Pnhp.org

Characteristics of Recipients of Free Prescription Drug Samples: A Nationally Representative Analysis | Sarah L. Cutrona, MD, MPH, Steffie Woolhandler, MD, MPH, Karen E. Lasser, MD, MPH, David H. Bor, MD, Danny McCormick, MD, MPH, and David Free prescription drug samples are used widely Objectives. Free prescription drug samples are used widely in the United States.
in the United States. The retail value of drug We sought to examine characteristics of free drug sample recipients nationwide.
samples distributed in the United States totaled Methods. We analyzed data on 32 681 US residents from the 2003 Medical Ex- more than $4.9 billion in 1996 and climbed to penditure Panel Survey (MEPS), a nationally representative survey.
Results. In 2003, 12% of Americans received at least 1 free sample. A higher pro- Controversy surrounds the use of free sam- portion of persons who had continuous health insurance received a free sample ples.4 Studies have described potential safety (12.9%) than did persons who were uninsured for part or all of the year (9.9%; P<.001).
The poorest third of respondents were less likely to receive free samples than were problems,5,6 health professionals who divert those with incomes at 400% of the federal poverty level or higher. After we con- samples for self-administration or resale,7–10 trolled for demographic factors, we found that neither insurance status nor income the influence of pharmaceutical representa- were predictors of the receipt of drug samples. Persons who were uninsured all or tives who distribute samples,11–13 and the con- part of the year were no more likely to receive free samples (odds ratio [OR]=0.98; tribution of samples to rising drug and health 95% confidence interval [CI]=0.087, 1.11) than those who were continuously insured.
insurance costs.14–16 In addition, numerous Conclusions. Poor and uninsured Americans are less likely than wealthy or in- studies suggest that free samples may influ- sured Americans to receive free drug samples. Our findings suggest that free ence the prescribing behavior of physicians drug samples serve as a marketing tool, not as a safety net. (Am J Public Health. and trainees.7,14,16–22 In its most recent report, 2008;98:284–289. doi:10.2105/AJPH.2007.114249) the Institute of Medicine has called for furtherinvestigation of sample use, citing concernsover patient safety, provider prescribing hab- Component. MEPS is a nationally representa- its, and consumer adherence to prescribed tive longitudinal survey of the civilian nonin- participants to name all filled prescriptions re- stitutionalized US population. The MEPS co- ceived in conjunction with a hospital discharge, Nonetheless, many physicians believe that emergency department visit, or medical outpa- samples allow them to give free medications previous year’s National Health Interview tient visit. Surveyors then ask respondents to to their neediest patients.10,15 This view is also Survey, conducted by the National Center for name any medications purchased or received Health Statistics at the Centers for Disease that have not already been listed. The sur- Manufacturers of America, whose vice presi- Control and Prevention. The National Health veyor then asked: “Since [the last interview] dent wrote in the New York Times, “many Interview Survey uses a stratified, multistage did [you] get any free samples of prescribed uninsured and low-income patients benefit probability cluster sampling design with an medicines from a medical or dental provider from these free samples, which often serve as oversampling of Blacks and Hispanics.27 The that we have not yet talked about?”28 MEPS defines free samples as “limited amounts of a However, few data are available on recipi- prescription medication which are given out by ents of free samples. Although a few studies doctors to patients free of charge, sometimes in have looked at the receipt of free samples in over 2.5 years. Interviewers travel to the lieu of a written or verbal prescription.”28 selected populations,6,25,26 no national study homes of respondents and conduct in-person, If a respondent answers “yes” to this ques- has examined this issue. We analyzed the re- computer-assisted interviews. The MEPS sur- tion, the names of any medicines received as ceipt of free samples using nationally repre- veyors collect detailed information on health sentative data from the United States in 2003 care expenditures, health care utilization, We were interested in 3 questions that re- to determine the characteristics of free sam- health insurance, and sociodemographic char- quired us to analyze the complex MEPS data acteristics, as well as information on all outpa- set in different ways: (1) Are free drug sam- ples more frequently given to uninsured and low-income persons than to insured and afflu- search and Quality provides weights that ad- ent persons? (2) Does type of drug coverage just for the complex sample design and sur- influence the likelihood of receiving free sam- Research and Quality’s 2003 Medical Ex- vey nonresponse and facilitate extrapolation 284 | Research and Practice | Peer Reviewed | Cutrona et al. American Journal of Public Health | February 2008, Vol 98, No. 2 initially including our income and insurance atory multivariate analysis to evaluate the variables in the logistic regression model. We whether free drug samples were given more role of potential intermediary variables re- then entered into the model all demographic frequently to uninsured and low-income per- lated to access to care: site of usual medical variables that were significant on bivariate sons. First, we analyzed bivariate associations care (hospital based vs office vs no usual site analysis (P < .1) or that we considered to be between receipt of at least 1 free sample in of care) and total number of prescription clinically significant. We ran a secondary mul- 2003, and insurance status and income, re- medications received including refills.
tivariate model that included income, insur- spectively. For this analysis, we classified re- ance, and all significant demographic vari- spondents as “insured all year” if they indi- ables and added (1) site of usual medical care We then examined the effect of prescription and (2) total number of prescription medica- drug coverage on sample receipt. For this anal- medical insurance, or both, for every month ysis we focused on a single round of MEPS in- of 2003. We classified respondents as “unin- We used the SAS version 9.1 (SAS Institute terviews that collected data for the preceding sured part or all year” if they reported having Inc, Cary, NC). To account for sample design 2 to 6 months. We conducted this single-round insurance during some but not all months or effects, we used SAS survey commands that analysis to identify as accurately as possible if they had no insurance during any month of make it possible to estimate confidence intervals the type of drug coverage at the time any sam- 2003. We excluded individuals for whom in- in the presence of stratification and clustering.
ple was received. We analyzed bivariate associ- ations between type of drug coverage during available for all of 2003 (2.8% of respon- the interview round and receipt of at least 1 those with family incomes less than 200% of the federal poverty line, middle-income per- health insurance and no drug coverage, (2) Medicaid at any point in the round, (3) non- received at least 1 drug as a free sample.
Medicare private insurance with drug coverage, income persons as those whose incomes were (4) non-Medicare private insurance without 400% of the poverty line or higher. In 2003, health care use characteristics of sample re- drug coverage, (5) Medicare with supplemental the federal poverty line was set at $18 400 cipients and nonrecipients. Low-income re- drug coverage, and (6) Medicare without sup- spondents who were uninsured all or part of plemental drug coverage. We chose the inter- 2003 were less likely to receive free samples view round for our analysis by selecting the of other demographic features on the rela- than were high-income and insured respon- only MEPS interview that collected data for a tion between receipt of free drug samples dents. Among persons who were insured all and insurance status or income. We devel- year, 12.9% received a sample, versus 9.9% of those uninsured for part or all of the year For our estimate of the most frequently dis- same definitions for outcome (receipt of at (P < .001). Of all persons who received a sam- tributed drugs, we reviewed the names of all least 1 free sample in 2003), insurance clas- ple, 82.1% were insured all year; only 17.9% medications given as samples during calendar sification, and income as were used in our of sample recipients were uninsured for all or year 2003. To provide a comparison, we re- bivariate analyses. We examined the effect part of the year. Similarly, of all sample recipi- of insurance and income on receipt of free ents, 71.9% had an income 200% or more of data. Because the MEPS data do not indicate samples and we controlled for demographic the federal poverty line, whereas 28.1% had features including age, gender, race, His- patient received, we were able to estimate the panic ethnicity, place of birth, education line. The poor were the least likely to receive level, and language spoken. Information on free samples, whereas individuals in the high- ications but were not able to establish an all demographic features, including ethnicity est income category were the most likely to and race, was provided by the respondents receive free samples (10.8% of low-income through the survey questionnaire. To deter- persons received at least 1 sample vs 12.3% mine Hispanic ethnicity, respondents were asked to characterize themselves as either received free drug samples in 2003 as a per- higher-income persons; P < .001 for ordered centage of all respondents and as a percentage race, respondents were asked to character- of all those taking 1 or more prescription Other races, Hispanics, non-English speak- drugs. We used the χ2 test to study the bivari- Indian/Alaska Native, Asian, Native Hawaiian/ ate association between categorical predictors States were less likely to receive a free sample Pacific Islander, or multiple races. For our States, respectively. Respondents who usually February 2008, Vol 98, No. 2 | American Journal of Public Health Cutrona et al. | Peer Reviewed | Research and Practice | 285 received their medical care in an office were TABLE 1—Percentage of Respondents Who Received at Least 1 Free Prescription Drug
Sample in 2003, by Demographic Group: Medical Expenditure Panel Survey, 2003
tal clinics or hospital emergency depart- provider (6.3%; P < .001). Persons who Total persons who received prescription drugs (n = 19 848) of those insured continuously; P < .001) medical care in an office (14.5% of unin- sured part or all year vs 85.5% of contin- uously insured; P < .001). They were also less likely to report receiving medication < 200% of poverty line and uninsured part or all year sured continuously; P < .001).
≥200% of poverty line or insured all year Table 2 presents the results of our multi- variate analyses of sample receipt. In our principle model, we analyzed income and in- surance as predictors of the receipt of free drug samples; we also controlled for age, gen- der, race, Hispanic versus non-Hispanic eth- nicity, place of birth (United States vs foreign born), education level, and language spoken.
Persons who were uninsured for part or all of the year were no more likely to receive free samples (odds ratio [OR] = 0.98; 95% confi- dence interval [CI] = 0.87, 1.11) than were those continuously insured. Likewise, being in the lowest income group was not a significant predictor of sample receipt (OR = 1.05; 95% demographic variables and adding 2 inter- health care: the number of prescriptionmedications received and the site of usual were uninsured for part or all of the year were more likely than those insured contin- uously to receive free samples (OR = 1.25; 95% CI = 1.10, 1.43). The association be- In our bivariate analysis of type of drug coverage and receipt of free drug samples Hospital (clinic or emergency department) (Table 3), respondents with Medicaid at any point in the round had the lowest likelihood of receiving a sample (4.12%), followed bythose with no insurance coverage (4.66%).
286 | Research and Practice | Peer Reviewed | Cutrona et al. American Journal of Public Health | February 2008, Vol 98, No. 2 pills received as samples and, therefore, we TABLE 1—Continued
were unable to determine the percentage of No. of prescription medications in 2003 (by quartile for those who total medications represented by free sam- ples. Such information would be useful to ob- tain in future studies. We may have underesti- have forgotten to report samples that they re- ceived for brief durations earlier in the inter-view reference period, although the relatively Note. CI = confidence interval.
aWeighted percentages are representative of the noninstitutionalized US civilian population. Totals may not add to 100 short duration of interview reference periods (ranging from 2- to 6-month intervals) should bP < .001, for χ2 analysis measuring difference between categories for this variable.
have minimized recall bias. Poor or uninsuredrespondents may have perceived receipt of free samples as shameful or embarrassing and named free drug samples in 2003 were: (1) gests that the relation of health coverage and underreported these events. It is not our ex- Lipitor (atorvastatin), (2) Allegra (fexofenadine), affluence to sample receipt is mediated by 2 perience, however, that free samples carry and (3) Advair diskus (fluticasone/salmeterol).
access-related factors: site of usual medical such a stigma. Free samples obtained directly The 3 most frequently prescribed drug samples care and total use of pharmacotherapy. Office- from manufacturers by mail order may have in 2002 were: (1) Vioxx (rofecoxib), (2) Lipitor been undercounted, but as of 2002, the ma- (atorvastatin), and (3) Celebrex (celecoxib).
likely to have received at least 1 free drug jority of such programs required that applica- sample. If we include site of medical care in tions be filled out by a physician and (in ap- our multivariate model, uninsured persons ap- pear more likely to receive a free sample than delivered to the doctor’s office.29 Hence, we To our knowledge, ours was the first popu- do insured persons. We interpret this finding believe that many, perhaps most, of such free lation-based study of free drug sample distri- to reflect office-based practitioners’ sincere ef- medications would be classified as free sam- bution. We found that 12% of US residents fort to give free samples to their neediest pa- received free samples during 2003 but less tients. Unfortunately, these efforts do not ap- than one third of all sample recipients were pear to compensate for larger access barriers lowed by filled prescriptions within a single low income and less than one fifth of all sam- that prevent uninsured and other disadvan- ple recipients were uninsured at any point taged patients from consulting physicians who undercounted, because the interviewer asked during the year. Indeed, the poor were less are office based. People who were uninsured respondents about free samples received only likely to receive free samples than were those in 2003 were more likely to use hospital clin- after reviewing filled prescriptions. If 2 pa- ics or hospital emergency departments or to tients were each given a free sample along were less likely to receive free samples than report no usual source of care and were less with prescriptions to be filled, the patient with likely to have purchased or received medica- lower income and no insurance is probably Several other vulnerable groups, including tion compared with people who were insured.
less likely to fill the prescription because of Other races, Hispanics, non-English-speakers, Previous studies have looked at receipt of and persons born outside the United States were free samples in selected populations and gen- therefore be more likely to report having re- also less likely to receive a free sample. In a erated similar findings. Stevens et al.25 found ceived a sample in our survey design; if so, study of Medicare patients in Hawaii, Taira et al.
that insured adults with asthma were more our study may understate the relation of so- similarly found that being White was associated likely to receive samples than were their unin- with a greater likelihood of receiving a drug sured counterparts. A survey of elderly en- sample.26 Although overt discrimination might rollees in a single health insurer in Hawaii26 likely to receive free samples than those explain the racial and ethnic disparities, we sus- found that 50% to 60% had received a free pect that they reflect unmeasured differences in sample in the previous 12 months. That study, with insurance coverage. Although physicians overall access to care. Persons from these minor- like ours, found that race, ethnicity, and age ity groups may also be seeing providers who dis- were associated with likelihood of receiving a enter their offices, these individual efforts fail tribute fewer samples. We found that women sample. Lack of drug coverage among insured to counteract society-wide factors that deter- and older persons had a greater likelihood of re- persons was also associated with greater likeli- mine access to care and selectively direct free ceiving samples, which was possibly a reflection samples to the affluent. Our findings suggest of increased use of health care services by these Our study had several limitations. We did that free drug samples serve as a marketing not have information on the total number of February 2008, Vol 98, No. 2 | American Journal of Public Health Cutrona et al. | Peer Reviewed | Research and Practice | 287 TABLE 2—Multivariate Odds of Free Drug Sample Receipt in 2003: Medical Expenditure
All of the authors are with the Department of Medicine, Panel Survey, 2003
Cambridge Health Alliance, Cambridge, Mass, and theHarvard Medical School, Cambridge. Requests for reprints should be sent to Sarah L. Cutrona, Department of Medicine, Cambridge Hospital, 1493 Cambridge St, Cambridge, MA 02139 (e-mail: This article was accepted May 22, 2007. S.L. Cutrona, S. Woolhandler, and D.U. Himmelstein per- formed the statistical analysis. D.H. Bor provided supervi- sion and obtained funding. S.L. Cutrona has had full ac-cess to all the data in the study and has final responsibility for the decision to submit for publication. All of the au- thors participated in designing the study, analyzing and interpreting the data, writing and revising the article. This work was supported by a National Research Ser- We are indebted to Amy Cohen, Department of In- formation Technology, and E. John Orav, Department of Biostatistics, both at the Harvard School of Public Health, for advice on statistical programming. We are also indebted to Neal S. LeLeiko, from the Departmentof Gastroenterology and Nutrition at Hasbro Children’s Hospital for valuable discussions and careful reading of This study was deemed exempt from review by theCambridge Health Alliance institutional review board.
Donohue JM, Cevaso M, Rosenthal MB. A decade of direct-to-consumer advertising of prescription drugs.
N Engl J Med. 2007;357(7):673-681.
Rosenthal MB, Berndt ER, Donohue JM, Frank RG, Epstein AM. Promotion of prescription drugs to consumers. N Engl J Med. 2002;346:498–505.
Ma J, Stafford RS, Cockburn IM, Finkelstein SN. A statistical analysis of the magnitude and composition of drug promotion in the United States in 1998. ClinTher. 2003;25:1503–1517.
Charatan F. Hospital bans free drug samples. West Dill JL, Generali JA. Medication sample labeling practices. Am J Health Syst Pharm. 2000;57: Backer EL, Lebsack JA, Van Tonder RJ, Crabtr BF. The value of pharmaceutical representative visits and medication samples in community-based family practices. J Fam Pract. 2000;49:811–816.
Morelli D, Koenigsberg MR. Sample medication Note. OR = odds ratio; CI = confidence interval.
dispensing in a residency practice. J Fam Pract. 1992; aThis model included insurance status and income, and controlled for age, gender, race, Hispanic ethnicity, place of birth, education level, and language spoken.
Westfall JM, McCabe J, Nicholas RA. Personal use bThis model controlled for the factors controlled for in the principle model as well as site of usual medical care (hospital of drug samples by physicians and office staff. JAMA.
1997;278:141–143.
based vs office vs no usual site of care) and total number of prescription medications received including refills.
Tong KL, Lien CY. Do pharmaceutical representa- OR for each incremental increase of 1 drug.
tives misuse their drug samples? Can Fam Physician.
1995;41:1363–1366.
288 | Research and Practice | Peer Reviewed | Cutrona et al. American Journal of Public Health | February 2008, Vol 98, No. 2 TABLE 3—Percentage of Respondents Receiving at Least 1 Free Drug Sample in a Single
TABLE 4—Most Frequently Reported
Interview Round in 2003, by Detailed Insurance Coverage: Medical Expenditure Panel
Free Drug Samples: Medical
Survey, 2003
Expenditure Panel Survey 2002–2003
Received at Least 1 Free Sample, % (95% CI) Non-Medicare private insurance with drug coverage Non-Medicare private insurance without drug coverage Medicare with private supplemental drug coverage 5. Nexium (esomeprazole) 4. Celebrex (celecoxib) Medicare without supplemental drug coverage Note. CI = confidence interval.
8. Toprol XL (metoprolol) 7. Toprol XL (metoprolol) 10. Peterson MC, Ebbert TL, Edwards MW, Willmore 23. Institute of Medicine. Preventing Medication Errors: J. Disposition of pharmaceutical samples from a private Quality Chasm Series. Washington, DC: National Acade- medical clinic. J Am Pharm Assoc. 2004;44:397–398.
11. Andaleeb SS, Tallman RF. Relationships of physi- 24. Johnson K. Drug samples to doctors. New York cians with pharmaceutical sales representatives and pharmaceutical companies: an exploratory study.
25. Stevens D, Sharma K, Kesten S. Insurance status Health Mark Q. 1996;13:79–89.
and patient behavior with asthma medications. J 12. Avorn J, Soumerai SB. Improving drug-therapy de- cisions through educational outreach. A randomized 26. Taira DA, Iwane KA, Chung RS. Prescription controlled trial of academically based “detailing.” N drugs: elderly enrollee reports of financial access, re- Engl J Med. 1983;308:1457–1463.
ceipt of free samples, and discussion of generic equiva- 13. Soumerai SB, Avorn, J. Economic and policy anal- lents related to type of coverage. Am J Manag Care. ysis of university-based drug “detailing.” Med Care. 27. Cohen SB, DiGaetano R, Goksel H. Estimation 14 Chew LD, O’Young TS, Hazlet TK, Bradley KA, procedures in the 1996 Medical Expenditure Panel Maynard C, Lessler DS. A physician survey of the ef- Survey Household Component. Rockville, Md: Agency fect of drug sample availability on physicians’ behavior.
for Health Care Policy and Research; 1999. Available J Gen Intern Med. 2000;15:478–483.
at: http://207.188.212.220/mepsweb/data_files/publi-cations/mr5/mr5.shtml. Accessed September 12, 15. Maguire P. Samples: cost-driver or safety net? ACP Observer (online); 2001. Available at: http://www.acponline.org/journals/news/ jan01/drugsam- 28. Medical Expenditure Panel Survey. HC-077A: 2003 Prescribed Medicines. Rockville, MD: Agency for 16. Adair RF, Holmgren LR. Do drug samples influ- Healthcare Research and Quality; 2005. Available at: ence resident prescribing behavior? A randomized trial.
http://www.meps.ahrq.gov/mepsweb/survey_comp/ Am J Med. 2005;118:881–884.
hc_survey/2003/PM-p7r5.pdf. Accessed September 8,2006.
17. Hall KB, Tett SE, Nissen LM. Perceptions of theinfluence of prescription medicine samples on prescrib- 29. Johnson D. Free (or Almost Free) Prescription Med- ing by family physicians. Med Care. Apr 44:383–387, ications: Where and How to Get Them. San Francisco, Calif: Robert D. Reed Publishers; 2002.
18. Shaughnessy AF, Bucci KK. Drug samples andfamily practice residents. Ann Pharmacother. 1997;31:1296–1300.
19. Brewer D. The effect of drug sampling policies onresidents’ prescribing. Fam Med. 1998;30:482–486.
20. Hodges B. Interactions with the pharmaceuticalindustry: experiences and attitudes of psychiatry resi-dents, interns and clerks. Can Med Assoc J. 1995;153:553–559.
21. Reeder M, Dougherty J, White LJ. Pharmaceuticalrepresentatives and emergency medicine residents: anational survey. Ann Emerg Med. 1993;22:1593–1596.
22. Boltri JM, Gordon ER, Vogel RL. Effect of antihy-pertensive samples on physician prescribing patterns.
Fam Med. 2002;34:729–731.
February 2008, Vol 98, No. 2 | American Journal of Public Health Cutrona et al. | Peer Reviewed | Research and Practice | 289

Source: http://www.pnhp.org/PDF_files/DrugSampAJPH08.pdf