatic increases in the use of herbals/supplements has prompted similar concerns about those products, both alone and in combination with FDA-regulated pharmaceuticals. However, verification and quantification of the clinical, research, and public health implications of these concerns has been limited by the absence of comprehensive information on exposures to the full range of medications in the general population. This survey was developed to provide such information.

Our findings confirm that large numbers of US adults take large quantities of medications. In 1998-1999, more than 80% of the adult population took at least 1 medication in the preceding week, and 25% took at least 5. These rates translate into 169 million and 52 million individuals, respectively. Even among the group with the lowest rate of use, men 18 to 44 years of age, more than two thirds took at least 1 medication during a given week. Use of prescription drugs is also widespread: half of US adults took at least 1 drug and 7% took at least 5. For the Medicare population, women 65 years or older had the highest prevalence of medication use in our data: 94% took at least 1, 57% took 5 or more, and 12% took 10 or more, translating to 19, 12, and 1.4 million older women, respectively. Use of prescription drugs was also highest among this group, with rates of 81% and 23%, respectively, for at least 1 and at least 5 drugs. Men 65 years or older also reported high prevalence of medication use, although not quite as high as among older women.

Among FDA-regulated drugs, OTC analgesics were the most frequently used individual products, taken by 17% to 23% of the population. Prescription drugs predominated among the remaining commonly used compounds. The most common prescription drugs among women were conjugated estrogens and thyroid supplements, while among men they were cardiovascular drugs and diuretics. Antiulcer agents and cholesterol-lowering drugs were also commonly used prescription drugs. Among OTC agents, decongestants and antihistamines followed analgesics in frequency of use.

Forty percent of the population took 1 or more vitamin or mineral supplements (the same prevalence reported from the National Health and Nutrition Examination Survey [NHANES] III, a national survey conducted in 1988-1994¹⁶). The most commonly used products, by far, were multivitamins (26%).

Herbals/supplements are widely used in the United States: 14% of the survey respondents had used at least 1 of these products in the preceding week. The most common individual herb, ginseng, was taken by 3.3% of the population. In a survey conducted in 1997, Eisenberg et al⁷ observed a prevalence of 12% for use of any herbal preparation in a 1-year period. That represented a substantial increase from the 2.5% prevalence reported by the same investigators from a similar survey conducted in 1990.⁶ The figures reported by Eisenberg et al are not directly comparable to those in this survey, since the exposure period was 1 year rather than 1 week, and the definition of "herbal preparations" may not have included some of the other natural supplements that we considered. More directly comparable are the NHANES III data from 1988-1994, in which the 1-month prevalence of use of any supplement other than vitamins or minerals was 3.6%,¹⁷ substantially lower than our current observation of a 14% 1-week prevalence from a decade later.

While information on indications for drug use is typically derived from medical sources or inferred from the product name, our data provide the users' own perceptions of reasons they are taking medications. The distribution of stated reasons for use varied across type of medication, but our examination of selected prescription drugs showed that, with few exceptions, patients report indications that are compatible with the prescribers' likely intent. For vitamins/minerals, which are commonly self-prescribed, almost half of all use was accounted for by "health/good for you" and "diet supplement." "Health/good for you" was also a common reason for taking herbals/supplements, although the large majority of users were taking these products to achieve more specific health goals. For all 3 classes of medications, 2% to 3% of users did not know or could not specify a reason they were taking the product.

As increasing numbers of reports describe clinically serious interactions between prescription drugs and herbals/supplements and as more is learned about the pharmacologic actions and metabolic pathways of the latter compounds, concern has grown that the problem of interactions may be substantial but unrecognized.⁸ Our data document that patients taking prescription drugs frequently take herbals/supplements concurrently, lending support and direction to those concerns. This finding also emphasizes the importance of clinicians inquiring about their patients' use of herbals/supplements, since knowledge of concurrent use can help identify patients at risk for potential interactions.

To our knowledge, this is the first population-based survey to provide comprehensive information on use of the broad range of medications, including herbals/supplements, in the United States. Until now, most information on use of regulated drugs has been derived from sales data or reports of dispensed prescriptions. These sources are typically proprietary and unpublished, and they do not measure use by individual patients. Overall rankings of specific drug products by sales are readily available,^18,19 but that information does not necessarily reflect actual use of the drugs. Sales data do not provide information on characteristics of users or details such as perceived indications or concurrent medications that have direct clinical relevance to compliance and drug interactions. The product sales rankings are not directly comparable to our results, which provide information on the actual use of drug entities that may be contained in multiple products.

In recent years, limited population-based surveys have focused on use of selected drugs in subpopulations,^20-24 and a few have provided a wider range of information on subpopulations, particularly the elderly.^25-27 There have also been intermittent surveys of the general population regarding specific classes of drugs or supplements, eg, dietary supplements in 1990⁶ and again in 1997.⁷ The NHANES III^16,17,28-30 was a more general survey in which a nationally representative population was interviewed during 1988-1994, with information obtained on use in the prior month of all prescription drugs and dietary supplements. However, only selected OTC drugs were included.^16,17,28

The survey population in this report is similar to the US population based on comparisons with US Census data. Most differences identified were small. Our study included fewer individuals of Hispanic origin than the general population, but otherwise the ethnic distribution of the survey population was similar to the US Census figures. The survey population was generally representative in terms of household size, except for somewhat fewer single-person households, which may reflect the greater difficulty in reaching a household occupied by only 1 person. The largest difference was a deficit of survey subjects from lower socioeconomic levels. Both this underrepresentation and the slight excess of women participants are known aspects of RDD surveys.³¹ The overall results after adjustment for education and sex were similar to the unadjusted findings.

The possibility of bias due to differences in medication use between participants and nonparticipants cannot be excluded. However, our participation rate of 72% is quite high for such studies: participation in academic RDD surveys has been reported to range from 22% to 70%, with a median of 50%.³² The information that was available on the characteristics of the nonparticipants indicated that the only material difference was a modest excess of men. The overall sex-adjusted usage rates were similar to the unadjusted rates, and most results were provided separately for men and women.

Another potential source of error is differential reporting of the use of specific products. The interview process was carefully designed to minimize this problem. All interviewers were rigorously trained to conduct the interviews in a consistent manner, the exposure interval of interest was brief and recent, and subjects were asked to confirm the names of as many medications as possible from containers. Thus, we believe that the data from the Slone Survey provide an accurate picture of medication use in the US population.

In conclusion, our data document that in any given week, the large majority of US adults take at least 1 prescription or OTC medication, and substantial numbers take multiple products. Forty percent take at least 1 vitamin/mineral, and 1 in 7 adults takes at least 1 herbal/supplement. One in 6 patients taking prescription drugs is concurrently taking 1 or more herbals/supplements, which poses potential risks of interactions. Identification of usage patterns for the full range of medications, including the reasons for use, provides a basis for improving safety and reducing risks associated with their consumption.