Abstract
BACKGROUND:
Days supply values reported in large administrative claims databases are commonly used to estimate drug exposure and quantify adherence and persistence with prescribed therapy. In recent claims database studies assessing treatment patterns for biologic therapies, a high frequency of 28-31–days supply values has been observed for therapies with label-recommended maintenance dosing intervals longer than 4 weeks. Such inconsistencies suggest potential inaccuracy of days supply data.
OBJECTIVES:
To confirm the existence and describe the extent of inconsistencies in the reported days supply values and the documented fill intervals among prescription claims from administrative claims databases for 2 different biologics with label-recommended maintenance dosing intervals longer than 4 weeks and 2 biologics with intervals less than or equal to 4 weeks.
METHODS:
Using data from 2 large US administrative claims databases (IBM MarketScan Commercial Claims and Encounters and the Optum Clinformatics Data Mart Socio-economic Status [SES]), the reported days supply values and associated intervals between consecutive fills for 2 biologics with maintenance dosing intervals longer than 4 weeks (guselkumab and ustekinumab) and 2 with intervals less than or equal to 4 weeks (adalimumab and ixekizumab) were described. For all fill pairs with reported days supply values of 28-31 days, the percentage with inconsistent fill intervals (defined as >45 days or >60 days) was calculated.
RESULTS:
Across all datasets, the proportions of fill pairs with inconsistent days supply values and fill intervals (ie, days supply values of 28-31 days but fill intervals of >45 days) were 41.8%-73.4% for guselkumab, 33.4%-59.4% for ustekinumab, 8.5%-9.5% for adalimumab, and 7.3%-11.4% for ixekizumab. The same trend was observed across these biologics when using more than 60 days to define an inconsistent fill interval. Unlike adalimumab and ixekizumab, a wide distribution of fill intervals was observed among guselkumab and ustekinumab fill pairs with 28-31 days supply values, with peaks evident at approximately 28-31 days as well as around the label-recommended maintenance dosing intervals for these therapies (56 or 84 days).
CONCLUSIONS:
This study demonstrated a large discrepancy between days supply values and fill intervals reported in administrative claims data for biologics with label-recommended maintenance dosing intervals longer than 4 weeks (ie, guselkumab and ustekinumab), potentially suggesting widespread underestimation of days supply values for these therapies. Such inconsistencies in the reported days supply values may lead to underestimation of treatment adherence and persistence for these biologics, which could be mitigated by systematic data imputation.
Plain language summary
Some biologic drugs should be taken less often than once every 4 weeks. In prescription claims for these drugs, the recorded number of days that a prescription should cover did not match well with the number of days until the next prescription was filled. This may have been done so patients would not have their claims denied by health plans. This could affect results of studies that assess whether patients take their medications as directed.
Implications for managed care pharmacy
In administrative claims databases, inconsistencies between the reported days supply values and prescription fill intervals were observed for biologics with maintenance intervals of more than 4 weeks. Days supply may be incorrectly reported as 1 month or less because of 30-day supply restrictions implemented by some health plans. Such underestimation of days supply could bias adherence/persistence estimates derived from claims data. Researchers and health care decision makers who use claims data in their assessments should consider this potential data limitation. label-recommended maintenance dosing intervals longer than 4 weeks and 2 biologics with intervals less than or equal to 4 weeks.
Medication adherence, defined as the extent to which a patient acts in accordance with the prescribed dosing regimen, and persistence, defined as the time from treatment initiation to its discontinuation, are important determinants of treatment outcomes and cost of care.1 Broadly speaking, poor adherence/persistence with prescribed medications is associated with poor clinical outcomes and increased health care costs.1-5 Further, adherence and persistence measures, such as those reported by the Healthcare Effectiveness Data and Information Set, are sometimes considered in the calculation of indicators used by policy-makers to assess performance of health care plans and quality of care in the United States.6-8 Across the specialty pharmacy industry, these measures are also increasingly used by accrediting bodies, payers, pharmacies, and drug manufacturers to assess quality differences or clinical benefit.9 Administrative claims databases are frequently used to assess real-world persistence and adherence to medications, including biologic therapies.10-19 The days supply field in the pharmacy claims, which indicates the intended duration of each prescription fill, is commonly used to estimate drug exposure and quantify adherence and persistence with prescribed therapy.20 However, potential inaccuracies in reported days supply values for various therapies, including biologics, have previously been reported, which could have important implications in calculations of adherence and persistence measures using administrative claims data.6,17,18,21
In some recent studies assessing adherence, discontinuation, and/or persistence of biologic therapies among patients with psoriasis, it was noted that a significant proportion of days supply values for guselkumab and ustekinumab fills documented in US claims databases was 30 days,17,18,22 which is not consistent with the label recommended maintenance dosing intervals for these therapies (8 weeks and 12 weeks, respectively)23,24 or the corresponding fill intervals (ie, the time between prescription refills) reported in the database.17,18 Further discussions with pharmacists and payer reimbursement experts indicated that inaccurate reporting may be due to restrictions on the maximum days supply (typically 30 days) imposed by some health plans, regardless of the actual drug quantity dispensed to the patient or label-recommended maintenance dosing intervals. This potential system-level factor could influence data entry and may have methodological implications for quantifying adherence and persistence using administrative claims data.21 More specifically, if the days supply values are consistently underestimated for therapies with label-recommended maintenance dosing intervals longer than 4 weeks, then using the reported number of days supply could substantially underestimate adherence and persistence with these therapies.
The goal of this study was to confirm the existence and describe the extent of inconsistencies in the reported days supply values and the documented fill intervals using data from 2 large US claims databases. Data were assessed for 2 biologics with label-recommended maintenance dosing intervals longer than 4 weeks (guselkumab and ustekinumab) and for 2 biologics with label-recommended maintenance dosing intervals of less than or equal to 4 weeks (adalimumab and ixekizumab) across all adult indications.
Methods
DATA SOURCES
This retrospective, observational, descriptive study was conducted using data from 2 large administrative claims databases: the IBM MarketScan Commercial Claims and Encounters (CCAE) database (MarketScan Commercial) and the Optum Clinformatics Data Mart Socio-economic Status (SES) database (OptumInsight Life Sciences, Inc). Data were analyzed separately for each data source.
The IBM MarketScan CCAE administrative database captures patient-level health care claims data (inpatient medical, outpatient medical, and outpatient pharmacy services) of individuals, as well as their spouses and dependents, enrolled in US employer-sponsored insurance health plans.25 It also captures performed laboratory tests for a subset of the covered lives. This administrative claims database includes a variety of fee-for-service health plans, preferred provider organizations, and capitated health plans. The population is nationally representative of all 10 US census regions. All data are fully deidentified and comply with the Health Insurance Portability and Accountability Act of 1996.
The Optum Clinformatics Data Mart is an adjudicated administrative health claims database that includes members with private health insurance fully insured in commercial plans or in administrative services only, Medicare Advantage (Medicare Advantage Prescription Drug coverage starting January 2006), or Legacy Medicare Choice Lives (prior to January 2006).26,27 Optum SES also contains socioeconomic status and location information for members with both medical and pharmacy coverage. Patient-level data from Optum SES are mainly representative of US commercial claims patients (aged 0-65 years), with some Medicare patients (aged 65+ years). The data are collected from administrative claims processed from inpatient and outpatient medical services, dispensed prescriptions, and outpatient laboratory tests. All data are deidentified. In the present study, data from the Optum SES commercial and Medicare Advantage datasets were analyzed separately.
STUDY POPULATION
The analyses were conducted at the prescription-fill level. The reported days supply of a prescription and the fill interval between that prescription and the subsequent prescription were described. The inclusion and exclusion criteria were first applied to patients and then to prescription fills. Patients with 2 or more biologic prescriptions during the intake period for any of the therapies of interest (adalimumab, ixekizumab, guselkumab, or ustekinumab) were considered for inclusion. Female patients who were pregnant at any time during the study period were excluded. All prescriptions for the biologics of interest filled between April 1, 2017, and September 30, 2020, for which (1) the patient was aged 18 years or older at the time of the first fill, and (2) there was a subsequent refill within 6 months were included (Figure 1). The fills for which the cost was equal to $0 in the claims database were excluded because zero cost could indicate reversed (patient withdrew) or rejected (payer rejected) pharmacy claims. There was no continuous enrollment criterion for the index date (the time before the fill date of the earlier prescription in each pair). Continuous enrollment for at least 6 months post-index (ie, after first prescription) was required.
FIGURE 1.
Study Design
STUDY DESIGN
This study examined the reported days supply values and fill intervals (calculated as fill date of subsequent prescription minus the fill date for index prescription) for 4 commonly prescribed biologic therapies: 2 biologics with label-recommended maintenance dosing intervals of less than or equal to 4 weeks (adalimumab and ixekizumab)28,29 and 2 biologics with label recommended maintenance dosing intervals of more than 4 weeks (guselkumab and ustekinumab)23,24 (Supplementary Table 1 (762.2KB, pdf) , available in online article). The analysis focused on prescription fills with reported days supply values of 28-31 days, which would be consistent with the label-recommended induction dosing interval but not with the label recommended maintenance dosing intervals for guselkumab and ustekinumab. The drug codes used to identify prescription fills for the therapies of interest in the claims databases are provided in Supplementary Table 2 (762.2KB, pdf) .
Because dosing intervals for ustekinumab vary by indication, additional separate analyses were conducted for this biologic specifically examining claims from patients with diagnoses of plaque psoriasis or psoriatic arthritis (PsO/PsA) and those with Crohn disease or ulcerative colitis (CD/UC). The diagnostic codes used to identify claims from patients with PsO/PsA and with CD/UC are provided in the Supplementary Table 3 (762.2KB, pdf) .
Consecutive prescription fills were paired such that each pair had an index date (date of the first fill in the pair) and a refill date (date of the second fill in each pair) (Figure 1). For example, if a patient had 3 prescriptions for the same biologic during the study period, there were 2 pairs included in the study: for the first pair, the first prescription date was the index date and the second prescription was the refill date, whereas for the second pair, the second prescription fill date was the index date and the third prescription fill date was the refill date.
Among the fill pairs with reported days supply values of 28-31 days, inconsistencies between the days supply values and fill intervals were defined as fill intervals of more than 45 days (ie, ≥ 50% longer than the indicated days supply) or more than 60 days (ie, ≥ 2 times the indicated days supply).
STATISTICAL ANALYSIS
Descriptive statistics were used to summarize the patient characteristics, proportions of fill pairs with days supply values of 28-31 days, and inconsistencies between the days supply values and fill intervals. Counts and percentages were reported for categorical variables, and means and SDs were used for continuous variables. The distributions of fill intervals for fill pairs with days supply values of 28-31 days were displayed in histograms. No statistical comparisons were made and there were no adjustments for possible confounders. Statistical analyses were performed using SAS Enterprise Guide version 7.15 (SAS Institute Inc.).
Results
STUDY POPULATION
A total of 1,739,569 prescription pairs met the inclusion/exclusion criteria for this analysis. These prescriptions were filled by a total of 100,612, 56,944, and 12,275 unique patients in the IBM CCAE database, Optum SES Commercial dataset, and Optum SES Medicare Advantage dataset, respectively. The demographic characteristics of these patients are summarized by treatment in Table 1. Across the 3 datasets, the average number of fill pairs identified per patient was 10-11 (range of 4-11 per patient depending on the specific biologic and dataset). Most patients included in the study (85%-91% overall across datasets) had at least 1 fill pair with a reported days supply value of 28-31 days, approximately half of whom (45.2%-51.2%) had at least 1 fill pair with a fill interval of more than 45 days. Refer to Supplementary Table 4 (762.2KB, pdf) for more details.
TABLE 1.
Demographic Characteristics of the Study Cohort (N = 169,831)
| IBM CCAE (N = 100,612) | Optum SES Commercial (N = 56,944) | Optum SES Medicare Advantage (N = 12,275) | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ADA | IXE | GUS | UST | ADA | IXE | GUS | UST | ADA | IXE | GUS | UST | |
| Number of patients, Na | 79,495 | 6,182 | 3,429 | 14,689 | 44,770 | 1,688 | 2,515 | 9,018 | 10,182 | 593 | 366 | 1,525 |
| Sex, n (%) | ||||||||||||
| Female | 43,092 (54.2) | 2,939 (47.5) | 1,564 (45.6) | 7,308 (49.8) | 21,336 (47.7) | 718 (42.5) | 1,053 (41.9) | 4,181 (46.4) | 7,117 (69.9) | 373 (62.9) | 217 (59.3) | 933 (61.2) |
| Male | 36,403 (45.8) | 3,243 (52.5) | 1,865 (54.4) | 7,381 (50.2) | 23,434 (52.3) | 970 (57.5) | 1,462 (58.1) | 4,837 (53.6) | 3,065 (30.1) | 220 (37.1) | 149 (40.7) | 592 (38.8) |
| Age at index (years), n (%) | ||||||||||||
| 18-34 | 17,788 (22.4) | 826 (13.4) | 551 (16.1) | 3,295 (22.4) | 10,115 (22.6) | 221 (13.1) | 483 (19.2) | 2,075 (23.0) | 126 (1.2) | 7 (1.2) | 2 (0.5) | 30 (2.0) |
| 35-44 | 16,585 (20.9) | 1,337 (21.6) | 795 (23.2) | 3,428 (23.3) | 9,929 (22.2) | 387 (22.9) | 600 (23.9) | 2,121 (23.5) | 433 (4.3) | 31 (5.2) | 12 (3.3) | 118 (7.7) |
| 45-54 | 21,967 (27.6) | 2,054 (33.2) | 1,061 (30.9) | 4,078 (27.8) | 12,032 (26.9) | 526 (31.2) | 747 (29.7) | 2,494 (27.7) | 1,138 (11.2) | 83 (14.0) | 41 (11.2) | 184 (12.1) |
| 55-64 | 22,368 (28.1) | 1,916 (31.0) | 1,009 (29.4) | 3,837 (26.1) | 10,976 (24.5) | 469 (27.8) | 608 (24.2) | 2,026 (22.5) | 2,300 (22.6) | 135 (22.8) | 81 (22.1) | 290 (19.0) |
| 65-74 | 787 (1.0) | 49 (0.8) | 13 (0.4) | 51 (0.3) | 1,589 (3.5) | 77 (4.6) | 76 (3.0) | 272 (3.0) | 4,576 (44.9) | 248 (41.8) | 165 (45.1) | 696 (45.6) |
| 75-79 | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 84 (0.2) | 5 (0.3) | 1 (0.0) | 20 (0.2) | 996 (9.8) | 58 (9.8) | 43 (11.7) | 125 (8.2) |
| ≥80 | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 45 (0.1) | 3 (0.2) | 0 (0.0) | 10 (0.1) | 613 (6.0) | 31 (5.2) | 22 (6.0) | 82 (5.4) |
| Mean (SD) | 45.3 (12.6) | 47.8 (10.9) | 46.6 (11.4) | 44.7 (12.3) | 45.4 (12.7) | 48.2 (11.3) | 45.9 (11.8) | 44.8 (12.6) | 64.6 (10.9) | 63.6 (11.3) | 65.3 (10.3) | 63.1 (12.1) |
| Region, n (%) | ||||||||||||
| Northeast | 11,838 (14.9) | 879 (14.2) | 607 (17.7) | 2,883 (19.6) | 4,062 (9.1) | 155 (9.2) | 291 (11.6) | 1,066 (11.8) | 1,178 (11.6) | 83 (14.0) | 64 (17.5) | 257 (16.9) |
| Midwest | 17,083 (21.5) | 1,291 (20.9) | 642 (18.7) | 2,916 (19.9) | 12,702 (28.4) | 504 (29.9) | 543 (21.6) | 2,408 (26.7) | 1,643 (16.1) | 97 (16.4) | 49 (13.4) | 230 (15.1) |
| South | 38,486 (48.4) | 3,131 (50.6) | 1,749 (51.0) | 6,755 (46.0) | 20,233 (45.2) | 793 (47.0) | 1,260 (50.1) | 3,977 (44.1) | 5,464 (53.7) | 323 (54.5) | 194 (53.0) | 716 (47.0) |
| West | 11,587 (14.6) | 806 (13.0) | 374 (10.9) | 2,009 (13.7) | 7,679 (17.2) | 230 (13.6) | 418 (16.6) | 1,555 (17.2) | 1,877 (18.4) | 88 (14.8) | 58 (15.8) | 322 (21.1) |
| Unknown | 501 (0.6) | 75 (1.2) | 57 (1.7) | 126 (0.9) | 94 (0.2) | 6 (0.4) | 3 (0.1) | 12 (0.1) | 20 (0.2) | 2 (0.3) | 1 (0.3) | 0 (0.0) |
| Commercial plan type, n (%) | ||||||||||||
| HMO | 8,358 (10.5) | 564 (9.1) | 373 (10.9) | 1,511 (10.3) | 4,636 (10.4) | 240 (14.2) | 207 (8.2) | 886 (9.8) | 2,265 (22.2) | 132 (22.3) | 83 (22.7) | 281 (18.4) |
| PPO | 41,787 (52.6) | 3,147 (50.9) | 1,644 (47.9) | 7,531 (51.3) | 378 (0.8) | 18 (1.1) | 7 (0.3) | 51 (0.6) | 815 (8.0) | 50 (8.4) | 39 (10.7) | 164 (10.8) |
| POS | 8,274 (10.4) | 719 11.6) | 416 (12.1) | 1,660 (11.3) | 34,566 (77.2) | 1,198 (71.0) | 1,955 (77.7) | 7,026 (77.9) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) |
| OTH | 21,076 (26.5) | 1,752 (28.3) | 996 (29.0) | 3,987 (27.1) | 5,190 (11.6) | 232 (13.7) | 346 (13.8) | 1,055 (11.7) | 7,102 (69.8) | 411 (69.3) | 244 (66.7) | 1,080 (70.8) |
aThe total for all 4 biologics of interest may not correspond to the unique patient count indicated for each database because a patient may have switched from 1 therapy of interest to another during the study period.
ADA = adalimumab; CCAE = Commercial Claims and Encounters; GUS = guselkumab; HMO = health maintenance organization; IXE = ixekizumab; OTH = other; POS = point of service; PPO = preferred provider organization; SES = Socio-economic Status; UST = ustekinumab.
PERCENTAGE OF FILL PAIRS WITH DAYS SUPPLY VALUES OF 28-31 DAYS
For adalimumab and ixekizumab, the majority of fill pairs (87.4%-97.8% and 94.6%-98.8%, respectively) had reported days supply values of 28-31 days in all 3 datasets (ie, IBM CCAE, Optum SES Commercial, and Optum SES Medicare Advantage) (Table 2), consistent with the label-recommended maintenance dosing intervals for these biologics. For guselkumab and ustekinumab, the proportions of fill pairs with reported days supply values of 28-31 days were 19.9%-89.6% and 16.1%-34.1%, respectively, with greater variability across the datasets than that observed with adalimumab and ixekizumab (Table 2). As mentioned above, a days supply value of 28-31 days is not consistent with the label-recommended maintenance dosing intervals for guselkumab (56 days) and ustekinumab (56 days for CD/UC and 84 days for PsO/PsA).
TABLE 2.
Percentage of Fill Pairs With Inconsistency in Days Supply Values and Fill Intervals
| Claims database | Total pairs of fills, N | Fill pairs with days supply of 28-31 days, n (%) | Fill pairs with 28-31 days supply and fill interval of >45 days, n (%)a | Fill pairs with 28-31 days supply and fill interval of >60 days, n (%)a |
|---|---|---|---|---|
| Adalimumab | ||||
| IBM CCAE | 844,050 | 757,907 (89.8) | 72,183 (9.5) | 30,900 (4.1) |
| Optum SES Commercial | 524,285 | 512,903 (97.8) | 48,902 (9.5) | 21,720 (4.2) |
| Optum SES Medicare Advantage | 117,260 | 102,528 (87.4) | 8,687 (8.5) | 4,025 (3.9) |
| Ixekizumab | ||||
| IBM CCAE | 54,109 | 51,421 (95.0) | 5,474 (10.6) | 2,328 (4.5) |
| Optum SES Commercial | 15,370 | 15,178 (98.8) | 1,724 (11.4) | 845 (5.6) |
| Optum SES Medicare Advantage | 5,884 | 5,567 (94.6) | 407 (7.3) | 178 (3.2) |
| Guselkumab | ||||
| IBM CCAE | 18,107 | 6,105 (33.7) | 3,758 (61.6) | 1,857 (30.4) |
| Optum SES Commercial | 13,552 | 12,141 (89.6) | 8,909 (73.4) | 4,240 (34.9) |
| Optum SES Medicare Advantage | 2,230 | 443 (19.9) | 185 (41.8) | 88 (19.9) |
| Ustekinumab: All | ||||
| IBM CCAE | 85,187 | 29,037 (34.1) | 16,789 (57.8) | 12,929 (44.5) |
| Optum SES Commercial | 50,691 | 14,846 (29.3) | 8,812 (59.4) | 6,738 (45.4) |
| Optum SES Medicare Advantage | 8,844 | 1,422 (16.1) | 475 (33.4) | 358 (25.2) |
| Ustekinumab: PsO/PsA | ||||
| IBM CCAE | 47,040 | 15,180 (32.3) | 11,488 (75.7) | 10,684 (70.4) |
| Optum SES Commercial | 30,037 | 7,713 (25.7) | 5,694 (73.8) | 5,307 (68.8) |
| Optum SES Medicare Advantage | 5,187 | 674 (13.0) | 326 (48.4) | 280 (41.5) |
| Ustekinumab: CD/UC | ||||
| IBM CCAE | 32,858 | 12,128 (36.9) | 4,468 (36.8) | 1,689 (13.9) |
| Optum SES Commercial | 17,279 | 6,074 (35.2) | 2,597 (42.8) | 1,137 (18.7) |
| Optum SES Medicare Advantage | 3,018 | 645 (21.4) | 121 (18.8) | 54 (8.4) |
aThe number of fill pairs with days supply of 28-31 days was used as the denominator for the calculations of percentages of such fill pairs with fill intervals of more than 45 days or more than 60 days.
CCAE = Commercial Claims and Encounters; CD = Crohn disease; PsA = psoriatic arthritis; PsO = plaque psoriasis; SES = Socio-economic Status; UC = ulcerative colitis.
PERCENTAGE OF FILL PAIRS WITH INCONSISTENCIES BETWEEN DAYS SUPPLY VALUES AND FILL INTERVALS
The reported days supply values of 28-31 days were generally inconsistent with the corresponding fill intervals for guselkumab and ustekinumab, whereas the fill intervals and reported days supply were generally consistent for adalimumab and ixekizumab.
Specifically, in all 3 datasets, the proportions of fill pairs with inconsistent days supply values and fill intervals (ie, days supply values of 28-31 days but fill intervals of >45 days) were 41.8% 73.4% for guselkumab, 33.4%-59.4% for ustekinumab, 8.5%-9.5% for adalimumab, and 7.3%-11.4% for ixekizumab (Table 2; Figure 2). For ustekinumab, higher proportions of inconsistent days supply values and fill intervals were observed for prescription pairs associated with diagnostic codes for PsO/PsA (48.4%-75.7%; maintenance interval of 12 weeks) than for those associated with diagnostic codes for CD/UC (18.8%-36.8%; induction and maintenance interval of 8 weeks).
FIGURE 2.
Percentage of Fill Pairs With Fill Intervals of More Than 45 Days Among Pairs of Fills With 28-31 Days Supply
Comparing the 3 datasets, Optum SES Medicare Advantage had the lowest proportions of fill pairs with inconsistent values; this observed difference between the datasets was more pronounced for guselkumab and ustekinumab than for adalimumab and ixekizumab. Nonetheless, the proportions of inconsistent values for longer-interval therapies (ie, days supply values of 28-31 days but fill intervals of >45 days) in the Optum SES Medicare Advantage dataset were still relatively high: 41.8% for guselkumab and 33.4% for ustekinumab overall (48.4% for PsO/PsA, 18.8% for CD/UC).
Similar results were obtained when the defining threshold for an inconsistent fill interval was increased to more than 60 days (Table 2; Supplementary Figure 1 (762.2KB, pdf) ). The proportions of fill pairs with days supply values of 28-31 days but fill intervals of more than 60 days were 19.9%-34.9% for guselkumab, 25.2%-45.4% for ustekinumab, 3.9%-4.2% for adalimumab, and 3.2%-5.6% for ixekizumab.
DISTRIBUTION OF FILL INTERVALS
The inconsistency between days supply values and fill intervals for biologics with label-recommended maintenance dosing intervals longer than 4 weeks was also apparent when examining the distribution of fill intervals among all pairs of fills with reported days supply values of 28-31 days (Supplementary Figure 2 (762.2KB, pdf) ). Histograms displaying the distribution of fill intervals in each dataset are presented in Supplementary Figures 3-5 (762.2KB, pdf) . Data from all 3 datasets indicate that for the biologics with label-recommended maintenance dosing intervals of less than or equal to 4 weeks (ie, adalimumab and ixekizumab), the fill intervals had relatively narrow distributions, with single peaks centered around approximately 4 weeks. In contrast, the distributions of fill intervals for therapies with longer label-recommended maintenance dosing intervals (ie, guselkumab and ustekinumab) were wide, typically with apparent peaks at approximately 4 weeks (which corresponds to the induction dosing interval of guselkumab and ustekinumab for PsO/PsA but not ustekinumab for CD/UC) and at around the time that a refill would be expected based on label-recommended maintenance dosing intervals for these therapies (56 days for guselkumab and ustekinumab for CD/UC, 84 days for ustekinumab for PsO/PsA).
Discussion
Large administrative claims databases are valuable, convenient, and informative data sources frequently used to assess real-world adherence and persistence to prescribed medication regimens.20,30 However, such assessments typically rely on the accuracy of the days supply field recorded in claims data.6,20,21 In the present analysis of data from the IBM CCAE and Optum SES claims databases, we found that a considerable proportion of index fills for guselkumab and ustekinumab with a reported days supply value of 28-31 days had an interval to a subsequent fill of longer than 45 days. Longer fill intervals are consistent with the label-recommended maintenance dosing intervals for guselkumab (8 weeks) and ustekinumab (8 or 12 weeks depending on the indication). In contrast, the reported days supply values and fill intervals were generally consistent for adalimumab and ixekizumab, both of which have label-recommended maintenance dosing intervals of less than or equal to 4 weeks. More specifically, the majority of adalimumab and ixekizumab fills (approximately 90%) that had a reported days supply value of 28-31 days also had a consistent fill interval of less than 45 days.
A potential reason for the observed discrepancy between days supply values and fill intervals for therapies with longer label-recommended maintenance dosing intervals like guselkumab and ustekinumab has been speculated to be short-cycle policies imposed by payers that restrict certain medications to a 30-day maximum.17,18,31 As with other forms of coverage restrictions, both private and public health plans can impose quantity or days supply limitations, which are intended to ensure patient safety and control medication wastage and health care costs.32,33 For example, to contain Medicaid pharmacy costs, nearly all states set limits on medication days supply.33 Most states have a dispensing limit of 34 days, and only 13 states allow up to a 90 day supply for some medications.33 Such policies may influence how pharmacists enter the days supply for medications with label-recommended maintenance dosing intervals exceeding the maximum limit, which has also been speculated by others.21 In the case of prescriptions for guselkumab and ustekinumab assessed in the present study, it is possible that some days supply values may have been intentionally reported as 28-31 days to ensure reimbursement (ie, avoid claim rejections).
It is critical to understand how factors like reimbursement policies may influence data entry in administrative claims databases, as this may have methodological implications for quantifying adherence and persistence using these data sources.20,21 Two commonly used adherence metrics are the medication possession ratio and the proportion of days covered, both of which use data estimated from the days supply field in their calculations.4,6,9,20 Therefore, for both of these measures, the accuracy of days supply values reported in claims data has been recognized as critical to obtaining valid estimates.21 Underreported days supply values at large scale for therapies with dosing intervals longer than 4 weeks would result in a substantial underestimation of adherence and persistence estimates. Development of data cleaning, adjustment, and imputation approaches for the days supply field may be needed to ensure accurate estimation of adherence and persistence, particularly for medications with longer label-recommended maintenance dosing intervals.
Methodological challenges associated with using data from large administrative claims databases to estimate adherence/persistence, including the potential for exposure misclassification when using the days supply field, have been reported by others. For example, a Canadian claims database study by Burden and colleagues found potential errors in days supply reporting for osteoporosis medications with fixed dosing intervals, identifying that a large proportion of values did not match the expected dosing intervals and noting that an underestimation of days supply was particularly common among prescriptions dispensed to long term care residents.34 The authors hypothesized that inaccurate reporting could be the result of unintentional errors and/or strategies to avoid claim rejections. A subsequent study by the same authors demonstrated that not accounting for potential misclassification of days supply values can lead to underestimation of both adherence and persistence with these medications, whereas using dose-specific and database-specific data-cleaning algorithms improved adherence and persistence estimates.21
Potential limitations with using the days supply values reported in claims data have also been noted for biologics. In a recent analysis estimating persistence with PsO therapies using the Truven Health MarketScan claims database, Hendrix and colleagues noted that the validity of persistence estimates for ustekinumab that relied on days supply data was undermined by the high frequency of reported 30-day days supply values, the reason for which they could not distinguish from the claims data alone.22 The authors emphasized that defining ustekinumab’s permissible gap based on days supply values for persistence estimates is therefore of uncertain validity. A recent study by Shah and colleagues assessing medication adherence among patients with inflammatory bowel disease also noted that the actual prescribed days supply for ustekinumab often exceeds the 30-day supply limit mandated by some third-party benefit managers and, therefore, opted to use the prescribed days supply reported in electronic medical records instead of that reported in claims data.31 Lastly, in 2 previous studies by our group that used the IBM CCAE health care claims database to assess adherence with 7 different biologic medications among patients with psoriasis, the days supply value was imputed to 56 days for guselkumab and to 84 days for ustekinumab to account for the high frequency of 30-day days supply values that were inconsistent with the reported fill intervals among pharmacy claims for these therapies.17,18
LIMITATIONS
The results of this study should be interpreted with consideration of certain limitations. Although claims data are valuable for research purposes, they are collected for the purpose of facilitating payment for health care services and, therefore, lack definitive diagnoses, context for treatment choice and utilization, and information on actual usage of medications between fill dates. For this reason, it is not possible to confirm the exact reason for the inconsistency between the days supply values and fill intervals in each case. For example, inconsistencies between the reported days supply value and fill intervals could be caused by a variety of factors, including unintentional errors, actual noncompliance, the tendency of patients to discontinue and switch biologics based on therapy effectiveness, or purposeful misrepresentation to avoid claims rejections when health plans have restrictions on days supply values. It should also be noted that medications provided as a free sample by physicians or directly purchased by the patient would not be reflected in insurance claims data. As noted by others, another limitation is that it is not possible to reliably differentiate between induction and maintenance dosing based on claims data alone.22 For example, patients may have initiated a therapy of interest through trial programs and free samples, which, as mentioned above, would not be captured in the database, or patients may have initiated the therapy while under one health plan but then transferred to another. Therefore, assuming that the first few prescriptions filled by a patient are an induction regimen would be inaccurate. As with all retrospective claims database analyses, other limitations associated with the use of such data should also be recognized, including missing codes, coding limitations leading to potential misclassification, and coding errors. Finally, this study includes patients covered by IBM MarketScan Commercial and Optum health coverage; the results may not be generalizable to patients with other insurance types or to those without health insurance coverage. These results may also not be generalizable to claims database studies outside of the United States because of differences in coverage policies and restrictions.
Conclusions
Overall, the results of this descriptive study demonstrate inconsistencies between the days supply values and fill intervals for guselkumab and ustekinumab in 2 large US administrative claims databases. These observations suggest that using the days supply values reported in claims databases to quantify adherence and persistence may underestimate these outcomes for biologics with label-recommended maintenance dosing intervals longer than 4 weeks. Based on these findings, it is recommended that investigators using administrative claims data, as well as health care decision makers who consider such studies in their assessments, take into account these potential limitations of claims data. As noted by others, best practices for handling inaccuracies in days supply reporting, as well as for methodological transparency in any research using such data, should be developed and used.20,21 Consistent and clearly reported data cleaning, adjustment, and imputation approaches for the days supply field would improve the accuracy of adherence and persistence measures derived from claims data. Future research should aim to address these unmet needs and continue to examine other potential confounding factors that may influence the results of analyses that use data from claims databases.
ACKNOWLEDGMENTS
The authors would like to thank Aidan Mangan of Cobbs Creek Healthcare LLC and Cullen Seal of TechData Service LLC for programming support, as well as Dr Irene Zahirovic and Dr Ari Mendell of EVERSANA for their medical writing and editorial assistance in the development of this manuscript.
REFERENCES
- 1.Cramer JA, Roy A, Burrell A, et al. Medication compliance and persistence: Terminology and definitions. Value Health. 2008;11(1):44-7. doi:10.1111/j.1524-4733.2007.00213.x [DOI] [PubMed] [Google Scholar]
- 2.DiMatteo MR, Giordani PJ, Lepper HS, et al. Patient adherence and medical treatment outcomes: A metaanalysis. Med Care. 2002;40(9):794-811. doi:10.1097/00005650-200209000-00009 [DOI] [PubMed] [Google Scholar]
- 3.Roebuck MC, Liberman JN, Gemmill-Toyama M, et al. Medication adherence leads to lower health care use and costs despite increased drug spending. Health Aff (Millwood). 2011;30(1):91-9. doi:10.1377/hlthaff.2009.1087 [DOI] [PubMed] [Google Scholar]
- 4.Iuga AO, McGuire MJ. Adherence and health care costs. Risk Manag Healthc Policy. 2014;7:35-44. doi:10.2147/RMHP.S19801 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Walsh CA, Cahir C, Tecklenborg S, et al. The association between medication nonadherence and adverse health outcomes in ageing populations: A systematic review and meta-analysis. Br J Clin Pharmacol. 2019;85(11):2464-78. doi:10.1111/bcp.14075 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Campagna EJ, Muser E, Parks J, et al. Methodological considerations in estimating adherence and persistence for a long-acting injectable medication. J Manag Care Spec Pharm. 2014;20(7): 756-66. doi:10.18553/jmcp.2014.20.7.756 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.National Committee for Quality Assurance. HEDIS & performance measurement. Accessed March 8, 2022. http://www.ncqa.org/HEDISQualityMeasurement.aspx
- 8.Lafeuille M-H, Frois C, Cloutier M, et al. Factors associated with adherence to the HEDIS quality measure in Medicaid patients with schizophrenia. Am Health Drug Benefits. 2016;9(7):399-410. [PMC free article] [PubMed] [Google Scholar]
- 9.Canfield SL, Zuckerman A, Anguiano RH, et al. Navigating the wild west of medication adherence reporting in specialty pharmacy. J Manag Care Spec Pharm. 2019;25(10):1073-7. doi:10.18553/jmcp.2019.25.10.1073 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Dommasch ED, Lee MP, Joyce CJ, et al. Drug utilization patterns and adherence in patients on systemic medications for the treatment of psoriasis: A retrospective, comparative cohort study. J Am Acad Dermatol. 2018;79(6):1061-8.e1. doi:10.1016/j.jaad.2018.06.053 [DOI] [PubMed] [Google Scholar]
- 11.Doshi JA, Takeshita J, Pinto L, et al. Biologic therapy adherence, discontinuation, switching, and restarting among patients with psoriasis in the US Medicare population. J Am Acad Dermatol. 2016;74(6):1057-65.e4. doi:10.1016/j. jaad.2016.01.048 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Lee S, Xie L, Wang Y, et al. Evaluating the effect of treatment persistence on the economic burden of moderate to severe psoriasis and/or psoriatic arthritis patients in the US Department of Defense population. J Manag Care Spec Pharm. 2018;24(7):654-63. doi:10.18553/jmcp.2018.24.7.654 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Lofland JH, Johnson PT, Ingham MP, et al. Shared decision-making for biologic treatment of autoimmune disease: Influence on adherence, persistence, satisfaction, and health care costs. Patient Prefer Adherence. 2017;11:947-58. doi:10.2147/ppa.S133222 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Long MD, Cohen RD, Smith TW, et al. Retrospective database analysis: Dose escalation and adherence in patients initiating biologics for ulcerative colitis. Dig Dis. 2022;40(5):553-64. doi:10.1159/000521299 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Murage MJ, Anderson A, Casso D, et al. Treatment patterns, adherence, and persistence among psoriasis patients treated with biologics in a real-world setting, overall and by disease severity. J Dermatolog Treat. 2019;30(2):141-9. doi:10.1080/09546634.2018.1479725 [DOI] [PubMed] [Google Scholar]
- 16.Walsh JA, Cai Q, Lin I, et al. Treatment persistence and adherence among patients with psoriatic arthritis who initiated targeted immune modulators in the US: A retrospective cohort study. Adv Ther. 2021;38(5):2353-64. doi:10.1007/s12325-021-01687-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Xu C, Teeple A, Wu B, et al. Drug adherence and persistence of patients with moderate to severe psoriasis treated with biologic medications in a US commercially insured population. Dermatology. 2022;238(3):438-47. doi:10.1159/000519176 [DOI] [PubMed] [Google Scholar]
- 18.Xu C, Teeple A, Wu B, et al. Treatment adherence and persistence of seven commonly prescribed biologics for moderate to severe psoriasis and psoriatic arthritis in a US commercially insured population. J Dermatolog Treat. 2022;33(4):2270-7. doi:10.1080/09546634.2021.1950600 [DOI] [PubMed] [Google Scholar]
- 19.Ding Z, Obando C, Muser E, et al. Real-world persistence, maintenance dosing, and pre-post corticosteroid and opioid use among Crohn’s disease patients with prescription claims for ustekinumab in the USA. Drugs Real World Outcomes. 2021;8(4):565-75. doi:10.1007/s40801-021-00264-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Peterson AM, Nau DP, Cramer JA, et al. A checklist for medication compliance and persistence studies using retrospective databases. Value Health. 2007;10(1):3-12. doi:10.1111/j.1524-4733.2006.00139.x [DOI] [PubMed] [Google Scholar]
- 21.Burden AM, Paterson JM, Gruneir A, et al. Adherence to osteoporosis pharmacotherapy is underestimated using days supply values in electronic pharmacy claims data. Pharmacoepidemiol Drug Saf. 2015;24(1):67-74. doi:10.1002/pds.3718 [DOI] [PubMed] [Google Scholar]
- 22.Hendrix N, Marcum ZA, Veenstra DL. Medication persistence of targeted immunomodulators for plaque psoriasis: A retrospective analysis using a US claims database. Pharmacoepidemiol Drug Saf. 2020;29(6):675-83. doi:10.1002/pds.5021 [DOI] [PubMed] [Google Scholar]
- 23.Stelara. Prescribing information. Janssen Biotech Inc; 2019. Accessed March 8, 2022. https://www.accessdata.fda.gov/drugsatfda_docs/label/2019/761044s003lbl.pdf [Google Scholar]
- 24.Tremfya. Prescribing information. Janssen Biotech Inc; 2020. Accessed March 8, 2022. https://www.accessdata.fda.gov/drugsatfda_docs/label/2020/761061s007lbl.pdf [Google Scholar]
- 25.IBM Watson Health. IBM MarketScan Research Databases for life sciences researchers. 2018. Accessed March 8, 2022. https://www.ibm.com/downloads/cas/0NKLE57Y
- 26.Arnold JG. Pharmacoeconomics: From Theory to Practice. 2nd ed. CRC Press, 2021. [Google Scholar]
- 27.Optum. Clinformatics® Data Mart. 2017. Accessed March 8, 2022. https://www.optum.com/content/dam/optum/resources/productSheets/Clinformatics_for_Data_Mart.pdf
- 28.Humira. Prescribing information. Abbvie Inc; 2021. Accessed March 8, 2022. https://www.accessdata.fda.gov/drug-satfda_docs/label/2021/125057s417lbl.pdf [Google Scholar]
- 29.Taltz. Prescribing information. Eli Lilly and Co; 2021. Accessed March 8, 2022. https://www.accessdata.fda.gov/drug-satfda_docs/label/2021/125521s014lbl.pdf [Google Scholar]
- 30.Schneeweiss S, Avorn J. A review of uses of health care utilization databases for epidemiologic research on therapeutics. J Clin Epidemiol. 2005;58(4):323-37. doi:10.1016/j.jclinepi.2004.10.012 [DOI] [PubMed] [Google Scholar]
- 31.Shah NB, Haydek J, Slaughter J, et al. Risk factors for medication nonadherence to self-injectable biologic therapy in adult patients with inflammatory bowel disease. Inflamm Bowel Dis. 2020;26(2):314-20. doi:10.1093/ibd/izz253 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Centers for Disease Control and Prevention. Prescription drug time and dosage limit laws. Public Health Law. 2015. Accessed March 8, 2022. https://www.cdc.gov/phlp/docs/menu_prescription-limits.pdf [Google Scholar]
- 33.Taitel M, Fensterheim L, Kirkham H, et al. Medication days’ supply, adherence, wastage, and cost among chronic patients in Medicaid. Medicare Medicaid Res Rev 2012;2(3):E1-13. doi:10.5600/mmrr.002.03.a04 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Burden AM, Huang A, Tadrous M, et al. Variation in the days supply field for osteoporosis medications in Ontario. Arch Osteoporos. 2013;8:128. doi:10.1007/s11657-013-0128-1 [DOI] [PubMed] [Google Scholar]