Abstract
Background
In the elderly (those aged ≥65 years), retail pharmacy claims are used to study drug use among the uninsured after drug policy changes, to prevent drug drug interactions and duplication of therapy, and to guide medication therapy management. Claims include only prescriptions filled at one pharmacy location or within one pharmacy chain and do not include prescriptions filled at outside pharmacies, potentially limiting research accuracy and pharmacy-based safety interventions.
Objectives
The aims of this study were to assess elderly patients’ pharmacy loyalty and to identify predictors of using multiple pharmacies.
Methods
Patients enrolled in the Pharmaceutical Assistance Contract for the Elderly pharmacy benefit program with corresponding Medicare claims in the state of Pennsylvania comprised the study cohort. Among patients with pharmacy claims from all pharmacies used in 2004–2005, a primary pharmacy was defined as the pharmacy where >50% of a patient’s prescriptions were filled. The number of pharmacies/chains used and prescriptions filled in 2005 was calculated. Predictors of using multiple pharmacies in 2005 were age, gender, race, urban residency, comorbidities, number of unique medications used, and number of prescriptions, which were all assessed in 2004.
Results
In total, pharmacy claims data from 182,235 patients (147,718 [81.1%] women; mean [SD] age 78.8 [7.1] years; 168,175 white; 76,580 residing in an urban zip code area) were included. In 2005, patients filled an average of 59.3 prescriptions, with 57.0 (96.1%) prescriptions having been filled at the primary pharmacy. Compared with patients who used <5 unique medications in 2004, patients who used 6 to 9 unique medications had 1.39 times (95% CI, 1.34–1.44), and patients who used 15 unique medications had 2.68 times (95% CI, 2.55–2.82) greater likelihood of using multiple pharmacies in 2005. Patients aged ≥85 years were 1.07 times (95% CI, 1.03–1.11) as likely to use multiple pharmacies compared with patients aged 65 to 74 years.
Conclusions
This study found that patients aged ≥65 years were loyal to their primary pharmacy, offering reassurance to researchers and pharmacists who use retail pharmacy claims to evaluate and/or to improve safe and appropriate medication use among the elderly. Care should be used in analyzing claims from or managing the drug regimens of patients using many medications or patients aged ≥85 years; they are more likely to use multiple pharmacies and thus are more likely to have missing prescription information.
Keywords: retail pharmacy, medication errors, drug utilization, loyalty, missing data
INTRODUCTION
With increasing frequency, researchers and pharmacists have relied on retail pharmacy claims data for information about patients’ drug utilization. Researchers use retail pharmacy claims to study drug use among patients who do not have drug insurance.1 Retail pharmacy claims have also been particularly helpful in assessing patient drug utilization before and after the implementation of Medicare Part D (Part D),1–4 which expanded the federal government’s role in providing drug coverage to elderly (aged ≥65 years) Americans.5 Pharmacists use retail pharmacy claims to assist in the prevention of drug–drug interactions6 and duplication of therapies,7,8 and to identify other medication-related errors,9 improve adherence to medication regimens,10 and implement medication therapy management programs (MTMPs).11 Because the Medicare Modernization Act of 2003 mandates that Part D plans have such programs in place, MTMPs have become increasingly prevalent.12,13 The success of both safety protocols and MTMPs depends on the completeness of pharmacy claims data.
Despite the utility of retail pharmacy claims data, these claims include only prescriptions filled at one pharmacy location or within one retail pharmacy chain, but does not include prescriptions filled at nonaffiliated pharmacies.1 Missing or incomplete prescription information may reduce the accuracy of research findings and expose patients to medication-related adverse events. The completeness of retail pharmacy prescription drug data among patients aged ≥65 years is of particular interest. Elderly patients comprise the largest population enrolled in Part D,14 and before 2006, they were less likely to have drug insurance than were adults aged 18 to 64 years.15–18 Many MTMPs and safety programs6–11 for the elderly population rely on complete pharmacy data. To quantify the completenessof retail pharmacy prescription data, we undertook a validation study among the population aged ≥65 years who had complete pharmacy claims information.
The purpose of this study was to describe elderly patients’ loyalty to a particular pharmacy and/or pharmacy chain. A secondary aim was to identify patient demographic, health care, and pharmacy characteristics that predict the use of multiple nonaffiliated pharmacies and thus an increased likelihood of missing prescription information.
PATIENTS AND METHODS
Study Population
The study cohort was assembled using complete 2004–2005 eligibility and pharmacy claims for patients enrolled in the Pharmaceutical Assistance Contract for the Elderly (PACE) program. PACE pharmacy claims were then linked with Medicare Parts A and B health care claims. The PACE program provides drug insurance coverage for low to moderate income patients aged ≥65 years in Pennsylvania. PACE beneficiaries in 2004 and 2005 were eligible for the study. To ensure consistent PACE benefit use, the final study cohort included only those patients who filled ≥1 prescription for any medication during each calendar quarter of 2005.
The study was approved by the Brigham and Women’s Hospital Institutional Review Board in Boston, Massachusetts.
Pharmacy Definitions
To evaluate patients’ loyalty to a particular pharmacy location, each patient’s primary pharmacy was defined. The primary pharmacy was the unique pharmacy identifier (ID) (ID = pharmacy name + geographic location) where at least 50% of prescriptions were filled during 2005. If no pharmacy ID met this definition, then the patient’s primary pharmacy was the first pharmacy ID used in 2005. If 2 pharmacy IDs each had 50% of filled prescriptions for a patient, then the patient’s primary pharmacy was the first pharmacy ID used in 2005.
We examined whether pharmacy loyalty differed between those patients who used larger pharmacies versus those who used smaller chain or single location pharmacies. If larger chain pharmacy patrons are significantly more loyal than smaller chain/single-location pharmacies patrons or vice versa, then both researchers and pharmacists would need to consider these differences in their research and clinical practice. To categorize patients as patrons of larger chain pharmacies or not, we determined the top 5 pharmacy chains in our data based on the number of unique geographic locations in Pennsylvania. Pharmacy names and locations were manually grouped. When there was a question whether 2 pharmacies were affiliated with the same chain, these pharmacies were coded as unrelated (not in the same chain). Patients whose primary pharmacy was affiliated with 1 of the top 5 pharmacy chains were referred to as large pharmacy chain patients.
Baseline Measures
Baseline demographics, including age, gender, and race, were assessed for each patient using PACE enrollment files and pharmacy claims (data from 2004). The number of unique medications used and prescriptions filled in 2004 was assessed, as well as the patient’s 2004 Chronic Disease Score,19 a summary score of health status based on medication use. The score has been found to be a stable predictor of chronic disease status and subsequent hospitalizations and mortality.20 Patients’ zip codes were matched to census data on population density per square mile, and an urban zip code was defined as 1 in which there were ≥1000 persons per square mile.21 Using 2004 Medicare data, we calculated a baseline Charlson comorbidity index score, based on the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes.22 This score is predictive of mortality in populations aged ≥65 years.23 The Charlson comorbidity index includes 19 clinical important comorbidities that were found to be predictors of 1-year mortality and assigns them weights of 1, 2, 3, or 6 based on the magnitude of their individual association with 1-year mortality.23 Because patients with medical diagnoses regularly requiring the concurrent use of multiple drugs might have different pharmacy loyalty behavior, we noted whether patients had ≥1 diagnosis of cardiovascular disease (ICD-9 codes 410.xx–414.xx; 420.xx–429.xx), diabetes (249.xx–250.xx), and/or cancer, excluding nonmelanoma skin cancer (140.xx–172.xx; 174.xx–208.xx). Patients with psychiatric diagnoses (290.xx–319.xx) might have different pharmacy loyalty behavior because of potential difficulty in managing their medications, so these patients were also identified. Comorbidity indicators for these diseases were included in predictor models (see below).
Statistical Analysis
Baseline demographic and health care characteristics for all PACE patients and the subset of large pharmacy chain patients were analyzed and expressed as mean (SD) or frequency (percent). To measure prescription filling behavior in 2005, the mean (SD) as well as the median interquartile range for the number of prescriptions filled, number filled at the primary pharmacy, and number of unique pharmacy locations used were calculated. Because a patient might have different pharmacy loyalty behavior when filling a prescription for an acute condition than when filling a prescription for a chronic condition, we determined the total number of antibiotic prescriptions filled and the number of those that were filled in the primary pharmacy for each month and for the entire year 2005. These analyses were repeated, stratified by month, to detect seasonal variation. For large pharmacy-chain patients, we calculated the number of prescriptions filled within the primary pharmacy chain.
Univariate and multivariate logistic regression analyses were conducted to identify baseline correlates of using multiple pharmacies in filling prescriptions during 2005. Using a complete information model, which utilized pharmacy and health care claims from all pharmacies and providers in 2004, potential predictors included age (65–74, 75–84, ≥85 years), female gender, white race, residence in an urban zip code, number of unique generic medications (<5, 6–9, 10–14, ≥15), number of prescriptions filled (<25, 25–49, 50–74, ≥75), Charlson comorbidity index score (0, 1–3, ≥4), and diagnosis of cancer (excluding nonmelanoma skin cancer), cardiovascular disease, diabetes, and/or psychiatric disease.
We created a “primary pharmacy” regression model that consisted of only those data and predictors available to the primary pharmacy/pharmacy chain, which included age, female gender, white race, residence in an urban zip code, number of unique generic medications used within the primary pharmacy/pharmacy chain, and number of prescriptions filled within the primary pharmacy/pharmacy chain. In a separate analysis, we added the patients’ 2004 Charlson comorbidity index score to the primary pharmacy model to determine the predictive value of diagnostic information. The complete information model and the primary pharmacy model were run for all PACE patients and for large pharmacy chain patients.
To assess the ability of the models to differentiate between those patients who used multiple pharmacies and those who did not, a C-statistic24 comparison was calculated for each model. A Nagelkerke pseudo R2 was also calculated. Similar to the R2 value obtained in a linear regression, which estimates the proportion of variation in the outcome that can be attributed to the model predictors, the Nagelkerke pseudo R2 estimates the variation attributed to the predictors in a logistic regression model.25 Models were tested for collinearity using variance inflation factor tests.26
RESULTS
In total, pharmacy claims data from 182,116 patients (147,718 [81.1%] women; mean [SD] age 78.8 [7.1] years; 168,175 white; 76,580 residing in an urban zip code area) were included (Table I). Among all PACE patients, 75,413 patients’ (41.4%) primary pharmacy was among the top 5 pharmacies in Pennsylvania, defining them as large pharmacy chain patients. On average, large pharmacy chain patients were more likely to reside in urban areas compared with all PACE patients ( [49.5%] vs [42.1%], respectively), filled a similar number of prescriptions in 2004 (50.9 [29.9] versus 52.1 [31.0]), and had fewer comorbidities, with a Charlson comorbidity index score of 0 ( [51.4%] vs [46.97%]).
Table I.
All PACE Patients (N =182,116) | Patients with Primary Pharmacy in the Top 5 (N =75,413) | |
---|---|---|
Age as of January 1, 2005, y | 78.8 (7.1) | 78.7 (6.9) |
Female | 147,718 (81.1%) | 61,802 (82.0%) |
White race | 168,175 (92.3%) | 67,571 (89.6%) |
Urban residence | 76,580 (42.1%) | 37,300 (49.5%) |
Population density (persons/sq mile) | 2903.0 (4940.1) | 3475.6 (5314.6) |
<500 | 76,210 (41.8%) | 25,786 (34.2%) |
500–999.99 | 29,326 (16.1%) | 12,327 (16.3%) |
1000–1499.99 | 8699 (4.8%) | 3902 (5.2%) |
≥1500 | 67,881 (37.3%) | 33,398 (44.3%) |
Chronic disease score for 2004 | 3.9 (2.7) | 3.9 (2.7) |
0 | 20,952 (11.5%) | 8714 (11.6%) |
1–3 | 66,420 (36.5%) | 27,272 (36.2%) |
≥4 | 94,744 (52.0%) | 39,427 (52.3%) |
Unique medications used in 2004 | 9.1 (5.1) | 8.9 (5.0) |
≤5 Medications | 48,570 (26.7%) | 20,656 (27.4%) |
6–9 Medications | 61,494 (33.8%) | 25,894 (34.3%) |
10–14 Medications | 46,661 (25.6%) | 19,097 (25.3%) |
≥15 Medications | 25,391 (13.9%) | 9766 (13.0%) |
Prescriptions filled in 2004 | 52.1 (31.0) | 50.9 (29.9) |
<25 Prescriptions | 34,294 (18.8%) | 14,912 (19.8%) |
25–49 Prescriptions | 63,720 (34.7%) | 26,563 (35.2%) |
49–74 Prescriptions | 45,718 (25.1%) | 18,881 (25.0%) |
≥75 Prescriptions | 38,384 (21.1%) | 15,057 (20.0%) |
Charlson comorbidity index score* in 2004 | 1.4 (1.8) | 1.2 (1.8) |
0 | 85,499 (46.9%) | 38,788 (51.4%) |
1–3 | 74,024 (40.7%) | 28,285 (37.5%) |
≥4 | 22,593 (12.4%) | 8340 (11.1%) |
Diagnosis of cardiovascular disease in 2004 | 70,697 (38.8%) | 26,912 (35.7%) |
Diagnosis of psychiatric disease in 2004 | 30,284 (16.6%) | 10,744 (14.3%) |
Diagnosis of diabetes in 2004 | 42,208 (23.2%) | 15,973 (21.2%) |
Diagnosis of cancer in 2004 | 15,953 (8.8%) | 6174 (8.2%) |
The scale can be used to predict 1-year mortality.
Table II describes the 2005 prescription-filling behavior for PACE patients. . These patients used an average of 1.3 (0.6) unique pharmacy locations to fill prescriptions. Of the mean (SD) 59.3 (33.1) prescriptions filled by PACE patients in 2005, 57.0 (96.1%) were filled at the primary pharmacy. Patients who resided in a rural zip code area filled an average of 58.3 (98.3%) prescriptions at their primary pharmacy, whereas patients who resided in an urban area filled an average of 55.2 (93.1%) at the primary pharmacy. Large pharmacy chain patients’ prescription-filling behavior was nearly identical, with 96.1% of prescriptions being filled at the primary pharmacy and 97.4% within the primary pharmacy chain. Loyalty to the primary pharmacy for antibiotic prescription fills was high, with an average of 0.4 of 0.5 antibiotic prescriptions among all PACE patients and an average of 0.4 of 0.4 antibiotic prescriptions among large pharmacy chain patients filled at the primary pharmacy. For both PACE patients and large pharmacy chain patients, analysis of prescription fills by month found minimal variation in prescription-filling behavior over the calendar year (data not shown).
Table II.
All PACE patients (N = 182,116) | PACE enrollees whose primary pharmacy was in the top 5 pharmacies in Pennsylvania(N = 75,413) | |||
---|---|---|---|---|
Mean (SD) | Median (IQR) | Mean (SD) | Median (IQR) | |
Unique pharmacy locations where prescriptions were filled, no. | 1.3 (0.6) | 1 (1–1) | 1.3 (0.6) | 1 (1–1) |
Unique medications used, no. | 9.8 (5.5) | 9 (6–13) | 9.5 (5.3) | 9 (6–12) |
Prescriptions filled, no. | 59.3 (33.1) | 54 (35–78) | 57.4 (31.6) | 52 (34–76) |
Prescriptions filled at the primary pharmacy, no. | 57.0 (32.4) (96.1%) | 52 (33–76) | 55.3 (31.2) (96.3%) | 50 (32–74) |
Residing in an urban zip code, no. | 55.2 (31.7) (93.1%) | 50 (31–73) | 53.4 (30.3) (93.0%) | 48 (30–71) |
Residing in a rural zip code, no. | 58.3 (32.8) (98.3%) | 53 (34–77) | 57.1 (31.9) (99.5%) | 52 (33–76) |
Prescriptions filled at a nonprimary pharmacy, no. | 2.3 (7.8) | 0 (0–0) | 2.2 (7.3) | 0 (0–0) |
Prescriptions filled within the same pharmacy chain, no. | – | – | 55.9 (31.3) (97.4%) | 51 (32–74) |
Antibiotic prescriptions filled, no. | 0.5 (1.1) | 0 (0–1) | 0.4 (1.1) | 0 (0–1) |
Antibiotic prescriptions filled at the primary pharmacy, no. | 0.4 (1.1) | 0 (0–0) | 0.4 (1.1) | 0 (0–0) |
IQR = interquartile range.
The number of pharmacy locations used in 2005 by all PACE patients is shown in Figure 1A. In 2005, a majority of patients (142,544 [78.3%]) used 1 pharmacy, whereas 31,161 (17.1%) used 2 pharmacies. Figure 1B shows that again, 78.3% of PACE patients filled all of their prescriptions in 2005 at the primary pharmacy. An additional 20,022 patients (11.0%) filled ≥90% of their prescriptions at the primary pharmacy in the same year. On average, 96.1% of prescriptions were filled at the primary pharmacy.
Based on the complete information multivariate regression model (Table III), PACE patients aged 75 to 84 years were less likely to use multiple pharmacies in 2005 (odds ratio [OR] = 0.96; 95% CI, 0.94–0.99) compared with PACE patients aged 65 to 74 years. In contrast,, PACE patients aged ≥85 years were more likely to use multiple pharmacies (OR = 1.07; 95% CI, 1.04–1.11). Compared with PACE patients who used ≤ 5 unique medications in 2004, patients who used 6 to 9 unique medications in 2004 were 1.38 times (95% CI, 1.34–1.43) more likely to use multiple pharmacies in 2005; patients who filled ≥15 prescriptions unique medications in 2004 were 2.66 times (95% CI, 2.53–2.80) more likely to use multiple pharmacies in 2005. After controlling for the number of unique medications used and other covariates, the greater the number of prescriptions filled, the lesser the likelihood that a PACE patient used multiple pharmacies. PACE patients who filled 25 to 49 prescriptions in 2004 had a reduced likelihood of using multiple pharmacies compared with patients who filled <25 prescriptions in 2004 (OR = 0.87; 95% CI, 0.84–0.90). While PACE patients with a Charlson comorbidity index score of 1 to 3 or those with a score ≥4 were more likely to use multiple pharmacies than were PACE patients with a Charlson score of 0 (OR = 1.07; 95% CI, 1.04–1.11; and OR = 1.20; 95% CI, 1.14–1.27, respectively), the specific diagnoses of cancer, cardiovascular disease, diabetes, and psychiatric disease were not significant predictors. Large pharmacy chain patients’ model findings were similar in magnitude and direction, and CIs largely overlapped those in a model that included nonlarge pharmacy chain patients (data not shown), suggesting no effect-measure modification by pharmacy type (large chain vs small chain or single business).
Table III.
All PACE patients (N = 182,116). | Patients whose primary pharmacy was among the top 5 pharmacies in Pennsylvania (N =75,413). | |||
---|---|---|---|---|
Univariate | Multivariate | Univariate | Multivariate | |
C statistic = 0.597 | C statistic = 0.599 | |||
Pseudo R2 = 0.0304 | Pseudo R2 = 0.0323 | |||
Odds ratio (95% CI) | ||||
Baseline covariates assessed in 2004 | ||||
Age, y, as of January 1, 2005 | ||||
65–74 | Reference | Reference | Reference | Reference |
75–84 | 0.96 (0.94–0.99) | 0.96 (0.94–0.99) | 0.92 (0.88–0.96) | 0.93 (0.89–0.97) |
≥85 | 1.05 (1.02–1.09) | 1.07 (1.04–1.11) | 0.90 (0.86–0.95) | 0.93 (0.88–0.98) |
Female gender | 1.06 (1.03–1.09) | 1.00 (1.00–1.00) | 1.05 (1.00–1.09) | 1.00 (1.00–1.00) |
White race | 0.64 (0.61–0.66) | 1.00 (1.00–1.00) | 0.64 (0.61–0.67) | 1.00 (1.00–1.00) |
Residence in an urban zip code | 1.27 (1.24–1.30) | 1.00 (1.00–1.00) | 1.34 (1.30–1.39) | 1.00 (1.00–1.00) |
Number of unique medications in 2004 | ||||
5 or less | Reference | Reference | Reference | Reference |
6–9 Medications | 1.29 (1.25–1.33) | 1.38 (1.34–1.43) | 1.27 (1.21–1.33) | 1.37 (1.30–1.45) |
10–14 Medications | 1.64 (1.59–1.69) | 1.83 (1.75–1.90) | 1.60 (1.52–1.68) | 1.81 (1.70–1.93) |
≥15 Medications | 2.35 (2.27–2.44) | 2.66 (2.53–2.80) | 2.38 (2.25–2.52) | 2.78 (2.57–3.01) |
Number of prescriptions filled in 2004 | ||||
<25 Prescriptions | Reference | Reference | Reference | Reference |
25–49 Prescriptions | 1.07 (1.04–1.11) | 0.87 (0.84–0.90) | 1.06 (1.01–1.12) | 0.86 (0.82–0.91) |
49–74 Prescriptions | 1.26 (1.22–1.31) | 0.82 (0.78–0.86) | 1.23 (1.16–1.30) | 0.81 (0.76–0.87) |
≥75 Prescriptions | 1.52 (1.47–1.57) | 0.77 (0.74–0.81) | 1.45 (1.37–1.53) | 0.74 (0.69–0.80) |
Charlson comorbidity index score* in 2004 | ||||
0 | Reference | Reference | Reference | Reference |
1–3 | 1.08 (1.06–1.11) | 1.07 (1.04–1.11) | 1.04 (1.00–1.08) | 1.05 (1.00–1.11) |
≥4 | 1.45 (1.41–1.50) | 1.20 (1.14–1.27) | 1.43 (1.35–1.51) | 1.18 (1.08–1.29) |
Diagnosis of cardiovascular disease in 2004 | 1.14 (1.12–1.17) | 1.00 (1.00–1.00) | 1.13 (1.09–1.17) | 1.00 (1.00–1.00) |
Diagnosis of diabetes in 2004 | 1.12 (1.09–1.15) | 1.00 (1.00–1.00) | 1.13 (1.09–1.18) | 1.00 (1.00–1.00) |
Diagnosis of psychiatric disease in 2004 | 1.34 (1.30–1.38) | 1.00 (1.00–1.00) | 1.26 (1.21–1.33) | 1.00 (1.00–1.00) |
Diagnosis of cancer (excluding nonmelanoma skin cancer) in 2004 | 1.11 (1.07–1.15) | 1.00 (1.00–1.00) | 1.12 (1.06–1.19) | 1.00 (1.00–1.00) |
The index can be used to predict 1-year mortality.
In the primary pharmacy model (Table IV), the same predictors of using multiple pharmacies were observed. PACE patients aged ≥85 years were 1.09 times (95% CI, 1.06–1.13) more likely to use multiple pharmacies in 2005 than were patients aged 65 to 74 years. Among all PACE patients, patients who used ≥15 unique medications in 2004 were 1.92 times (95% CI, 1.82–2.03) more likely to use multiple pharmacies in 2005 compared with patients who used ≤5 unique medications in 2004. After controlling for the number of unique medications used in 2004 and all other covariates, patients who filled 25 to 49 prescriptions in 2004 were half as likely (OR = 0.51; 95% CI, 0.49–0.53) to use multiple pharmacies compared with patients who filled <25 prescriptions, and the likelihood of using multiple pharmacies further decreased as the number of prescriptions filled increased. In comparison with the complete information models in Table III, the C statistic and pseudo R2 were not meaningfully reduced when only the primary pharmacy’s data were used in the model for all PACE patients (C = 0.595; pseudo R2 = 0.0291). Large pharmacy chain patient model findings were similar to those among all PACE patients and to those among nonlarge pharmacy chain patients (data not shown), and CIs overlapped, indicating no effect-measure modification by pharmacy type. Adding the Charlson comorbidity index score to the primary pharmacy model did not meaningfully change other predictor effect estimates, and minimally improved the C statistic and pseudo R2 in the model with all PACE patients (C = 0.601; pseudo R2 = 0.0329).
Table IV.
All PACE patients (N = 182,116). | Patients whose primary pharmacy was among the top 5 pharmacies in Pennsylvania (N =75,413). | |||
---|---|---|---|---|
Univariate | Multivariate | Univariate | Multivariate | |
C statistic = 0.595 | C statistic = 0.597 | |||
Pseudo R2 = 0.0291 | Pseudo R2 = 0.0304 | |||
Odds ratio (95% CI) | ||||
Baseline covariates assessed in 2004 | ||||
Age on January 1, 2005, y | ||||
65–74 | Reference | Reference | Reference | Reference |
75–84 | 0.96 (0.94–0.99) | 0.99 (0.96–1.01) | 0.92 (0.88–0.96) | 0.95 (0.92–0.99) |
≥85 | 1.05 (1.02–1.09) | 1.09 (1.06–1.13) | 0.90 (0.86–0.95) | 0.95 (0.91–1.00) |
Female | 1.06 (1.03–1.09) | 1.00 (1.00–1.00) | 1.05 (1.00–1.09) | 1.00 (1.00–1.00) |
White race | 0.64 (0.61–0.66) | 1.00 (1.00–1.00) | 0.64 (0.61–0.67) | 1.00 (1.00–1.00) |
Residing in an urban zip code | ||||
1.27 (1.24–1.30) | 1.00 (1.00–1.00) | 1.34 (1.30–1.39) | 1.00 (1.00–1.00) | |
Unique medications in 2004 | ||||
≤5 | Reference | Reference | Reference | Reference |
6–9 | 0.70 (0.68–0.71) | 1.01 (0.98–1.05) | 0.75 (0.72–0.79) | 1.03 (0.98–1.08) |
10–14 | 0.83 (0.80–0.85) | 1.36 (1.30–1.42) | 0.90 (0.86–0.95) | 1.39 (1.31–1.48) |
≥15 Medications | 1.10 (1.06–1.14) | 1.92 (1.82–2.03) | 1.25 (1.19–1.33) | 2.07 (1.91–2.24) |
Prescriptions filled in 2004 | ||||
<25 | Reference | Reference | Reference | Reference |
25–49 | 0.55 (0.54–0.57) | 0.51 (0.49–0.53) | 0.57 (0.54–0.59) | 0.52 (0.50–0.55) |
49–74 | 0.62 (0.60–0.64) | 0.50 (0.48–0.52) | 0.64 (0.61–0.67) | 0.52 (0.49–0.55) |
≥75 | 0.71 (0.69–0.74) | 0.47 (0.45–0.49) | 0.73 (0.69–0.77) | 0.48 (0.44–0.51) |
Includes only data available to the primary pharmacy.
DISCUSSION
To assess the completeness of prescription information in retail pharmacy claims data, this study examined loyalty to a particular pharmacy or pharmacy chain in a population aged ≥65 years. Of all prescriptions filled, 96.1% of prescriptions were filled at a single pharmacy location. This high degree of pharmacy loyalty was replicated among patients who patronize large pharmacy chains, suggesting no differences in prescription-filling behavior between patients who use major pharmacy chains and those who do not.
The present study also assessed patient factors that were predictive of using multiple unrelated pharmacies. In all multivariate models, the number of unique medications used in 2004 was a strong predictor of using multiple pharmacies in 2005. From this perspective, patients who used ≥15 unique medications in 2004 were twice as likely to use multiple pharmacies compared with those who used ≤5 unique medications. There appeared to be a greater potential for drug exposure misclassification (missing prescriptions) when using retail pharmacy data to evaluate drug use in patients who used a greater number of unique medications. In contrast, after controlling for the number of unique medications used, the more prescriptions a patient filled, the less likely it was that the patient used multiple pharmacies. These conclusions might suggest that patients who were more adherent to their medications were less likely to use multiple pharmacies, but more dedicated research is needed to confirm this finding. Among all PACE patients, those aged ≥85 years had a 7% increased odds of using multiple pharmacies compared with patients aged 65 to 74 years. Mobility restrictions may play a contributory factor in the increased likelihood of using multiple pharmacies, as patients might have to rely on adult children or caregivers to fill prescriptions.
The complete information regression models and the primary pharmacy models both identified the same predictors of using multiple pharmacies, with nearly identical C statistic and pseudo R2 values. Researchers and pharmacists who use retail claims data exclusively, without complementary access to health services claims, can be reassured that among the population aged ≥65 years, the number of missing prescriptions is minimal, and that prescription data completeness can be predicted with similar accuracy with or without diagnostic information.
The findings from the present study also suggest that while patients are, for the most part, loyal to a particular pharmacy and even more so to a particular pharmacy chain, patients with more complicated drug regimens and those aged ≥85 years merit additional attention from researchers and pharmacists. At minimum, researchers should consider what effect missing prescriptions will have on the findings of their studies. For the practicing pharmacist, a first step toward minimizing safety concerns related to missing prescriptions is to ask the oldest patients and/or patients who use many unique medications about their use of other pharmacies. Questions such as, “What other medications are you taking that you do not fill here at our pharmacy?”; and “Does someone else fill prescriptions for you at a different pharmacy?” are currently being asked at some pharmacies on a routine basis. Other pharmacies might benefit from adopting these or similar safety questions. The oldest patients and those using numerous medications may therefore gain the most from MTMPs, because a more thorough evaluation of medication use is employed. In fact, a patient’s use of multiple medications is a criterion for MTMP intervention under Part D.12,27
Limitations
Our study described pharmacy loyalty among a population of elderly patients (aged ≥65 years). This population is largely retired, suggesting that they might have had more time and flexibility to patronize a single pharmacy location. It is unclear whether the high level of pharmacy loyalty found in this study is generalizable to a younger, working-age population who might fill some prescriptions at a pharmacy near their workplace, some prescriptions at an unrelated pharmacy near their home, and/or use mail-order prescription services. The patients in the present study all had drug insurance coverage. It is unknown whether patients with no drug insurance coverage have different loyalty behaviors, but such information would only be available from survey data, which, to our knowledge, has not been published. Since the time period 2004–2005 under study, some pharmacies have introduced drug discount programs (e.g., $4 generics, free antibiotics), but our data do not allow us to comment on these programs’ impact. Finally, while our regression models found predictors of using multiple pharmacies, their predictive ability was low. Other factors must be enumerated to better clarify why elderly patientsuse multiple pharmacies.
Nevertheless, the findings of the present study provide reassurance to researchers who use retail pharmacy claims data to evaluate drug use by the elderly and to pharmacists who use these data to improve safe and appropriate medication use in this population. This study also highlights predictive variables that can be used to address the implications of missing prescription data in research and clinical applications.
CONCLUSIONS
In the present study, patients aged ≥65 years displayed a high level of loyalty to their primary pharmacy, with 96.1% of prescriptions filled at a single pharmacy location. Missing prescriptions are most common among the oldest patients and those using more unique medications; both are more likely to use multiple pharmacies.
Acknowledgments
Dr. Shrank is the principal investigator for and has research funding from CVS Caremark (Richardson, Texas) and Express Scripts (St. Louis, Missouri). Dr. Schneeweiss is a coinvestigator with CVS Caremark and receives research funding from the CVS Caremark grant. These sponsors had no role in the design or analysis of the present study, nor did they participate in any capacity in the preparation of the paper. Ms. Polinski and Ms. Levin have no financial disclosures or conflicts of interest to report.
The present study received grant support from the National Institute on Aging (Bethesda, Maryland), grant no. T32 AG000158 (Ms. Polinski), and the National Institute of Mental Health (Bethesda, Maryland), grant no. R01 5U01MH079175-02 (Dr. Schneeweiss).
Footnotes
The authors have indicated that they have no other conflicts of interest regarding the content of this article.
References
- 1.Schneeweiss S, Patrick AR, Pedan A, et al. The effect of Medicare Part D coverage on drug use and cost-sharing in seniors without prior drug benefits. Health Aff (Millwood) 2009;28:w305–w316. doi: 10.1377/hlthaff.28.2.w305. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Shrank WH, Patrick AR, Pedan A, et al. The effect of transitioning to medicare part d drug coverage in seniors dually eligible for medicare and medicaid. J Am Geriatr Soc. 2008;56:2304–2310. doi: 10.1111/j.1532-5415.2008.02025.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Yin W, Basu A, Zhang JX, et al. The effect of the Medicare Part D prescription benefit on drug utilization and expenditures. Ann Intern Med. 2008;148:169–177. doi: 10.7326/0003-4819-148-3-200802050-00200. [DOI] [PubMed] [Google Scholar]
- 4.Lichtenberg FR, Sun SX. The impact of Medicare Part D on prescription drug use by the elderly. Health Aff (Millwood) 2007;26:1735–1744. doi: 10.1377/hlthaff.26.6.1735. [DOI] [PubMed] [Google Scholar]
- 5.United States Congress. Medicare Prescription Drug, Improvement and Modernization Act of 2003. [Accessed October 7, 2008];Public Law. 2003 December 8;:108–173. http://frwebgate.access.gpo.gov/cgi-bin/getdoc.cgi?dbname=108_cong_public_laws&docid=f:publ173.108.pdf.
- 6.Malone DC, Abarca J, Skrepnek GH, et al. Pharmacist workload and pharmacy characteristics associated with the dispensing of potentially clinically important drug-drug interactions. Med Care. 2007;45:456–462. doi: 10.1097/01.mlr.0000257839.83765.07. [DOI] [PubMed] [Google Scholar]
- 7.Wolters Kluwer Health - Medi-Span. [Accessed April 30, 2009];Duplicate therapy database. http://www.medispan.com/duplicate-therapy-database.aspx.
- 8.First Data Bank. [Accessed September 1, 2009];National drug data file plus: Duplicate therapy module. http://www.firstdatabank.com/Products/national-drug-file.aspx.
- 9.Warholak TL, Rupp MT. Analysis of community chain pharmacists’ interventions on electronic prescriptions. J Am Pharm Assoc. 2009;49:59–64. doi: 10.1331/JAPhA.2009.08013. [DOI] [PubMed] [Google Scholar]
- 10.Bieszk N, Patel R, Heaberlin A, et al. Detection of medication nonadherence through review of pharmacy claims data. Am J Health Syst Pharm. 2003;60:360–366. doi: 10.1093/ajhp/60.4.360. [DOI] [PubMed] [Google Scholar]
- 11.Barnett MJ, Frank J, Wehring H, et al. Analysis of pharmacist-provided medication therapy management (MTM) services in community pharmacies over 7 years. J Manag Care Pharm. 2009;15:18–31. doi: 10.18553/jmcp.2009.15.1.18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Department of Health & Human Services (DHHS), Centers for Medicare and Medicaid Services (CMS) [Accessed April 30, 2009];CMS Manual System: Chapter 7, Medication Therapy Management and Quality Improvement Program. http://www.cms.hhs.gov/PrescriptionDrugCovContra/Downloads/R3PDBv2.pdf.
- 13.Touchette DR, Burns AL, Bough MA, Blackburn JC. Effective health care research report No. 1. (Prepared by University of Illinois-Chicago DEcIDE Center under contract No. HSA29020050038I.) Rockville, MD: Agency for Healthcare Research and Quality; Survey of Medicare Part D plans’ medication therapy management programs. [Google Scholar]
- 14.The Henry J. Kaiser Family Foundation (KFF) Medicare Fact Sheet. [Accessed on April 30, 2009];Medicare Part D Plan Characteristics. 2007 http://www.kff.org/medicare/upload/7426-03.pdf.
- 15.Felland LE, Reschovsky JD. More nonelderly Americans face problems affording prescription drugs. Track Rep. 2009:1–4. [PubMed] [Google Scholar]
- 16.Long SH. Prescription drugs and the elderly: Issues and options. Health Aff (Millwood) 1994;13:157–174. doi: 10.1377/hlthaff.13.2.157. [DOI] [PubMed] [Google Scholar]
- 17.Safran DG, Neuman P, Schoen C, et al. Prescription drug coverage and seniors: Findings from a 2003 national survey. Health Aff (Millwood) 2005:W5–152. W5–166. doi: 10.1377/hlthaff.w5.152. (Suppl Web Exclusives) [DOI] [PubMed] [Google Scholar]
- 18.Thaler RH, Sunstein CR. Nudge: Improving Decisions About Health, Wealth, and Happiness. New Haven, Conn: Yale University Press; 2008. [Google Scholar]
- 19.Von Korff M, Wagner EH, Saunders K. A chronic disease score from automated pharmacy data. J Clin Epidemiol. 1992;45:197–203. doi: 10.1016/0895-4356(92)90016-g. [DOI] [PubMed] [Google Scholar]
- 20.Schneeweiss S, Seeger JD, Maclure M, et al. Performance of comorbidity scores to control for confounding in epidemiologic studies using claims data. Am J Epidemiol. 2001;154:854–864. doi: 10.1093/aje/154.9.854. [DOI] [PubMed] [Google Scholar]
- 21.United States Census. Census 2000 Urban and Rural Classification. 2002 (updated 2008) [Google Scholar]
- 22.National Center for Health Statistics. [Accessed September 1, 2009];International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) 2009 Available at: http://www.cdc.gov/nchs/about/otheract/icd9/abticd9.htm.
- 23.Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: Differing perspectives. J Clin Epidemiol. 1993;46:1075–1079. doi: 10.1016/0895-4356(93)90103-8. Discussion 1081–1090. [DOI] [PubMed] [Google Scholar]
- 24.Hosmer DW, Lemeshow S. Applied Logistic Regression. 2. New York, NY: Wiley; 2000. [Google Scholar]
- 25.Nagelkerke N. A note on a general definition of the coefficient of determination. Biometrika. 1991;78:691–692. [Google Scholar]
- 26.Armitage P, Berry G. Statistical Methods in Medical Research. 3. Oxford; Boston, Mass: Blackwell Scientific Publications; 1994. [Google Scholar]
- 27.Centers for Medicare and Medicaid Services. Medicare Program; Medicare Prescription Drug Benefit; Final Rule, 42 CFR 70, no. 18, 4460 Federal Register. 2005.