variation in the days supply field for osteoporosis medications in ontario

5
SHORT COMMUNICATION Variation in the days supply field for osteoporosis medications in Ontario Andrea M. Burden & Anjie Huang & Mina Tadrous & Suzanne M. Cadarette Received: 12 October 2012 / Accepted: 12 February 2013 # International Osteoporosis Foundation and National Osteoporosis Foundation 2013 Abstract Summary We examined pharmacy claims for osteoporosis medications dispensed in the community (78 %) and long- term care (LTC) to determine if days supply values matched expected dosing intervals. Results identify potential reporting errors that can have implications for drug exposure misclassification, particularly in LTC where only 59 % of reported values matched expected values. Introduction The days supply field is commonly used to examine patterns of drug utilization and classify drug expo- sure, yet its accuracy has received little attention. We sought to describe the days supply reported for osteoporosis drugs and examine if values matched expected therapeutic dosing intervals. Methods We examined days supply values for osteoporosis medications submitted to the Ontario Drug Benefits pro- gram for seniors, 19972011. Days supply values were evaluated by dosing regimen and setting (community or long-term care [LTC]) and compared to pre-defined expected values. We defined expected days supply by the therapeutic dosing interval: daily in 7- or 30-day intervals, or as 100 days; weekly in 7- or 30-day intervals; monthly and daily nasal spray in 28- or 30-day intervals; and cyclical etidronate as a 90-day supply. Results We identified 17,615,404 osteoporosis prescrip- tions, with 78 % dispensed in the community. Most daily oral prescriptions were dispensed by an expected therapeutic dosing interval (97 %). Annual IV zoledronic acid was most commonly dispensed as a 1-day supply (62 %). Distinct differences in agreement were observed for other regimens, with the expected days supply more commonly reported in community versus LTC: cyclical etidronate (86 % vs. 40 %), weekly (91 % vs. 60 %), monthly (94 % vs. 35 %), and nasal spray (84 % vs. 40 %). Conclusions Results suggest that inaccuracies in the days supply field exist, particularly among prescriptions dispensed in LTC. Inaccurate reporting may have significant implica- tions for osteoporosis drug exposure misclassification. Keywords Data quality . Data reporting . Osteoporosis . Pharmacoepidemiology Introduction Pharmacoepidemiologic studies that leverage healthcare uti- lization data are essential to estimate the real-world safety and effectiveness of osteoporosis therapy [1]. Pharmacy claims databases are commonly used to examine drug utili- zation and define drug exposure when studying drug effects on health outcomes [2]. Within pharmacy claims, the days supply field is often used to define exposure status and quantify measures of adherence [24]. The accuracy of the days supply field is thus key for drug exposure classification in pharmacoepidemiology. Prior studies identify less than 0.5 % missing data for days supply or quantity dispensed values [5, 6] and less than 0.1 % error in drug identification [3]. In addition, several studies have found high agreement between self-reported use and pill counts or pharmacy claims [79]. However, little attention has been given to the validity of the days supply field that is used to define periods of drug exposure. We therefore sought to describe days supply values reported for osteoporosis drugs and A. M. Burden : M. Tadrous : S. M. Cadarette Leslie Dan Faculty of Pharmacy, University of Toronto, Toronto, ON, Canada A. Huang : S. M. Cadarette Institute for Clinical Evaluative Sciences, Toronto, ON, Canada S. M. Cadarette (*) Leslie L. Dan Pharmacy Building, University of Toronto, 144 College Street, Toronto, ON M5S 3M2, Canada e-mail: [email protected] Arch Osteoporos (2013) 8:128 DOI 10.1007/s11657-013-0128-1

Upload: mina-tadrous

Post on 13-Dec-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Variation in the days supply field for osteoporosis medications in Ontario

SHORT COMMUNICATION

Variation in the days supply field for osteoporosismedications in Ontario

Andrea M. Burden & Anjie Huang & Mina Tadrous &

Suzanne M. Cadarette

Received: 12 October 2012 /Accepted: 12 February 2013# International Osteoporosis Foundation and National Osteoporosis Foundation 2013

AbstractSummary We examined pharmacy claims for osteoporosismedications dispensed in the community (78 %) and long-term care (LTC) to determine if days supply values matchedexpected dosing intervals. Results identify potentialreporting errors that can have implications for drug exposuremisclassification, particularly in LTC where only 59 % ofreported values matched expected values.Introduction The days supply field is commonly used toexamine patterns of drug utilization and classify drug expo-sure, yet its accuracy has received little attention. We soughtto describe the days supply reported for osteoporosis drugsand examine if values matched expected therapeutic dosingintervals.Methods We examined days supply values for osteoporosismedications submitted to the Ontario Drug Benefits pro-gram for seniors, 1997–2011. Days supply values wereevaluated by dosing regimen and setting (community orlong-term care [LTC]) and compared to pre-definedexpected values. We defined expected days supply by thetherapeutic dosing interval: daily in 7- or 30-day intervals,or as 100 days; weekly in 7- or 30-day intervals; monthlyand daily nasal spray in 28- or 30-day intervals; and cyclicaletidronate as a 90-day supply.Results We identified 17,615,404 osteoporosis prescrip-tions, with 78 % dispensed in the community. Most daily

oral prescriptions were dispensed by an expected therapeuticdosing interval (97 %). Annual IV zoledronic acid was mostcommonly dispensed as a 1-day supply (62 %). Distinctdifferences in agreement were observed for other regimens,with the expected days supply more commonly reported incommunity versus LTC: cyclical etidronate (86 % vs. 40 %),weekly (91 % vs. 60 %), monthly (94 % vs. 35 %), and nasalspray (84 % vs. 40 %).Conclusions Results suggest that inaccuracies in the dayssupply field exist, particularly among prescriptions dispensedin LTC. Inaccurate reporting may have significant implica-tions for osteoporosis drug exposure misclassification.

Keywords Data quality . Data reporting . Osteoporosis .

Pharmacoepidemiology

Introduction

Pharmacoepidemiologic studies that leverage healthcare uti-lization data are essential to estimate the real-world safetyand effectiveness of osteoporosis therapy [1]. Pharmacyclaims databases are commonly used to examine drug utili-zation and define drug exposure when studying drug effectson health outcomes [2]. Within pharmacy claims, the dayssupply field is often used to define exposure status andquantify measures of adherence [2–4]. The accuracy of thedays supply field is thus key for drug exposure classificationin pharmacoepidemiology. Prior studies identify less than0.5 % missing data for days supply or quantity dispensedvalues [5, 6] and less than 0.1 % error in drug identification[3]. In addition, several studies have found high agreementbetween self-reported use and pill counts or pharmacyclaims [7–9]. However, little attention has been given tothe validity of the days supply field that is used to defineperiods of drug exposure. We therefore sought to describedays supply values reported for osteoporosis drugs and

A. M. Burden :M. Tadrous : S. M. CadaretteLeslie Dan Faculty of Pharmacy, University of Toronto, Toronto,ON, Canada

A. Huang : S. M. CadaretteInstitute for Clinical Evaluative Sciences, Toronto, ON, Canada

S. M. Cadarette (*)Leslie L. Dan Pharmacy Building, University of Toronto,144 College Street, Toronto, ON M5S 3M2, Canadae-mail: [email protected]

Arch Osteoporos (2013) 8:128DOI 10.1007/s11657-013-0128-1

Page 2: Variation in the days supply field for osteoporosis medications in Ontario

examine if values matched an expected days supply basedon the therapeutic dosing interval.

Methods

We identified all osteoporosis medications (alendronate,nasal calcitonin, cyclical etidronate, raloxifene, risedronate,and annual zoledronic acid infusion) dispensed through theOntario Drug Benefit (ODB) program for seniors from April1, 1997 to March 31, 2011. Ibandronate is not available inCanada, and denosumab and teriparitide were not listed onthe ODB formulary at the time of analysis [10, 11], andtherefore, these medications were not examined. The ODBdatabase includes a variable that identifies if patients residein long-term care (LTC). This is a mandatory field complet-ed by the pharmacist at time of medication dispensing. Dayssupply values were summarized by dosing regimen (oral:daily, weekly, monthly and cyclical; daily nasal spray; orannual infusion) and setting (community vs. LTC), andcompared to pre-defined expected values. We definedexpected days supply by the therapeutic dosing interval:oral daily in 7- or 30-day intervals, or as 100 days; oralweekly in 7- or 30-day intervals; oral monthly and dailynasal spray in 28- or 30-day intervals; and cyclical etidro-nate as a 90-day supply. The ODB caps the days supplyvalue at 100, and thus, the expected days supply for annualinfusion (365 days) was not examined. Trends over timewere examined by plotting median days supply values andproportion of expected days supply by year and dosingregimen, separately for community and LTC patients. Trendlines were fitted using linear regression.

Results

We identified 17,615,404 prescriptions dispensed for anosteoporosis medication, 78 % in community. Overall, mostprescriptions (82 %) were dispensed for an expected dayssupply value. In both community and LTC settings, oraldaily prescriptions were most commonly dispensed in 7-or 30-day intervals, or as 100 days (96 % community and89 % LTC), and annual infusion was most commonly dis-pensed as a 1-day supply (62 %). However, we identifieddistinct differences between pharmacy settings for otherregimens, with fewer expected days supply values reportedin LTC compared to community, Table 1.

Among community dispensed prescriptions, 90 % had anexpected days supply value, with most prescriptions dis-pensed in 1- or 3-month intervals, Fig. 1a. No significanttrend in the proportion of expected days supply was notedover time in the community (data not shown). However,there were differences in the median days supply over time

for daily and weekly oral regimens between 1998 and 2007,Fig. 2a. Fewer medications dispensed to LTC were capturedby an expected days supply value (59 %), with most pre-scriptions dispensed with a 1-week supply, Fig. 1b. In LTC,the proportion of expected days supply values reportedincreased over time for weekly and daily regimens, yetdecreased over time for the longer dosing interval medica-tions nasal spray calcitonin and cyclical etidronate (data notshown). Since 2001, a 7-day supply was the median valuedispensed for most agents in LTC, Fig. 2b.

Discussion

The days supply field is commonly used to identify osteo-porosis drug exposure in large observational studies of drugutilization, safety, and effectiveness [12]. The accuracy ofthe days supply field is thus essential to assess drug adher-ence and drug effects (safety and effectiveness). Our resultsidentify considerable difference in the days supply fieldcompared to expected dosing intervals, and therefore, resultssuggest potential inaccuracies in the days supply field. Inboth community and LTC settings, daily oral medicationswere commonly dispensed according to our pre-defineddefinition of expected days supply (96 % community and89 % LTC). However, given that any variation from 1 to100 days is possible with a daily oral medication, our abilityto identify potential errors is limited. Indeed, a prior studythat compared electronic information submitted to theODB with written information on prescriptions found that94 % of prescriptions were dispensed as written: 3 %were reduced based on ODB claim restrictions and 1.5 %

Table 1 Proportion of days supply values that matched pre-definedexpected values based on dosing regimen, April 1997–March 2011

Dosing regimen Community(N=13,793,168)

Long-term care(N=3,822,236)

Total(N=17,615,404)

Oral dailya 96.1 89.2 95.2

Oral weeklyb 90.9 59.7 82.1

Oral monthlyc 93.9 34.8 82.7

Oral cyclicald 86.1 40.0 83.0

Nasal sprayc 84.5 39.7 82.7

Total 90.1 59.2 82.4

a Defined as 7- or 30-day intervals, or as 100-day supply; includesalendronate (5 mg, 10 mg), risedronate (5 mg), and raloxifene (60 mg)b Defined as 7- or 30-day intervals, includes weekly alendronate(70 mg) and weekly risedronate (35 mg)c Defined as 28- or 30-day intervals, includes daily nasal calcitonin(200 U) and monthly risedronate (150 mg)d Defined as 90 days, includes cyclical etidronate (400 mg etidronatedaily for 14 days+500 mg calcium for 76 days)

128, Page 2 of 5 Arch Osteoporos (2013) 8:128

Page 3: Variation in the days supply field for osteoporosis medications in Ontario

were partially filled medications due to drug availabilityin the pharmacy [3].

Fortunately, the correct days supply for scheduled dosinginterval medications can be estimated, for example, one pillfor a weekly formulation covers 7 days of therapy, and thus,days supply values should be divisible by 1-week intervals.Among osteoporosis medications with scheduled dosingintervals (oral: weekly, monthly, and cyclical; and dailynasal spray), only 81 % of community and 53 % of LTCprescriptions were captured by an expected days supplyvalue. These results identify differences in potential mea-surement error between community and LTC settings. Ofinterest, most medications in LTC are dispensed as a 7-daysupply. The reasons for the shorter days supply in LTCrequires further investigation but are likely attributable to anumber of structural and process factors within the LTCsetting. We therefore encourage researchers to carefully

consider the dosing interval and setting when analysingpharmacy data to define drug exposure and utilization.

Further research is needed to carefully consider potentialdays supply reporting errors on drug exposure misclassificationand recommend strategies for data cleaning and/or imputation.Researchers may simply impute a value based on an assump-tion of logical clerical error. For instance, a days supply of 1 or4 for a weekly medications may indicate the number of weeksof drugs supplied, and thus, imputing a days supply value of 7and 28, respectively, may be considered. Another strategy mayconsider the current dosing interval and the subsequent dis-pensing date. However, the latter strategy is only feasible ifanother prescription is dispensed [2]. More empirical data areneeded to support data cleaning strategies. Although errors inentries for weekly, monthly, cyclical, and annual medicationsare easier to detect, great attention must be given whenselecting an imputation strategy. The choice of strategy must

Fig. 1 a Days supply values reported for osteoporosis medicationsdispensed in community, by dosing regimen, April 1997–March 2011(N=13,793,168). b Days supply values reported for osteoporosis

medications dispensed in long-term care, by dosing regimen, April1997–March 2011 (N=3,822,236)

Arch Osteoporos (2013) 8:128 Page 3 of 5, 128

Page 4: Variation in the days supply field for osteoporosis medications in Ontario

consider database and drug specific information, and howoverestimation versus underestimation of exposure will affectthe outcome of interest.

A few study limitations are worth noting. First, we cannotidentify the reasons for potential reporting errors. Inaccuratereporting could be the result of a number of unintentionalerrors, including strategies to avoid claim rejections; however,the potential impact on drug exposure misclassification re-mains. Second, we are limited by not having written prescrip-tions or pill counts to validate the days supply values and thuscannot confirm errors among daily dispensed medications.Third, given the nature of the ODB database that specifies amaximum 100-day supply, annual infusions could not be

accurately captured. This emphasizes the importance of under-standing database restrictions and drug formulations whenutilizing pharmacy claims data to identify drug exposure. Forexample, in Ontario, imputing a 365-day supply for an annualzoledronic acid infusion and 90-day supply for cyclical etidro-nate may be appropriate because each infusion of zoledronicacid covers a 365-day period and cyclical etidronate is dis-pensed as a 90-day package of 14 days of active drug and76 days of calcium. However, imputation for weekly ormonth-ly formulations may be more challenging and requires consid-erable methodological justification.

Despite the noted limitations, our study has severalstrengths. First, we were able to report results separately for

Fig. 2 a Median days supply values reported for osteoporosis medi-cations dispensed in community, by dosing regimen and fiscal year,April 1997–March 2011 (N=13,793,168). b Median days supply

values reported for osteoporosis medications dispensed in long-termcare, by dosing regimen and fiscal year, April 1997–March 2011(N=3,822,236)

128, Page 4 of 5 Arch Osteoporos (2013) 8:128

Page 5: Variation in the days supply field for osteoporosis medications in Ontario

prescriptions dispensed in community and LTC, and foundsignificant differences in days supply reporting. Second,ODB data capture all prescription records for medicationsdispensed through the ODB for seniors, permitting a largesample of more than 17 million dispensing records. Third,we examined reporting for three different routes of adminis-tration (oral, nasal, infusion) and five distinct dosing intervals(daily, weekly, monthly, cyclical, yearly). We therefore expectthat our findings for osteoporosis medications in Ontario pointto potential errors in days supply reporting for other med-ications in Ontario, as well as in other pharmacy databases.Pharmacoepidemiologic applications must consider potentialerrors in the days supply field when examining drugutilization and exposure status.

Conclusion

Pharmacy claims data are rich sources to identify real-worlddrug utilization and classify drug exposure. While pharmacydata are thought to be reliable since reimbursement is di-rectly related to the drug dispensed, we identify possibleerrors in days supply reporting, particularly in LTC. Re-searchers should remain mindful of the limitations of phar-macy data. We encourage further research to consider bestpractice in handling days supply reporting errors and em-phasize the importance of improving methodological trans-parency in pharmacoepidemiologic research. At minimum,we encourage pharmacoepidemiologic applications to clar-ify methods used to clean or standardize days supply values.Researchers may currently take steps to clean their databased on a logical and validated method, yet publicationstypically do not report on data cleaning. Without accuratedays supply reporting or consistent methodological strate-gies to adjust for potential misclassification errors, we can-not accurately assess real-world utilization or estimate drugeffects. Future research is needed to clarify the extent ofpotential exposure misclassification related to the days sup-ply value in pharmacoepidemiologic applications and sup-port improved transparency in drug exposure data cleaning.

Acknowledgments This research was supported by research grants toDr. Cadarette from the Canadian Institutes of Health Research (CIHR,DSA-10353) and the Ontario Ministry of Research and Innovation(OMRI, Early Researcher Award ER09-06-043). Dr. Cadarette issupported by a CIHRNew Investigator Award in Aging and Osteoporosis(MSH-95364), Andrea Burden is supported byOntario Graduate Scholar-ships for doctoral research, and Mina Tadrous is supported by a CIHRDoctoral Award—Frederick Banting and Charles Best Canada GraduateScholarship (GSD-11342). Authors acknowledge Brogan Inc. for provid-ing access to drug identification numbers used to identify eligible drugs

and Usama El-Bbayoumi, PharmD, for insightful discussions about LTCpharmacy. The Institute for Clinical Evaluative Sciences (ICES) is a non-profit research corporation funded by the Ontario Ministry of Health andLong-Term Care. As a Prescribed Entity under Ontario's privacy law(Section 45 of the Personal Health Information Protection Act and Reg-ulation 329/04, Section 18), ICES is legally permitted to receive and usepersonal health information for health services research. To have thisprivilege, ICES maintains policies, practices, and procedures that areapproved and regularly audited by the Information and Privacy Commis-sioner of Ontario (www.ipc.on.ca). The opinions, results, and conclusionsare those of the authors and are independent from the funding sources. Noendorsement by CIHR, ICES, OMRI, or the Ontario Ministry Health andLong-Term Care is intended or should be inferred.

Conflicts of interest None.

References

1. Lindsay R (2007) Beyond clinical trials: the importance of largedatabases in evaluating differences in the effectiveness ofbisphosphonate therapy in postmenopausal osteoporosis. Bone40:S32–S35

2. Schneeweiss S, Avorn J (2005) A review of uses of health careutilization databases for epidemiologic research on therapeutics. JClin Epidemiol 58:323–337

3. Levy AR, O'Brien BJ, Sellors C et al (2003) Coding accuracy ofadministrative drug claims in the Ontario Drug Benefit database.Can J Clin Pharmacol 10:67–71

4. West SL, Savitz DA, Koch G et al (1995) Recall accuracy forprescription medications: self-report compared with database in-formation. Am J Epidemiol 142:1103–1112

5. Eurich DT, Majumdar SR, Marrie TJ (2012) Population-basedcohort study of outpatients with pneumonia: rationale, design andbaseline characteristics. BMC Infect Dis 12:135

6. Tamblyn R, Lavoie G, Petrella L et al (1995) The use ofprescription claims databases in pharmacoepidemiological re-search: the accuracy and comprehensiveness of the prescrip-tion claims database in Quebec. J Clin Epidemiol 48:999–1009

7. Cadarette SM, Jaglal SB, Raman-Wilms L et al (2011) Osteoporosisquality indicators using healthcare utilization data. Osteoporos Int22:1335–1342

8. Grymonpre R, Cheang M, Fraser M et al (2006) Validity of aprescription claims database to estimate medication adherence inolder persons. Med Care 44:471–477

9. Steiner JF, Prochazka AV (1997) The assessment of refill compli-ance using pharmacy records: methods, validity, and applications.J Clin Epidemiol 50:105–116

10. Cadarette SM, Carney G, Baek D et al (2012) Osteoporosis med-ication prescribing in British Columbia and Ontario: impact ofpublic drug coverage. Osteoporos Int 23:1475–1480

11. Ontario Ministry of Health and Long-Term Care (http://www.health.gov.on.ca/english/providers/program/drugs/odbf_eformulary.html) Drugs funded by Ontario Drug Benefit (ODB)Program—Search e-Formulary. Accessed January 15, 2012

12. Cadarette SM, Burden AM (2010) Measuring and improving ad-herence to osteoporosis pharmacotherapy. Curr Opin Rheumatol22:397–403

Arch Osteoporos (2013) 8:128 Page 5 of 5, 128