Patient characteristics associated with multiple pharmacy use in the U.S. population: Findings from the Medical Expenditure Panel Survey

Research in Social and Administrative Pharmacy 11 (2015) 507–516

Original Research

Patient characteristics associated with multiple pharmacy use in the U.S. population: Findings from the Medical Expenditure Panel Survey Kevin A. Look, Pharm.D., Ph.D.* Social and Administrative Sciences Division, University of Wisconsin School of Pharmacy, 777 Highland Ave., Madison, WI 53705-2222, USA

Abstract Background: Multiple pharmacy use (MPU) is an important safety and quality issue, as it results in fragmented patient care. However, few studies have examined patient characteristics predicting the use of multiple pharmacies, and the findings have been inconsistent. Objectives: To identify patient characteristics associated with MPU using national data. Methods: Data were obtained from the 2011 U.S. Medical Expenditure Panel Survey. The dependent variable was MPU, or the use of more than one pharmacy. The Andersen Behavioral Model of Health Service Use was used to guide the selection of independent variables, which were categorized as predisposing, enabling, and medical need related characteristics. Multivariable logistic regression analysis was conducted to identify the relationships between predisposing, enabling, and need variables and MPU in a hierarchical fashion. Point estimates were weighted to the U.S. non-institutionalized population, and to adjust standard errors to account for the complex survey design. Results: MPU was common, with a national prevalence of 41.3%. Individuals aged 40–64 and adults 65 and older were significantly less likely to use multiple pharmacies as patients aged 18–39 years (40–64 years OR: 0.67, CI: 0.58–0.77; R65 years OR: 0.49, CI: 0.41–0.58). Females were significantly more likely to use multiple pharmacies than males (OR: 1.16, CI: 1.05–1.29). Individuals lacking health insurance were more likely to use multiple pharmacies as individuals with private health insurance (OR: 1.42, CI: 1.16–1.73); in contrast, individuals having drug insurance were more likely to use multiple pharmacies (OR: 1.25, CI: 1.06–1.47) relative to those without drug insurance. Any mail order use was the strongest predictor of MPU (OR: 6.94, CI: 5.90–8.18). Conclusions: Pharmacists and other health care providers need to be aware that their patients may be using multiple pharmacies, especially younger patients, those lacking access to health insurance, or those using mail order pharmacies. The findings from this study can be used to identify patients that may need additional monitoring to ensure safe and appropriate drug therapy, and has important implications as health care continues to shift toward performance-based reimbursement and quality ratings. Ó 2015 Elsevier Inc. All rights reserved. Keywords: Multiple pharmacy use; Pharmacy patronage; Medical expenditure panel survey

* Corresponding author. Tel.: þ1 608 890 0367; fax: þ1 608 262 5262. E-mail address: [email protected]. 1551-7411/$ - see front matter Ó 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.sapharm.2014.10.004

508

Look / Research in Social and Administrative Pharmacy 11 (2015) 507–516

Background Multiple pharmacy use (MPU) is an important safety and quality issue for patients, pharmacists, and prescribers. It results in fragmented patient care, because an individual pharmacy’s records may not contain medications dispensed at other pharmacies, which decreases the pharmacist’s ability to monitor for drug interactions and provide appropriate patient consultation.1–4 Of particular concern for prescribers, hospitals, and health plans, MPU has been linked to many undesirable outcomes affecting patient safety or quality of care, including inappropriate drug use and adverse drug reactions,4–6 decreased patient adherence,5,7,8 increased mortality,3 and an increase in ambulatory care and hospital expenditures.2,3,9 The use of multiple pharmacies is common, with 38.1% of older adults with Medicare Part D using multiple pharmacies in 2009.5 It is also becoming increasingly prevalent on a national level, with the prevalence of MPU among adults 18 years and older increasing from 36.4% in 2003 to 43.2% in 2009, a relative increase of 18.7%.10 Although several studies have identified potential associations between MPU and demographic characteristics, these findings have been inconsistent among the few studies that have been conducted. A 1984 national survey found that younger age, male gender, higher family income, and higher levels of education were associated with higher prevalence of MPU.11 Male gender was significantly associated with higher prevalence of MPU in a second survey of elderly adults in Texas.12 A 2003 national survey of non-institutionalized Medicare beneficiaries found significantly lower prevalence of MPU among uninsured and low income beneficiaries, and significantly higher prevalence among beneficiaries with multiple chronic medical conditions.13 MPU was found to be higher among patients 85 years and older compared with patients aged 65–74 years in a study of Medicare beneficiaries in Pennsylvania enrolled in a pharmaceutical assistance program.14 Additionally, the authors found that patients who used R15 unique medications were twice as likely to use multiple pharmacies compared with those who used %5 unique medications. Finally, a recent study of Medicare Part D beneficiaries found that multiple pharmacy users were significantly younger, used more unique medications, had a higher prevalence of all chronic conditions assessed, had more prescribers, and were more likely to use mail order as their primary pharmacy than single pharmacy

users.4 Rural-urban differences in MPU have been inconsistent, with one study finding higher prevalence in central city pharmacies compared to rural pharmacies,15 and one study finding no difference between individuals living in urban and rural areas.16 Much of the current evidence for characteristics associated with MPU comes from dated studies conducted over a broad period between 10 and 30 years ago,4,11–13,15,16 with several of these studies being conducted outside of the U.S.4,15 As such, the purpose of this study was to identify characteristics associated with MPU using data from a 2011 U.S. national survey of health care use. Of particular interest were differences in MPU between younger and older adults. This information will help pharmacists, prescribers, and health plans identify patients who may be at the highest risk of MPU, so that they can better manage or coordinate patient care to ensure safe and appropriate medication use. Methods Data source This study used data from the 2011 U.S. Medical Expenditure Panel Survey (MEPS), which is a nationally representative survey sponsored by the Agency for Healthcare Research and Quality that is conducted annually. The MEPS uses an overlapping panel design, where data are collected for individuals via 5 interviews over a 2-year period.17 The MEPS provides nationally representative estimates of health care use, expenditures, sources of payment, and health insurance coverage for the U.S. civilian non-institutionalized population. This study utilizes the following MEPS components: the Household Component, which contains demographic information, and the Prescribed Medicines data file which contains self-reported information about prescribed medication purchases. More information about the MEPS can be found at http://meps.ahrq.gov. Study population The sample included community-dwelling adults age 18 years or older in 2011. Children younger than 18 years were excluded because parents or caregivers likely dictate their pharmacy selection. The sample was restricted to respondents who had 2 or more medication fills dispensed (i.e., had the opportunity to use multiple pharmacies). A medication dispensation was

Look / Research in Social and Administrative Pharmacy 11 (2015) 507–516

defined as each acquisition of a new prescription or a refill of an existing prescription. Dependent variable MPU was determined using the pharmacy type variables in the Prescribed Medicines file. MEPS respondents were asked to report the name and location of each pharmacy from which they obtained prescription medications. Each unique pharmacy location used by the respondent is included in the Prescribed Medicines file as a separate variable (numbered sequentially from 1 to n), such that each different location of a pharmacy organization is treated as a separate pharmacy. That is, a person who obtained medications from 3 separate locations within the same community pharmacy chain would have 3 separate variables in the dataset. These pharmacy type variables (PHARTP1 to PHARTPn) represented all of the pharmacy providers from whom a patient’s medications were obtained during the survey year, and included the following options: “mail-order”, “another store”, “HMO/clinic/ hospital”, “drug store”, and “on-line”. Because every pharmacy location is entered as a separate variable, the number of PHARTP variables with non-missing values (i.e., not coded as “inapplicable”) were defined as the total number of pharmacies used by that respondent during the year. Thus, a dichotomous variable indicating MPU was created, which classified MPU as occurring when a respondent used more than one pharmacy location in 2011.10 Independent variables Covariates were identified using the Andersen Behavioral Model of Health Service Use, which includes three domains: predisposing, enabling, and medical need related characteristics.18,19 Predisposing characteristics included age, gender, race/ethnicity, rurality of residence, census region of residence, and education. Age was categorized as 18–39 years old, 40–64 years old, and 65 years and older. Race/ethnicity consisted of 4 mutually exclusive groups: Caucasian, African-American, Hispanic, and other. Any respondent who identified him/herself as Hispanic was categorized as Hispanic, regardless of race. Rurality of residence was measured as a dichotomous variable to indicate whether an individual lived in a Metropolitan Statistical Area (MSA, more urban) or non-MSA (more rural) area. Census regions included Northeast, Midwest, South, and West.

509

Education was categorized as 0–11 years of education, 12–15 years of education, and R16 years of education. Enabling characteristics included family income, health insurance type, prescription drug insurance status, usual source of health care, and mail order pharmacy use. Family-level total income was measured as a percentage of the Federal poverty level (FPL), and was categorized as poor (!100% FPL), near poor or low income (100%– 199%), and middle to high income (R200%). Health insurance type was categorized as public insurance, private insurance, and uninsured. Because prescription drug insurance status in the MEPS is determined only for respondents reporting private health insurance, prescription drug insurance status in this study was calculated using the 12 expenditure variables in the Prescribed Medicines component. A dichotomous variable was created classifying respondents as lacking drug insurance if all drug expenditures were entirely paid by self or family, or as having drug insurance if there was any non-zero amount from another source of payment. A dichotomous usual source of care variable was created indicating whether there was a particular doctor’s office, clinic, health center, or other place the respondent usually goes if they are sick or need advice about their health. A dichotomous variable indicating any mail order pharmacy use was created using the pharmacy type variables. Medical need variables included perceived health status, number of unique medications, total number of medication fills, and number of medical conditions. Perceived health status was classified into 2 categories: good to excellent, and fair to poor. Number of unique medications filled in 2011 was determined using the MEPS-created drug-level ID in the prescribed medicines file, which identifies all fills of the same drug product (e.g., lisinopril) obtained by a patient. [The druglevel ID is consistent across all fills of the same drug product, even if there are multiple National Drug Codes (NDCs).] Total number of medication fills was determined as a count of all prescribed medicine purchases during 2011 (including initial purchases and refills). Finally, MEPS participants were asked whether they were ever diagnosed with the following priority medical conditions: high blood pressure, heart disease (including coronary heart disease, angina, myocardial infarction, and other unspecified heart disease), stroke, emphysema, high cholesterol, cancer, diabetes, arthritis, and asthma. The total

510

Look / Research in Social and Administrative Pharmacy 11 (2015) 507–516

number of self-reported chronic diseases was determined for each respondent. Because MEPS respondents were asked the same questions multiple times during the year, the independent variables used the MEPS-constructed 2011 year-end summary measures. The only exception was the usual source of care measure, which respondents were asked only one time during the year. Data analysis Analyses were conducted using Stata 13 (Stata Corp, College Station, TX). Descriptive statistics were examined for all variables, and t-tests and chi-square tests were used to test for differences between single and multiple pharmacy users. Multivariable logistic regression analysis was conducted to identify the relationships between predisposing, enabling, and need variables and MPU in a hierarchical fashion.19 MEPS personlevel weight, stratum, and primary sampling unit variables were used to adjust the standard errors to account for the complex survey design and to weight estimates to the U.S. noninstitutionalized population. A series of sensitivity analyses were conducted to test for robustness. First, the inclusion criteria were relaxed to include respondents who had 1 or more prescription fills dispensed. Second, age was examined as a continuous variable, with an age2 term to test for a nonlinear trend in MPU. Results Weighted characteristics of the study sample are described in Table 1. A total of 12,591 (weighted sample of 138,874,118) individuals 18 years and older were identified that had at least 2 medication fills in 2011. Approximately onequarter of the sample was age 65 or older and 58.3% were female. The sample was primarily composed of non-Hispanic White individuals (74.2%), with the majority residing in urban areas (82.9%). In total, 71.3% had private health insurance, 21.5% had public health insurance only, and 7.2% were uninsured. Approximately 85% of the sample had drug insurance, with 18.5% reporting any mail order usage. On average, sample individuals used 5 unique medications and had 22 total prescription fills during the year. MPU was common, with a national prevalence of 41.3% (Table 2). Overall, 58.7% of the sample used only 1 pharmacy, with 30.6% using 2 pharmacies and 10.7% using 3 or more pharmacies.

Table 1 Weighted characteristics of sample: U.S. adults 18 and older Unweighted n ¼ 12,591 Weighted N ¼ 138,874,118 Predisposing variables Age, years 18–39 40–64 R65 Female Race/ethnicity White, non-Hispanic Black, non-Hispanic Hispanic, any race Other Urban residency MSA Non-MSA Region Northeast Midwest South West Education, years 0–11 12–15 O16 Enabling variables Family income Poor Low to near poor Middle to high Health insurance Any private Public only Uninsured Has drug insurance Has usual source of care Any mail order usage Need variables Perceived health status Good to excellent Fair to poor Unique medications used (mean) Total medication fills (mean) Medical conditions (mean)

%

95% CI

26.8 47.1 26.1 58.3

25.5–28.2 45.7–48.4 24.6–27.7 57.5–59.2

74.2 10.2 10.2 5.4

72.4–76.0 9.0–11.4 9.1–11.5 4.5–6.5

82.9 17.1

79.9–85.5 14.5–20.1

18.1 23.1 37.8 21.1

16.5–19.7 21.5–24.7 35.7–39.9 19.6–22.7

14.1 54.0 31.9

13.2–15.0 52.5–55.4 30.4–33.5

12.0 17.0 70.9

11.2–12.9 16.0–18.2 69.4–72.4

71.3 21.5 7.2 85.2 89.1 18.5

69.7–72.7 20.2–22.9 6.5–7.9 84.4–86.0 88.3–89.9 17.4–19.6

82.2 17.8 5.1

81.2–83.2 16.8–18.8 5.0–5.2

21.7 2.0

21.0–22.4 1.9–2.0

CI, 95% confidence interval; MSA, Metropolitan Statistical Area. Medical conditions included high blood pressure, coronary heart disease, angina, myocardial infarction, other unspecified heart disease, stroke, emphysema, diabetes, arthritis, and asthma.

Individuals who were age 65 or older had a slightly higher prevalence of MPU, although this difference was not statistically significant.

511

Look / Research in Social and Administrative Pharmacy 11 (2015) 507–516 Table 2 Prevalence of number of pharmacies used

1 Pharmacy 2 Pharmacies R3 Pharmacies

Total n ¼ 12,591 N ¼ 138,874,118

Under 65 n ¼ 9292 N ¼ 102,608,792

65 or older n ¼ 3299 N ¼ 36,265,326

58.7% 30.6% 10.7%

59.4% 30.3% 10.3%

57.0% 31.5% 11.5%

n, unweighted sample size; N, weighted sample size.

Table 3 describes the unadjusted and adjusted odds ratios (ORs) from the single and multiple logistic regression results. Without adjusting for other characteristics, several variables had significant relationships with MPU, including female gender, black and Hispanic race/ethnicity, South census region, drug insurance status, mail order usage, and all variables measuring medical need. When only predisposing characteristics were included, female gender, urban residence, and South census region were positively associated with MPU; in contrast, black and Hispanic adults were significantly less likely to use multiple pharmacies. When enabling characteristics were added to the model, race/ethnicity and urban residency were no longer significantly related to MPU. However, age became significantly related to MPU, such that individuals age 40–64 and age 65 or older were significantly less likely to use multiple pharmacies relative to those 18–39 years old. Additionally, residence in the Northeast census region was associated with the lowest likelihood of MPU. Among the enabling variables, lacking health insurance, having drug insurance, and any mail order usage were positively associated with MPU. In the fully adjusted model, individuals 40–64 years old remained less likely to use multiple pharmacies relative to individuals 18–39 years old (OR ¼ 0.67, 95% confidence interval [CI] ¼ 0.58–0.77), while adults 65 years and older were the least likely to use multiple pharmacies (OR ¼ 0.49, CI ¼ 0.41–0.58). Females were 1.16 (CI ¼ 1.05–1.29) times as likely to use multiple pharmacies as males. Individuals residing in the South and the Midwest were 1.40 (CI ¼ 1.18– 1.66) and 1.21 (CI ¼ 1.02–1.43) times as likely to use multiple pharmacies as individuals residing in the Northeast. Individuals lacking health insurance were 1.42 (CI ¼ 1.16–1.73) times as likely to use multiple pharmacies as individuals with private health insurance; similarly, those having drug

insurance were more likely to use multiple pharmacies (OR ¼ 1.25, CI ¼ 1.06–1.47) relative to those without drug insurance. Mail order usage was the strongest predictor of MPU (OR ¼ 6.94, CI ¼ 5.90–8.18). Although the number of unique medications was positively related to MPU (OR ¼ 1.15, CI ¼ 1.12–1.17), the total number of medication fills had a small negative relationship with MPU (OR ¼ 0.99, CI ¼ 0.99–0.99). Results from the sensitivity analyses showed the findings from the primary analysis to be robust to the inclusion criteria and model specification. When the inclusion criterion requiring 2 or more medication fills was relaxed to include all individuals receiving 1 or more medication fills, the findings for the fully adjusted model were identical to those in the primary analysis, with only slightly larger ORs. In total, 95% of those with only 1 medication fill during the study period were under 65 years of age, and the vast majority were for medications to treat acute conditions (e.g., antibiotics, analgesics, dermatological agents, etc.). When tested as a continuous variable in the fully adjusted model, age was significantly related to MPU in a negative linear fashion (fully adjusted OR ¼ 0.97, CI ¼ 0.95–0.99, P ! 0.001); however, no significant non-linear trend was identified. Discussion Multiple pharmacy use (MPU) was common at a national level in 2011, with an overall rate of 41.3% among adults 18 years and older. The MPU prevalence seen in this study was comparable with the national prevalence reported in 2009 for both the general adult population and among older adults.5,10 This suggests that the increasing MPU trend seen over the past decade may have abated somewhat between 2009 and 2011.10 Pharmacists play an important role in ensuring safe and appropriate drug therapy for their patients; however, they need to be aware that

Variable

Unadjusted OR

95% CI

Reference 0.91 1.04

0.81–1.03 0.91–1.19

Reference 0.92 1.05

0.82–1.05 0.92–1.21

Reference 0.75*** 0.62***

0.66–0.86 0.53–0.74

Reference 0.67*** 0.49***

0.58–0.77 0.41–0.58

Reference 1.17***

1.07–1.29

Reference 1.18***

1.07–1.29

Reference 1.20***

1.08–1.33

Reference 1.16**

1.05–1.29

Reference 0.84** 0.86* 0.89

0.74–0.94 0.75–0.97 0.74–1.07

Reference 0.79*** 0.86* 0.90

0.70–0.89 0.75–0.99 0.75–1.07

Reference 0.88 0.98 0.99

0.78–1.01 0.84–1.15 0.82–1.21

Reference 0.90 1.02 1.03

0.79–1.03 0.88–1.19 0.85–1.26

1.13 Reference

0.98–1.30

1.18* Reference

95% CI

1.02–1.36

Adjusted OR predisposing þ enabling variables

1.10 Reference

95% CI

0.95–1.27

Adjusted OR predisposing þ enabling þ need variables

1.10 Reference

95% CI

0.95–1.29

Reference 1.11 1.27** 1.07

0.94–1.32 1.08–1.49 0.91–1.26

Reference 1.13 1.31*** 1.08

0.95–1.33 1.11–1.54 0.92–1.28

Reference 1.28** 1.48*** 1.23*

1.08–1.51 1.25–1.76 1.03–1.48

Reference 1.21* 1.40*** 1.17

1.02–1.43 1.18–1.66 0.98–1.40

Reference 1.11 1.12

1.00–1.24 0.99–1.27

Reference 1.09 1.09

0.97–1.22 0.95–1.24

Reference 1.05 1.03

0.92–1.19 0.87–1.21

Reference 1.05 1.07

0.92–1.20 0.90–1.27

Reference 1.15 1.06

0.98–1.35 0.93–1.21

Reference 1.11 0.91

0.94–1.31 0.79–1.05

Reference 1.14 0.99

0.96–1.36 0.85–1.15

Reference 0.99

0.88–1.12

Reference 1.20*

1.02–1.41

Reference 1.15

0.97–1.36

Look / Research in Social and Administrative Pharmacy 11 (2015) 507–516

Predisposing variables Age, years 18–39 40–64 R65 Female Male Female Race/ethnicity White, non-Hispanic Black, non-Hispanic Hispanic, any race Other Urban residency MSA Non-MSA Region Northeast Midwest South West Education, years 0–11 12–15 O16 Enabling variables Family income Poor Low to near poor Middle to high Health insurance Any private Public only

Adjusted OR predisposing variables

512

Table 3 Logistic regression results of patient characteristics and predictors of multiple pharmacy use

1.08–1.11 1.00–1.01 1.09–1.15

0.70–0.89

0.79*** Reference 1.10*** 1.00*** 1.12***

Uninsured Has drug insurance Has usual source of care Any mail order usage Need variables Perceived health status Good to excellent Fair to poor Unique medications used Total medication fills Medical conditions

OR, odds ratio; CI, 95% confidence interval; MSA, Metropolitan Statistical Area. Medical conditions included high blood pressure, coronary heart disease, angina, myocardial infarction, other unspecified heart disease, stroke, emphysema, diabetes, arthritis, and asthma. *P ! 0.05, **P ! 0.01, ***P ! 0.001.

1.12–1.17 0.99–0.99 0.97–1.06

0.85–1.15

0.99 Reference 1.15*** 0.99*** 1.02

0.86–1.17 1.54–2.04 0.89–1.24 5.61–7.64 1.01 1.77*** 1.05 6.55***

1.44*** 1.56*** 0.96 7.51***

1.18–1.75 1.33–1.84 0.80–1.14 6.40–8.80

1.42*** 1.25** 0.87 6.94***

1.16–1.73 1.06–1.47 0.74–1.04 5.90–8.18

Look / Research in Social and Administrative Pharmacy 11 (2015) 507–516

513

nearly half their patients use at least one other pharmacy. Thus, they may have incomplete information about their patients’ medication use. Pharmacists and prescribers may encourage their patients to use one pharmacy whenever possible to improve patient safety and ensure continuity of care. Pharmacists can attempt to identify patients most at risk for MPU, particularly when their patients use a mail order pharmacy or take multiple unique medications. Additionally, pharmacists are advised to employ open-ended questions (e.g., “What other medications are you taking?”) to elicit information from their patients, to ensure they have more complete medication information to help reduce the risk of potential drug–drug interactions when dispensing prescription medications or making over-the-counter recommendations. Several important predisposing, enabling, and medical need related characteristics identified using the Andersen Model were found to be significantly related to MPU, indicating that multiple factors may influence a person’s decision to use multiple pharmacies. Although older adults (age 65 and older) had a slightly higher prevalence of MPU than those under 65 in this study, older age was found to have a strong negative relationship (OR ¼ 0.49) with MPU once enabling and medical need characteristics were controlled. Younger adults may be more likely to use multiple pharmacies due to a lack of loyalty to one pharmacy or because of pharmacy locations that are conveniently located to home, work, school, etc. at the time a prescription is needed. Pharmacy convenience is consistently rated among the most important reasons for selecting a pharmacy.11,20– 25 In contrast, older adults have long been considered to be more loyal to one pharmacy.14,25,26 Three factors may have contributed to the higher unadjusted MPU rate seen among older adults in this study. First, individuals with high utilization of health care services (regardless of age) have more opportunities to use multiple pharmacies due to the use of multiple prescription drugs or providers. Second, some older adults live part of the year in another state (so-called “snowbirds”), and may therefore have two primary pharmacies. This could result in sequential multiple pharmacy use; that is, they fill prescriptions at 2 or more pharmacies without overlapping time periods throughout the year. Sequential MPU has been associated with significantly lower exposure to drug–drug interactions than the concurrent use of multiple pharmacies.5

514

Look / Research in Social and Administrative Pharmacy 11 (2015) 507–516

Third, the prevalence of mail order pharmacy use was nearly 2.5 times greater among adults 65 and older compared to adults 18–64 (25% and 12%, respectively). Mail order use was the strongest predictor of MPU likely because individuals who use mail order to obtain chronic medications must also utilize a second local pharmacy to obtain acute medications. However, once factors such as medical need and mail order usage were controlled for, MPU was less likely for older adults. This suggests that they may be more loyal to one pharmacy, or that they may be loyal to one local pharmacy and have to use mail order as a second pharmacy. Alternatively, this may be a sign that older adults are more aware of the potential implications of using multiple pharmacies. Interviewing older adults may be useful to determine the reasons for why they use one or more pharmacies. A potential implication of these findings is that pharmacists may need to stress the importance of using a single pharmacy to younger individuals and/or use innovative approaches to encourage their loyalty. Additionally, individuals using mail order pharmacies and a large number of unique medications may be useful targets for pharmacists to screen for drug–drug interactions through services such as medication therapy management. Female gender was found to be significantly associated with higher prevalence of MPU in this study. Previous research on gender and MPU has been inconclusive, although two studies found positive associations between male gender and MPU.11,12 Two possible explanations for this finding exist. First, one study found that the majority of people picking up prescriptions were female.27 Thus, males may have fewer opportunities for MPU if females obtain prescriptions for the entire family. Secondly, MEPS household respondents report information for the entire family, and the majority of household respondents for the MEPS are female. Thus, the female head of the household would have full knowledge of the pharmacies they use to fill their own prescriptions, but they may not be as aware of all pharmacies used by a spouse or other household member. Thus, there may be underreporting of MPU for males in the MEPS data. Few previous pharmacy patronage studies have included measures of race or ethnicity, and their effects on pharmacy selection decisions are unclear.28 Although significant differences in MPU by race or ethnicity were seen in the unadjusted results, these differences disappeared once

enabling and need characteristics were controlled, suggesting other characteristics may be more important. Urban residency was not associated with MPU, which is similar to the findings of Ranelli and Coward, who found no significant difference in the use of a single pharmacy between individuals living in urban and rural areas.16 However, regional differences in MPU did exist, which may reflect differences in pharmacy density and access in these areas. Lacking health insurance and having prescription drug insurance were two significant predictors of increased MPU in this study. Although individuals lacking health insurance are younger and have fewer chronic medical conditions than the privately and publicly insured, they also report lower incomes than those with employer sponsored insurance.29,30 Thus, individuals without health insurance may be more price sensitive and shop around at multiple pharmacies for the lowest medication prices. In contrast, prescription drug insurance has reduced the impact of price as a barrier to medication use, and having drug insurance has previously been associated with higher prevalence of MPU.13 However, drug insurance may be a better predictor of MPU than health insurance due to high overlap between individuals lacking both types of coverage. A post-hoc analysis showed that individuals with any private and public only health insurance had prescription drug insurance prevalence of 86% and 95%, respectively; in contrast, 48% of individuals without health insurance had assistance with prescription drug costs, likely reflecting the use of coupons or prescription discount cards. Future research on the impact of prescription drug insurance should be conducted using data that can better differentiate between true insurance coverage and the use of discount cards. Additionally, future research should be performed to examine how characteristics of a drug insurance benefit impact MPU. The results of this study hold important implications for hospitals and health plans as health care shifts towards performance-based reimbursement and quality ratings. Hospitals face financial penalties from Medicare for hospital readmissions and poor quality care, and Medicare Advantage plan payments are linked to their quality ratings. Additionally, Medicare Part D plans are rated on quality measures such as use of high risk medications, appropriateness of hypertension and diabetes treatments, and medication adherence. The findings from this study can be used by hospitals and health plans to screen

Look / Research in Social and Administrative Pharmacy 11 (2015) 507–516

patients for potential MPU, particularly when historical prescription drug claims are unavailable (e.g., new patients or enrollees). Once identified, patients at high risk of MPU can be followed using case management procedures to coordinate care for these individuals and improve their care quality. Identification and management of potentially harmful MPU patterns may in turn reduce undesirable outcomes affecting patient safety.3–9 Limitations This study had several limitations. First, because much of the MEPS data are self-reported, they could be subject to recall bias (e.g., not remembering all pharmacies or the types of pharmacies used) and social desirability bias (e.g., perceptions that using multiple pharmacies is bad). Although no previous study has evaluated the impact of these biases on MPU measurement or behavior, these biases may lead to reduced reporting of the number of pharmacies used by respondents and a potential underreporting of MPU. Another limitation is the inability to link a specific medication fill with a particular pharmacy, which precludes analyses of patterns of multiple pharmacy use. Additionally, it is not possible to determine whether prescriptions were filled within a chain of pharmacies that share patient information and medication profiles, which may enable better patient care than when prescriptions are transferred between different pharmacy organizations that do not share this information. However, previous research has shown that patient use of pharmacies that have the potential to share medication information electronically is low among multiple pharmacy users.10 The MPU measure used in this study has not been previously validated. Because the information related to the original pharmacies from which prescriptions were obtained are removed from MEPS data (R. Kuntz, personal communication, May 2011), and due to the inability to identify individual respondents to the MEPS survey, it is not possible to validate the MPU measure used with actual patterns of multiple pharmacy use. However, because each unique pharmacy location used by the respondent is included in the Prescribed Medicines file as a separate variable, this measure should reasonably reflect actual MPU as reported by the MEPS respondent. Although it is unknown to what extent this measure may vary from actual MPU behaviors, the MPU prevalence found in a previous national study using this

515

measure was similar to another national study using prescription drug claims data among older adults with Medicare Part D in the same year (43.2% and 38.1%, respectively).5,10 Because prescription drug insurance status in the MEPS is determined only for respondents reporting private health insurance, it was selfcalculated by the author in this study. This may have overestimated the prevalence of drug insurance, as the use of prescription discount cards or coupons may appear as if an individual was insured when they actually may not be. Additionally, characteristics of the prescription drug benefit may be related to MPU, such as mandatory mail order plans, managed care or restricted pharmacy networks, and cost sharing amounts; however, the MEPS does not contain any detailed characteristics of a respondent’s prescription drug insurance. Finally, health system characteristics that may impact MPU were not accounted for in this analysis. For example, a recent study that evaluated MPU before and 18 months after implementation of e-prescribing found a large but non-significant increase in MPU from 20.0% to 30.1%, a relative increase of 50%.31

Conclusion MPU remains common on a national level, and is a complex decision influenced by multiple factors. Pharmacists and other health care providers need to be aware that their patients may be using multiple pharmacies, especially younger patients, those lacking access to health insurance, or those using mail order pharmacies. The findings from this study can be used to identify patients that may need additional monitoring to ensure safe and appropriate drug therapy, and has important implications as health care continues to shift toward performance-based reimbursement and quality ratings. Acknowledgments The author would like to acknowledge the helpful comments of David Mott, PhD on earlier versions of this manuscript.

References 1. Xu KT. Choice of and overall satisfaction with pharmacies among a community-dwelling elderly population. Med Care 2002;40:1283–1293.

516

Look / Research in Social and Administrative Pharmacy 11 (2015) 507–516

2. McCombs JS, Nichol MB, Sclar DA. Single-pharmacy patronage and adverse drug reactions: a case study of cimetidine. J Res Pharmaceut Econ 1993;5:3–24. 3. Gupta S, Rappaport HM, Bennett LT. Inappropriate drug prescribing and related outcomes for elderly Medicaid beneficiaries residing in nursing homes. Clin Ther 1996;18:183–196. 4. Tamblyn RM, McLeod PJ, Abrahamowicz M, Laprise R. Do too many cooks spoil the broth? Multiple physician involvement in medical management of elderly patients and potentially inappropriate drug combinations. CMAJ 1996;154:1177–1184. 5. Marcum ZA, Driessen J, Thorpe CT, Gellad WF, Donohue JM. Effect of multiple pharmacy use on medication adherence and drug-drug interactions in older adults with Medicare Part D. J Am Geriatr Soc 2014;62:244–252. 6. Wu C, Bell CM, Wodchis WP. Incidence and economic burden of adverse drug reactions among elderly patients in Ontario emergency departments: a retrospective study. Drug Saf 2012;35:769–781. 7. Monane M, Bohn RL, Gurwitz JH, et al. The effects of initial drug choice and comorbidity on antihypertensive therapy compliance: results from a population-based study in the elderly. Am J Hypertens 1997;10:697–704. 8. Cocohoba JM, Althoff KN, Cohen M, et al. Pharmacist counseling in a cohort of women with HIV and women at risk for HIV. Patient Prefer Adherence 2012;6:457–463. 9. Blake SG. The effect of the Louisiana Medicaid lockin on prescription drug utilization and expenditures. Drug Benefit Trends 1999;11:45–55. 10. Look KA, Mott DA. Multiple pharmacy use and types of pharmacies used to obtain prescriptions. J Am Pharm Assoc 2013;53:601–610. 11. Laverty RF. Patients loyal but need more Rx information. Drug Top 1984;128:35–40. 42–45. 12. Shepherd MD, Crawford SY. An investigation of what factors are important to the elderly in selecting a pharmacy and purchasing drug products. J Pharm Mark Manage 1987;2:63–82. 13. Safran DG, Neuman P, Schoen C, et al. Prescription drug coverage and seniors: findings from a 2003 national survey. Health Aff 2005;(suppl web exclusives). W5-W152–W5-W166. 14. Polinski JM, Schneeweiss S, Levin R, Shrank WH. Completeness of retail pharmacy claims data: implications for pharmacoepidemiologic studies and pharmacy practice in elderly patients. Clin Ther 2009;31: 2048–2059. 15. Ryan K, Norris P, Becket G. Capturing data on medicines usage: the potential of community pharmacy databases. N Z Med J 2005;118:U1677. Available at: http://journal.nzma.org.nz/journal/118-1223/1677/. Accessed 05.13.2014. 16. Ranelli PL, Coward RT. Residential differences in the use of pharmacies by older adults and their

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

communication experiences with pharmacists. J Rural Health 1996;12:19–32. Ezzati-Rice TM, Rohde F, Greenblatt J. Sample Design of the Medical Expenditure Panel Survey Household Component, 1998–2007. Methodology report no. 22, Rockville, MD: Agency for Healthcare Research & Quality; 2008. Andersen RM. Revisiting the behavioral model and access to medical care: does it matter? J Health Soc Behav 1995;36:1–10. Wang CC, Farley FJ. Patterns and predictors of antipsychotic medication use among the U.S. population: findings from the Medical Expenditure Panel Survey. Res Social Adm Pharm 2013;9:263–275. Wiederholt JB. Development of an instrument to measure evaluative criteria that patients use in selecting a pharmacy for obtaining prescription drugs. J Pharm Mark Manage 1987;1:35–59. Stergachis A, Maine LL, Brown L. The 2001 National Pharmacy Consumer Survey. J Am Pharm Assoc 2002;42:568–576. Schommer JC. The effect of age on pharmacy patronage: is locational convenience a lurking variable? J Pharm Mark Manage 1994;9:33–46. Metge CJ, Hendricksen C, Maine L. Consumer attitudes, behaviors, and perceptions about pharmacies, pharmacists, and pharmaceutical care. J Am Pharm Assoc 1998;38:37–47. Franic DM, Grauer DW, Law AV. An evaluation of patients’ attitudes toward pharmacy selection: a determinant attribute approach. J Pharm Mark Manage 2005;17:35–50. Gagnon JP. Factors affecting pharmacy patronage motives – a literature review. J Am Pharm Assoc 1977;17:556–559. 566. Lipowski EE. Predictors of community pharmacy preference: a logistic regression analysis of consumers’ socioeconomic and shopping characteristics. J Pharm Mark Manage 1991;6:43–61. Ryan K, Becket G, Norris P. Who goes where and why? The patronage of community pharmacies in New Zealand. Aust Pharm 2002;21:46–51. Shannon GW, Cromley EK, Fink JL 3rd. Pharmacy patronage among the elderly: selected racial and geographical patterns. Soc Sci Med 1985;20: 85–93. Abraham JM. How might the Affordable Care Act’s coverage expansion provisions influence demand for medical care? Milbank Q 2014;92:63–87. Blumberg LJ, Holahan J. Health Status of Exchange Enrollees: Putting Rate Shock in Perspective. Timely Analysis of Immediate Health Policy Issues. Washington, DC: Urban Institute, 2013. Available at: http:// www.urban.org/publications/412859.html. Accessed 05.13.2014. Bergeron AR, Webb JR, Serper M, et al. Impact of electronic prescribing on medication use in ambulatory care. Am J Manag Care 2013;19:1012–1017.